Peptides in the regulation of glucagon secretion

Glucose homeostasis is maintained by the glucoregulatory hormones, glucagon, insulin and somatostatin, secreted from the islets of Langerhans. Glucagon is the body's most important anti-hypoglycemic hormone, mobilizing glucose from glycogen stores in the liver in response to fasting, thus maintaining plasma glucose levels within healthy limits. Glucagon secretion is regulated by both circulating nutrients, hormones and neuronal inputs. Hormones that may regulate glucagon secretion include locally produced insulin and somatostatin, but also urocortin-3, amylin and pancreatic polypeptide, and from outside the pancreas glucagon-like peptide-1 and 2, peptide tyrosine tyrosine and oxyntomodulin, glucose-dependent insulinotropic polypeptide, neurotensin and ghrelin, as well as the hypothalamic hormones arginine-vasopressin and oxytocin, and calcitonin from the thyroid. Each of these hormones have distinct effects, ranging from regulating blood glucose, to regulating appetite, stomach emptying rate and intestinal motility, which makes them interesting targets for treating metabolic diseases. Awareness regarding the potential effects of the hormones on glucagon secretion is important since secretory abnormalities could manifest as hyperglycemia or even lethal hypoglycemia. Here, we review the effects of each individual hormone on glucagon secretion, their interplay, and how treatments aimed at modulating the plasma levels of these hormones may also influence glucagon secretion and glycemic control.


Introduction
More than 150 years ago, the endocrine structures of the pancreasthe islets of Langerhans -were discovered. The islets contain several different hormone secreting cell types, whose primary function is to regulate metabolism and maintain glucose homeostasis. Among these cells is the α-cell, responsible for the secretion of glucagon -a hyperglycemic hormone that acts upon hepatocytes to stimulate during fasting the release of glucose derived from stored glycogen. For many years, insulin was the protagonist in type 2 diabetes (T2D), until Unger and Orci presented the bihormonal hypothesis and suggested that hyperglycemia in T2D was a result of the combined effect of hypoinsulinemia and hyperglucagonemia [1]. This for a while turned the attention towards the glucagon secreting α-cell. Now, almost 50 years after the bihormonal hypothesis was first proposed, it is finally clear that hyperglucagonemia plays an important role in the pathogenesis of diabetes by contributing to hyperglycemia [2], with some studies even suggesting that dysregulated glucagon secretion, rather than insufficient insulin secretion, is the primary cause for hyperglycemia in diabetes [3].
However, in spite of the diabetogenic effects of hyperglucagonemia, positive effects of glucagon agonism, including increases in energy expenditure, inhibition of food intake, and reductions in liver steatosis are currently being explored for obesity treatment as part of co-agonism with GLP-1, whereby the diabetogenic effects can be mitigated [4]. Glucagon secretion is influenced by many factors, including sympathetic and parasympathetic neural signals [5], changes in blood glucose [6], amino acids [7] and fatty acids [8], but also both endocrine and paracrine factors [9] (Fig. 1). The use of medications for treating diabetes is increasing, alongside the development of novel drugs for this. In addition to exogenous insulin, pharmacological agonism of glucagon-like peptide-1 (GLP-1) receptors, inhibition of dipeptidylpeptidase-4 (DPP-4), inhibitors of sodium-glucose transporter 2 (SGLT2), and sulfonylureas, all influence pancreatic endocrine secretion and in this way improve glycemic control. The goal of most treatments is to lower glucose via increases in insulin secretion, but many of these drugs are also potential regulators of glucagon secretion [10]. Considering the importance of well-regulated glucagon secretion for the maintenance of plasma glucose homeostasis, understanding of this is essential. Here we review the latest advances in glucagon research with a focus on its regulation by various peptides released from the pancreatic islets, the thyroid, the gastrointestinal tract (GIT) and the hypothalamus, and how the regulation may be altered in diabetes, and by various forms of diabetes treatment.

Insulin
In an attempt to explain the hypersecretion of glucagon in T2D, it was hypothesized that the systemic insulin resistance in T2D might also apply to the α-cell, and that missing inhibition by insulin on glucagon secretion was the reason for the unfavorable, high glucagon levels in T2D [11]. For insulin to have a direct inhibitory effect on glucagon secretion, it is required that the α-cell expresses the insulin receptor (IR).
In spite of its importance, this has only been investigated in a few studies [12]. In one of the most recent, Franklin et al. used rat α-cells isolated by fluorescence activated cell sorting (FACS) to show that IR expression levels in the α-cells were at equal levels to the IR expression found in the liver [13]. In another study, IR protein was detected by western blots on α-TC6 cells [12] (but these cells are not necessarily representative of natural α-cells), and lastly, our lab recently collected IR expression data from human α-cells (based on publicly available sequencing data from the European Bioinformatics Institute), showing high expression of the IR in human α-cells from both healthy and T2D donors [14]. Thus, α-cells expression of the IR seems well established, although there is inadequate data to show that the expression is associated with the presence of functional IR on the cell surface. Insulin is the most abundant pancreatic hormone. Its role as an inhibitor of glucagon secretion has long been discussed, and it is still unclear to what extent and how insulin influences glucagon secretion. Supporting insulin's role as a regulator of glucagon secretion, several studies have shown that exogenous insulin decreases glucagon levels, in both non-diabetic [13,15] and diabetic models [16][17][18]. Furthermore, neutralization of insulin with specific insulin antisera was reported to increase glucagon secretion during hyperglycemia [19], but in vivo studies in mice showed only a tendency towards increased glucagon using the specific peptide-based, competitive IR antagonist S961 [20]. Many of the previous studies were done in vitro, and while the demonstrated mechanisms may have relevance in the setting of an incubated cell or islet, the mechanisms may not translate directly to in vivo conditions. Suspecting an important effect of insulin on glucagon secretion, Kawamori and colleagues utilized an α-cell specific IR knockout mouse (αIRKO). The αIRKO mice showed postprandial glucose intolerance, hyperglycemia and hyperglucagonemia, supporting an important role of insulin to restrain glucagon levels in the fed state. The authors, however, also found that during a hyperinsulinemic clamp, glucagon levels in wild type mice were still strongly stimulated by hypoglycemia, suggesting that in spite of insulins potential inhibitory actions, it is not the only factor responsible for the inhibition and stimulation of glucagon by changing glucose levels [21]. For instance, Vergari and colleagues suggested that rather than acting directly on the α-cell, insulin inhibits glucagon secretion indirectly, by increasing SGLT2 facilitated Na + -glucose transport in δ cells [22], thus stimulating somatostatin secretion [23] (somatostatin is an important regulator of glucagon, as reviewed in 2.2). Not only did they find that somatostatin is strongly stimulated by insulin, but also that its inhibitory effect on glucagon secretion was abolished by addition of the specific IR antagonist S961 to islets. Additionally, both somatostatin receptor (SSTR) antagonists and SGLT2 transporter inhibitors (dapagliflozin) prevented insulin from stimulating somatostatin, and inhibiting glucagon [23]. The authors were, however, unable to show an inhibitory effect of exogenous insulin in perfusion studies on wildtype mice, and in our lab, we have not been able to reproduce the finding that IR antagonist S961, or exogenous insulin stimulates glucagon secretion in perfused pancreas models [14,24]. It is important to note that the S961 concentrations in the study by Vergari et al. and in the study conducted in our lab, were very high (100 nM -10 μM), but it is possible that even higher levels of S961 may be required to completely block the actions of endogenous insulin, as intra-islet insulin concentrations could reach as high as 1μM [25]. Antagonist concentrations higher than that may not be meaningful, however, and are doomed to cause unspecific effects. In a study of dapagliflozin from our lab, concentrations were markedly lower (0.5 μM vs. 12.5 μM) than those employed in the study by Vergari, et al. but still matched peak plasma concentrations in humans receiving therapeutic doses of dapagliflozin [26], and were 500 times above dapagliflozin's IC 50 of 1 nM. Importantly, at 12.5 μM, dapagliflozin also inhibits SGLT1 [23], and the potential involvement of this glucose transporter in regulating glucagon secretion remains uncertain [27,28]. SGLT2 is interesting in the scope of glucagon regulation as SGLT2 inhibitors are widely used in diabetes treatment, and studies have associated SGLT2 inhibitors with an increase in glucagon secretion [29][30][31]. Speaking against a contribution of SGLT2s in the regulation of glucagon secretion, a recent study from our lab showed no expression of SGLT2 in rodents or human alpha cells. Furthermore, neither phloridzin (a blocker of both SGLT1 + 2) or dapagliflozin (a selective SGLT2 blocker) had effects on glucagon or insulin or somatostatin secretion, when administered to the isolated perfused pancreas [27]. In agreement with these findings, a recent study in diabetic patients showed no changes in glucagon during treatment with the selective SGLT2 inhibitor, empagliflozin [32]. In a study following children with new-onset type 1 diabetes (T1D) for 60 months, the progressive deterioration of insulin secretion (measured as C-peptide), was negatively associated with plasma glucagon (increased by 160 %) and HbA 1c levels, suggesting that insulin deficiency may in part explain the hyperglucagonemia in these patients [33]. Curiously, insulin may also play a permissive role on glucagon secretion. In dogs with alloxan-induced T1D, the glucagonostatic effects of glucose, and its ability to stimulate somatostatin was lost, but regained upon repletion of insulin [34]. In T1D, continuous insulin infusions may rescue diabetic hyperglucagonemia and normalize glucagon levels [35], suggesting that insulins permissive effects may be responsible for maintaining normal glucagon levels in diabetes.
Newer studies of the hyperglucagonemia of T2D indicate that the responsible mechanisms is a disruption of the liver-alpha cell axis, a negative feed-back cycle, according to which amino acids stimulate glucagon secretion while glucagon lowers the plasma concentrations of amino acids by stimulating ureagenesis in the liver [36,37]. According to this theory, the hepatic steatosis characterizing most patients with T2D impairs the action of glucagon on amino acid turnover (in a manner similar to hepatic insulin resistance), leading to increased plasma levels of glucagon [38]. Indeed, it is only the patients with diabetes that also have non-alcoholic fatty liver (NAFLD) that have hyperglucagonemia, and conversely, non-diabetic individuals with NAFLD also have hyperglucagonemia [39]. This would indicate that the hyperglucagonemia of T2D is unrelated to any pathology of the pancreatic islets and also does not depend on a feed-back relationship with insulin secretion or action.
An important factor that is rarely discussed in the context of insulin's inhibition of glucagon secretion, is the signaling pathway initiated by insulin when it presumably binds to its receptor on the alpha cell, and the temporal aspects of the signal transduction events. Because the IR is a receptor tyrosine kinase, it initiates complex signaling pathways by activating PI3-kinase and Akt, that result in regulation of transcription factors, protein synthesis and receptor/transporter translocation [40]. Supporting a role for PI3-kinase in the regulation of glucagon, activation of the PI3-kinase and Akt pathway may inhibit glucagon expression [41]. Induction of this pathway in α-cells was reported to be increased by insulin, while inhibition of PI3-kinase with the non-specific inhibitor wortmannin, prevented the attenuation of glucagon secretion by insulin [42]. Further, knockdown of IR blunted the α-cell response to hypoglycemia. This inability to react to lowered glucose was associated with a decrease in PI3-kinase activity. Importantly, in spite of IR knock down, α-cell secretion was still inhibited by hyperglycemia [12]. An interesting observation, is that the inhibition of glucagon by insulin, occurs much slower (~10 min) compared to the inhibition elicited by high glucose in e.g. the isolated perfused rat pancreas (~2 min). So while IR signaling may be involved in the long term inhibition of glucagon secretion, the inhibition of glucagon by hyperglycemia occurs much more rapidly than what is likely accounted for by the IR initiated pathways. It is likely that other mechanisms are accountable for the acute inhibition of glucagon by hyperglycemia, but these pathways are currently unknown.
Thus, there is evidence for insulin as an inhibitor of glucagon secretion, but only during certain conditions. A part of the explanation for the discrepancies found, could be differences between species and between the various rodent models. A general finding that challenges the notion of insulin as the primary regulator of glucagon secretion, is that glucagon secretion is inhibited at glucose concentrations below the threshold for glucose stimulated insulin secretion (GSIS) [14,43,44].
This strongly argues against a direct, acute inhibitory β to α-cell axis.
Importantly, in several studies, based on isolated perfused pancreas models, which are particularly physiologically relevant because of the preserved islet architecture and vascular supply, it has been impossible to demonstrate an acute inhibition of glucagon secretion by insulin in concentrations up to 10 μmol/L [14,23], perhaps owing to the already very high interstitial insulin concentrations surrounding the αcells.
Thus, there are many unanswered questions regarding insulin's effect on glucagon secretion.

