The “ductal origin” hypothesis suffered a strong setback in 2004, when Dor and co-workers, using a pulse–chase strategy in a mouse transgenic model (see next section), established that adult islet regeneration occurs through self-replication rather than differentiation from non-insulin-producing pancreatic progenitors (see below). However, a very recent report using a similar lineage-tracing experimental design (in which transgenic mice with the ductal-specific carbonic anhydrase II promoter driving Cre recombinase are mated with floxed beta-galactosidase reporter mice) suggests that ductal cells do indeed give rise to new islets and acini both during normal islet turnover and after injury (ductal ligation). This would be in line with our recent finding that the expression of both Pdx1 and insulin was activated in the ductal epithelium of transplanted human pancreata upon recurrence of autoimmunity. However, these cells still retained a hybrid ductal–beta cell phenotype and might just represent an attempt at compensating for the loss of beta cell mass, possibly stimulated by hyperglycemia and chronic inflammation. In a recent study, Hao et al. explored the ability of non-endocrine epithelial cells from the adult pancreas to give rise to endocrine cells. The pancreas is mostly made of two cell types, namely mesenchymal and epithelial. The latter include ductal, acinar, and islet endocrine cell types. Among the former are pancreatic fibroblasts, endothelial cells, vascular smooth muscle cells, and stellate cells. Mesenchymal cells, in general, tend to take over the culture when pancreatic tissues are plated in conditions that favor adherence. However, treatment with the drug G418 is effective at getting rid of mesenchymal cells. The above investigators cultured the byproduct of islet isolation procedures, which were largely devoid of both endocrine (due to the mechanical separation of the islets) and mesenchymal cells (due to G418 treatment). When co-transplanted with fetal islet-like clusters in recipient immunodeficient mice, some of these CK-19-positive “non-endocrine pancreatic cells” differentiated into insulin-, glucagon-, and (more rarely) somatostatin-producing cells. Both the origin (ductal or acinar?) of the cells with this potential and the nature ( bona fide beta cells, or insulin-positive cellular byproducts?) of the differentiated progeny remain to be ascertained.

Do New Beta Cells Arise from the Islet?

A number of studies have pinpointed the origin of new beta cells to stem cells contained within islets. Thus, using the STZ model of regeneration, Fernandes et al. identified a population of somatostatin + /Pdx1 + cells inside the damaged islets. Follow-up of these cells led to the observation that they ended up turning into insulin-positive cells. These putative precursors were similarly observed in nonobese diabetic mice, where beta cell destruction is mediated by an autoimmune response. Similar findings were later reported by Guz et al., who documented islet regeneration in STZ-treated mice that received supportive insulin administration. Beta cell neogenesis was detected during the first week after the restoration of normoglycemia, and two putative beta cell progenitors were identified (Glut2 + and Ins + /somatostatin + ).

These results appear to be in contradiction with those of Dor et al., who also identified the islet as the source of new islets, but through a completely different mechanism .

Do New Islets Arise from the Bone Marrow?

The migration of transplanted bone marrow cells to many different tissues (particularly in response to insults or pathological conditions) is a phenomenon commonly observed both in animalsand humans. This apparent “transdifferentiation” potential of bone marrow cells led to the early hypothesis that they could be the basis of a universal self-repair mechanism – even if it is not normally active under physiological conditions . However, this idea suffered an important setback in 2002 with the publication of two studies showing that multipotent cells can fuse with differentiated ones, therefore adopting their phenotype. This was the case in a experimental setting where wild-type bone marrow transplantation rescued the liver of FAH −/− mice, which are a model of fatal hereditary type I tyrosinemia. Further investigation on the mechanisms behind the rescue revealed that donor bone marrow cells had migrated to the defective liver and fused with resident cells. The ensuing cells were indistinguishable from the local hepatocytes, but the complementation with the wild-type gene of the fused bone marrow cell resulted in a stronger hybrid with a selective proliferative advantage over the non-fused cells. These “corrected” cells eventually took over the liver, restoring function. The burden of the proof was now on those researchers claiming that bone marrow-derived cells could indeed differentiate into the target derivatives. Thus, Ianus et al. transplanted bone marrow cells from male transgenic INS2-EGFP mice into irradiated wild-type female recipients. Up to 3% of the cells within each islet exhibited EGFP expression 4–6 weeks after the procedure, most of them expressing insulin and Pdx1. This could be explained either by direct transdifferentiation of bone marrow cells into beta cells (which would activate the insulin promoter and therefore the reporter) or fusion to resident cells resulting in reprogramming of the donor ones. To rule out the latter, the authors transplanted the bone marrow of male INS2-Cre mice into ROSA-stoplox-EGFP female recipients. The rationale behind this approach was that any cell fusion event would be manifested by the Cre-mediated removal of the stop codon preventing EGFP expression.

