The physiological role of β-cell heterogeneity in pancreatic islet function

Abstract

Endocrine cells within the pancreatic islets of Langerhans are heterogeneous in terms of transcriptional profile, protein expression and the regulation of hormone release. Even though this heterogeneity has long been appreciated, only within the past 5 years have detailed molecular analyses led to an improved understanding of its basis. Although we are beginning to recognize why some subpopulations of endocrine cells are phenotypically different to others, arguably the most important consideration is how this heterogeneity affects the regulation of hormone release to control the homeostasis of glucose and other energy-rich nutrients. The focus of this Review is the description of how endocrine cell heterogeneity (and principally that of insulin-secreting β-cells) affects the regulation of hormone secretion within the islets of Langerhans. This discussion includes an overview of the functional characteristics of the different islet cell subpopulations and describes how they can communicate to influence islet function under basal and glucose-stimulated conditions. We further discuss how changes to the specific islet cell subpopulations or their numbers might underlie islet dysfunction in type 2 diabetes mellitus. We conclude with a discussion of several key open questions regarding the physiological role of islet cell heterogeneity.

Cellular heterogeneity is a general feature of biological systems and can arise as a result of stochastic gene expression, differences in genetic programmes underlying cell specification or the context of the cell within a tissue. Cell heterogeneity can be important to the function of the whole tissue or organ. For example, specialized pacemaker cardiomyocytes set the rhythm of periodic contractions that underlie cardiac function. The altered presence or dysfunction of these pacemakers in disease can underlie arrhythmias that disrupt cardiac function¹. Thus, understanding the role of cell heterogeneity is important to understanding the function and pathogenic dysfunction of a given tissue.

The pancreatic islets of Langerhans provide the sole source of circulating insulin and glucagon to regulate the homeostasis of glucose and other energy-rich nutrients. Endocrine cells within the islet of Langerhans have long been known to be heterogeneous in terms of transcriptional profile, protein expression and the regulation of hormone release². Detailed molecular analyses have been conducted to understand the basis of this heterogeneity and to gain clues as to how this heterogeneity influences islet function. The proportion of heterogeneous cell populations also changes during diabetes mellitus. However, how this heterogeneity affects the regulation of hormone release within the islets of Langerhans is not fully clear.

In this Review, we first describe the general features of heterogeneous β-cells and the characteristics of defined β-cell subpopulations in terms of their regulation of insulin release. We then overview how β-cells and other endocrine cells within the islet can interact and how functional subpopulations of β-cells exert control over islet function. Finally, we discuss additional aspects of functional heterogeneity among β-cells and other endocrine cells of the islet and address open questions.

What is β-cell heterogeneity?

In islets of Langerhans, β-cells secrete insulin upon direct in vitro exposure to elevated levels of glucose (>8 mM in mice^3,4, >5 mM in humans⁵). Histological analysis of mouse or human islets reveals little morphological distinction between β-cells, although some variation occurs in granularity^2,6. Although these findings imply that β-cells are homogeneous, a number of landmark studies revealed substantial variation in insulin release between individual β-cells dissociated from rodent or human islets^7–10. Some β-cells released substantial levels of insulin even at low concentrations of glucose, whereas others did not secrete much insulin despite elevated concentrations of glucose (FIG. 1a). These differences in hormone release therefore indicate β-cells to be functionally heterogeneous.

Fig. 1 |. Heterogeneity in insulin secretion by β-cells.

a | Histogram of the insulin released by dissociated human β-cells at low (2 mM) and high (22 mM) concentrations of glucose. Insulin release is measured by reverse haemolytic plaque assay⁹. Note the high spread in release levels: at the low concentration of glucose, there are cells that secrete more insulin than the average seen at the high concentration of glucose. Conversely, at the high concentration of glucose, there are cells that secrete less insulin than the average seen at the low concentration of glucose. b | Schematic histogram for glucose sensitivity of the β-cell population when dissociated from the islet². The x-axis shows the concentration of glucose at which β-cells become activated, indicated by elevated metabolic activity measured by NAD(P)H autofluorescence, protein synthesis, insulin biosynthetic activity or elevation in calcium concentration ([Ca²⁺]). The y-axis indicates the number of cells. Cells that are activated by ≤6 mM glucose are considered activated. c | Schematic histograms for the glucose sensitivity of the β-cell populations within the islet, based on published studies^55,70,74,76. Due to intra-islet communication, β-cells that are less responsive to glucose can suppress more-responsive β-cells (solid arrow, left panel). In addition, β-cells that are more glucose responsive can recruit less-responsive β-cells (solid arrow, right panel). Further, more subtle heterogeneity exists within the islet beyond these two defined populations (not shown). We speculate that inhibitory δ-cell somatostatin release might contribute to the suppression of glucose-responsive β-cells (dashed arrow, left panel). We also speculate that stimulatory α-cell glucagon release might contribute to the recruitment of non-excitable β-cells (dashed arrow, right panel). Figure 1a is reprinted from REF.⁸, Springer Nature Limited.

Heterogeneity in the functional properties that underlie insulin release has also been observed (FIG. 1b). For example, heterogeneity was observed in the autofluorescence of β-cells dissociated from rodent islets (caused by variable amounts of the redox cofactors FAD or NADH and NADPH, which together are referred to as NAD(P)H), which indicates differences in metabolic activity^7,11,12. Furthermore, β-cells dissociated from rodent islets show different levels of overall protein synthesis or specifically insulin synthesis^7,13. Finally, differences were observed in the pattern of electrical activity and elevation of calcium concentration ([Ca²⁺]) in β-cells dissociated from rodent islets^14–17. Although the molecular basis for this heterogeneity is not fully clear, some of the key factors that underlie the regulation of insulin release have long been known to be heterogeneously expressed; for example, factors such as glucokinase¹⁸ and insulin granularity⁷.

High-throughput single-cell approaches have also demonstrated islet cell heterogeneity. For example, single-cell RNA sequencing (scRNAseq) and mass cytometry have separated β-cell subpopulations defined by different gene expression or protein expression profiles^19,20. Related reporter-based studies have separated functional subpopulations, as discussed later (BOX 1). How these and other functional subtypes of β-cells might be specified falls outside the scope of this Review^21,22. Herein, we focus on discussing how these differing β-cell populations might be affecting islet function, the regulation of insulin release and glucose homeostasis (FIG. 1c).

Box 1 |. What defines a β-cell subpopulation?

When talking about β-cell heterogeneity, we must consider the methods by which we define specific subpopulations. A β-cell is defined as a cell that expresses and releases insulin. Some might apply further definitions, such as releasing insulin under nutrient stimulation or expressing specific transcription factors such as PDX1, NKX6.1 or MAFA¹³⁴. Nevertheless, the definition of a β-cell subpopulation is much looser. A variety of approaches have been taken to separate β-cells into defined pools, including the expression of specific genes as indicated by a genetic reporter, the expression of specific surface markers or combinations thereof, or the expression of specific gene expression profiles (as revealed by dimensionality reduction approaches and clustering)^{19,25–27,29,33,34}. In addition, specific functional characteristics, such as protein synthesis, metabolic activity or Ca²⁺ dynamics, might be used^{7,56,88,89,93}. Of key importance is to validate such signatures. For instance, is the gene expression profile consistent across samples or does it result from a specific instance of dimensionality reduction? Is the functional characteristic used consistent over time or is it some random occurrence? Is the specific genetic reporter or surface protein marker level arbitrarily separating a continuous distribution of expression levels associated with stochastic variation within a single population? Thus, care must be taken to avoid defining subpopulations based on single observations without verifying that the separation occurs robustly.