Somatostatin
For many years, hallmarks of T2D pathophysiology have included an inadequate insulin secretion (relative to insulin sensitivity), and dysregulated postprandial and fasting glucagon secretion, but in recent years it has become clear that islet somatostatin may also play a role and that its inhibitory actions on α-and β-cells -and disruption of these actionscould contribute, making somatostatin an increasingly interesting target in diabetes [45,46].
Inhibitory actions of exogenous somatostatin on glucagon secretion have been shown consistently, both in vivo, in vitro and in perfusion experiments, in humans [47], dogs [48] and rodents [49][50][51][52]. Somatostatin acts directly on the α-cell through binding to one or more of four (SSTR1, SSTR2, SSTR3 or SSTR5) different receptor sub-types, all of which may be expressed in α-cells [14,52,53], with SSTR2 usually being considered predominant with regards to glucagon secretion [14,54,55]. The somatostatin receptors are G-protein coupled receptors (GPCR), coupled to G i/o proteins that inhibit adenylyl cyclase (AC) and cyclic adenosine monophosphate (cAMP) production and prevent cell depolarization and secretion by reducing the activity of voltage-gated calcium channels [52,56,57]. cAMP is involved in the regulation of glucagon secretion [58][59][60], and studies show that somatostatin may effectively inhibit glucagon secretion independent of both membrane potential and intracellular Ca 2+ , by reducing cAMP in α cells [58]. While recognizing these effects, lowering of cAMP also occurs in α cells in response to increased glucose when somatostatin signaling is inhibited [59]. This suggests that mechanisms intrinsic to the α-cell also controls cAMP levels, but the exact mechanisms are yet to be elucidated. Other potential targets through which somatostatin may inhibit glucagon secretion, is the G-protein coupled inwardly rectifying potassium (GIRK) channels [55], and K ATP channels. Though a connection between glucagon secretion and Katp channel activity and α-cell membrane potential is well established, the area is surrounded by controversy [61][62][63][64]. This primarily comes down to deviating results (in some studies glucagon is stimulated by K ATP inhibition, in others glucagon is stimulated by K ATP activation [61,65,66]). Different experimental setups are likely the reason for these discrepancies. Some clarity may be found in human studies, where the SUs enhance glucagon secretion [67], without causing hyperglycemia [68], suggesting that inhibition of K ATP activity stimulates glucagon secretion.
The potent inhibitory effects of somatostatin on the α-cell are apparent even at hypoglycemia, when endogenous somatostatin secretion is not considered stimulated. Nevertheless, under these conditions, glucagon secretion increases 2-3 fold when somatostatin's actions are prevented with specific receptor antagonists [14,24], which would be consistent with tonic inhibition. Increasing glucose, stimulates somatostatin release [69]. Whether the threshold for somatostatin release lies precisely at the threshold for glucagon inhibition by glucose is, however, not entirely clear. Some investigators find the rise in somatostatin and the fall of glucagon to occur simultaneously, others find that somatostatin secretion is only stimulated at glucose concentrations above the threshold for inhibition of glucagon [14,34]. It has also been suggested that different mechanisms are responsible for the inhibition of glucagon at normoglycemia and hyperglycemia, and that the glucagonostatic effects of glucose are independent of somatostatin at hypoglycemia, but dependent on somatostatin at hyperglycemia [52]. This obviously contrasts to the powerful effects of the SSTR antagonists at low glucose. Because somatostatin is rapidly degraded and small amounts reach the systemic circulation, it is generally considered impossible to estimate pancreatic somatostatin secretion from measurements in peripheral plasma. Thus, in in vivo models, intra-islet somatostatin levels are not reflected by peripheral fluctuations. Interestingly, in studies involving the isolated perfused rat pancreas model, it was observed that the glucagonostatic effects of glucose were completely lost when somatostatins actions were prevented by SSTR2 antagonism at physiological glucose levels (increasing glucose from 3.5 mM to 5 mM) [14]. Another interesting observation, is that the α-cell is more sensitive to somatostatin-14 (the predominant molecular form produced by δ cells) than the β cell. This may in part explain the unique inhibitory effect on glucagon at normo-and hyperglycemia [66,70,71]. Together, these observations suggest that glucose, rather than exerting a direct effect on α-cell metabolism and ion channel activity, elicits its inhibitory actions through a direct stimulation of somatostatin release, or by other mechanisms involving stimulated somatostatin secretion (as discussed next in 2.3). This would mean that glucagon secretion is always stimulated (possibly driven by other circulating nutrients, such as amino acids [72,73]), and that somatostatin acts as a brake on the α-cell in the post-prandial state when blood glucose rises. In glucagon receptor knockout mice (GCGR-KO) (which causes massive α-cell hyperplasia), the glucagonostatic effect of glucose is preserved [24]. Assuming that somatostatin is the primary inhibitor of glucagon secretion at normoand hyperglycemia, it could be hypothesized that the glucagonostatic effects of glucose would be lost, or much less pronounced, in these mice, due to a lower δ:α-cell ratio. But this is not the case. The explanation for this seems to be, that GCGR-KO mice show a concurrent increase in both glucagon and somatostatin content (and cell number), compared to wildtype mice (glucagon: 542 ± 125fmol/mg vs. 4933 ± 1450fmol/mg. somatostatin: 1452 ± 451 fmol/mg vs 4270 ± 1021 fmol/mg) [74].
Thus, α-cell hyperplasia of the GCGRKO mice is accompanied by δ cell hyperplasia, which could explain that the response to glucose is preserved in this model. However, there is also evidence from islet studies suggesting a less prominent role for somatostatin. In one study, the authors studied the relationship between glucagon and somatostatin in two models that would exclude effects of somatostatin: 1: in somatostatin knockout (SSTKO) mice, and 2: in islets treated with pertussis toxin (which inhibits G i GPCR signaling). An increase in baseline glucagon secretion occurred with both these approaches, supporting a tonic, inhibitory role for somatostatin. However, using both perifused islets and perfused pancreas, glucose was found to retain its ability to regulate glucagon secretion, suggesting that other signaling pathways or mechanisms intrinsic to the α-cell, that may include regulation of K ATP channel activity, were driving the regulation [52,61]. In experimental diabetes, the diminished counter-regulatory glucagon response to hypoglycemia, was proposed to be caused in part by dysregulated somatostatin signaling. Thus, α-cells from mice on a high fat diet (mimicking T2D) displayed decreased sensitivity towards exogenous somatostatin that failed to adequately suppress glucagon secretion [75], possibly related to changes in SSTR expression [76]. In contrast, in two rat models of T1D, SSTR2 antagonists corrected the well-known loss of glucagon response to hypoglycemia [77,78]. If all of these findings are confirmed in humans, somatostatin is likely to play an important role in orchestrating the hormonal glucagon response to changes in glycemia in healthy people, while these regulatory effects may be compromised in diabetes. Currently, somatostatin-based diabetes treatment is not available, but somatostatin analogs pasireotide and octreotide, are used to treat other somatostatin related diseases, such as Cushing's disease, acromegaly and neuroendocrine tumors [79]. Both short-and long-term (3 months) pasireotide treatment is associated with hyperglycemia, and development of diabetes, likely caused by a decrease in insulin secretion and inhibition of GLP-1 and glucose-dependent insulinotropic polypeptide (GIP) [79]. In healthy humans, pasireotide infusions likewise caused pronounced inhibition of insulin and incretin levels, but had only minor inhibitory effects on glucagon secretion [80]. This differentiation can be explained by pasireotide working primarily by binding SSTR5, which is involved in the regulation of insulin from the β cell [81], as opposed to the α-cell secretion which is primarily regulated by the delta cells via SSTR2 [14,82]. Indeed, octreotide, which has high affinity for SSTR2, powerfully inhibits glucagon secretion in diabetes [83][84][85][86]. Thus, diabetes may develop in some individuals receiving pasireotide treatment [87], while the effect on glycemic control varies in individuals treated with octreotide [88,89]. At any rate, the numerous inhibitory actions of somatostatin and its analogs on e.g. growth hormone secretion, the GI tract, and the bile system (causing gallstones) do not render somatostatin therapy for metabolic diseases particularly attractive.