Abundant cells containing the Y chromosome were found in the pancreas of the recipient, but none of them was fluorescent. Since forced in vitro fusion of these two types of genetically modified cells did indeed result in EGFP expression, it was concluded that bone marrow cells can contribute to the endocrine pancreas in a fusion-independent fashion. However, two reports published shortly thereafter found little or no evidence of bone marrow transdifferentiation into pancreatic beta cells. Using again a GFP-labeled donor population, the authors of the first study observed fluorescent cells in the islets of the recipient animals, but none of them co-expressed insulin, either in healthy or in STZ-treated animals. The second group extended these studies to another model of pancreatic regeneration (partial pancreatectomy). Despite substantial contribution of the donor cells to blood, lymphatic, and interstitial cells in the pancreas, they could find only two cells positive for GFP in a screening of more than 100,000 beta cells – which turned out to be in control animals. They concluded, therefore, that the bone marrow does not significantly contribute to the endocrine component of the pancreas. A third study confirmed these findings but provided additional evidence that bone marrow-derived endothelial progenitor cells were recruited to the pancreas in response to islet injury, which could be theoretically exploited to improve vascularization and/or endogenous regeneration of injured beta cells.

If bone marrow (BM) cells contributed to islet regeneration, BM derived from GFPpositive donor mice could be tracked upon transplantation into wild-type animals and found in the recipient’s islets. However, this approach does not account for cell fusion.

An alternative approach to rule out cell fusion is the transplantation of BM cells from Ins-Cre mice into recipients in which GFP will not be expressed unless there is a Cre-mediated excision of a stop codon. Cells with a Y chromosome that express insulin within the islets would provide evidence of BM-mediated regeneration. If cell fusion occurred, GFP-positive cells would be detected. In the absence of GFP fluorescence, it could be concluded that the observation is not due to cell fusion.

Also, ES cells transfected with Pax4 were induced to express insulin at much higher levels than untransfected controls. Recent evidence indicates that Pax4 and Arx are mutually repressed, and that the balance between the two determines a ( Arx ) or beta ( Pax4 ) specification from Ngn3 + progenitors. Nkx2.2

The NK2 homeobox 2 (Nkx2.2) gene belongs to a family of genes involved in the differentiation of many tissues.Nkx2.2 expression is observed both in the ventral CNS and in mature alpha, beta, and PP cells. The gene is activated from very early on during pancreatic specification ( ~ e8), but – in a pattern similar to that of Pdx1 – becomes restricted to specific islet cell types later in development. Nkx2.2 mutant mice lack beta cells and have lower amounts of other islet endocrine cell types. Further analyses of these islets, however, show a sizeable population of hormone-negative undifferentiated cells. While Pdx1 expression remains largely unaffected in Nkx2.2 knockouts at the onset of pancreatic specification ( ~ e8.5), the relative strength of its signal is significantly reduced during the secondary transition ( ~ e13.5 onward). This led to the hypothesis that Nkx2.2 might be required for the terminal differentiation of beta cells.

Nkx6.1

Nkx6.1 is another member of the NK homeodomain family, which can be found both in the pancreas and the CNS.  Nkx6.1 expression is first detected from e10.5 in the pancreatic epithelium. At e15.5, it colocalizes either with Pdx1 + cells (postmitotic beta cells, mostly) or with Ngn3 (immature progenitors). At later developmental stages, as well as in the adult organ, Nkx6.1 is exclusively restricted to beta cells. Knockouts display a marked deficiency of beta cells, which could be traced back to the initiation of the secondary transition, during which beta cell neogenesis was severely impaired. Nkx2.2 expression was not affected, suggesting that this gene is upstream of Nkx6.1. This was further confirmed by double Nkx2.2/Nkx6.1 knockout experiments, whose phenotype was identical to that of Nkx2.2 −/− .