Molecularly defined β-cell subpopulations

Given that insulin release is highly variable between β-cells, at low concentrations of glucose, some cells secrete substantial amounts of insulin whereas, at elevated concentrations of glucose, some cells secrete fairly low amounts of insulin. By selecting rodent β-cells with low or high NAD(P)H autofluorescence, populations of cells that show differences in glucose responsiveness and insulin release could be isolated²³. For example, cells with high NAD(P)H autofluorescence were more responsive and showed an increased density of insulin granules. This early work demonstrated how functional subpopulations could be isolated. Interestingly, a 2017 study demonstrated that β-cells with low granularity could be identified by low side-scattering and that these cells are resistant to cell death²⁴.

Although many factors that underlie the regulation of insulin release (FIG. 2a) are heterogeneous, molecular markers that distinguish functional subpopulations have been characterized. We describe a number of these markers in this section, which consist of either genetically encoded fluorescent reporters or cell surface markers (TABLE 1). A common observation is that these markers or reporters generally separate two functionally distinct populations of β-cells: one that shows robust insulin release and another that shows more modest insulin release. The former population is often described as being ‘more mature’, showing elevated glucose metabolism or mitochondrial function and sometimes with differing expression of ion channels and G protein-coupled receptors^25–28. By contrast, the latter population sometimes shows signatures of being ‘immature’, including decreased expression of transcription factors important for the β-cell specification^26–29. Although some overlap exists between these populations, this overlap is not exact and smaller subsets of cells than those already described probably exist.

Fig. 2 |. β-Cell intrinsic and extrinsic mechanisms affecting glucose-stimulated insulin release.

a | Schematic of the glucose-stimulated insulin secretion pathway in β-cells. Indicated are the mechanisms common to the mouse and human β-cell (purple) or that are specific to one species (mouse pink; human yellow). Glucokinase (GCK) phosphorylates glucose, which is further metabolized to generate ATP and alter the ATP to ADP ratio (ATP/ADP) and close ATP-sensitive potassium (K_ATP) channels. This closure prevents K⁺ export, depolarizing the cell and increasing its membrane potential (V_m), which activates voltage-gated Ca²⁺ channels (Ca_V) (and in humans, voltage-gated Na⁺ channels; Na_V). This activation leads to [Ca²⁺] influx, which drives the exocytosis of insulin granules. In addition, several G protein-coupled receptors (GPCRs) enhance (for example, glucagon receptor and glucagon-like peptide 1 (GLP1) receptor) or suppress (for example, somatostatin receptor and adrenergic receptors) granule trafficking to the plasma membrane. b | Schematic indicating the role of connexin 36 (Cx36) gap junctions in the islet in controlling islet excitability. The presence of gap junction-mediated electrical currents (indicated by the arrows in the bottom cells) between β-cells makes V_m more uniform, such that hyperpolarized β-cells depolarize more and depolarized β-cells hyperpolarize more. c | Schematic indicating how EphA receptor–ephrin-A ligand bidirectional juxtacrine signalling regulates insulin granule trafficking and exocytosis in β-cells. At basal concentrations of glucose, EphA receptor forward signalling dominates (grey cell). At elevated concentrations of glucose, ephrin-A ligand reverse signalling dominates (pink cell). d | Paracrine factors released within the islet by different endocrine cells and their effect on other endocrine cells within the islet.

Table 1 |.

Role and relevance of markers of mature and functional β-cell subpopulations

Marker, reportera and expressionb	Proportion of positive β-cells	Species investigated	Maturity	Biological role in β-cell	Functional capacity	Ref.
MIP-GFP+	~90%	Mouse	Not studied	Insulin promoter activity	Increased GSIS	25
FLTP-Venus+	~80%	Mouse	Mature	Planar cell polarity effector	Increased GSIS, increased metabolism	27
UCN3+	~98%	Mouse	Mature	Paracrine factor co-secreted with insulin	Increased GSIS, increased Ca2+, increased metabolism	29
Psa-NCAM+	~50%c	Mouse	Not studied	Marker of cell activity (polysialyation of NCAM)	Increased GSIS, increased Ca2+, increased metabolism	33
E-cadherin+	~80%	Mouse	Not studied	Cell adhesion molecule	Increased GSIS	34
CD9-ST8SIA1−	~50%	Human	Mature	CD9 is a tetraspanin protein, ST8SIA1 transfers sialic acid	Increased GSIS	26
RBP4−	~50%	Human	Mature	Transporter protein for retinol	Increased exocytosis, Na+ and Ca2+ currents	37

GSIS, glucose-stimulated insulin secretion; MIP-GFP, mouse insulin promoter-driven GFP; Psa-NCAM, polysialylated neural cell adhesion molecule; UCN3, urocortin 3.

^aReporter protein only listed if relevant.

^bExpression positive (⁺) or negative (⁻).

^cPsa-NCAM⁺ was defined as the 50% of β-cells with the highest expression of psa-NCAM in the islet.

A mouse reporter strain expressing mouse insulin promoter-driven GFP (MIP-GFP) was used to examine heterogeneity in insulin promoter activity^25,30. Cells with a medium to high level of expression of MIP-GFP (~90% of GFP⁺ β-cells in adult MIP-GFP mouse islets) showed increased glucose-stimulated insulin secretion (GSIS); these cells are more granular and express greater levels of insulin. Conversely, cells with a low level of expression of MIP-GFP (~10% of GFP⁺ β-cells) are less granular and show lower GSIS.

FLTP is a planar cell-polarity effector. FLTP⁺ β-cells²⁷ show increased GSIS but normal basal insulin release. These cells make up ~80% of β-cells in the adult mouse islet and are less proliferative and more mature than FLTP⁻ β-cells. FLTP⁺ β-cells show increased density of mature dense-core insulin granules, increased mitochondrial coverage and greater expression of glucokinase and connexin 36 (Cx36), whereas FLTP⁻ β-cells (~20% of β-cells in the adult mouse islet) conversely show lower GSIS, reduced mitochondrial coverage and glucokinase expression, and are more proliferative.

Urocortin 3 (UCN3; encoded by UCN3) is co-secreted with insulin from β-cells and also serves as a maturity marker^31,32. Insulin-positive, UCN3 lineage-negative virgin β-cells (1–2% of β-cells in the adult mouse islet)²⁹ have been suggested to be a pool of ‘progenitor’ β-cells that originate from α-cells. These cells show low glucose uptake and minimal Ca²⁺ currents. They also show low expression of glucokinase, of genes involved in the Krebs cycle and oxidative phosphorylation, of genes encoding Ca²⁺ and K⁺ channel subunits, and of receptors such as the glucagon-like peptide 1 (GLP1) receptor.

