Conflicting Views About Interactions Between Pancreatic α-Cells and β-Cells

The possibility that optimal pancreatic insulin secretion depends on β-cells being stimulated by paracrine secretion of glucagon from neighboring α-cells is of great current interest (1–10). Interactions between β-cells, α-cells, and δ-cells can be paracrine, which implies that cells must be close to each other. However, the concept is complicated by the postulation over 30 years ago that islet blood flows from the β-cell–rich core to the α-cells and δ-cells contained in the islet mantle. While this core–mantle anatomical arrangement would facilitate insulin’s suppressive effect on downstream α-cells in the mantle, there could be effects of glucagon very locally on the adjacent peripheral β-cells (11–14). However, the majority of the β-cells would not be influenced by intraislet glucagon, since glucagon would have to move upstream. The case for somatostatin differs, since the δ-cell filipodia extend the δ-cell’s reach (15,16). These possible relationships between islet cells have been studied by numerous laboratories with a variety of technical approaches, but clear answers have not emerged.

The authors of this perspective have been observers and participants in the field for decades and favor the view that blood flow from core to mantle is operative and important for islet function. We are surprised that this mechanism is receiving so little attention, because the data supporting its importance have considerable strength. In particular, the perfused pancreas technique maintains the normal anatomical relationships between cells and blood vessels and lends itself to passive immunization studies with antiserum or antagonists against whatever secretory product is being studied. While the local paracrine model of α-cells directly influencing β-cell function has its appeal, we are concerned that some of the methods used to study the problem disrupt normal anatomical relationships, making the results difficult to interpret.

A major reason for writing this perspective is to express our concerns about the lack of appreciation of the problems and artifacts associated with the experimental use of isolated islets. However, some researchers have used pancreas perfusions (17) and even isolated islets to generate data that support the paracrine hypothesis. We look forward to future use of pancreas perfusions and a variety of new tools to study blood flow.

In 1973, Gerich et al. (18) published the striking finding that in type 1 diabetes, the blood glucagon response to insulin-induced hypoglycemia disappears. Because of the proximity of β-cells and α-cells, investigators wondered if local insulin secretion has a suppressive influence on α-cells that is lost with diabetes. Therefore, for people without diabetes, falling glucose levels shut down insulin secretion, which removes its local inhibition of glucagon secretion. Several investigators designed experiments to learn more about the local effects of insulin on glucagon secretion (19,20). It was important to understand whether the α-cells were a unique target for insulin, since insulin within the islet must be at enormously high concentrations compared with the circulating levels. This question was addressed with the perfused rat pancreas in 1976 (19), and it was found that glucagon secretion could be suppressed by an insulin concentration of 20,000 μU/ml (800 ng/ml) but not by 2,000 μU/ml (80 ng/ml). Normal fasting plasma insulin levels are in the range of 10 μU/ml (400 pg/ml). A little later, studies of the islet vasculature described a pattern whereby small arterioles penetrated gaps in the islet mantle of non–β-cells and formed capillaries in the islet core that then carried blood back through the mantle to exit the islet (11). This fits with the possibility that insulin secreted from the β-cell–rich islet core would bathe the mantle cells in high concentrations of insulin. Later, it was proposed that β-cells influence local glucagon by means of products other than insulin (21), including γ-aminobutyric acid, glutamate, ghrelin, zinc, and serotonin (22). However, as the field progresses, insulin and somatostatin remain the most attractive candidates.

These findings do not tell us if insulin is acting directly on α-cells or indirectly through some other mechanism. The hypothesis about the importance of local insulin secretion received a boost from experiments by Maruyama et al. (13) in 1984, when they employed the perfused rat pancreas to show a very large increase in glucagon secretion when insulin antiserum was infused. This hormone neutralization approach was then used to perform antegrade and retrograde perfusions of pancreases from rats (Samols et al. [14]), dogs (Stagner et al. [23]), nonhuman primates (Stagner et al. [24]), and humans (Stagner and Samols [25]). These experiments led to the conclusion that blood flow in all these species goes from β-cells to α-cells to δ-cells (11,12,14). It is not possible to measure the flow of interstitial fluid, but it seems likely that it goes in the same direction as blood flow. Thus, α-cells and δ-cells were thought to be downstream from β-cells, indicating that the flow of blood, and probably interstitial fluid, would impede the movement of α-cell and δ-cell secretory products toward β-cells. The retrograde perfusion technique of Stagner and Samols is novel, and the field could benefit from further analysis and confirmation of those studies.

