Pancreatic α-Cell Specific Deletion of Mouse Arx Leads to α-Cell Identity Loss

The specification and differentiation of pancreatic endocrine cell populations (α-, β-, δ, PP- and ε-cells) is orchestrated by a combination of transcriptional regulators. In the pancreas, Aristaless-related homeobox gene (Arx) is expressed first in the endocrine progenitors and then restricted to glucagon-producing α-cells. While the functional requirement of Arx in early α-cell specification has been investigated, its role in maintaining α-cell identity has yet to be explored. To study this later role of Arx, we have generated mice in which the Arx gene has been ablated specifically in glucagon-producing α-cells. Lineage-tracing studies and immunostaining analysis for endocrine hormones demonstrate that ablation of Arx in neonatal α-cells results in an α-to-β-like conversion through an intermediate bihormonal state. Furthermore, these Arx-deficient converted cells express β-cell markers including Pdx1, MafA, and Glut2. Surprisingly, short-term ablation of Arx in adult mice does not result in a similar α-to-β-like conversion. Taken together, these findings reveal a potential temporal requirement for Arx in maintaining α-cell identity.


Introduction
During development, the pancreas organizes into two distinct compartments: the exocrine acinar cells, which secrete digestive enzymes, and the hormone producing endocrine cells organized into islets of Langerhans [1]. These islets contain a core of insulinproducing b-cells with a surrounding mantle of a, d, e, and PPcells, which produce the hormones glucagon, somatostatin, ghrelin, and pancreatic polypeptide, respectively [2]. Islet band a-cells are the two key endocrine cell populations involved in maintaining glucose homeostasis [3]. Disruption of this homeostasis through b-cell loss or dysfunction leads to diabetes mellitus, a common metabolic disorder manifested at all ages.
Given the limited supply of functioning b-cells in diabetics, one potential treatment avenue is cell-replacement therapy [4]. Considerable effort has been invested in identifying alternative b-cell sources through either directed differentiation from embryonic/induced pluripotent stem cells or reprogramming from other differentiated cell types [5]. Due to the close lineage relationship between aand b-cells, the reprogramming potential of an a-cell to adopt a b-cell fate has been recently investigated [3]. In one study, new b-cells were generated from glucagon-producing a-cells through a glucagon + insulin + bihormonal intermediate state after a near-total b-cell loss [6]. Moreover, an a-to-b-cell lineage conversion was observed when Pax4, a pro-b-cell transcription factor, was expressed in pancreatic endocrine progenitors or acells [7]. Similarly, forced expression of Pdx1 in endocrine progenitors leads to an increase in b-cells and a decrease in acell number [8]. Although the a-cell population is mostly postmitotic, these studies collectively illustrate that a-cell fate can be plastic and is able to be reprogrammed to adopt b-cell fate. However, the extent of this plasticity during different stages of an animal's life is currently unknown.
One transcription factor capable of altering plasticity in endocrine cells is the Aristaless-related homeobox gene (Arx). In the mouse pancreas, Arx is expressed in a subset of endocrine progenitors and then restricted to glucagon-producing a-cells where it is expressed throughout the life of the animal [9,10]. When misexpressed in the developing pancreas, Arx is sufficient to force endocrine progenitors or b-cells to adopt an a-cell fate [11]. These results demonstrate that Arx is sufficient for b-to-a-cell reprogramming during development.
Although much is known regarding factors necessary and sufficient for endocrine development, the factors required to maintain the identity of mature a-cells during different stages are less clear. Mice with Arx null mutations in the germ-line, pancreatic progenitors, or endocrine progenitors all display a complete loss of a-cells with a concurrent increase in band d-cells in the pancreas [9,10,12]. Moreover, a-cell loss has been reported in patients with null mutations in ARX [13]. However, none of the existing mouse models are suitable for determining the function of Arx in maintaining (as opposed to establishing) mature a-cell identity. Further, lineagetracing experiments have not yet been performed to determine if loss of Arx leads directly to an a-to-b-cell conversion.
Here we show that Arx is required for a-cell lineage maintenance in the neonatal pancreas, but not in the adult pancreas. During the neonatal period, ablation of Arx results in loss of glucagon expression and activation of insulin and b-cell markers through an insulin + glucagon + bihormonal intermediate. In contrast, short-term Arx ablation in the adult pancreas does not result in either a loss of glucagon expression or an activation of b-cell marker expression. These data suggest that Arx may act in a stageand context-specific manner in maintaining a-cell identity and reveal potential differential plasticity between fetal and adult acells. When taken together, these findings have important implications for the potential use of a-cells for the purpose of bcell replacement therapy.