Polysialylated neural cell adhesion molecule (psa-NCAM) is a cell surface marker of cell activity in neurons and in β-cells. In rodent islets, psa-NCAM-high β-cells³³ show increased GSIS and basal insulin release. These cells also show increased Ca²⁺ influx, increased cAMP levels and increased ATP generation. Conversely, psa-NCAM-low β-cells show minimal GSIS (but some amino acid stimulation of insulin secretion) and little Ca²⁺ influx³³. E-cadherin-positive rodent β-cells³⁴ also show increased GSIS but normal basal insulin release.

CD9⁻ST8SIA1⁻ β-cells²⁶ show both increased GSIS and reduced basal insulin release, thus releasing insulin with a high dynamic range. These cells make up ~50% of β-cells in the adult human islet and show increased expression of many genes associated with the regulation of insulin release such as genes encoding K⁺ or HCN channels (of note, HCN channels are critical to cardiac pacemaker function). By contrast, CD9⁺ST8SIA1⁺ β-cells show a very low dynamic range of insulin release and represent ~10% of human islet β-cells²⁶.

scRNAseq measurements have several caveats primarily associated with low abundance transcripts³⁵ and do not provide a functional output. However, one observed population from scRNAseq studies is RBP4⁺ β-cells^19,22, which represent ~50% of β-cells dissociated from adult human islets. “Patch-seq”³⁶ (scRNAseq combined with electrophysiological measurements) in human β-cells identified increased RBP4 expression to be correlated with decreased exocytosis and Na⁺ currents but with normal Ca²⁺ currents³⁷. Thus, RBP4⁺ β-cells represent a subpopulation with low functional competency.

Other markers of subpopulations correlate with factors such as endoplasmic reticulum stress and a more in-depth discussion of single-cell-defined markers can be found elsewhere³⁸. Owing to β-cell heterogeneity, we can expect that some β-cells contribute to the level of insulin release to a greater degree than others. However, many of the studies described above examine the function of β-cells in isolation or as a pure population within an aggregate and not in the context of the intact mouse or human islet.

Cell–cell communication and heterogeneity

When discussing the concept of functional heterogeneity within the islet and its effects on the regulation of hormone secretion, we must consider that cells within the islet interact in multiple ways (FIG. 2b–d). Indeed, the interaction between endocrine cells within the islet is critical to the regulation of insulin release^9,39 and glucose homeostasis^32,40–43. These pathways include gap junction channels, which mediate ionic currents between cells within the islet and thus enable cells to coordinate and influence their electrical activity^44–47. Furthermore, neurotransmitters, hormones and other factors are either co-released with insulin^32,48 or released via other means⁴⁹; these paracrine factors diffuse within the islet and activate a number of signalling pathways in islet endocrine cells. Finally, several cell contact-dependent pathways (juxtracrine communication) include cell adhesion molecules and EphA receptor–ephrin-A ligand bidirectional signalling^41,50–53.

Given that β-cells (and other endocrine cells) are functionally heterogeneous, we should consider whether heterogeneity exists in the way that these cells can influence each other within the islet⁵⁴. If no interactions occurred between endocrine cells within the islet, then the total insulin release would be the sum of that from the constituent β-cells. The presence of smaller subpopulations of β-cells would have little effect on overall hormone release, irrespective of their function. However, with cell interactions, a range of behaviours can emerge, which is not always intuitive. Some β-cells seemingly show a disproportionately high level of control over the function of the islet^55,56. This behaviour has been well studied when considering gap junction-mediated electrical communication between β-cells. As such, the role of β-cell subpopulations that exert control over islet function via gap junction electrical coupling will be a major focus of the following sections. The way in which subpopulations of β-cells and other endocrine cells influence islet function via paracrine communication or juxtacrine communication has received less attention than gap junctions and will be discussed later. Of note, β-cell populations have been suggested to influence the state of other β-cells in the islet and not simply the function of those cells⁵⁷.

Cell polarity influences the way β-cells interact such as the organization of gap junctions, interactions with the vasculature and basement membrane, and the formation of primary cilia^58,59. Of note, primary cilia have also been suggested to act as a signalling platform to mediate cell–cell communication within the islet⁶⁰. The islet architecture also influences how cells interact. For example, in rodent islets, extensive interactions occur between β-cells in the core of the islet and fewer interactions occur between these β-cells and α-cells and δ-cells that are located on the mantle. However, this structural organization differs between species⁶¹. For example, human islets also show a more complex and diverse architecture, with increased numbers of heterotypic contacts between β-cells and non-β-cells but fewer homotypic contacts between β-cells^62,63. Changes in homotypic interactions between β-cells might have important physiological consequences⁶⁴. The mechanisms by which cells can interact within the islet also differ between rodent and human islets. For example, human islets show a more diverse expression of gap junction proteins and more diverse paracrine factors than mouse islets^32,65–67.

Islet heterogeneity under basal conditions

As discussed earlier, in dissociated β-cells at basal concentrations of glucose, a population of β-cells exists that secretes substantial amounts of insulin as well as β-cells that secrete very low amounts of insulin. This difference indicates substantial functional heterogeneity manifesting at basal levels of glucose.

High basal-secretors.

The reverse haemolytic plaque assay has been applied to β-cells dissociated from islets showing a wide range in the levels of insulin secretion at low levels of glucose (~2 mM). For example, in β-cells dissociated from adult human islets, ~20% of β-cells account for >90% of the total insulin secreted at 2 mM (REF.⁸). The same study found that, at 2 mM glucose, ~5% of β-cells secrete more insulin per cell than the average amount secreted by β-cells at a high level of glucose (~20 mM). Similar results have also been reported in β-cells dissociated from adult mouse islets¹⁰. Furthermore, β-cells dissociated from the islet display much greater heterogeneity in their glucose thresholds for elevated [Ca²⁺] than β-cells in intact islets^54,68. NAD(P)H autofluorescence measurements and insulin biosynthetic activity measurement have also shown that 10–20% of dissociated mouse β-cells are metabolically active at low levels of glucose (<3 mM)^7,13. Therefore, a population of cells exists that is highly secretory at low levels of glucose. However, these early studies relied on the dissociation of β-cells from the intact islet.

Using multiplexed single-molecule imaging of fluorescent in situ hybridization, a population of β-cells was discovered within the intact mouse islet that showed high levels of insulin mRNA (more than twofold increase)⁶⁹. These β-cells, termed ‘extreme cells’, had an elevated expression of genes involved in insulin processing and were also elevated in number in islets from the db/db mouse model of type 2 diabetes mellitus (T2DM). The authors suggested that these extreme cells are engaged in elevated release of insulin at basal concentrations of glucose, thus showing similarities with those subpopulations of β-cells identified in dissociated preparations. As such, these cells might contribute disproportionately to the levels of insulin released at low levels of glucose.

Non-responsive β-cells can suppress basal islet activity.