Most of the methods used for studying islet function do not provide useful information about paracrine effects. Isolated islets are an example of the problem. Most studies cited as providing secretion data supporting the paracrine mechanism used static incubation or perifusion of islets (26–28). One obvious concern is central hypoxia. Islets in tissue culture that are over 150 μm in diameter in culture will have hypoxic or sometimes even necrotic tissue in the center of the islet. In addition, without normal blood flow the vascular channels collapse and within 24 h in culture are no longer visible (29). This results in a more compact islet in which glucagon will diffuse and accumulate throughout the islet, creating a very different environment than what happens in vivo. Perifused pancreatic slices would be expected to have the same problems (30). Whatever paracrine relationships that might have been operative in vivo would be obliterated. In contrast, with the pancreas perfusion, the anatomic connections are intact. The perfusate is usually an oxygenated modified Krebs-Ringer buffer with dextran (for oncotic pressure) to which one adds different glucose concentrations, hormones, drugs, neutralizing antibodies, or molecules of interest. The use of such a defined perfusate removes confounding factors normally contained in blood, which can be seen both as a benefit and a fault. It is regrettable that the isolated perfused pancreatic islet is now used so rarely.

Since the 1970s, there has been intermittent interest in the existence and potential importance of this specialized core–mantle vascular system, but in recent years has often been often ignored or dismissed. The results described above using the perfused pancreas technique indicate that α-cells are downstream from β-cells, suggesting that insulin secretion is not dependent upon paracrine secretion from α-cells.

From studies mainly performed on rodents, we know that islets receive more blood flow than the rest of the pancreas (31) and have a higher vascular density than the exocrine pancreas (32). The islet blood flow is at least partially regulated by glucose (33), neural input (34), and pericytes (22) and can increase when blood flow to the rest of the pancreas does not (35). As we described based on corrosion casts and reconstructions of serially sectioned islets after infusion with India ink (36), short arterioles enter the islet directly into the β-cell area through gaps in the discontinuous non–β-cell mantle and immediately branch into a glomerulus-like structure of tortuous capillaries that pass through the islet before coalescing into a cage-like network of postcapillary venules. Efferent vessels fuse into large venules at the edge of large islets, but those from smaller islets course through exocrine tissue for 100 μm or so before coalescing. Thus, there is direct venous drainage in some islets (mainly the larger islets) as well as a distinct islet-acinar portal system in others; there was also direct arteriolar blood flow to the acinar tissue. This pattern of vasculature has been reported by Ohtani and associates (37–39) but with the arteriole branching into capillaries in the “cortical area,” which they concluded was the non–β-cell mantle. However, without clear identification of the cell types with immunostaining, one cannot state that the vessels go from mantle to core. We and others have shown that the mantle is discontinuous, and the Hara group (40) reported that, theoretically, the mantle of non–β-cells would not cover the whole β-cell core and even should not be called a mantle.

The vascular structure we defined suggests that blood enters the islet core before spreading outward across the islet, a pattern reported long ago with in vivo imaging by Liu et al. (41). In vivo imaging of blood flow has improved over the years. Even so, for interpretation, seeing the location of the afferent arteriole (i.e., the orientation of the islet) is necessary but is not always easy with a depth of imaging of only 50–100 μm. Measuring dynamic blood flow using mouse insulin promotor–green fluorescent protein (MIP-GFP) mice to visualize the islets and rhodamine-dextran injected into the blood system, Nyman et al. (42) reported that the islet blood flow pattern was inner to outer in 12 of 20 islets. A top-to-bottom pattern (that they acknowledged could be a reoriented inner-to-outer pattern) was seen in 7 of 20 islets, and only 1 of 20 islets had an outer-to-inner pattern. Based on their data, they concluded that “it is unlikely that secreted products of α-cells influence β-cells.” In a subsequent study using fluorochrome-labeled red blood cells (RBCs) (43), they reported that “the islet core of β-cells usually perfused first.”

In contrast, recently the Hara group (40) used similarly labeled RBCs and MIP-GFP mice to argue for the presence of a bidirectional open blood flow between the islet and acinar tissue. However, they did not focus on the filling of the islet but rather on the ongoing blood flow, and no clear afferent arterioles are visible in their published images and videos of mouse islets to give orientation of the islet. The tracking of RBCs to follow direction of blood flow was thus shown only for the efferent vessels. Moreover, they did not acknowledge that their findings were inconsistent with past literature on blood flow done with microspheres or in vivo imaging as well as ultrastructural data on islet capsule and the findings of corrosion casts. For these reasons, this work does not provide strong evidence against the core–mantle hypothesis.