Ethics Statement
The Children's Hospital of Philadelphia's Institutional Animal Care and Use Committee (IACUC) approved all animal experiments under the protocol number 2011-10-756. CLM monitored all animal studies.

Animals and Breeding Strategy
The derivation of the Arx L / Y and Glucagon-Cre transgenic lines has previously been described [14,15,16]. To generate Arx L / Y ;Glucagon-Cre mice, Arx L / + ;Glucagon-Cre and Arx L / Y mice were mated on a BL6 background. Male and female Arx L / Y or Arx L / L; Glucagon-Cre mice were phenotypically indistinguishable in terms of their islet morphology, size, body size and weight. All mutants used in our analysis were compared to their sex-matched controls. Arx + / Y ;Glucagon-Cre, Arx + / Y , Arx L / + and Arx L / + ;Glucagon-Cre mice were used for controls with no observable phenotypic differences in the islets between any of them. The reporter Rosa26 YFP / YFP was mated into this line in either heterozygosity or homozygosity for lineage tracing studies, which yielded the same result in all experiments [17]. The generation of pCAGG-CreER animals has been previously described [18]. Arx L / Y or Arx L / L ;pCAGG-CreER animals were generated by crossing Arx L / + ;pCAGG-CreER females to Arx L / Y males. Male and female mutants were phenotypically indistinguishable in the endocrine pancreas and both were used in this study.

Immunohistochemistry and Histology
All dissections were performed in cold 16PBS and tail or toe snips collected for genotyping. Tissues were fixed in cold 4% paraformaldehyde overnight at 4uC, embedded in paraffin, and 8 mm sections collected. Antigen retrieval was performed in 10 mmol citric acid buffer (pH 6.0) and endogenous peroxidase, avidin D, and biotin activity blocked with 3% H 2 O 2 (Sigma) and Avidin/Biotin Blocking Kit (Vector), respectively. Endogenous protein was blocked with CAS-Block reagent (Invitrogen). Slides were incubated in primary antibody overnight at 4uC. Primary antibodies used were: Insulin (MS 1:400, Thermo Scientific and GP 1:1000, Abcam), Glucagon (1:1000, Millipore), and Chromogranin A (1:3000, DiaSornin). After rinsing in PBS, appropriate secondary antibodies were added for two hours at room temperature. Immunohistochemical detection was performed with the VECTASTAIN ABC kit (Vector Laboratories) and diaminobenzidine tetrahydrochloride (DAB) as the substrate. Immunofluorescence utilized secondary antibodies conjugated to Cy3, Cy2 or Cy5. All images were obtained using a Leica DM6000B microscope.

Real-Time PCR Analysis
Total RNA was extracted in TRIZOL (Invitrogen) using the protocol provided with reagent. Oligo-dT, Superscript II, and additional required reagents were used to synthesize cDNA (Invitrogen). PCR reactions were performed using Brilliant SYBR Green PCR Master Mix (Agilent) in the Stratagene Mx3005P realtime PCR machine. All PCR reactions were performed in duplicate for each sample with at least 3 animals per group analyzed with reference dye normalization. Primer sequences are available upon request.

Hormone Cell Quantification
Hormone-positive cells from pancreatic sections were counted, averaged, and normalized to either total pancreatic area or total endocrine cell number. Three separate regions of each pancreas were used for quantification in both control and mutant mice. At least three animals for each group were used for quantification in all analyses. To determine hormone cell mass, hormone-positive area as well as pancreatic area was measured using the Aperio Image Analysis System. These areas as well as weight of the pancreas was used to determine hormone cell mass. For specific hormone cell number, hormone positive cells were counted and normalized to total endocrine cell number, which was determined by combining counts for all endocrine hormones (insulin, glucagon, somatostatin, and PP). Over 10,000 total endocrine cells were counted for each analysis consisting of over 5,000 insulin + , over 1,000 glucagon + and somatostatin + , and over 500 PP + cells.