At basal levels of glucose, insulin release from the intact islet is lower than that of dissociated β-cells when expressed as a percentage of insulin content^70,71. Similarly, at basal levels of glucose, β-cells within the islet are largely quiescent but dissociated β-cells show frequent [Ca²⁺] elevations^68,70. A population of β-cells exists that is metabolically active and highly secretory at low levels of glucose when studied in isolation. However, in the intact islet, these functional characteristics have not been directly observed. As such, properties of cell–cell contacts within the intact islet are largely although not completely suppressing these high basal-secretors. Several studies have indicated that Cx36 gap junction-mediated electrical communication determines this suppression^70,72–74. For example, following a global genetic knockout of the gene expressing Cx36 in mice and complete loss of intercellular electrical conductance, a population of ~50% of the β-cells in adult mouse islets show spontaneous elevations in Ca²⁺ (REF.⁷⁰). Further rodent studies subsequently suggested that a population of less-excitable β-cells — those that are not responsive at low levels of glucose — are suppressing electrical activity and insulin release from these high basal-secretor-like β-cells via Cx36 gap junction communication^55,75 (FIG. 3).

Fig. 3 |. β-Cell subpopulations suppress electrical activity and basal insulin release in other β-cells across the islet.

a | Schematic of three β-cells within the islet with different ATP-sensitive potassium (K_ATP) conductance. Cell 1 has high K_ATP conductance (indicated by the thick arrow), which hyperpolarizes the cell despite an elevated level of glucose, rendering it inexcitable and maintaining a low intracellular calcium concentration ([Ca²⁺]). Cell 3 has low K_ATP conductance (indicated by the thin arrow), which depolarizes the cell at an elevated level of glucose, rendering it excitable and elevating [Ca²⁺]. Gap junction coupling causes an electrical current (indicated by arrows) to flow from cell 2 to cell 1, which hyperpolarizes cell 2, whereas gap junction coupling of cell 2 with cell 3 depolarizes cell 2. In turn, cells 1 and 3 are, respectively, depolarized and hyperpolarized. b | In the absence of gap junction coupling, cell 1 (inexcitable, hyperpolarized) secretes less insulin, whereas cell 3 (excitable, depolarized) secretes more insulin. In the presence of gap junction coupling, β-cells secrete insulin more uniformly, but are influenced more by the inexcitable cell. Cx36, connexin 36.

To understand the role of heterogeneity in suppressing the basal release of insulin, one approach has introduced defined populations of inexcitable β-cells into the mouse islet, each of which has genetic mutations or knockout of key factors involved in the regulation of insulin release, leading to differing functional characteristics. For example, if ~20% of β-cells within mouse islets express ATP-insensitive mutant K_ATP channels and thus have high K_ATP conductance, elevations in Ca²⁺ and GSIS are completely suppressed in β-cells that express wild-type K_ATP channels^55,75. However, if islets lack Cx36 gap junctions, this suppression of β-cells that express wild-type K_ATP channels is absent. Similarly, at low levels of glucose, <30% of non-responsive β-cells are capable of suppressing highly excitable cells that show minimal K_ATP conductance⁷⁴. Thus, a small population of ~20% of β-cells with high K_ATP conductance within an islet can suppress spontaneous elevations in [Ca²⁺] and insulin release. However, when introducing β-cells with deficient glucokinase into mouse islets, >40% of the cells in the islet must lack glucokinase to suppress elevations in [Ca²⁺] and GSIS⁷⁶. Taken together, these findings indicate that small populations of inexcitable β-cells can suppress spontaneous elevations in Ca²⁺ and insulin release within islets but these β-cells are probably characterized by having increased K⁺ channel conductance rather than decreased metabolic activity. Ex vivo experiments with pharmacological activators and inhibitors of K_ATP channels and glucokinase supported that inexcitable β-cells are characterized by elevated K_ATP channel conductance rather than by diminished glucokinase and glucose metabolism^55,76. Further supporting this concept, in rodent β-cells sorted for lower or higher NAD(P)H autofluorescence (lower or higher metabolic activity, respectively), insulin release under low (~2 mM) levels of glucose is not different²³. This finding indicates that differences in metabolic activity do not underlie functional heterogeneity at low concentrations of glucose.

Taken together, the findings suggest that cell populations with increased K⁺ conductance (for example, increased K_ATP channel activity due to a decreased ratio of ATP to ADP or an increased number of K_ATP channels) effectively clamp the islet in a hyperpolarized condition that suppresses [Ca²⁺] and insulin release. A hyperpolarizing current is transmitted via gap junction channels that prevent more excitable cells such as high basal-secretors releasing insulin at low levels of glucose (FIG. 3). An exception to this concept will be if the high basal-secretor β-cells lack gap junction coupling. Cx36 gap junction coupling is heterogeneous within the mouse islet in terms of levels of Cx36 and gap junction permeability⁷⁷. Some molecularly defined mouse β-cell subpopulations show differences in GJD2 expression (encoding Cx36)²⁷. If high basal-secretor β-cells lack gap junction coupling, they will not be suppressed by inexcitable β-cells and will determine basal insulin release. However, differences in Cx36 gap junction coupling have not yet been examined in β-cells that show increased electrical activity or insulin release at low concentrations of glucose.

Islet heterogeneity under glucose stimulation

Following increases in blood levels of glucose, insulin is released in a rapid first phase and again in a subsequent pulsatile second phase. Most but not all rodent and human β-cells secrete elevated levels of insulin in response to elevated concentrations of glucose⁸. However, within the large population of glucose-responsive β-cells, several excitable subpopulations have been reported to control different aspects of insulin release dynamics. In some cases, these populations show overlapping properties and their role is under debate.

β-Cell hubs and highly excitable cells.

Gap junction communication between β-cells within the mouse islet is heterogeneous⁷⁷. Following analogous studies in neuronal systems and biological networks^78–80, analysis of islet Ca²⁺ dynamics revealed heterogeneity in the apparent strength of connections between β-cells in mouse and human islets. Functional connections (that is, ‘connectivity’) were implied by the similarity of Ca²⁺ oscillations^56,81,82. A population of β-cells were identified that showed the highest connectivity with other β-cells in the islet, termed hubs or hub cells (FIG. 4). Further analysis of the 1–10% of β-cells in mouse islets that showed this ‘hub’ behaviour revealed important properties related to islet function⁵⁶. These hub cells showed immature signatures (low PDX1 and insulin) but increased glucokinase levels, thereby suggesting increased metabolic activity. Importantly, strong optogenetic-mediated hyperpolarization of these hub cells in mouse islets (FIG. 4) elicited a suppression of both islet glucose-stimulated [Ca²⁺] and insulin release (as measured by Zn²⁺ co-secretion)⁵⁶. This finding suggested that a specialized population of β-cells is required to maintain elevated Ca²⁺ oscillations and insulin release in response to glucose stimulation.

