Heterogenous impairment of α-cell function in type 2 diabetes is linked to cell maturation state

SUMMARY

In diabetes, glucagon secretion from pancreatic α-cells is dysregulated. The underlying mechanisms, and whether dysfunction occurs uniformly amongst cells, remain unclear. We examined α-cells from human donors and mice using electrophysiological, transcriptomic, and computational approaches. Rising glucose suppresses α-cell exocytosis by reducing P/Q-type Ca²⁺ channel activity, and this is disrupted in type 2 diabetes (T2D). Upon high-fat-feeding of mice, α-cells shift towards a ‘β-cell-like’ electrophysiologic profile in concert with indications of impaired identity. In human α-cells we identify links between cell membrane properties and cell surface signalling receptors, mitochondrial respiratory complex assembly, and cell maturation. Cell type classification using machine learning of electrophysiology data demonstrates a heterogenous loss of ‘electrophysiologic identity’ in α-cells from donors with T2D. Indeed, a sub-set of α-cells with impaired exocytosis is defined by an enrichment in progenitor and lineage markers, and up-regulation of an immature transcriptomic phenotype, suggesting important links between α-cell maturation state and dysfunction.

Graphical Abstract

eTOC:

In diabetes, glucagon secretion from pancreatic α-cells is dysregulated. Dai et al. examined electrical and transcriptomic α-cell phenotypes and find that dysfunction in type 2 diabetes is linked to cell maturation state and impaired α-cell identity. Notably, a sub-set of α-cells enriched for lineage markers appears uniquely susceptible to dysfunction.

INTRODUCTION

In concert with reduced insulin secretion from pancreatic β-cells in type 2 diabetes (T2D), disrupted glucagon secretion from α-cells contributes to hyperglycemia and impaired hypoglycemia counter-regulation (Girard, 2017). While insulin and glucagon secretion are both dependent upon electrical excitability and Ca²⁺-dependent exocytosis, the nature of the ion channels involved and their roles, and the impact of glucose-stimulation, differ in α- and β-cells. Insulin granule exocytosis is linked to Ca²⁺ entry via L-type Ca²⁺ channels, whereas glucagon secretion is coupled to P/Q-type Ca²⁺ channels (Marinis et al., 2010). Also, Na⁺ channels play a more prominent role in glucagon secretion (Barg et al.,2000; Göpel et al., 2000; Ramracheya et al., 2010) and differ in their regulation between these cell types (Zhang et al., 2014). In rodents such differences can distinguish α- from β-cells, either by Na⁺ current properties (Zhang et al., 2014) or ‘electrophysiological fingerprints’ (Briant et al., 2017).

Similar to β-cells the excitatory and secretory machinery in α-cells, including ion channel activities (Huang et al., 2011a), Ca²⁺ responses (Marchand and Piston, 2010; Reissaus and Piston, 2017; Shuai et al., 2016), glucagon content (Zadeh et al., 2020) and exocytotic capacity (Huang et al., 2011b) are heterogeneous. Indeed, the intracellular Ca²⁺ response of α-cells varies, with some suppressed by glucose and others activated (Shuai et al., 2016). This suggests a heterogeneity among these cell types, which is supported by single-cell transcriptomics (Camunas-Soler et al., 2020; Korsunsky et al., 2018). Recent reports suggest a sub-population of α-cells (or ‘α-like-cells’) that are proliferative. In mice these are identified by Slc38a5, which encodes an amino acid transporter (Kim et al., 2017). In humans these may be identified by the presence of ARX and cytosolic Sox9 (Lam et al., 2018). These cells could account for α-cell hyperplasia upon glucagon-receptor antagonism and may be a source of new β-cells (Meulen et al., 2017).

While little evidence so far suggests that α-cells dedifferentiate in diabetes, they can transdifferentiate in rodents following severe β-cell loss (Thorel et al., 2010) or genetic manipulation of transcription factor expression (Chakravarthy et al., 2017; Matsuoka et al., 2017). Interestingly, α-cells may show more plasticity than β-cells (Bramswig et al., 2013) as they appear to exist in distinct states characterized by chromatin accessibility at promotors for GCG, functional genes such as ABCC8, and at sites enriched in motifs for transcription factors of the RFX, GATA and NEUROD families, among others (Chiou et al., 2021). In type 1 diabetes, the expression of α-cell markers is reduced (Brissova et al., 2018) and in T2D α-cells express an immature transcriptomic profile (Avrahami et al., 2020).

We hypothesised that this plasticity influences α-cell membrane function, contributing to dysfunction in T2D. We used correlated electrophysiological and single-cell RNA-seq (patch-seq) of α-cells from human donors and mice to define a cell-autonomous glucose-regulation of α-cell Ca²⁺ channel activity and exocytosis that is associated with α-cell maturation state, and identify putative regulators of glucagon secretion. These include the mitochondrial respiratory chain complex and numerous cell surface receptors. In mice, high fat-feeding prompts some α-cells to adopt a ‘β-cell-like’ electrical profile and impaired identity. Similarly, in human T2D, α-cells enriched for markers of mitochondrial function and endocrine lineage such as NEUROD1, ISL1, NKX2-2, and ARX exhibit a selective induction of an immature transcriptional profile, impaired electrophysiological phenotype and dysregulated exocytosis. This suggests an important link between α-cell maturation, identity and dysfunction in T2D.