Islet Isolation
For P21 and adult RNA analysis, islet isolation was performed by injecting 5 mL Collagenase P (Roche) in HBSS with 0.02% BSA (Sigma) into the clamped pancreatic duct to inflate the pancreas. Once inflated and removed, pancreatic tissue was incubated in 15 mL CollagenaseP/HBSS at 37uC at 50 rpm for 16 minutes to digest exocrine tissue. After spin down and rinse, islets were isolated from remaining exocrine tissue in HBSS. Upon isolation, islets were placed in TRIZOL for RNA extraction.

Tamoxifen Induction
Two-month-old male and female mice, matched with littermate controls, were injected intraperitoneally (IP) with 50 mg/g body weight of 10 mg/ml tamoxifen (Sigma) solution, which consisted of 10% ethanol and 90% sunflower seed oil (Sigma). Injections were performed for three consecutive days followed by a two-week chase. After the chase period, pancreatic tissue was removed and processed for either immunostaining or RNA analysis.

Statistical Analysis
All values are presented as average 6 standard error of the mean. Significance was determined using a two-tailed Student's ttest. p-values less than or equal to 0.05 were considered significant.

Arx Removal in Neonatal Glucagon-producing Cells
Arx is expressed in endocrine progenitors, a-cell precursors, and mature glucagon-producing cells of the pancreas [9,10]. To investigate its role in the neonatal a-cell, we generated mice with Arx ablation in glucagon + cells (Arx L / Y or Arx L / L ;Glucagon-Cre; referred to as GKO hereon). First, a Rosa-YFP reporter was used to assess the Cre-mediated recombination efficiency in our GKO model. In P5 control mice (including Arx + / Y ;Glucagon-Cre, Rosa-YFP and Arx L / + or + / + ; Glucagon-Cre;Rosa-YFP), Arx expression was found in all glucagon + cells; however, only 13% of glucagon + cells co-expressed Arx and YFP (Fig. 1A, C; white bar in control). All male and female control animals utilized in these experiments were phenotypically identical according to their islet morphology and were compared to their sex-matched GKO animals. These observations indicate a low Cre-mediated recombination rate, which is in agreement with what others have previously reported using this Glucagon-Cre transgenic mouse [19]. In GKO;Rosa-YFP mutant mice, Arx protein was removed in all glucagon + YFP + cells, which equated to about 12% of glucagon + cells (Fig. 1B, C; grey bar in GKO). This result suggests that YFP expression faithfully marks cells that have undergone Arx ablation. The low Cremediated recombination frequency was also observed in P21 animals ( Fig. S1 and data not shown). Real-time PCR analysis from P5 and P21 animals further showed Arx mRNA levels were decreased by 20% and 50% in GKO animals, respectively, although not significantly (Fig. 1D). In addition, in P5 control animals, while 90% of YFP + cells co-expressed glucagon, approximately 10% of the YFP + cells were positive for insulin staining due to some leakiness of the Cre (see below and data not shown).