Fig. 4 |. Dynamics underlying different β-cell subpopulations within the islet.

a | The first-responder β-cell within the islet is that which has the earliest response to glucose (upper diagram). Ca²⁺ dynamics are differently influenced in the presence of first-responder cells (middle graph) or in their absence (lower graph). b | Ca²⁺ wave propagation and the identification and influence of leader β-cells within the islet (also referred to as the wave origin region). Wave propagation differs in the presence or absence of leader cells. c | Cell connectivity and the identification and influence of hub β-cells with the islet. Lines between cells indicate the high level of correlation between Ca²⁺ time courses in the connected cells, indicating a ‘functional connection’. Hub cells show the greatest numbers of ‘functional connections’. Connectivity differs in the presence or absence of hub cells. d | Identification of excitable β-cells in mouse islets by use of the expression of the light-activated cation channel ChR2 (an optogenetic construct)⁸⁸. Blue light illumination over a single-cell region depolarizes the cell and elevates [Ca²⁺] (cell 1). This depolarization spreads to a number of neighbouring cells, recruiting elevated [Ca²⁺] (cell 2 but not cell 3). e | Identification of hub β-cells in mouse islets by use of the expression of the light-activated Cl⁻ pump eNpHr3 (an optogenetic construct)⁵⁶. Yellow light illumination over a single-cell region hyperpolarizes the cell to suppress [Ca²⁺] elevation (cell 1). This hyperpolarization spreads to a number of neighbouring cells, suppressing elevated [Ca²⁺] (cell 2 but not cell 3).

Some criticism has been directed at the concept of hub cells (BOX 2)^83–85. Nevertheless, in silico studies have reproduced some of the experiments in question. For example, hyperpolarizing a small population (~10%) of β-cells within a simulated mouse islet suppresses the response of the whole islet to glucose^86,87; this effect was also reproduced in simulated human islets⁸⁶. Despite these findings, in silico studies have also shown that gap junction coupling is insufficient for a small population of excitable β-cells to maintain electrophysiological control over the whole mouse islet⁸⁷. This result suggests that other interactions are needed for islet-wide maintenance. Alternatively, a population as low as 10% of the total number of β-cells is unable to depolarize and elevate glucose-stimulated [Ca²⁺] and insulin secretion across the islet⁸⁷. Thus, the degree to which islet-wide [Ca²⁺] and insulin release is maintained by a small population of β-cells and the mechanism by which they act have not been fully clarified.

Box 2 |. The discussion regarding hub cells.

Hub cells are β-cells that show the greatest number of ‘links’ with other cells within the islet. ‘Links’ are defined by an algorithm that assesses the similarity of the Ca²⁺ oscillations in a cell with other cells^78,81. Studies in mouse islets demonstrated that hyperpolarization of hub cells (via light-activation of the Cl⁻ pump eNpHr3) silenced Ca²⁺ oscillations across the whole islet, whereas hyperpolarization of other β-cells had minimal effect⁵⁶.

The implication that one β-cell can maintain electrical control over the whole islet has been criticized as being incompatible with electrophysiological principles^83–85. For instance, upon patch clamping of a mouse islet, applying a hyperpolarized voltage clamp to a single β-cell in the islet did not result in islet-wide quiescence⁸⁵. Furthermore, the majority of β-cells (>80%) are capable of elevated [Ca²⁺] at high levels of glucose and do not need to be activated via a specialized cell. Most β-cells are at the threshold to fire and will do so irrespective of some signal from another cell⁷⁰. Nevertheless, the implication that one cell can maintain electrical control over the whole islet is perhaps a misinterpretation or misrepresentation of the originally published data^56,83. In these studies, a region local to a hyperpolarized ‘hub cell’ lost coordination rather than the entire islet being silenced. This observation is reasonably consistent with electrophysiological principles, where injection of an electrical current into a cell is capable of inducing membrane potential changes (and [Ca²⁺] changes) that spread several (approximately three) cells⁸⁵.

Mathematical modelling of islet electrophysiology has successfully described the originally published experiments on hub cells^56,83. Hyperpolarization of 10% of β-cells can silence the whole islet in multiple computational models that have differing assumptions^86,87, although these assumptions do question the underlying role of hub cells or the mechanisms by which they act. For example, hub cells might exert an effect over islet function via diffusible factors^85,87. Despite these ongoing discussions, a population of β-cells clearly exists that exerts a disproportionate effect over the recruitment and elevation of [Ca²⁺] and insulin release. However, the size of the subpopulation needed and the mechanism by which these cells act are still under debate.

Despite these details being under debate, the concept whereby a population of β-cells (small or large) disproportionally acts to elevate glucose-stimulated [Ca²⁺] and insulin release is supported by other studies. For example, a population of β-cells exists in mouse islets that, when depolarized via optogenetic stimulation, recruits elevations in [Ca²⁺] in many neighbouring cells⁸⁸ (FIG. 4). These β-cells show increased NAD(P)H, indicating increased metabolic activity and thus overlap with the properties of hub cells. As such, there is a large β-cell population (10–40% of total β-cells) with increased metabolic activity that disproportionally elevates [Ca²⁺] across the islet and is important for increasing the triggering of insulin release.

Hub cells show signatures of reduced maturity, including reduced PDX1 (REF.⁵⁶). When immature β-cells low in PDX1 or MAFA are made absent in mouse islets and human islets following selective overexpression of PDX1 and/or MAFA, glucose-stimulated [Ca²⁺] and insulin release in response to glucose paradoxically worsen⁵⁷. This finding might indicate an absence of hub cells and their influence over islet function. However, this finding might also indicate how immature β-cells positively influence the state of other β-cells within the islet via some poorly understood cell non-autonomous mechanism.

Leader and first-responder β-cells.

Further analysis of islet [Ca²⁺] dynamics has revealed additional signatures of functional subpopulations of β-cells that disproportionately control islet [Ca²⁺] and insulin release dynamics. ‘Leader’ β-cells have been described in mouse islets that show glucose-stimulated [Ca²⁺] oscillations that elevate earlier than the rest of the islet⁸⁹ (FIG. 4). In vivo work has shown that these leader cells spatially overlap with hub cells in mouse islets and are thus expected to be important to elevate glucose-stimulated [Ca²⁺] and insulin release. However, the role of leader cells in mice has principally been associated with the second phase of insulin release, during which [Ca²⁺] and insulin release is pulsatile^90,91. As pulsatile insulin is physiologically important to enhance insulin action⁹², a reduction in insulin pulse height upon the loss of leader cells would probably diminish insulin action. However, this hypothesis remains to be tested.

Analogous to leader cells, preliminary findings in a pre-print paper have identified a population of first-responder cells, which are β-cells that show elevations in [Ca²⁺] initially following glucose stimulation⁹³ (FIG. 4). Their role in mice is principally associated with the first phase of insulin release. Ablation of first-responder cells in mouse islets reduced the amplitude and coordination of first-phase [Ca²⁺], similar to the phenotype observed following a loss of Cx36 gap junction coupling in mouse islets⁴⁰ and a reduction in gap junction coupling in human islets⁹⁴. Therefore, we expect that these first-responder cells are important to enhance and coordinate first-phase [Ca²⁺] and insulin release.