Overall, while more work is clearly needed, the weight of evidence at present supports the pattern of core–mantle blood flow in rodent islets.

It is commonly said that the ratio of β-cells to non–β-cells is much lower in human than in rodent islets (44,45), which leads some in the field to dismiss results obtained from rodents as being irrelevant. However, a thorough review of published work casts doubts on this conclusion. Our quantification of the composition of 33 isolated human islets preparations (29) found that 73.6% of the human islet cells were β-cells, which is similar to levels in rodent islets. Our review of the literature (Table 6 of the article by Pisania et al. [29]) found similar values in many, but not all, reports. There is evidence that the larger islets have a greater proportion of α-cells (46,47), which could lead to selection bias if only a few islets were analyzed. The lack of standardization of the measurement techniques makes it difficult to draw definitive conclusions, but it seems likely that the relative number of β-cells in humans does not differ greatly from that of other mammals.

Additionally, interactions between the cell types in human islets are said to be different than those in rodents because there was no distinct organization of the cell types, with α-cells and δ-cells in the middle of large islets (44,45). However, when looked at carefully, human islets most often consist of subunits that have a core-to-mantle relationship that is very similar to that of rodent islets. In our 2015 article (47), pancreases of seven nondiabetic adult human donors were examined in detail. The number of islets examined in each ranged from 50 to 275, and of these, over 85% had obvious β-cell core subunits. Using high-resolution imaging and fixed human pancreatic slices (150 μm thick), Cohrs et al. (48) reported that in about 80% of human islets, β-cells were contiguous as one or two clusters, and Cottle et al. (49) commented on the general core of β-cells and a mantle (α-cells) organization. Using serial paraffin sections Bosco et al. (50) described human islets as having “insulin-expressing cells seemed clustered into discrete ovoid areas surrounded by α-cells” but with no CD34⁺ blood vessels penetrating the β-cell cores. This lack of penetrating blood vessels into β-cell cores is likely a technical problem, since the combination of anti-CD34 and anti-smooth muscle actin showed vessels lined by β-cells in other studies (44,47) and human β-cells show polarity in relation to capillaries (49).

Human islets were found to have less dense vascularity than mouse islets, reported as twofold (48) or fivefold (51) less islet area, and their capillaries were less tortuous. Nonetheless, most β-cells were seen to be within 10 μm of the capillaries (52). Importantly, Cohrs et al. (52) found 85% of the β-cells were within 10 μm of α-cells. An important caveat to this study is that only 25 total islets (four donors) were used in the analysis, and larger islets have a greater proportion of α-cells (46,47). Nonetheless, such close proximity to α-cells by most β-cells could allow more paracrine effects. However, the pattern of blood flow or even the organization of the islet vasculature in humans is still unknown. In one of the first three-dimensional reconstructions of human pancreas using modern whole-mount techniques, Fowler et al. (53) found a large vessel (alternatively described as a large capillary or a smooth-muscle actin-positive afferent arteriole) entering the islet and then branching extensively, much as we described in the rat. While tracing blood flow in transplanted human islets or in pancreatic slices is unlikely to reflect what happens in situ, new techniques continue to emerge, and we should expect this field to move forward.

Since the early studies of Wharton in 1932 (54), there has been the concept of an islet-acinar portal system in which blood flows from islets to the acinar tissue (31,37–39). In rodents, capillaries leaving the islet coalesce into postcapillary venules, which are thought to be leaky, so the adjacent acinar tissue would be bathed in high levels of islet hormones. This arrangement fits with the concept of there being an islet-acinar relationship postulated by Henderson and associates (55,56).