Urocortin-3 (UCN-3)
UCN-3 of the corticotropin-releasing factor family (CRF) of peptides, is co-secreted with insulin from the β cell, and is, in spite of its suggested effects on both glucose-stimulated insulin secretion and glucagon secretion, understudied -and perhaps underappreciated -as a glucoregulatory peptide. Owing to this, less than a handful of studies have investigated UCN-3 ′ s role in glucose metabolism, and of these, even fewer have had glucagon secretion as an end point. Chien Li and colleagues found almost 20 years ago that UCN-3 stimulated insulin and glucagon secretion after i.v. injection in rats as well as in isolated rat islets [90]. However, the authors decided not to pursue the glucagonotropic effects of UCN-3 further, but to focus attention on UCN-3 ′ s insulinotropic effects. In 2015, Meulen et al. suggested that UCN-3 acts as a negative regulator of insulin secretion that upon secretion ensures timely inhibition of insulin via stimulation of somatostatin secretion [91]. This contrasts the insulinotropic actions reported by Chien Li, but as Meulen and collegues discuss, the model proposed by Chien Li involving direct effects of UCN-3 on the β cell is incompatible with the finding that β cells do not express the UCN-3 receptor (corticotropin-releasing hormone receptor 2 (CRHR2)), whereas delta cells show expression (until the paper by Meulen et al. in 2015, the islets were known to express CRHR2, but the cell type was unknown). As reviewed in section 2.2, somatostatin is a potent regulator of glucagon, meaning that the indirect inhibitory actions of UCN-3 on insulin via stimulation of δ-cells, is likely to also affect glucagon secretion. To gain a better understanding, our lab recently set out to further explore the effects of UCN-3 on islet hormone secretion. Thus far, data from experiments with the isolated perfused pancreas support the notion that UCN-3 is a potent stimulator of somatostatin secretion. We also confirmed CRHR2 expression exclusively on delta cells in rats as well as in humans, who show the same receptor expression in both healthy and T2D individuals. We furthermore found that UCN-3 stimulated somatostatin secretion was associated with a concurrent 50 % inhibition of glucagon secretion, which was abolished by addition of SSTR antagonists. These findings position UCN-3 as a key regulator of glucagon secretion, but since the effects of UCN-3 on pancreatic secretion are clearly understudied, the role of UCN-3 in the pathophysiology of diabetes as well as the regulation of glucagon secretion remains unclear.

Pancreatic polypeptide (PP)
PP is primarily found in PP-cells in the pancreatic islets [92], and is released in response to meal ingestionin particular by meals containing protein and fat [93], but also by hypoglycemia by a mechanism involving efferent vagal activity [94]. PP is a member of the neuropeptide Y family (also including neuropeptide Y (NPY) and peptide YY (PYY) [95]), and signals through binding to the Y family of inhibitory GPCR's, of which five different subtypes exist (Y 1 , Y 2 , Y 4 , Y 5 and Y 6 (pseudogene in humans)) [96]. Of these, PP has the highest affinity for the Y 4 subtype (PPYR1) [97]. PP is known to inhibit cellular activity by inhibiting intracellular cAMP formation. It is believed to play a role as an anorectic hormone in humans [98] although the evidence for this is sparse. The mechanisms behind the anorectic effects are not fully understood, but may include both central mechanisms via vagal afferent nerves and delay of gastric emptying and reduction of intestinal motility [99]. In addition to these effects, PP have been reported to influence the secretion of glucagon [100], insulin [101] and somatostatin [97,102]. There is however limited data to support these effects and reports on the distribution of PP receptors diverge. Thus, one study used immunostaining techniques to show PPYR1 (Y 4 ) protein exclusively on δ cells [97], while another immunohistochemical study (with different antibodies), showed exclusively PPYR1 protein on α-cells. In the latter study, the authors supplemented their immunostaining with quantitative reverse transcription polymerase chain reaction (RT-PCR) on a FACS sorted non-β-cell fraction (with high glucagon expression), and again found PPY1R expression [100]. In another study, the 'translating ribosome affinity purification' technique was used to amplify β cell expression of receptor subtype Y 1 (for which PP has a lower affinity compared to Y 4 ), suggesting some level of endogenous receptor Y 1 expression in β cells [103]. The inconsistency in these receptor studies clearly warrants further investigation. Studies on the effects of PP on islet hormone secretion show a similar inconsistency. While PPYR1 expression in δ-cells is not clear, administration of PP (at doses ranging from ~20 pM -2μM) to both isolated rat islets and the perfused rat pancreas, was reported to stimulate somatostatin secretion, during both hypo-and hyperglycemia [102]. However, in another study PP was identified as a potent inhibitor of somatostatin secretion from human islets, but the physiological relevance of this remains unclear, as these effects were observed at very high, clearly supraphysiological PP concentrations of 100 nM and 1μM [97]. Glucagon secretion would be expected to be influenced by the PP-induced changes in somatostatin secretion, but could also be directly regulated by PP. Islets incubated with physiological levels of PP showed a decrease in glucagon secretion in line with the inhibitory pathways initiated by PPYR1 (somatostatin could also be involved, but measurements of somatostatin were not included in the study), with no concomitant effect on insulin secretion [102]. In contrast, infusion of PP in anesthetized rats slightly raised basal glucagon levels [104]. Because all receptor subtypes Y 1 -Y 6 are expressed by the hypothalamus [95], the effects of PP on glucagon in vivo could also result from both direct effects on the α-cell and centrally mediated effects, similarly to the stimulation of insulin secretion by centrally administered NPY [105]. More research is needed in order to define PP's actions on glucagon secretion (if any).