Importantly, first-responder β-cells are more excitable than other β-cells within the islet, which is linked to decreased K_ATP conductance rather than differences in metabolic activity. Nevertheless, first-responder cells do not spatially or functionally overlap with hub cells or leader subpopulations of β-cells. Furthermore, there is a functional hierarchy where other early-responding β-cells were able to ‘take over’ from laser-ablated first-responder cells in mouse islets⁹³. Similarly, in zebrafish islets, laser ablation of β-cells that respond first to increased concentrations of glucose leads to diminished elevations in [Ca²⁺] within the islets⁸⁹. This finding demonstrates the in vivo physiological relevance of the subpopulation of β-cells that respond first to glucose. We note that, in this study, these cells in zebrafish were termed “leader” cells; however, they responded to glucose stimulation earlier during the first phase of [Ca²⁺] elevation rather than during the second phase under which leader cells are identified in mouse. Thus, β-cells that respond earlier to nutrient stimulation enhance the first-phase response, with less effect during the second phase. We infer from these studies that the lack of first-responder β-cells and the consequent diminished first-phase [Ca²⁺] and insulin release would cause glucose intolerance.

Although conceptually similar, first-responder cells are associated with the first phase of Ca²⁺ elevation and insulin release, whereas leader cells and hub cells are associated with the oscillatory second phase of Ca²⁺ and insulin release. The lack of functional overlap between these cell populations indicates how they each regulate Ca²⁺ and thus insulin during their respective phases. Indeed, distinct patterns of [Ca²⁺] activity have been suggested to underlie activation or deactivation and second-phase Ca²⁺ oscillations⁹⁵, supporting the concept that first-responder cells are distinct from hub cells and leader cells.

Rhythmic pacemaker cells.

In organ systems such as the heart, a small group of specialized cells controls the pace of electrical propagation across the organ^1,96,97. The existence of these rhythmic pacemaker cells has led to speculation that such cell populations exist in the islet. In the heart, both the atrioventricular node and Purkinje fibres have intrinsic membrane potential oscillatory capability (known as automaticity), such that, if the sinoatrial node rhythmic pacemakers are absent or blocked, other cells can take over pacemaking, albeit at a slower pace⁹⁸. The intrinsic [Ca²⁺] oscillatory frequency of β-cells is variable^15,99. As β-cells are gap junctional coupled, an equivalent scenario might reasonably exist by which a rhythmic pacemaker β-cell population controls the islet oscillatory dynamics. Of note, the term ‘pacemaker’ has also been loosely applied to refer to hub cell, leader cell and first-responder cell populations that control [Ca²⁺] elevations as described earlier. Here we specifically use the term rhythmic pacemaker cells to refer to β-cells that entrain the oscillatory [Ca²⁺] dynamics across the islet.

Studies have sought to define whether mechanisms intrinsic to the β-cell^100–103 or extrinsic factors^104–106 determine islet oscillations. In addition, the existence of a specific rhythmic pacemaker population^88,107 has been suggested as the β-cell population that ‘leads’ islet [Ca²⁺] oscillations. β-cells that lead the [Ca²⁺] oscillations within mouse islets have been indirectly shown to have a faster intrinsic oscillation frequency^88,107, consistent with a rhythmic pacemaker concept. As such, the leader β-cell population^88,89,107 might be important for the proper regulation of pulsatile insulin release during the second phase (FIG. 4).

In silico studies have suggested an alternative scenario. In simulated mouse islets, following the removal of β-cells that ‘lead’ the [Ca²⁺] oscillations (that is, show the earliest [Ca²⁺] oscillations of all β-cells in the islet at elevated levels of glucose), the overall islet [Ca²⁺] oscillation was affected to a lesser degree than by removal of those β-cells that ‘lag’ the oscillations (that is, show the most delayed oscillations)⁸⁷. In this model, those ‘lagging’ β-cells show the slowest [Ca²⁺] oscillations compared with other β-cells and these slow oscillations result from increased metabolic activity that sustains the active phase. These results imply that a population of β-cells with the slowest [Ca²⁺] oscillations in the islet, which might functionally overlap with hub cells, disproportionately drags the oscillatory pace of the islet^87,108. This concept makes physiological sense given that slow oscillations of insulin release are physiologically important and fast oscillations would be unlikely to be maintained in circulation and thus not enhance insulin action. However, findings from a 2020 study show that increased pyruvate kinase activation within mouse islets increases the ratio of ATP to ADP, which closes K_ATP channels but also increases [Ca²⁺] oscillation frequency in β-cells. These findings suggest an alternative scenario, where cells with increased pyruvate kinase activation serve as rhythmic pacemakers¹⁰⁹. Testing this concept will be important to understand precisely how β-cell subpopulations pace the rhythm of pulsatile insulin release.

Glucose-non-responsive β-cells.

Early single-cell studies discovered that some β-cells have decreased levels of insulin release at elevated concentrations of glucose. For example, in β-cells dissociated from adult human islets, ~10% of β-cells secrete less insulin at 20 mM glucose than the average amount secreted by all β-cells at a low concentration of glucose (~2 mM)⁸. Similarly, a different study found that some β-cells do not show elevated NAD(P)H autofluorescence except at very high levels of glucose (>11 mM). Therefore, a small population of glucose-non-responsive β-cells exists. The effect of these β-cells on the regulation of hormone release from islets has been poorly studied. We do know that a population of poorly responsive cells (low expression of glucokinase and/or high K_ATP conductance) can dampen [Ca²⁺] elevations and insulin release at elevated concentrations of glucose via gap junction-mediated hyperpolarization (see earlier sections^55,76). As such, these non-responsive β-cells might similarly be diminishing the responses in other cells, dampening insulin release. Further investigation of these β-cell populations within the intact islet is needed. For example, are they represented by CD9⁺ST8SIA1⁺ β-cells, which in human islets show reduced GSIS^20,26?

Heterogeneity in other islet endocrine cells

The presence of heterogeneity in other endocrine cells of the islet, such as glucagon-producing α-cells and somatostatin producing δ-cells, has started to receive attention. Based on single-cell studies, such as scRNAseq or cell surface markers, mouse and human α-cells are heterogeneous in terms of gene transcription^19,110,111. As with β-cells, the effect of this heterogeneity on cell function is only just starting to be considered.

α-Cell heterogeneity.

Although β-cells are gap junctional coupled and show synchronized behaviour, α-cells are not gap junction coupled⁷⁷ and do not show synchronized behaviour^112,113. However, α-cells do interact with other cells in the islet via paracrine and juxtracrine communication. Furthermore, when dissociated from the islet, α-cells show fundamentally different glucose-stimulated glucagon release¹¹⁴. As such, there is some complexity in interpreting the link between α-cell heterogeneity and the regulation of hormone release.

Several studies have demonstrated that α-cells are functionally heterogeneous in terms of electrical activity. For example, wide variation exists in the electrophysiological parameters of α-cells within intact mouse islets^115,116. Patch-seq analysis has further correlated functional heterogeneity with gene expression in α-cells dissociated from human islets: a subpopulation of transcriptionally more mature α-cells (increased ARX and MAFB expression) exists that shows increased Na⁺ currents and exocytosis³⁷. Conversely, a population of less mature α-cells also exists in human islets, with decreased Na⁺ currents and exocytosis. The change in [Ca²⁺] is also highly heterogeneous among α-cells in the islet: upon elevated levels of glucose, some α-cells are inhibited but others remain responsive¹¹⁴. Wide variation also occurs in glucagon content between α-cells within the islet¹¹⁷. This wide variation between cells is reminiscent of the heterogeneity among dissociated β-cells discussed earlier. In silico studies have further suggested that heterogeneity in the electrophysiological properties of α-cells is important for the regulation of glucagon release although paracrine communication with δ-cells is also important¹¹⁸.