Why would an islet–acinar tissue axis relationship be favored evolutionarily in mammals, and by extension, why did islets evolve to be scattered throughout the mammalian pancreas? In nonmammalian species, large portions of the endocrine pancreas are not integrated with the exocrine pancreas, with examples of this being the “principal islets” in bony fish (55). Additionally, in chickens there are four lobes of the pancreas, and the splenic lobe consists almost entirely of endocrine tissue. An explanation articulated by Henderson (55) was that there was an evolutionary advantage in having secretion from islet cells exerting trophic effects on exocrine cells, thus enhancing digestion and nutrient absorption to provide a survival advantage. While insulin clearly has anabolic effects, the effects of glucagon upon exocrine cells are less certain. The concept that the exocrine cells receive anabolic signals from the islets is supported by anatomically larger acinar cells close to islets (“peri-islet”) compared with those further away (“tele-islet”) (57). More recently, using single-molecule fluorescence in situ hybridization, a zonation pattern for gene expression around the islet was shown (58). Among many findings, the authors also showed evidence of increased cholecystokinin (CCK) and mTOR signaling in the peri-islet region. Their data that implicate CCK coming from islets do not exclude a similar major role for insulin.

Other data supporting the anabolic effects of islets on the exocrine pancreas come from the finding of atrophy of the exocrine pancreas in type 1 diabetes (59,60). Similar findings have been found in a variety of animal models.

One of our main points is that the choice of an experimental system to show paracrine effects is critical. Moens et al. (61) used the perfused pancreas to examine the effects of blocking glucagon-like peptide 1 receptors (GLP-1R) with exendin-(9-39)-NH₂ and glucagon receptors with [des-His¹,des-Phe⁶, Glu⁹]glucagon-NH₂ on glucose-induced insulin secretion (GSIS). Their striking result was that blocking GLP-1 and glucagon effects had no detectable effect on GSIS. These data fit with the concept that glucagon from α-cells in the mantle does not flow in meaningful amounts into the β-cell core. In contrast, with perifused islets glucagon had a strong paracrine effect, as shown by Cabrera et al. (62), who used a similar approach of blocking the effects of glucagon with the glucagon receptor antagonist (LY2786890) or a GLP-1R antagonist (exendin-[Ex9-39]). Thus, this effect of glucagon on GSIS was only seen when the normal compartmentalized relationship of the core–mantle anatomy was disrupted, as in isolated islets.

We have known for some time that both glucagon and GLP-1R are present on β-cells, and there has been considerable interest in the possibility that both GLP-1 and glucagon secreted from α-cells have a direct paracrine stimulatory effect on β-cells (63). After early debates, we can agree with the conclusions that GLP-1 secreted from the intestine is rapidly destroyed in the circulation and has no proven effect on pancreatic β-cells (3). While we know that α-cells contain large amounts of glucagon, there are doubts, expressed in recent reviews (3,64), as to whether α-cells contain enough GLP-1 to have any meaningful effects on β-cells. This is an important point, because GLP-1 has been reported to be contained in and secreted by α-cells (27,65–68). If we accept the concerns of Holst (3) and others (64) that the amounts of GLP-1 in islets are negligible, we are left with knowing that α-cells contain large amounts of glucagon, which could stimulate β-cells through GLP-1 and glucagon receptors (64). However, we cannot be sure that the glucagon produced within islets has an important intraislet influence on β-cells.

A study by Svendsen et al. (17) approached the question by using pancreas perfusion on genetic mouse models (global GLP-1R knockout, diphtheria toxin–induced proglucagon knockdown, and global [Gcgr^−/−] and inducible β-cell–specific glucagon receptor [Gcgr] knockout mice). They found that at high glucose concentrations (12 mmol/L), infused glucagon led to increased insulin secretion and that the loss of glucagon or the GLP-1R alone did not alter this; only the loss of both led to severe impairment. The findings are probably best explained by glucagon acting on both glucagon and GLP-1R on β-cells. These studies were very well done. The finding that Ex9 without exogenous glucagon can inhibit insulin secretion in the perfused pancreas of wild-type mice is of considerable interest; similar data have been found in humans (69). However, these are early days, and more must be done to clarify the role of GLP-1 and the specificity of Ex9. In addition, concerns can be raised about the chronicity of these models in that the deletion of the receptors can lead to a number of compensatory changes in both the structure (70) and function of the islets.