Peptide tyrosine tyrosine (PYY)
PYY is primarily secreted from intestinal endocrine L cells, but has also been reported to be present in lesser amounts in the stomach, as well as in α-, δ-and PP-cells of the pancreas [106] and in the brain [107].
PYY appears in two forms; full length PYY 1− 36 , and truncated PYY 3− 36 . PYY 3− 36 is formed as a consequence of DPP-4 cleavage of its N-terminus, and is the most abundant form circulating in human plasma [108]. PYY 1− 36 is agonistic towards all Y receptor subtypes (Y 1 , Y 2 , Y 4 , Y 5 ), whereas PYY 3− 36 primarily binds the Y 2 receptor, with lesser affinity towards Y 1 and Y 5 [109]. PYY 1− 36 stimulates food intake through Y 1 and Y 5 [110], while PYY 3− 36 , through Y 2 , induces weight loss and reduces appetite and food intake [111,112], without affecting energy expenditure [111,113]. The anorexic effects of PYY 3− 36 are driven by activation of Y 2 receptors on neurons in the arcuate nucleus, as demonstrated by the use of Y 2 R antagonists as well as Y 2 R knockout mice [114], [112], but the mechanism downstream from receptor activation that facilitates these effects is still being debated [115]. PYY 1− 36 also affects glucose homeostasis. As alluded to in section 2.4, Y receptor expression in islets is still not fully mapped. While studies report on expression of both Y 1 , Y 4 and Y 5 in whole islet preparations, the distribution on the cell types is unknown [116,117]. Notably, Y 2 expression is not evident from these studies, seemingly excluding direct effects of PYY 3− 36 . In line with the receptor expression, PYY 1− 36 (that may target both Y 1 , Y 4 and Y 5 ), affects islet secretion as reported in perfusion experiments and studies on rodent islet. Here, administration of PYY 1− 36 inhibited GSIS [118,119], and transiently (only for the first 3 min) reduced glucagon secretion [119]. Supporting an inhibitory role of PYY 1− 36 on insulin secretion, PYY KO mice are both hyperinsulinemic and susceptible to developing obesity [120]. It is however important to mention that i.v. infusion of PYY 1− 36 during a glucose challenge had no effect on GSIS in humans, suggesting that important species differences may exist. Plasma PYY concentrations are significantly lower in T2D [121] and obesity [122] suggesting the PYY is not involved in the impaired insulin secretion. After gastric bypass surgery, PYY 1− 36 levels are increased [123], and is thought to play a pivotal role for the metabolic improvements after this operation. In fact, while GLP-1 ′ s impact on improvement in metabolic parameters after bariatric surgery has been questioned by some [124,125], PYY 1− 36 effects have gained interest, and were suggested to be essential [126]. This was based on the reported ability of PYY 1− 36 to improve β-cell function, growth, and survival. In a model with increased PYY expression, islet and β-cell size increased, and GSIS was improved [127], while a PYY deficient model showed deteriorated β-cell function and hyperglycemia [106]. However, none of these studies investigated potential effects on glucagon secretion. Because of the reported beneficial effects of PYY 1− 36 on β-cell function, growth and survival without inhibition of GSIS (in humans), and because of the anorectic effects of PYY 3− 36 , several attempts have been made to produce stable PYY 1− 36 [128] and PYY 3− 36 analogs [129]. Unfortunately, frequent adverse effects are associated with PYY 3− 36 administration (nausea and vomiting) [130], and difficulties in generating analogs of PYY 1− 36 that are resistant to DPP-4 degradation, without losing efficacy of β-cell survival and growth [128] makes this ambition challenging. Novo Nordisk has developed a PYY 3− 36 analog (NC0165− 1273) with higher affinity for Y 2 compared to native PYY 3− 36 , enabling a more specific Y 2 receptor activation, with increased effectiveness (native PYY 3− 36 has some affinity for Y 1 and Y 5 which could counteract the anorectic effects of Y 2 ) [131]. In another approach, a stable analog was developed by conjugating a monoclonal antibody to a cyclized PYY 3− 36 analog (mAb-cyc-PYY). Administration of mAb-cycPYY to macaque monkeys over a 23 day period, induced weight loss of around 10 % with no incidence of vomiting. Interestingly, mAb-cycPYY treatment could be combined with liraglutide treatment (a GLP-1 analog), further enhancing the weight loss effects, up to 30 % at the highest dose [132]. These treatment forms are however focused on weight loss, and the effects on glucagon (which are most likely weak or none-existing) are not considered.

Calcitonin (CT)
CT is a hypocalcemic 32 amino-acid peptide hormone released from specialized parafollicular C-cells in the thyroid [133]. Here, it is stored in vesicles, and released to the circulation in response to elevated plasma calcium [134]. The CT receptor (CTR) is a class B GPCR capable of signaling through both G s and G q g-proteins [135], to activate cAMP [136,137], phospholipase C (PLC) [136,137] and phospholipase D (PLD) [138] pathways. Several splice-variants are known, and the pattern of expression is species-dependent. In humans, the most important splice-variants are characterized by the absence or presence of a 16-amino acid sequence inserted in the first intracellular loop; this influences the interaction with g-proteins, and absence impairs in particular the G q coupling [139]. In humans, CT primarily reduces plasma calcium by directly inhibiting osteoclast activity, but also by reducing calcium reabsorption in the kidneys [140]. Though these actions are well known, the physiological importance of CT has been questioned, as both thyroidectomised patients and patients with CT producing thyroid tumors (with up to 50-fold increases in plasma CT) display normal plasma calcium levels, with no apparent difference in bone structure [140,141]. Being a highly conserved hormone, found in several different species [142], it does seem highly unlikely that CT is redundant, and effects outside of calcium homeostasis have also been identified. Thus, CT may also regulate food intake [143,144], and is involved in the regulation of glucoregulatory hormones. CT has been described as a diabetogenic hormone [145], and while no increase in diabetes prevalence was reported in patients receiving CT treatments for Paget's disease [146], studies in humans confirmed that acute administration of CT (both the potent salmon derived CT (sCT) and porcine CT) inhibited insulin secretion and attenuated the inhibition of glucagon in response to an oral glucose tolerance test (OGTT) [147] and to i.v. glucose infusion [148]. However, the literature regarding CTR activation by CT in humans is scarce, and the physiological relevance of CT as a regulator of insulin and glucagon remains undetermined. There are, however, other beneficial effects of CTR activation which are highly relevant in the scope of diabetes and obesity, as activation of the receptor by the beta-cell derived peptide hormone, amylin, results in anorexia and improves glucose tolerance [149], which will be discussed in further detail in section 2.7.