The precise role of this α-cell functional heterogeneity has not been fully defined. However, based on the aforementioned studies, we speculate that functional heterogeneity is needed to retain glucagon release at elevated levels of glucose. At elevated concentrations of glucose, glucagon release is sufficiently reduced so as to reduce hepatic glucose production. However, a population of α-cells remains active and secretes glucagon in order to provide sufficient tone within the islet to amplify insulin release⁴². Although local glucagon release can amplify insulin release, a population of α-cells has also been identified that expresses and secretes GLP1 and can further contribute to the amplification of insulin release¹¹⁹. This population is also increased in number in patients with T2DM, which further points to distinct changes that can occur in α-cell heterogeneity¹²⁰. Indeed, following increased K_ATP conductance that can occur in neonatal diabetes mellitus, some mouse and human α-cells are still responsive at low levels of glucose and some remain glucose responsive by reducing glucagon release at elevated concentrations of glucose¹²¹, thus α-cell heterogeneity seems to provide some resistance to islet dysfunction due to changes in K_ATP channel conductance. We speculate that a means to preferentially stimulate the population of α-cells that remains active at high concentrations of glucose in order to increase β-cell tone but avoid stimulating hepatic glucose production would be beneficial for therapeutic purposes in T2DM.

δ-Cell heterogeneity.

The existence of δ-cell heterogeneity is still unclear. δ-cells have been suggested to be gap junctional coupled with β-cells and show synchronized [Ca²⁺] dynamics with β-cells in mouse islets^122,123. Furthermore, δ-cells engage in extensive paracrine communication and form projections that might mediate long-range communication within islets¹²⁴. As such, interpreting the link between δ-cell heterogeneity and the regulation of hormone release is a notably complex task.

Sodium–glucose cotransporter 2 (SGLT2) is expressed at low levels in δ-cells, with 30−50% of mouse or human δ-cells showing SGLT2 protein expression¹²⁵. Under elevated concentrations of glucose, insulin can stimulate somatostatin release from δ-cells via the action of SGLT2; this mechanism includes changes in δ-cell Na⁺ levels and the influx of [Ca²⁺] to trigger somatostatin release¹²⁶. Increased somatostatin release suppresses glucagon release from α-cells and thus prevents elevations in blood levels of glucose. As such, the SGLT2⁺ population of δ-cells might potentially have a key role in driving the release of somatostatin upon elevated levels of glucose and in the inhibition of glucagon release. The regulation of somatostatin secretion by insulin in the SGLT2⁺ δ-cell population might also have an important role in the regulation of insulin release itself given that somatostatin inhibits insulin release¹²⁷.

Under low concentrations of glucose, α-cell release of glucagon is partially inhibited by somatostatin in mouse islets¹²⁸. As such, characterizing whether a subpopulation of δ-cells is secreting somatostatin at low levels of glucose is warranted. We speculate that inhibiting such a population might be beneficial for the promotion of glucagon release under hypoglycaemia and of effective glucose counter-regulation. A population of δ-cells releasing somatostatin at low levels of glucose might also suppress spontaneous elevations in insulin release that can occur at low concentrations of glucose.

Further characterizing these populations of α-cells and δ-cells will be important to understand how they influence each other and β-cells via paracrine communication as well as how these actions affect glycaemia.

Other aspects of islet cell heterogeneity

There are several additional concepts that are important to consider when trying to understand how islet cell heterogeneity affects the regulation of hormone release.

Changes in subpopulation proportions in T2DM.

β-Cell dysfunction has long been considered a primary driver of T2DM. A number of subpopulations of β-cells, whether defined by molecular markers or by functional properties, differ in their relative numbers in patients with T2DM or other conditions of islet dysfunction compared with healthy individuals³⁸. For example, in β-cells purified from human islets, a substantial decrease is observed in the number of more functional CD9⁻ ST8SIA1⁻ β-cells (TABLE 1) in patients with T2DM (25%) compared with healthy individuals (~50%). Conversely, a substantial increase is seen in the number of less functional CD9⁺ST8SIA1⁺ β-cells in patients with T2DM (25%) compared with healthy individuals (~10%)²⁶. This finding is further supported by mass cytometry-based analysis of islets from patients with T2DM and healthy individuals²⁰.

In mice fed a high-fat diet, a decline is observed in the number of more functional FLTP⁺ β-cells in islets, with an increase in the number of less functional FLTP⁻ β-cells²⁷ (TABLE 1). Furthermore, mouse islets treated with pro-inflammatory cytokines also show a decrease in the presence of hub β-cells⁵⁶. The decrease in the number of more functional subpopulations of β-cells in these studies suggests that an overall reduction in insulin release would follow. However, when considering the effects of cell–cell communication, the reduction of functional subpopulations could have a notable effect in eliminating the recruitment of less functional β-cells. Nevertheless, many of the factors important for insulin release have a different role in T2DM. For example, genes that negatively correlate with insulin release in healthy individuals positively correlate with insulin release in patients with obesity and T2DM³⁷. This concept still requires further investigation.

How much control over islet function do β-cell subpopulations have?

An underappreciated factor in islet heterogeneity is how many β-cells within a subpopulation are needed to affect islet function. Some studies have suggested that 1−10% of mouse islet β-cells must be hub cells to maintain elevated [Ca²⁺] and insulin release⁵⁶. Alternatively, only 15–20% of inexcitable β-cells in mouse islets are required to dampen [Ca²⁺] elevations⁵⁵. Another study has suggested that subpopulations of cells must have cell numbers exceeding 30% of all β-cells in the islet to have such an effect⁸⁷. The role of a β-cell population might not be physiologically important in circumstances under which they are present in numbers below those needed for their action. As discussed earlier, such situations could occur in pathological conditions such as in T2DM, where a reduction in the numbers of an excitable β-cell subpopulation might cause islet dysfunction.

Proliferative or immature β-cells and their potential effect on function.

Several mouse and human islet studies using genetic reporters or cell surface markers have defined novel β-cell populations that are more resistant to cell damage, act as a source of new β-cells or are more proliferatively competent^24,27,29; these populations are discussed in depth in other Review articles^22,62–64. Common to these studies is that these novel proliferative or immature β-cell populations are poorly responsive to elevated levels of glucose, showing diminished insulin release, decreased [Ca²⁺] elevations or reduced glucose metabolism. These cells also make up a minority of β-cells (~10%). Whether these subpopulations correspond to those β-cells that only respond (if at all) to very high levels of glucose is unclear^7,8. If these cells are gap junction coupled to other β-cells, their effect might be substantial; for example, they might hyperpolarize and limit excitability and hormone release in other endocrine cells and thus might have deleterious effects over islet function. However, if these populations are not gap junction coupled, the reduction of insulin release would be more marginal. For example, in mouse islets, FLTP⁻ β-cells show reduced GJD2 expression²⁷ and UCN3⁻ β-cells show dis-coordinated [Ca²⁺] dynamics²⁹. However, the functional effect of these populations, beyond their role on survival or proliferation, is still to be fully tested.