Arx Ablation in the GKO Mice Results in an Emergence of Glucagon + Insulin + Co-Expressing Cells
To determine whether loss of Arx in the a-cells of GKO mice may have resulted in a change of cell fate in a small subset of cells, we performed double immunostaining for glucagon and other endocrine hormones. Given that only 12% of the glucagonproducing cells have lost Arx expression, we did not expect to observe any significant changes in the number of glucagon cells or the localization of these cells in P5 GKO mice. Indeed, immunostaining analyses confirmed this anticipated result ( Fig. 2A-H). Real-time PCR analysis also revealed no significant changes in the mRNA levels of glucagon transcript between control and GKO mice (Fig. 2I). Although hormone cell numbers were not significantly altered, close examination revealed a small population of glucagon + insulin + bihormonal cells in the pancreas of GKO mice (Fig. 2B). These bihormonal cells were only found in the GKO mice. There was no overlap or significant differences in the expression of glucagon with somatostatin or PP between P5 GKO and control mice (Fig. 2C-F). Endocrine cells expressing glucagon and ghrelin have been reported in the developing and neonatal pancreas [20,21,22]. The number and location of these glucagon + ghrelin + cells were comparable between P5 control and GKO mice (Fig. 2G-H). Real-time PCR analysis did not reveal any significant changes in hormone expression between P5 control and GKO animals ( Fig. 2I-L). Together, this data indicates that loss of Arx in glucagon + cells results in misexpression of the b-cell hormone insulin in glucagon + a-cells.
To determine the fate of this bihormonal population we evaluated glucagon, insulin, somatostatin, PP, and ghrelin expression in the pancreata of P21 control and GKO mice by immunostaining. Again, there was no significant change in the cell mass or distribution associated with these endocrine populations ( Fig. S2A-B). Interestingly, while a few glucagon + insulin + cells could still be found in the P21 GKO mice, the frequency of this bihormonal population was dramatically reduced by this age relative to P5. Instead, many of the remaining bihormonal cells in P21 GKO mice have reduced glucagon staining in cells readily expressing insulin (Fig. S2A-B). These cells are likely in the later stage of their a-to-b-like cell fate conversion. These observations suggest that upon ablation of Arx, insulin expression is activated in the glucagon-producing a-cell, which then gradually loses glucagon expression. Finally, we evaluated mRNA levels for bcell (Pdx1 and Nkx6.1) and a-cell (MafB and Brn4) markers in the islets isolated from P21 control and GKO mice. From the realtime PCR analysis, we observed a significant upregulation of Pdx1 mRNA and an upward trend of Nkx6.1 levels (Fig. S2J). Conversely, we detected a significant reduction in Brn4 with a small downward trend in MafB expression (Fig. S2J). While the significant changes in the mRNA levels of Pdx1 and Brn4 data is surprising in the context of the small changes in hormone expression, this result could be due to direct Arx regulation. Arx could potentially directly repress Pdx1 and activate Brn4, which would result in a drastic increase of Pdx1 (and resulting decrease of Brn4) upon Arx ablation. When taken together, these data suggest that glucagon-producing cells require Arx to maintain a-cell identity and repress b-cell markers during neonatal life.

Arx-deficient Cells Fail to Maintain a-cell Identity
To directly determine the origin of the glucagon + insulin + cells seen in P5 GKO mice, lineage-tracing studies were performed in P5 GKO mice. Triple-immunostaining for glucagon, insulin, and YFP were performed in the pancreas of control;Rosa-YFP and GKO;Rosa-YFP mice. YFP expression was detected in only a subset of glucagon-producing cells at P5 (Fig. 3A-B), due to the low frequency of the Cre-mediated recombination in the Glucagon-Cre transgenic mice (Fig. 1). The majority of YFP + cells in the P5 control mice expressed glucagon (Fig. 3A, E and F) though a very small number of insulin-producing cells positive for YFP expression were found (blue), demonstrating relatively high, though not 100%, fidelity of the Cre-mediated recombination ( Fig. 3E and F). In P5 GKO mice, we noticed an emergence of glucagon + insulin + YFP + (purple) cells and an increase in the number of insulin + YFP + (blue) cells while the number of glucagon + YFP + (red) cells was reduced compared to controls ( Fig. 3A-B, E and F).
To follow up with our previous observations that the glucagon + insulin + cell number has dramatically reduced by P21, we evaluated the pancreata of control;Rosa-YFP and GKO;Rosa-YFP mice at P21 for glucagon, insulin, and YFP expression. Interestingly, corresponding to the previously described disappearance of bihormonal cells by P21 (Fig. S2A-B), the majority of YFP + cells in GKO;Rosa-YFP pancreata at this stage were insulin + (blue) with only a small percentage of YFP + cells expressing both insulin and glucagon (purple) or glucagon alone (red) (Fig. 3C-D  and G). These data demonstrate that Arx loss in glucagonproducing a-cells leads to a failure in maintaining a-cell identity and a conversion to a b-cell-like fate. Taken together, these lineage-tracing data indicate that the loss of neonatal Arx in glucagon-producing cells results in a cell fate conversion from a glucagon + a-cell into an insulin + b-cell-like fate through a bihormonal intermediate.

