We have shown that near-total α-cell loss in adult mice has little effect on glucagonemia and no apparent effect on β-cell function or glucose homeostasis. This reveals that 2 to 4% of the normal α-cell mass is astonishingly sufficient to produce enough glucagon to ensure glycemic control under basal conditions. Massive α-cell loss is associated with a rapid enhancement of glucagon sensitivity, which allows increased glucose mobilization, even if the decrease in glucagonemia is very mild.
A corollary of this observation is that the normal amount of pancreatic α-cells apparently exceeds the physiologic requirement for proper blood glucose homeostasis. Indeed, the total glucagon found in the pancreas when only 2% of the α-cells remain was sufficient to maintain normoglucagonemia. A simple calculation reveals that if all pancreatic glucagon found 1 week after α-cell loss (~7,400 pg; Supplementary Table 2
) were entirely released at once into the circulation, glucagonemia would be more than 120-fold higher than the normal value of about 60 pg (Supplementary Table 2
). The amount of glucagon measured in plasma after 98% α-cell ablation (~40 pg; Supplementary Table 2
) only represents 0.5% of the residual pancreatic glucagon content produced by the remaining 2% of α-cells (7,400 pg). This suggests that the intracellular glucagon store in the remaining α-cells is, in principle, sufficient to maintain basal glucagonemia. However, because plasma glucagon has a half-life of only 2 min (20
), sustained glucagon biosynthesis after α-cell ablation is mandatory to maintain glucagonemia and pancreatic content over time. Indeed, in absence of glucagon biosynthesis, all pancreatic glucagon stocks found after DT should theoretically be finished within ~6 h vs. ~17 days in control mice with normal α-cell mass (estimated considering the observed pancreatic glucagon contents and glucagon half-life). In agreement with this, we found that although the number of α-cells had not increased 1 month after the ablation, the glucagon content per cell had doubled. In this situation, the lack of response of the α-cell–ablated pancreata to arginine suggests that the few remaining α-cells have reached a maximum secretion capacity or that glucagon biosynthesis is rate limiting, or both. The inability of Glucagon-DTR
mice to ensure normoglucagonemia 1 week after DT, despite having sufficient pancreatic glucagon, indicates that glucagon secretion is limited; however, because two-thirds of normal plasma glucagon level is ensured by just 2% of normal α-cell mass, this indicates that α-cells display a huge secretory capacity. This property may contribute to the excessive glucagon secretion observed in diabetic patients. Alternatively, our observations also suggest that in normal conditions, with a normal α-cell mass, glucagon secretion (relative to α-cell number) is very low.
It is therefore intriguing to observe that α-cells are so numerous. They may somewhat be required for proper endocrine pancreas development. Alternatively, or in addition, their numbers may reflect the effect of selection pressure during evolution: most species must often thrive through long periods of sustained starvation, and glucagon facilitates glucose mobilization during food deprivation. In the latter perspective, the massive loss of α-cells, which we report here, was associated with a rapid enhancement of glucagon sensitivity, which allows an increased glucose mobilization likely through glycogenolysis, rather than gluconeogenesis, as revealed with glucose tolerance tests and a tendency to express higher levels of liver GcgR and GP.
Interestingly, we have seen that extreme α-cell removal in stressful situations, such as prolonged starvation and insulin-induced hypoglycemia, has no obvious consequences on fasting blood glucose and normoglycemia recovery. Furthermore, Unger and colleagues (6
) reported that inhibition of glucagon signaling prevented STZ-induced diabetes in mice. By contrast, we show here that near-total α-cell ablation does not prevent hyperglycemia, indicating that a minimal fraction of the α-cell mass is sufficient to mediate normal glucagon signaling. Together, these combined observations strongly suggest that a total α-cell loss, or glucagon-signaling blockade, would be required to prevent hyperglycemia and diabetes after massive β-cell destruction.
Cell–cell signaling between α- and β-cells is thought to be essential for proper blood glucose homeostasis (4
), but the direct, physical influence of α-cells on glucose-stimulated insulin release has never been studied in vivo. We know β-cell activity is independent of glucagon signaling because it remains unaltered upon treatment with GcgR antagonists or in GcgR−/−
), that is, in situations where α-cells are abundant or expanded. In DT-treated Glucagon-DTR
mice, however, almost all islets are totally devoid of α-cells, which prevents any paracrine intraislet direct interaction between these two cell types. Yet, glucose homeostasis in this situation is not affected under basal conditions, thus suggesting that local α-cell–β-cell interactions are dispensable for adequate β-cell function. This conclusion could not be reached in previous studies because the animal model reported here is the first in which adult mice have nearly normal glucagonemia, yet lack most α-cells.
α-Cell hyperplasia has been reported in patients bearing mutations in the GcgR gene (24
) and in mice exhibiting glucagon deficiency, such as in glucagon−/−
, and PC2 −/−
mice, as well as in wild-type mice treated with GcgR-neutralizing antibody (7
). In all these situations, the formation of new α-cells likely represents a compensatory response to deficient glucagon signaling.
The origin of newly formed α-cells in glucagon-deficient conditions is not known. The continuous emergence of α-cells from pancreatic ductal Neurog-3
–expressing cells was reported in mice expressing the β-cell–specific transcription factor paired box gene 4 (PAX4) in embryonic α-cells (30
). In these animals, α-cells reprogram toward the β-cell phenotype upon PAX4 misexpression, leading to a substantial reduction in postnatal α-cell numbers. The authors proposed that glucagon deficiency was the driving force for the continuous appearance of α-cells, yet glucagonemia was not measured in their study, and the α-cell deficit was similar to the one observed here in DT-treated Glucagon-DTR
mice. By contrast, there is no evidence for ductal α-cell neogenesis in Glucagon-DTR
mice: the rare α-cells found after DT were never preferentially located within or near ducts. This difference may be attributed to the perturbed metabolic status of mice over-expressing PAX4, which exhibit strong β-cell hyperplasia and very high insulinemia, likely resulting in an abnormally elevated insulin-to-glucagon balance.
We recently reported that adult α-cells reprogram to insulin production after extreme β-cell ablation (13
). Here, we explored whether β-cells can reprogram to produce glucagon after massive α-cell loss and failed to observe any such β- to α-cell conversion. However, the interconversion between α- and β-cells may occur in both directions when conditions are appropriate. Indeed, it has been shown that 1
) the ectopic expression of the α-cell–specific transcription factor ARX or 2
) the conditional inactivation of the DNA methyltransferase gene Dnmt1
in embryonic β-cells converts them to the α-cell phenotype (31
After β-cell destruction in RIP-DTR
mice, ~5% of adult α-cells spontaneously undergo cell reprogramming; this limited number of cells probably determines the extent of β-cell recovery in RIP-DTR
transgenic mice (13
). This observation alone makes us speculate that promoting the conversion of most α-cells to β-cells, if it could be achieved, might represent an attractive therapeutic strategy for diabetes, once we learn how to modulate autoimmunity. In this regard, it was unclear what the minimal α-cell mass is to ensure enough glucagon production should α-cell reprogramming become a therapy to treat diabetes. The present work reveals that 2 to 4% of functional α-cells is sufficient to guarantee the glycemic control, at least under basal conditions. Reduction of the α-cell pool by reprogramming to β-cells could have the additional beneficial effect of preventing the glucagon excess typical of type 1 diabetes, which enhances the risk for hyperglycemia and ketoacidosis in patients afflicted with the disease (5