Amylin
Amylin is a 37 amino acid peptide produced in the β cell, and secreted alongside insulin, in a molar ratio of ~1:100. The amylin receptor includes the CTR, but active amylin receptors (AMY 1− 3 ) are only formed when the CTR hetero-dimerizes with one of three receptor activity-modifying proteins (RAMP1− 3). Of these, RAMP1 and 3 (forming AMY 1 and AMY 3 ) have the greatest affinity for amylin, while the RAMP2 (forming AMY 2 ) affinity is less, compared to CTR [150]. The actions of specific AMY and CTR agonists on their receptors have proven difficult to test in controlled in vitro experiments. Because not all CTR are interacting with a RAMP protein at any given time, some "free" CTR will always be present [139]. Thus, in Cos-7 cells transfected with AMY1 or AMY3 receptors, amylin, but not CT, displaced the receptor bound [ 125 I]-amylin, suggesting specific interaction of amylin with AMY1 and AMY3 receptors, but in these cells, CT nevertheless increased intracellular cAMP levels. This occurs presumably because CT interacts with a population of CTR that has not heterodimerized with a RAMP [139], thereby retaining their affinity for CT. While this, and the many different splice variants of the receptor with potentially different signaling properties [139] makes it difficult to identify the exact effects of AMY 1− 3 activation, receptor activation is considered beneficial in diabetes. Indeed, both native and synthetic amylin have been reported to promote glucose tolerance in rodent models and human diabetes, by regulating gastric emptying and glucagon secretion [149,151].
It is currently unknown if the α-cell expresses CTR and/or RAMPs to mediate direct effects of amylin on glucagon secretion, but the effects have been studied in various models. In isolated islets, findings are inconsistent. Some report insulin regulating properties [152], while others report glucagon regulating properties [153], and others find no effects at all [154]. In perfusion studies, no direct effect of amylin on glucagon was found [154], but attenuation of arginine-stimulated glucagon secretion in vivo, is a recurring finding [153][154][155]. Interestingly, in freely fed mice, amylin infusions inhibited glucagon and improved glucose tolerance, while amylin in the setting of an OGTT had no effect on pancreatic hormone secretion or plasma glucose [156]. Thus, amylin may inhibit glucagon secretion in response to amino acids, but not in response to glucose alone, and a mechanism extrinsic to the pancreas, perhaps central, mediated by vagal efferent activity, seems to be essential for these effects [154]. At any rate, the physiological importance of amylin in regulating islet secretion in humans, seems negligible [157,158]. Because of amylins effects in vivo to delay gastric emptying, reduce appetite and glucagon secretion, a huge interest in developing stable amylin analogs has arisen. Currently, the only amylin analog approved for diabetes and obesity treatment is pramlintide, which may be used in concert with insulin in both T1D and T2D. Pramlintide reduces glucagon secretion in T1D [159], and promotes weight loss. The weight-loss effects are however modest, and in long-term treatment with pramlintide (up to 52 weeks) amounted to only 1-3 % of total body weight in the patients [160]. Pramlintide's short half-life (20-45 min [161]), requiring injections three times a day, and the fact that insulin and pramlitide must be prepared at different pH [162], thus requiring separate injection, diminishes patient compliance [163]. Therefore, more stable amylin analogs with higher efficacy have been developed, and the possibilities of co-administration with GLP-1 analogs is being explored. In particular, Dual Amylin and Calcitonin Receptor Agonists (DACRAs) have been shown to reduce fat mass. 9-week co-administration of a DACRA developed for weight loss, KBP-089, with liraglutide, reduced body-weight of rats on a high fat diet (HFD) by 21 %, and slightly lowered glucagon, though not significantly (the authors argue that a more pronounced effect on glucagon might have been present in a stronger T2D model) [164]. In comparison, weight-loss in monotherapy amounted to 15 % (KBP-089) and 7% (liraglutide) [164], revealing the potential benefits of combinations of amylin analogs and GLP-1 based incretin treatment. In another study with KBP-089, the authors found a pronounced weight loss in HFD rats (17 % over 8-weeks), and a significant reduction in insulin secretion, and increased glucose tolerance. The authors credit the paradoxical decrease in insulin and increase in glucose tolerance to delayed gastric emptying and potentially increased insulin sensitivity, but a potential influence of glucagon has not been explored [165]. In the same study, the effects of KBP-089 were investigated in Zucker Diabetic Fatty rats (ZDF). Here KBP-089 caused a significant reduction in plasma glucose compared to vehicle during an OGTT. Indeed, insulin sensitivity was increased in these animals, but the question whether some of the beneficial effects of KBP-089 on plasma glucose excursions could be driven by changes in glucagon (that is known to be dysregulated in T2D) was not answered [165]. KBP-089 showed a tendency towards lowering glucagon in HFD rats, but the effects in T2D models remain unexplored. Currently, a long-acting amylin analog developed by Novo Nordisk A/S, called cagrilintide, is showing promising effects on weight loss when co-administered with the GLP-1 analog semaglutide (from ~8%-~17 % bodyweight reduction depending on dosage), with a concomitant decrease in glucagon (that may also be caused by semaglutide) [166]. Likewise, Zealand Pharma is developing promising novel amylin analogs as potential treatment for obesity, but as their primary end-point is weight loss, data on their glucoregulatory properties are currently not available [167,168].

Glucagon-like peptide-1
In the intestinal L cells, proglucagon is cleaved by prohormone convertase 1/3 (PC1/3) to form glicentin and GLP-1 and GLP-2 in equimolar amounts. GLP-1 is an incretin hormone primarily known for its insulinotropic actions, but GLP-1 also strongly inhibits glucagon secretion [169]. Upon release, GLP-1 is rapidly degraded by the enzyme DPP-4, which cleaves off the two NH 2 -terminal amino acids, resulting in the inactive forms, GLP-1(9-36)-amide or GLP-1(9-37). The metabolites act as a rather weak competitive antagonists on the GLP-1 receptor, and the relevance of this in vivo seems minor [170]. The almost immediate degradation means that of the total amount of GLP-1 released from the gut, only a small percentage actually reaches the circulation and the pancreas [171]. In the pancreatic α-cells where the majority of proglucagon is cleaved by PC2 to form glucagon, glicentin-related polypeptide (GRPP) and the major proglucagon fragment (MPGF). A very small proportion may also be processed to GLP-1 [172], that may act locally to regulate insulin, somatostatin and glucagon secretion [169]. Most of the paracrine activity from the α-cell is exerted by the large amounts of secreted glucagon [24,173]. Thus, glucagon stimulates somatostatin in control mice, but not in GCGRKO mice. However, glucagon-stimulated somatostatin secretion is also attenuated by the GLP-1R antagonist exendin-9, suggesting that glucagon stimulates somatostatin secretion through both GCGR and GLP-1R on the δ cell [24]. The effect of exendin-9 on somatostatin (and insulin) secretion has been interpreted to support significant actions of GLP-1 derived from the alpha cells, but the actions of the much larger amounts of glucagon on both receptors support that glucagon is the primary activator of GLP-1R in islets.
The exact mechanisms behind GLP-1 ′ s inhibition of glucagon secretion are still debated, and it is unclear to what extent the α-cell expresses the GLP-1 receptor (GLP-1R). Only few studies find appreciable expression of GLP-1R on α-cells [174], and most often it is either not found [13,14,175,176], or found in very low amounts. In a study using the αTC-1 cell line (which may not be a reliable representative of the natural α-cell), 20 % of α-cells were reported to express the GLP-1R [177], while in another study it was expressed in only ~1% of a FACS sorted primary α-cell population [178]. Immunostainings of α-cells show a similar inconsistency, with between 0.3 %-9.5 % of α-cells staining positive for the GLP-1R [179,180]. In this context it is important to mention, that studies on GLP-1R localization have been limited by a lack of good antibodies, which has made it difficult to visualize and quantify the receptor. Recently, more specific and sensitive GLP-1R antibodies have been generated, providing a clearer picture [181]. In spite of finding low GLP-1R expression (0.3 %, that may be underestimated by subpar antibodies), Ramracheya, R., and colleagues claimed that GLP-1 acts directly on the α-cell, as the inhibitory effects of GLP-1 persisted even when the actions of somatostatin and insulin were prevented (seemingly excluding that GLP-1 could inhibit glucagon secretion through these), and could be prevented by specific GLP-1R antagonists. It does however seem unlikely that GLP-1 directly affects glucagon secretion to an appreciable extent, if only one in 300 α-cells has the receptor, a finding which is consistent with the expression analysis.
Thus, a role of GLP-1 on α-cells through the GLP-1R seems questionable. So if GLP-1 is not acting directly on the α-cell, then how does GLP-1 regulate glucagon secretion? A possible mechanisms is that GLP-1 instead works indirectly by stimulating secretion from δ cells of somatostatin, known to strongly inhibit α-cell secretion [14,182] as reviewed in section 2.2.
Several studies have shown expression of the GLP-1R on somatostatin secreting δ cells [179,180,183], and the finding that in isolated α-cells, GLP-1 paradoxically stimulated glucagon secretion (in stark contrast to in vivo studies where GLP-1 inhibits glucagon secretion), suggests that somatostatin may be the responsible factor [184]. Substantiating this theory, in the perfused pancreas, the inhibitory effect of GLP-1 was prevented by addition of somatostatin receptor 2 antagonists [185]. These findings conflict with the previously mentioned study in islets, where the glucagonostatic effects of GLP-1 persisted in spite of SSTR antagonism, but since these were incubation studies where both GLP-1 and antagonists may reach the α-and δ-cells via uncontrolled diffusion from the outside rather than via the circulation, the implications of these findings are uncertain. The notion that GLP-1-induced somatostatin secretion inhibits glucagon, but not insulin, may seem paradoxical, but has several explanations. First, GLP-1 has powerful insulinotropic effects on insulin secretion, that may counter the inhibitory effects of somatostatin. Second the β-cells are more resistant towards somatostatin, than α-cells [186,187], and as a result, SSTR antagonism does not change insulin levels [188]. Third, the compartmentalization of the islet, and the intra-islet blood flow could affect the exposure of β-cells to somatostatin, but some evidence questions the importance of this [6].
Due to the rapid degradation of GLP-1 by DPP-4, GLP-1(9-36) is the most abundant form of GLP-1 in the circulation [189], with some studies suggests that GLP-1(9-36), rather than being an inert product, may regulate islet secretion as a weak antagonist towards the GLP-1 receptor, but the physiological relevance of this seems minor, if at all. In one study, the affinity of GLP-1(9-36) was found to amount to ~1% of intact GLP-1, and in spite of circulating in concentrations that are ~10 fold higher compared to GLP-1 , the levels were insufficient for GLP-1  to confer any of its antagonistic effects [190]. Supporting this, a study in T2D humans showed that plasma glucose, insulin and glucagon levels after a meal was unaffected by altering GLP-1(9-36) levels, using a DPP-4 inhibitor [191]. In another study, i.v. infusion of GLP-1(9-36) during an i.v. glucose tolerance test, likewise showed no changes in glucagon levels [192]. In a human studies in which GLP-1(9-36) was infused prior to a mixed meal test, a minor increase in glucose tolerance was measured, suggesting some effect on glucose control, but these effects were not attributed to changes in either insulin or glucagon secretion [193], and so its effects on islet secretion are negligible.
Treatments with GLP-1 analogs and DPP-4 inhibitors (aimed at prolonging the half-life of released GLP-1), are both effective methods for treating T2D, both because of the glucose-dependent stimulation of insulin secretion (which means that hypoglycemia is avoided), but also because of the inhibitory actions on glucagon which is hypersecreted in T2D [194]. Also the GLP-1 analogs exenatide and liraglutide inhibit glucagon secretion in both healthy and T2D individuals [195,196]. Likewise, DPP-4 inhibitors decrease glucagon levels [197,198].