Pax4

Paired box-containing gene 4 ( Pax4 ) belongs to a multigene family that shares a conserved motif, termed “paired box.” Both Pax4 and Pax6 have, in addition, a homeodomain. Pax4 expression is first detected in the pancreas at around e9.5, and is later restricted to beta cells. Knockout mice lack both beta and delta cells, but alpha cells appear to compensate for their absence with a much higher than normal representation in the islet, which is suggestive of a “default” alpha cell differentiation pathway. Several lines of evidence point at Pax4 as a direct downstream target of Ngn3. As discussed in the main text, the balance between Pax4 and Arx might be critical for the adoption of alpha or beta cell fates.

Pax6

Perhaps best known for its role in eye development, Pax6 is also expressed both in the developing pancreas (scattered throughout the fore/midgut cells at e9.5; colocalizing with arising alpha cells at e9.5; with alpha or beta cells at e15.5) and with alpha , beta, or delta cells in the adult pancreas – but not in acinar tissue. Knockout mice show both a very significant reduction in all hormone- producing cells (with alpha cells largely absent) and an abnormal distribution of the remaining ones within the islet.

Arx

The aristaless-related homeobox gene ( Arx ) contains both a C-terminal stretch of amino acids known as the OAR/aristaless domain and a prd-class homeobox domain. Like many other genes involved in pancreatic development, Arx was first identified in the mouse CNS. Arx expression is observed in the proliferating epithelium of the pancreatic anlage at e9.5, and then is progres sively restricted to endocrine cell types. Ngn3 knockouts do not express Arx, indicating that the former is upstream of the latter. Arx −/− mutants lack alpha cells and display a concomitant increase in delta cells (and, to a lesser extent, beta cells).This effect is exacerbated in double Pax4/Arx mutants, where both alpha and beta cell subsets are largely replaced by delta cells. Arx, however, does not seem to be required for early alpha cell differentiation, because glucagonpositive cells can be readily detected until e12.5 in mutant embryos, which suggests that the main role of Arx is during the secondary transition. Taken together, these observations support a model where Arx promotes alpha cell differentiation and antagonizes that of beta cells (from e12.5 onward) and delta cells (from e14.5 onward). Because Pax4 has an opposing activity, the relative levels of each protein will likely dictate cell fate after e12.5 .

A model for alpha , beta, and delta cell specification. Experimental evidence suggests that the first insulin-positive and glucagon-positive cells that appear prior to the secondary transition are a developmental dead end. Relative levels of the mutually repressing proteins Arx and Pax4 will define two populations of endocrine progenitors during the secondary transition. Both are characterized by the expression of Nkx2.2 and Pax6. However, Pax4 expression is only detected in those cells that will give rise to both beta and delta cells. The effect of knocking out either Pax6 or Pax4 is indicated in dotted lines .

MafA/MafB

Maf transcription factors (containing a basic leucine zipper) have been associated with the regulation of multiple differentiation processes. The best-characterized Maf factors expressed in the pancreas are MafA and MafB, which are preferentially expressed in beta and alpha cells, respectively. The role of these factors has been difficult to ascertain. MafA knockout mice, for instance, display glucose intolerance and develop diabetes; however, there are no developmental effects associated with the mutation. Recent evidence suggests that a switch from MafB to MafA might be critical for the embryonic maturation and prolonged survival/function of beta cells. However, MafB has also been recently shown to be essential for the appropriate regulation of Pdx1, Nkx6.1, and GLUT-2 in the final stages of maturation of beta cells.

Foxa2 (HNF3beta)

The winged-helix transcription factor Foxa2 (also known as hepatocyte nuclear factor 3beta [HNF3beta]) is one of the first true markers of definitive endoderm. Despite its expression throughout the development of the pancreas, not many studies have looked into its role in this process. These studies are chiefly based on conditional loss of function. Thus, selective abrogation of Foxa2 expression in beta cells results in insulin hypersecretion, leading to profound hypoglycemia. In fact, Foxa2 has been shown to control Pdx1 expression in mature beta cells. However, the onset of Pdx1 expression and pancreatic specification is not affected in mice where the Foxa2 has been selectively inactivated in the developing endoderm. These animals, however, are severely hypoglycemic and die within days of birth, due to a dramatic reduction (>90%) in the number of alpha cells. They show impaired maturation, but not initial specification of alpha cells.