How much overlap is there between functional β-cell sub-populations?

The degree to which the various functional β-cell subpopulations overlap is still unclear. Of note, hub cells, leader cells and rhythmic pacemaker cells have several shared properties. For example, in mouse islets, leader cells have been shown to have greater connectivity than other β-cells, indicative of hub cells⁸⁹. Furthermore, leader-like cells show properties of higher intrinsic oscillation frequency, indicative of rhythmic pacemaker cells⁸⁸. However, β-cells that lead Ca²⁺ oscillations do not overlap with more metabolically active β-cells, which activate their neighbours more effectively⁸⁸, and thus these metabolically active cells share properties with hub cells. Importantly, β-cell subpopulations associated with the first-phase and second-phase [Ca²⁺] response were suggested to be distinct populations⁹⁵. In human islets, CD9⁻ST8SIA1⁻ β-cells show much greater insulin release than CD9⁺ST8SIA1⁺ β-cells and this functionally active population is diminished in number in T2DM²⁶. Furthermore, a ‘C1’ β-cell population that was also low in these CD9 and ST8SIA1 markers and was diminished in number in patients with T2DM was previously defined by mass cytometry²⁰. Thus, excitable β-cells might consist of a set of partially overlapping sub-states with differing degrees of overlap in terms of function and marker expression. However, some marker-defined populations do not overlap. For example, in mice, proliferative-competent FLTP⁻ β-cells express UCN3 (REF.²⁷), whereas UCN3 lineage-negative virgin β-cells express FLTP²⁹ and proliferate normally¹²⁹. Thus, although these two populations show both low functional competence and signs of immaturity, they are probably distinct. It will be important for future work to examine the relative overlap between different β-cell subpopulations.

Open questions

Although we have learned much regarding how heterogeneous β-cells control the regulation of hormone release, open questions remain that need addressing before we have a more complete picture.

Many human single-cell studies have identified β-cell subpopulations that show gene expression differences, which suggests differences in function. However, whether these represent true subpopulations is unclear given the limited overlap between studies³⁵. Of note, RBP4⁺ β-cells are one exception that has been replicated^19,37. Furthermore, low abundance transcripts often encode receptors or channels, which suggests that signalling molecules that affect function might not contribute notably to the dimensionality reduction analysis that forms these populations. As such, interpreting how subpopulations defined solely on gene expression contribute to islet function and the regulation of hormone secretion is speculative. Thus, validating singe-cell-based β-populations and determining whether they show functional differences will be important. Approaches such as Patch-seq that enable both cell electrophysiology and exocytosis to be linked to single-cell gene expression, as discusses above, have started to address this³⁷.

An important question is whether identified populations are stable over time. In studies where genetic markers have been used (for example, FLTP-Venus) these populations have been fairly stable²⁷. However, a pre-print study that tracked the first-responder functional population over time suggested them to be more transient⁹³. In the latter case, it is possible that the β-cell itself is stable but the way in which it interacts with the islet varies (for example, via gap junctions). Nevertheless, preliminary findings from a pre-print paper suggest that properties such as insulin gene expression can fluctuate, suggesting that other intrinsic properties of β-cells might also vary over time¹³⁰.

Cells within the islet can interact in many ways. Given the ease with which cellular electrical properties can be measured via chemical or genetically encodable Ca²⁺ indicators and given our robust knowledge of gap junction communication, electrical properties have naturally emerged as key characteristics of functional subpopulations. However, a number of paracrine factors are released by β-cells⁴⁸. It is conceivable that some β-cell populations might release different factors that could influence the function of other nearby β-cells and other islet cells. Understanding whether certain factors are released by β-cell subpopulations is thus a key goal for future work. This goal also applies to α-cells and δ-cells given that they principally influence β-cells via paracrine communication (that is, through the action of somatostatin and glucagon).

A need also exists for researchers to develop further ways to facilitate detailed molecular and functional analysis of β-cell subpopulations. Limited markers or labels are available for isolating live human β-cells beyond CD9 and ST8SIA1 or ENTPD3 (REFS^26,131). Furthermore, genetic reporters in mice do not always fully indicate differences in endogenous gene expression; for example, MIP-GFP versus insulin promoter-driven insulin expression²⁵. Patch-seq has integrated electrophysiological and scRNAseq analysis to provide detailed molecular and functional information^37,120. In addition, imaging-based methods can include biomarker measurements and functional analysis using fluorescent biosensors, as described above, but also multiplexed imaging of gene expression or protein levels^132,133. Nevertheless, improvements in fluorescent biosensors will be important, particularly for factors beyond Ca²⁺ such as for cAMP, redox state or exocytosis. Integrating these reporters and biomarkers with either pancreas slice preparations or re-aggregated pseudo-islets will be needed to effectively characterize the role of endocrine cell heterogeneity in the human islet.

Functional heterogeneity between β-cells generally considers differences in glucose responsiveness of metabolism, Ca²⁺ or hormone release. However, other nutrient secretagogues, such as fatty acids and amino acids, are important. Thus, it will be important to consider whether subpopulations of β-cells respond differently to fatty acids or GLP1 or whether α-cells respond differently to amino acids.

Conclusions

A wealth of information has been uncovered regarding β-cell heterogeneity through the use of new high-throughput single-cell technologies and via imaging and optogenetics. We are building a clearer picture as to how different β-cell populations contribute to the overall regulation of hormone release from the islet. Here, we have described the functional characteristics of different β-cell subpopulations. We further describe how they can communicate and influence the function of the rest of the islet under both basal and nutrient-stimulated conditions. This influence includes suppressing or elevating the release of insulin or regulating the pulsatile dynamics of insulin release. Furthermore, we are starting to discover functionally relevant populations of other endocrine cells within the islet. The availability of new technologies that can combine functional assessment and gene expression profiles by cell or that can assess gene expression in cells that have previously undergone imaging and/or optogenetic profiling will provide further details. Furthermore, assessing functional populations in vivo, including using novel animal models, will provide further details as to the role of functional subpopulations in the regulation of islet function.

Key points.

• Pancreatic β-cells are heterogeneous in terms of function (the regulation of insulin secretion) and their transcriptional profile.

• Functionally distinct subpopulations of β-cells can be identified by genetic and cell surface markers or based upon functional analyses by optogenetics and Ca²⁺ dynamics.

• Cell–cell communication allows functional subpopulations of β-cells to influence the regulation of insulin secretion across the rest of the islet.

• Under basal conditions, both excitable insulin secretory β-cells and suppressive inexcitable β-cells can be observed.

• Under glucose-stimulated conditions, a number of highly functional subpopulations of β-cells can be observed that influence the coordinated dynamics of islet [Ca²⁺] and insulin secretion.

• Changes in islet cell heterogeneity, loss of functional subpopulations or disruption to the communication between functional subpopulations may all underlie islet dysfunction in diabetes mellitus.