Glucagon-like peptide-2 (GLP-2)
GLP-2 is a well known intestinotrophic factor [199]. It regulates intestinal growth and promotes regeneration and healing of damaged intestinal mucosa [200,201], and GLP-2 based treatments for short bowel syndrome are currently available [202]. Although scarcely studied, it is also clear that GLP-2 acts beyond the intestinal tract. Intact GLP-2 is produced in, and secreted from, the pancreas, also in humans [203]. Its local actions are unknown, but anti-inflammatory actions have been proposed [204]. GLP-2 also has glucagonotropic activity [205]. Thus, an increase in glucagon is observed when GLP-2 is administered intravenously to humans [206], and in the perfused rat pancreas [207]. GLP-2 appears to stimulates glucagon secretion directly, by binding to GLP-2 receptors on the α-cell, which are expressed on both human and murine α-cells [207]. Of note, no concommitant changes in somatostatin are observed when GLP-2 is administered to the isolated rat pancreas, suggesting that the effects are indeed direct [207], in contrast to the effects of GLP-1, as discussed above. Interestingly, when equal concentrations of GLP-2 and GLP-1 were administered to the perfused pancreas in tandem, the glucagonotropic effects of GLP-2 was in fact overruling the glucagonostatic effects of GLP-1 [207]. This is an interesting observation given that in vivo, the two peptides are released in equal amounts. However, because GLP-2 has a longer half-life than GLP-1 (GLP-2: ~7min [208] vs GLP-1: ~2min [209]) the pancreas may be exposed to somewhat higher concentrations of GLP-2, the consequence of which may be a net increase in glucagon secretion. These effects appear to persist in T1D [210]. Importantly, while high-dose (10 pmol x kg x min) infusion of GLP-2 in humans during a mixed meal test significantly increased glucagon compared to low-dose (1 pmol x kg x min) and placebo [211], peak blood glucose was lowered, most likely due to delayed gastric emptying [212]. While these results speak against a hyperglycemic role in vivo, only a few studies have addressed this. Thus, the effects of GLP-2 and GLP-2 analogs on glucagon secretion should be borne in mind as the interest in using GLP-2 based treatments to treat intestinal disease increases. Glucagon is not included as an end-point in studies on the effects of GLP-2 based treatments (apraglutide, teduglutide) on short bowel syndrome, and thus their effect on glucagon secretion is currently unknown.

Oxyntomodulin (OXM)
OXM is, like the glucagon-like peptide hormones, derived from PC1/ 3 mediated processing of proglucagon [213,214], exclusively in intestinal L-cells [215,216]. OXM is a 37 amino acid (AA) peptide, consisting of the entire 29 AA glucagon peptide, extended c-terminally by an 8 AA fragment. Currently, no receptor has been identified for OXM [217], but it acts as an agonist for both the GCGR and the GLP-1R, albeit with lower affinity (~100-fold) than the receptors primary ligands, glucagon and GLP-1 [218], [219]. Because of its agonistic effects on the GLP-1R and GCGR, the effects of OXM on pancreatic islet secretion would be expected to mimic those of GLP-1 (as reviewed in section 3.1) and glucagon, but paradoxically, OXM appears to stimulate glucagon secretion in both the perfused pig pancreas and in humans, through a mechanism that is yet unknown [220,221]. In addition to its pancreatic effects, OXM may also activate GCGR on hepatocytes, stimulating glucose production, but this is counteracted by OXM's ability to also stimulate insulin secretion [222,223]. In humans, OXM appears in only small amounts, resulting from degradation of glicentin. As it is also susceptible to DPP-4 degradation [224], its physiological activity in healthy people is probably negligible [216]. In gastric bypass patients, OXM concentrations are increased 10-fold (reaching concentrations up to 100 pM) [225], which may contribute to the beneficial effects on weight loss and glycemic control associated with the surgery [226]. The dual agonist properties of OXM makes it therapeutically interesting. This stems from the unique ability of OXM to increase energy expenditure (via GCGR activation), and reduce food intake and stimulate insulin (both GCGR and GLP-1R activation), while decreasing liver fat (GCGR) [227]. Due to rapid degradation, treatment with OXM requires several daily injections [228], but there is currently great interest in developing stable OXM analogs. Thus, DPP-4 resistant OXM analogs were shown to reduce food intake, induce weight loss, and increase glucose tolerance in diet-induced obese (DIO) mice and rats. These analogs primarily targeted the GLP-1R, with low affinty for GCGR [229,230]. Several promising dual-agonists are currently in phase 1 or 2 trials [231] and are being tested for their ability to activate both receptors at a ratio that ensures the benefical effects associated with dual receptor activation, while limiting adverse effects. However, concrete data on their effectiveness is yet unavailable for many of these compounds. Of the compounds with available data, once-daily injection of Cotadutide (AstraZeneca) in humans, dose-dependently improved glucose tolerance (glucagon levels were not measured) and induced a weight loss comparable to liraglutide. In addition, improved hepatic parameters were reported, displaying the potential of GLP-1R/GCGR dual agonists not only as a treatment for diabetes and obesity, but also for non-alcoholic steatohepatitis and non-alcoholic fatty liver disease [232]. The incidence of adverse effects (nausea and vomiting) associated with Cotadutide treatment and placebo was greater than with liraglutide treatment (completion rates: 73-77 % for Cotadutide, 81 % for placebo, 94 % for liraglutide) [232]. Once daily administration of another dual agonist, SAR425899 (Sanofi), also improved glucose control [233], and induced a weight loss comparable to liraglutide [234], but the effects on glucagon secretion was not investigated. Quantative assessment of the binding properties of SAR425899 to GLP-1R and GCGR in humans, showed that interaction with GCGR was significantly lower than GLP-1R. Thus, although the affinity of SAR425899 towards the GLP-1R in vitro is 5-fold higher than for the GCGR, surprisingly little receptor interaction was found in vivo [235]. This could explain the similarities between weight loss outcomes between SAR425899 and liraglutide, and suggests that in vivo, SAR425899 may act mainly as a GLP-1R mono agonist, rather than a dual agonist. Lastly, daily injections over 12-weeks with the dual agonist MK8521 (Merck Sharp & Dohme) induced a weight loss similar that of liraglutide when given at the highest dose, but had slightly lower glucose lowering effect [226]. Again, effects on glucagon secretion were not investigated.