Markers Associated with Mature b-cells are Activated in Arx-deficient YFP + Cells
To further examine how closely these newly converted b-likecells were to true b-cells, expression of several known b-cell markers including Glut2, MafA, and Pdx1 were examined in control;Rosa-YFP and GKO;Rosa-YFP mice at P5 and P21. In P5 GKO;Rosa-YFP animals, there was a significant increase in the number of YFP + cells coexpressing Glut2, MafA, or Pdx1 (Fig. 4B, D, F, G, I, K). Similar increases were also seen in P21 mice with a further increase in the number of YFP + cells expressing Glut2 and Pdx1 in the pancreata of the GKO;Rosa-YFP mice (Fig. 4H, J, L,  Fig. S3). As expected, due to the leakiness of the Glucagon-Cre transgene, we did find a small, but not significant, percentage of YFP + cells that coexpressed Glut2, MafA, or Pdx1 in P5 or P21 control;Rosa-YFP mice (Fig. 4A, C, E, G, I, K). Taken together, these data demonstrate that a subset of the converted cells in the GKO;Rosa-YFP mice activate b-cell markers as well as insulin expression in the absence of Arx.

Short-term Ablation of Arx in Adult a-cells does not Lead to Loss of a-cell Identity
To explore the requirement for Arx in the maintenance of adult a-cell fate, we used a global, tamoxifen-inducible transgenic mouse model to ablate Arx in adult animals. Two-month-old control and Arx L / Y ;pCAGG-CreER (IKO) mice were injected with tamoxifen for three consecutive days and the animals sacrificed two weeks later for tissue analysis (Fig. 5A). Efficiency of Arx removal was evaluated by immunostaining in control and IKO mice. While Arx expression was found in glucagon + cells in control animals, all glucagon cells in IKO animals have lost Arx expression (Fig. 5B, B', C, C'; marked by arrows). To determine the impact of shortterm Arx ablation in adult a-cells, gene expression and immunostaining for endocrine hormones were examined in control and IKO animals (Fig. 5D-I). Real-time PCR analysis revealed no significant changes in the mRNA levels of hormone genes between control and IKO islets (Fig. 5K). We also did not detect any significant changes in the numbers of glucagon-, insulin-, somatostatin-, and PP-producing cells in the IKO mice compared to controls (Fig. 5D-J). Unlike P5 GKO mice in which a large proportion of Arxglucagon + insulin + cells were found ( Fig. 2A-B), we detected only a small number of bihormonal cells in adult IKO mice, which were not proportionally significant (less than 0.1%; data not shown). Additionally, analysis of aand b-cell factors including MafB, Brn4, Glut2, and Pdx1 also did not reveal any significant changes in transcriptional profile of IKO animals (Fig. 5L). Since the pCAGG-Cre is globally expressed, the IKO animals develop an intestinal phenotype that excludes any meaningful analysis to explore the long-term impact of Arx on adult a-cell (data not shown). Taken together, using our current mouse model with short-term Arx ablation, these findings demonstrate that Arx is likely dispensable in maintaining a-cell identity in adult mice. Future experiments utilizing an inducible acell specific Cre transgenic mouse will be required to study the long-term requirement for Arx in the maintenance of a-cell fate.