Another approach by Zhu et al. (5) used designer receptors exclusively activated by designer drugs (DREADD) to determine if secretory products from islet α-cells (glucagon and perhaps GLP-1) could enhance insulin secretion. This study used an inhibitory designer G protein–coupled receptor, hM4Di, which is a modified form of the human M4 muscarinic (hM4) receptor that can be quickly activated by the inert clozapine metabolite clozapine-N-oxide (CNO), engaging the Gi signaling pathway in α-cells. Within minutes after injection of CNO, glucagon and insulin levels fell, and glucose levels were higher after a glucose challenge. These in vivo results fit with the proposed paracrine interaction, so we hope there will be follow-up. One concern is that the specificity of CNO has been questioned, in that it binds to alternative targets at concentrations required for DREADD activation (71,72). The additional in vitro experiments by Zhu et al. (5) using perifused isolated islets to determine the effects of GLP-1R blockade with exendin-9 and the glucagon receptor inhibitor adomeglivant are problematic, because the normal relationships between α-cells and β-cells have been disrupted with islet isolation. Interestingly, they showed that glucagon enhancement of insulin secretion was exerted through the GLP-1 receptors.

The question of beneficial paracrine effects of α-cells on β-cells could be important now that islet cells are being generated from embryonic stem cells and induced pluripotent stem cells. If the technology were to generate pure β-cells, would the cells function well enough when transplanted, or will they need accompanying α-cells and δ-cells for optimal function?

It is not yet possible to obtain completely purified β-cells, but with flow cytometry of dispersed rat islet cells, a purity of 95% can be obtained. When we transplanted these cells (doses ranged from 1.2 × 10⁶ to 2.5 × 10⁶ cells) under the kidney capsule of streptozotocin-treated diabetic mice, they maintained glucose control for the 12-week period of study (73). The purified preparations consisted of 95% β-cells at the beginning and end of the study. When the distribution of cells in the transplant site was assessed, it turned out that 67% of the β-cells were more that 50 μm away from an endocrine non–β-cell, suggesting that the effects of paracrine secretion from α-cells were very limited. Similar results with transplanted purified rat islet cells (1.2 × 10⁶ cells) for even longer time periods were reported by Keymeulen et al. (74). However, their results were enhanced by adding purified α-cells. More work must be done to determine what mixture of cell types is best for transplants.

Another example of β-cells probably working well without help from α-cells is ob/ob mice, which have robust insulin secretion from very large islets consisting of almost 90% β-cells (75,76). The β-cells in the middle of these islets probably see very little paracrine glucagon or GLP-1 from the α-cells in the mantle.

As research has advanced, it was recently proposed that local insulin secretion exerts its effects on α-cells indirectly through paracrine secretion of somatostatin from δ-cells (6,77–79). While the hypothesis is attractive, it has been difficult to design definitive informative experiments. A variety of experimental approaches have been used, including the use of the peptide insulin receptor antagonist S961 and β-cell insulin receptor knockout mice. Other studies have used CYN154806, which inhibits somatostatin receptors SSTR2 and SSTR3 (78). Informative results from in vivo studies (intravenous injection of glucose or arginine) in somatostatin-null animals showed enhanced secretion of insulin and glucagon (6); these data support some paracrine effects of somatostatin.

Special mention should be made of the unique morphology of δ-cells, because they have filopodium-like structures that may be able to reach and influence cells some distance away (15,16). We can only speculate about the reach and density of these filopodium-like structures. Is somatostatin only released at the end of the filopodium-like structures? How far might these structures penetrate from the mantle into the islet core? δ-Cells may have a bigger role in regulating β-cells than previously thought.

However, one must ask questions about the differences between somatostatin-secreting cells found in different locations. There seems to be no reason for similarly appearing endocrine cells in the intestine or elsewhere to be responsive to insulin. Why would δ-cells in the islets evolve to respond to insulin?

At present, few experiments have been conducted, and it remains to be seen if the local effects of insulin on α-cells require paracrine secretion from δ-cells. There is excitement about ways in which somatostatin could be a major control mechanism, but we should be cautious about overinterpreting the limited data that are available.

The relationships between the three main islet cell types (α-cells, β-cells, and δ-cells) have been studied for decades, and major questions about their interactions continue to be debated. Our understanding of δ-cells and the effects of somatostatin has improved, but the conclusion that somatostatin is critical for glucose suppression of glucagon secretion is not yet on solid ground. The question of whether optimal insulin secretion is dependent on paracrine effects from α-cells is being intensively studied and gaining support. We also believe that the view of islet cells having a core–mantle relationship, with islet blood mainly flowing from β-cells to α-cells, has strength. The tools now available to islet biologists should provide definitive answers in the near future.

Funding. Funding for this work came from a grant from the National Institute of Diabetes and Kidney and Digestive Diseases, National Institutes of Health (DKO 0/36 836).

Duality of Interest. No potential conflicts of interest relevant to this article were reported.