Glucose-dependent insulinotropic polypeptide
GIP, along with GLP-1, are currently thought to be the most important known incretin hormones. GIP is secreted from specialized enteroendocrine K-cells that are located in high abundance in the proximal small intestine, decreasing more distally [236]. GIP secretion is stimulated by nutrient intake [237] with the key function to potentiate glucose-induced insulin secretion. In spite of its role as an incretin hormone, GIP has received far less attention as a potential target for treating diabetes than GLP-1, in part because it stimulates glucagon secretion [238], (rather than inhibiting like GLP-1, and therefore contribute to diabetic hyperglycemia), but also because the insulinotropic effects of GIP are weak or missing in T2D [239]. Though GLP-1 and GIP have similar signaling pathways, the difference in glucagon regulation by the two incretin hormones can most likely be explained by the fact that the α-cell does not express the GLP-1R (as reviewed in section 3.1), whereas expression of GIPR on the α-cell has been confirmed through both mRNA expression analysis, and protein analysis, and RNA sequencing [240][241][242]. The discovery that GIP has glucagonotropic properties in vitro, has caused some hesitation for using GIP to treat T2D, as it could cause potentially harmful hyperglucagonemia in patients receiving treatment [243]. Upon binding to its receptor on the α-cell, GIP increases intracellular cAMP levels [175], and activates cAMP/protein kinase A pathways to depolarize the cell, increase intracellular calcium levels and stimulate glucagon release [184]. In line with this, in vivo studies did indeed find a dose-dependent increase in glucagon levels when GIP was given intravenously to healthy male subjects in physiological concentrations, at euglycemia. However, the increase measured was only modest, and the peak concentration was increased by only ~3pM [244]. A comparable increase was seen in a study that investigated the glucose dependency of GIP on glucagon secretion. Here, the increase in glucagon was found at both hypoglycemia and euglycemia, but not at hyperglycemia, suggesting that prevailing glycemia affects the efficacy of GIP on glucagon secretion [245,246]. Importantly, loss of GIP's glucagonotropic effects at hyperglycemia means that GIP may not normally cause hyperglucagonemia at hyperglycemia, and thus the fear that GIP based treatments are potentially harmful, may be misplaced. Substantial effort is currently being invested into the development of dual GIPR and GLP-1R agonists for treating T2D that may exploit the many beneficial effects linked to both GIPR and GLP-1R activation, in particular inhibition of food intake with no concurrent increase in glucagonemia [247].

Neurotensin (NT)
NT is a 13 amino-acid neuropeptide originating from the central nervous system (CNS) and the intestines, from which it is formed through proteolytic cleavage of a larger precursor [248]. NT has distinct activities in both the CNS and GIT [249,250], but also regulates pancreatic hormone secretion. To our knowledge, there is no direct evidence suggesting NT receptor (NTSR) expression on α-cells, but, in a study by Brown and colleagues in 1976 it was first discovered that in vivo administration of NT in rats caused hyperglycemia. Since the hyperglycemia in this study was more pronounced compared to the response to exogenous glucagon, it was suggested that NT in addition to its glucagonotropic actions, also inhibited insulin secretion [251]. Not long after, another group studied the relationship between islet hormone secretion and NT in isolated rat islets, and could largely confirm the glucagonotropic actions of NT during short incubation periods (20 min) at 3 mM glucose, with a concurrent increase in insulin and somatostatin [252]. However, the stimulatory effect of NT was only apparent at hypoglycemia, whereas at hyperglycemia, NT dose-dependently inhibited glucagon secretion. This was accompanied by a concurrent inhibition of both insulin and somatostatin secretion, and NT was furthermore also found to blunt glucagon secretion in response to 20 mM arginine [252]. The glucagonotropic effect of NT, was suggested to involve signaling through β-adrenergic receptors, as the stimulation was attenuated by administration of propranolol, prior to NT [253]. Thus while NT is consistently found to regulate glucagon secretion, the effect seem highly dependent on the experimental conditions (e. g. in vivo vs in vitro). It appears that in vitro, the inhibitory actions of hyperglycemia on glucagon secretion overrule the potential stimulatory actions of NT, while in vivo this is not the case. This could stem from the fact that the hyperglycemic conditions from in vitro studies (23 mM glucose) are highly unlikely to be seen in healthy rat and dog models, and thus islet studies may reflect a response that is relevant to the α-cell in an isolated system, but with less physiological relevance. Some of the discrepancies could also be explained by exclusion of important neuronal signaling through β-adrenergic receptors in islet studies, the involvement of which was demonstrated in dogs receiving propranolol [253]. Because of its properties as a regulator of pancreatic hormone secretion, it was speculated that changes in NT levels could be associated with the development of diabetes. Indeed, diabetic rats and mice had increased immunoreactive NT in pancreatic extracts in one study [254], but no differences in NT were found between lean and obese healthy people, or in T2D, whether fasted or after a meal, suggesting that in humans, NT is not involved in the pathophysiology of T2D [255].
Paradoxically, studies position NT as both an anorexic and obesogenic peptide. Intraperitoneal infusions of a long-acting NT analog acutely decreased food intake in mice [256], while NT knockout (NTKO) mice displayed decreased fat absorption and decreased body weight, but no changes in food intake, compared to wildtype littermates when challenged with a high fat diet [257]. Additionally, the NTKO mice showed improved glucose tolerance, lower liver fat accumulation and maintained insulin sensitivity over 22 weeks [257]. The acute anorexic properties of NT are currently being investigated as a potential treatment for T2D. Long-acting NT analogs have shown promising synergistic effects when administered together with liraglutide over a 6 day period. Here, co-administration of NT and liraglutide had a synergistic effect, allowing for lower dosages while maintaining efficacy, thereby preventing common adverse effects such as nausea, and increasing tolerability in the patients [258]. This T2D treatment principle is primarily aimed at preventing obesity, and thereby also affects comorbidities such as diabetes. The treatment does not directly regulate pancreatic hormone secretion, and thus NT's position as a potential T2D treatment through regulation of glucagon secretion remains unclear. The observation that NT may attenuate amino acid-stimulated glucagon secretion is interesting, and should, considering the postprandial hyperglucagonemia in T2D, be studied further.

Ghrelin
Ghrelin, a growth hormone secretagogue, was discovered by Kojima and Kangawa in 1999 [259]. It is generally known as the hunger-inducing hormone, but is involved in processes in many different organs and tissues throughout the body, including regulation of energy homeostasis [260]. Thus, genetic knockout of the ghrelin activating enzyme, ghrelin O-acyltransferase (GOAT), in mice, leads to severe hypoglycemia, to the point of near-death after only 7 days on a calorie restricted diet. This was not caused by a lack of hunger sensation in the mice, but rather that ghrelin is essential to maintain blood glucose homeostasis when food is limited [261]. In this study there was no linkage between ghrelin signaling and glucagon levels (plasma glucagon concentrations and liver glycogen were similar in WT and GOAT KO mice). However, in another study of ghrelin deficiency, mice with a deletion of the ghrelin receptor (GHSR) in the hypothalamic arcuate AgRP neurons had lower blood glucose when challenged with a calorie-restricted diet. This response was normalized upon re-expression of GHSR, suggesting a neuronal pathway for ghrelin to maintain blood glucose levels during fasting, that may involve a stimulation of glucagon via ghrelin signaling on AgRP neurons [262]. A further role for ghrelin as a regulator of glucagon secretion is supported by studies showing GHSR expression in α-TC1 cells, and in immunostained mouse and rat islet α-cells [263][264][265]. The reported effects of ghrelin on glucagon secretion are not consistent. Thus, a careful study by Chuang, J.-C. and colleagues from 2011 concluded that in addition to GHSR expression and receptor protein on the α-cell, injections of physiological concentrations of ghrelin stimulated glucagon secretion dose-dependently in mice, and this effect was abolished in mice lacking the GHSR [263]. Furthermore, elevated glucagon levels were detected in mice after long-term (20 day) elevation of ghrelin (achieved by using osmotic pumps), and in a mouse strain overexpressing ghrelin [263]. In contrast, perfusion studies in rats and mice show either no effect of ghrelin on glucagon secretion [266], or a slight inhibitory effect, caused by the concurrent ghrelin-stimulated somatostatin increase [267], and ghrelin administered in randomized studies in humans, did not significantly change glucagon levels, regardless of the prandial state [268][269][270][271]. Likewise, a study investigating the effects of a GHSR inverse agonist in healthy humans reported no changes in plasma glucagon concentrations [272]. There is also the possibility that local ghrelin produced by ε-cells may regulate hormone secretion and plasma glucose. This was investigated in a recent study using Ghrelin KO and GHSRKO mice, that showed no involvement of locally produced ghrelin in maintaining plasma glucose during both a mixed meal or intraperitoneal glucose tolerance test, suggesting that the effects associated with ghrelin are not paracrine, but are instead driven by circulating ghrelin derived from the stomach [273] or intestines [274,275]. In rodent models of diet induced obesity, inhibition of ghrelin signaling increases the otherwise attenuated insulin secretion and enhances glucose tolerance, without affecting either food intake or weight [276]. Because GHSR is coupled to Gα q proteins, inhibitory effects of ghrelin on insulin would seem paradoxical, but the unique inhibitory properties of ghrelin may be explained by GHSR heterodimerizing with the SSTR5, enabling an inhibitory signaling pathway [277]. Thus antagonism of GHSR may relieve inhibitory tone on β cells, and restore glucose tolerance in diabetes [276]. Also, a stimulatory role of ghrelin on glucagon in diabetic rats has been reported [278]. However, as there are currently no human studies pointing towards ghrelin as a potential regulator of glucagon secretion, it is yet unclear whether these findings translate into humans. Thus, more work is needed to elucidate the potential of ghrelin as a treatment for diabetes in humans.