Discussion
This study demonstrates a requirement of Arx in a-cell fate maintenance. Ablation of Arx in neonatal glucagon + cells results in a loss of a-cell identity and conversion into an insulin-producing bcell-like fate (Fig. 6). Conversely, short-term ablation of Arx in adult animals did not result in a significant loss or conversion of acells or an increase in b-cells or b-cell markers (Fig. 6). Our findings from neonates and adults expand the previously defined role of Arx in the specification of a-cells. When taken together, Arx plays a role during specification as well as during early maintenance of a-cell fate, but appears not to be required in adult animals for its fate maintenance.
Others have shown that ablation of Arx at any stage of specification results in a complete loss of the a-cell lineage with a concomitant increase of band d-cells [9,10,12,23]. As there is no change in total endocrine mass reported in these studies, these acells likely undergo re-specification into band d-cell lineages. Our findings add significant support to these observations by using lineage tracing to directly demonstrate that Arx-ablated a-cells convert to b-like cells in neonatal animals. Interestingly, our data, while showing coexpression of glucagon and insulin, does not show coexpression of glucagon and somatostatin. Our initial hypothesis was that Arx-deficient a-cells would give rise to both somatostatin and insulin populations. It is possible that the inefficiency of the Glucagon-Cre did not enable us to detect a rare population of glucagon + somatostatin + cells. Alternatively, as the animal ages, the plasticity of cells among different endocrine fates could be altered.
Previous studies have demonstrated that endocrine cell fate is relatively undifferentiated during gestation such that ablation of single transcription factors results in loss of cell fate [2]. As endocrine cells mature, however, this plasticity drastically decreases, and more extreme measures are needed to convert one endocrine cell type to another [4,6,8]. While Pdx1 is normally restricted to b-cells, early overexpression in a-cells results in a postnatal loss of glucagon-expressing a-cells with a concomitant gain of insulin-producing b-cells demonstrating an a-to-b-cell fate conversion [8]. Conversely, overexpression of Pdx1 in adult a-cells does not result in a similar conversion; instead, these cells maintain proper cell identity [8]. The potential temporal requirement of Arx closely parallels the results obtained through Pdx1 overexpression in a-cells. Early in development, endocrine cell fate appears more plastic and subject to reprogramming. During later life, however, cell fate is more defined, and as a result, reprogramming is more difficult to achieve.
Interestingly, although overexpression of Pdx1 in a-cells results in a gain of insulin-producing cells, those cells did not appear to lose all markers of a-cell fate [8]. Examination of immunostaining for the expression of b-cell-specific factors in the Arx ablated neonatal animals demonstrates that YFP + cells are at least partially reprogrammed with the expression of Glut2, MafA, and Pdx1. While Arx is necessary to maintain a-cell fate during development, loss of Arx, even immediately after specification, may not be sufficient to fully reprogram cells into a functional b-cell fate. Due to the low efficiency of the Cre utilized in our study, functional analysis of the insulin-producing cells derived from Arx ablation in a-cells was not feasible.
As the animal ages, there could be epigenetic changes that have occurred during the process of specification or maturation that inhibit these Arx deficient cells from becoming functional b-cells under homeostatic conditions. In fact, epigenetic modification has been shown to play important roles in the differentiation and maintenance of cell types. A recent study demonstrates that a-tob-cell reprogramming could be promoted by manipulating the histone methylation signature in mammalian pancreatic islets [24]. Conditions of stress, however, may also make cell fate transitions more fluid. It has been shown that excessive loss of b-cell mass, induced by administration of a b-cell specific toxin, results in spontaneous reprogramming of a-cells into a b-cell fate [6]. Additionally, partial pancreatectomy in mice and rats can result in regeneration of b-cells through the conversion of duct cells or duct progenitor cells [25,26]. These studies demonstrate that while cell fate is more defined in adult animals, extreme conditions can force a non-b-cell into a b-cell fate. Future studies examining this possibility should be performed and will elucidate limits to cell fate maintenance in adult animals and how to overcome those limits. Particularly, the ability to utilize a-cells for conversion to functional b-cells could be a potential therapy for diabetes.
Finally, it is important to note that our current adult IKO mouse model does not allow for a complete investigation for the role of Arx in adult a-cells. Arx is required in early enteroendocrine cell development of the digestive tract [27]. Therefore mice with Arx removal in the intestine have alterations of specific enteroendocrine cell population, which lead to lipid malabsorption and diarrhea ( [27]; and unpublished observations). Since the adult IKO mouse model was generated using a global inducible Cre transgenic mouse, enteroendocrine cell populations were impacted (unpublished observations). It is important to note that we did notice a 0.1% increase in the number of bihormonal cells in IKO mice (data not shown). However, whether this small change is due to the direct impact upon Arx loss in a-cells or changes in the animal's physiology remains to be determined. An a-cell specific inducible Arx-deficient mouse model combined with lineage tracing studies will be required to precisely determine the role of Arx in adult a-cells.
In conclusion, the current study demonstrates a potential temporal requirement for Arx in maintenance of a-cell fate. Ablation of Arx in neonatal a-cells results in a loss of glucagon expression and a conversion of this cell population to adopt an insulin-producing b-cell-like fate. However, short-term loss of Arx in adult animals does not phenocopy this result but instead suggests that Arx is dispensable in maintaining a-cell fate in adulthood. These data expand the knowledge of the field not only related to the role of Arx in the endocrine a-cell but also in regards to global temporal restrictions for reprogramming endocrine cells. Future studies examining this temporal requirement, as well as perturbations to the cell that circumvent these restrictions, will help clarify this plasticity and bring understanding to endocrine cell fate specification, maintenance, and therapeutic potential.