Arginine-vasopressin (AVP)
Arginine-vasopressin (AVP) is an antidiuretic hormone produced in the hypothalamus, from where it is transported to the posterior pituitary to enter the systemic circulation [279]. It is best -known for its actions on promoting water retention by modulating blood flow [280], and increasing water reabsorption in the collecting ducts of the kidney [281]. AVP also has glucoregulatory effects, elicited through the different AVP receptor sub-types; V2R in the renal collecting ducts [282], V1aR located in the liver and vasculature [283,284], V1bR in the pancreas, and both V1aR and V1bR in the adrenals [285]. AVP is released in equimolar amounts with copeptin, a stable C-terminal fragment of the AVP precursor [282] for which robust assays have been developed. Elevated plasma copeptin is associated with obesity, metabolic syndrome, insulin resistance, and may be used as a predictor for development of diabetes [282,286,287]. In diabetes, plasma AVP is increased in both rodent models [288,289] and humans [290,291].
V1bR is expressed by mouse α-cells [292] and αTC1 cells [293], and administration of AVP stimulates glucagon in isolated islets, perfusion studies and in vivo [292,[294][295][296][297], with greater efficacy in streptozotocin induced diabetic rats [294]. AVP may also directly regulate insulin secretion, through V1bR expressed on β cells. The effects of AVP on both glucagon and insulin are highly dependent on prevailing glycemia. Thus, glucagon is stimulated at hypoglycemia, and insulin is stimulated at hyperglycemia, with glucagon stimulation occurring at lower AVP concentrations than insulin [298]. In addition to the hypothalamic source, AVP is also produced peripherally, and was detected in effluents from the perfused rat pancreas [294]. Supporting this, immunostainings detected AVP in the perivascular tissues of the islets [299], suggesting a possible role of locally produced AVP in the regulation of islet hormone secretion. Still, the relative contribution of central and peripheral AVP in the regulation of glucagon secretion is yet to be fully unveiled. Recently, Taveau and colleagues demonstrated that in acute in vivo experiments on rats, AVP injections increased glycemia dose-dependently. These effects were attenuated by prior administration of a V1aR specific antagonist, whereas V1bR antagonists had no effect, suggesting that the AVP driven hyperglycemia primarily stems from gluconeogenic effects on the liver [300][301][302] (that expresses V1aR). V1bR antagonism did however induce a more pronounced inhibition of glucagon than hyperglycemia alone, confirming a glucagonotropic effect of AVP [303]. Similar effects were observed in long-term experiments of continuous AVP infusion, where plasma glucose was raised by 1.1 mmol/L, with no concurrent change in insulin. This increase was normalized by V1aR antagonists [303]. Thus, AVP has glucagonotropic properties both in vivo and in vitro, but in vivo, the induction of gluconeogenic/glycogenolytic pathways by AVP in the liver, rather than dysregulation of glucagon secretion, appears to be the primary contributor to hyperglycemia. It is however important to consider that circulating plasma levels of AVP and oxytocin (reviewed in 4.2) in humans are much lower (low picomolar range [304]) than the concentrations utilized in these studies, and thus the reported effects are likely less pronounced in humans. A promising therapeutic potential of AVP analogs, targeting V1aR and V1bR was recently shown in mice on a high fat diet. Twice daily injection of the analog Ac3IV markedly improved several parameters of metabolic syndrome, including body weight, food intake, decreased circulating fat, and reduced plasma glucose. The latter appeared to involve a decrease in plasma glucagon levels, and an increase in insulin secretion and sensitivity [305].

Oxytocin (OT)
Oxytocin (OT) is a 9-amino acid neuropeptide produced in the supraoptic (SON) and paraventricular (PVN) nuclei of the hypothalamus, from which it is transported via axonal projections, to facilitate its central and peripheral effects [306], including modulation of social behavior and pair-bonding, the reward system, feeding, and energy control [306,307]. The OT receptor (OXTR) is expressed in many regions of the CNS [308], as well as in the enteric nervous system [309]. OXTR was also detected in islet α and β-cells by RT-PCR, immunostainings and autoradiography [310,311], supporting direct effects of OT in regulating both insulin and glucagon secretion [312][313][314]. OT is a glucagonotropic hormone, which is secreted, like glucagon, in response to hypoglycemia [315]. Thus, injections of OT in dogs increased plasma glucose, accredited to a stimulation of glucagon, as OT in this setting was unable to increase plasma glucose when glucagon and insulin levels were clamped [316]. Conversely, while similar experiments in humans also showed an increase in glucagon levels in response to OT infusion, OT retained its ability to increase blood glucose when endogenous hormone secretion was clamped, suggesting that in humans, OT can raise plasma glucose through effects on the liver [317,318]. To date only one type of OXTR has been identified. OXTR is a GPCR that signals through either G q or G i/ o-coupled g-proteins, the activation of which is highly dependent on prevailing OT concentrations. Thus the G q pathway is activated at low concentrations, and G i/o at higher concentrations [319], but further details on how OXTR activation regulates glucagon secretion are currently missing.
Because OT has anorexigenic properties, it is an interesting candidate for treating obesity and diabetes, and may because of its position as a down-stream effector of leptin signaling, be a possible treatment for obesity in leptin resistant individuals [320]. Using GLP-1R antagonists, it was found that OT depends on a functional GLP-1 system in order to fully elicit its anorexigenic effects [321], and a study on the effects of two novel OT analogs, showed that glucose tolerance was improved in response to an OGTT, compared to an i.p. glucose challenge, suggesting an involvement of incretins [322]. Substantiating the importance of a functional OT system, a rare variant of the OXTR gene is associated with the development of obesity [323], and OXTKO and OXTRKO mice showed decreased insulin secretion, increased insulin resistance and impaired glucose tolerance [324,325], morbidities that were improved by OT infusions in another study involving DIO mice [326]. Injections of OT are however not always beneficial, with some studies showing increases in blood pressure [327] corticosterone levels and worsening of glycemic control in obese mice [328]. AVP and OT are almost identical (both 9 AA peptides, with only 2 AA difference [329]), and their receptors are also highly homologous. Thus, OT may activate either of the three AVP receptors (V1aR, V1bR and V2R) [330], activation of which could explain the hyperglycemia and hypertension in these mice [327]. To avoid this, researchers have developed OT analogs with higher specificity for the OXTR that retain their ability to correct glucose tolerance, induce weight loss, improve their lipid profile, and inhibit gluconeogenic pathways [327,328]. Thus, injections of OXTR specific OT analogs in mice increased both insulin (although a stimulation of insulin was not found in human islets), glucagon and GLP-1, with improvements in both body weight, lipid profile and a lowering in expression of genes associated with gluconeogenesis. However, after a 2-week treatment glucose tolerance was not improved, which may be explained by the increased glucagon levels [327]. A recent study likewise tested two novel OT analogs in mice on a high fat diet, and both were found to improve several metabolic parameters, including both body weight, cholesterol, insulin sensitivity, glucagon and insulin secretion [322]. Interestingly, in stark contrast to the well-established glucagonotropic effects of OT, glucagon secretion was inhibited by both these analogs. Because of the unique coupling and preferential (depending on the conditions) activation of either G q and G i/ o g-proteins by the OXTR, a possible explanation for the effects of the OT analogs could be that they are either antagonistic towards the G q -coupling, or agonistic towards the G i/ o-coupling. This is not unlikely, as a similar mode of action was shown for atisoban [331], a clinically used OXTR antagonist.

Conclusion
As more and more hormones are being explored for their potential to treat diabetes and metabolic disorders, there is an increased need to understand the details of the individual pathways initiated by these, and the potential interplay that may be involved. Here, we have summarized the current literature on peptide hormones, analogs and inhibitors with the ability to regulate glucagon secretion, and the unique signaling pathways initiated by these. We furthermore present gaps in the knowledge surrounding the use of these (Table 1). These gaps are currently preventing firm conclusions to be drawn for many of these peptides regarding their effects on glucagon secretion and potential clinical application.

Declaration of Competing Interest
The authors report no declarations of interest.

Table 1
The effects on glucagon secretion in vitro, in vivo and in humans by native pancreatic, intestinal, thyroid and hypothalamic peptide hormones, their analogs and inhibitors.