RESULTS

Glucose-mediated suppression of α-cell exocytosis is disrupted in T2D

In β-cells, glucose metabolism amplifies Ca²⁺-triggered exocytosis and insulin secretion (Ferdaoussi et al., 2015; Gembal et al., 1992; Sato et al., 1992), which is linked to the activation of L-type Ca²⁺ channels (Barg et al., 2001; Bokvist et al., 1995; Wiser et al., 1999). We examined the impact of glucose on glucagon exocytosis in α-cells from donors with no diabetes (ND) or with T2D (Suppl Table 1) following dispersion to single cells and identification by glucagon immunostaining (Fig 1A). In ND α-cells, exocytosis was highest at low glucose (1 mM) and was suppressed by increasing glucose (to 5–20 mM) (Fig 1B; Suppl Fig 1). In α-cells from donors with T2D, exocytosis was ~75% lower at 1 mM glucose than in α-cells from ND donors, and elevating glucose levels exerted a modest increase (Fig 1B; Suppl Fig 1A). The differences between ND and T2D cells, also seen when grouped by donor (Suppl Fig 1B), are consistent with a recent live-cell imaging study (Omar-Hmeadi et al., 2020). In ND α-cells voltage-dependent Ca²⁺ channel activity was lower at elevated glucose than in α-cells from donors with T2D, here recorded using Ba²⁺ as a charge carrier (Fig 1C; Suppl Fig 1C), and the relationship between Ca²⁺ entry and exocytosis was not altered (Suppl Fig 1D).

Figure 1. Glucose suppresses human α-cell exocytosis in concert with P/Q-channel activity.

A) Schematic diagram illustrating that human islets were isolated, dispersed and cultured for 1–2 days prior to whole-cell patch-clamp and subsequent α-cell identification by glucagon immunostaining.

B) Total exocytosis upon a series of ten membrane depolarizations with increasing glucose in ND α-cells (grey; n = 42, 28, 24, 20 cells from 9 donors) and T2D α-cells (pink; n = 29, 16, 12, 27 cells from 6 donors).

C) Voltage-dependent Ca²⁺ channel activity, with Ba²⁺ as a charge carrier, in ND α-cells with increasing glucose (n = 31, 24, 27 cells from 7 donors).

D) Effect of the P/Q-type Ca²⁺ channel activator GV-58 (10 μM) on Ca²⁺ current inactivation (left; n = 24, 28, 22, 21 cells) and charge entry (middle; n = 22, 20, 20, 23 cells) during a 500 ms depolarization from 70 to 0 mV, and total exocytosis (right; n = 23, 22, 28, 34 cells) at 1 and 10 mM glucose (3 donors).

E) Effect of the P/Q-type Ca²⁺ channel blocker agatoxin (100 nM) and the L-type Ca²⁺ channel blocker isradipine (10 μM) on ND α-cells exocytosis at 1 mM glucose (n = 34, 30, 21 cells from 5 donors).

F) Voltage-dependent Ca²⁺ channel activity at 0 mV (Ba²⁺ as a charge carrier) with increasing glucose in ND α-cells, with agatoxin (100 nM), and in T2D α-cells (n = 31, 16, 24, 15, 27, 18 cells from 7 ND donors, and 34, 31, 27 cells from 5 T2D donors).

G) Effect of agatoxin (100 nM) and the L-type Ca²⁺ channel blocker isradipine (10 μM) on total exocytosis in T2D α-cells at 1 and 10 mM glucose (n = 30, 31, 29, 32, 31, 29 cells from 5 T2D donors).

H) Putative scheme for glucose-regulation of depolarization-induced exocytosis in α-cells, and the dysfunction seen in T2D.

*P < 0.05; **P < 0.01; and ***P < 0.001 by one-way ANOVA (D,E,G) or two-way ANOVA (B,C,F) and Tukey post test compared with 1 mM glucose control, versus the ND control (C) or as indicated.

The majority of the human α-cell Ca²⁺ current is mediated by L-type and P/Q-type channels (Ramracheya et al., 2010), and the latter are directly linked to glucagon exocytosis (Dai et al., 2014; Ramracheya et al., 2010, 2018). The low α-cell exocytosis at elevated glucose could be reversed by the P/Q-type Ca²⁺ channel activator GV-58 (Tarr et al., 2012), which delays Ca²⁺ current inactivation (Fig 1D). The P/Q-type channel blocker agatoxin inhibited exocytosis from ND α-cells at 1 mM glucose while the L-type channel blocker isradipine did not (Fig 1E). Finally, the greater P/Q-type Ca²⁺ channel activity seen in ND α-cells at 1 mM glucose was largely absent in T2D (Fig 1F; Suppl Fig 1A), but still contributed to the modest glucose-dependent increase in exocytosis (Fig 1G). Thus, in ND α-cells increasing glucose inhibits P/Q-type Ca²⁺ channels to limit exocytosis, an effect that requires intact mitochondrial function (Suppl Fig 1E,F). In T2D α-cells, Ca²⁺ channel activity at low glucose is reduced, and increasing glucose facilitates exocytosis (Fig 1H).

Patch-seq highlights a role for the mitochondrial respiratory chain in α-cell exocytosis, and suggests poor responsiveness in immature α-cells

We performed patch-seq in α-cells, some initially pre-incubated at low glucose for 1 hour. However, exocytosis was low after an extended time at 1 mM glucose, due to depletion of glucagon granules (Suppl Fig 2). These pre-incubated cells were excluded from further analysis, and we performed transcriptome-wide correlations with exocytosis in 400 human α-cells exposed acutely to 1 and 10 mM glucose, from 31 donors (Fig 2A). ND α-cells at 1 mM were enriched for positively-correlated genes while negatively-correlated genes were enriched at 10 mM glucose (Fig 2B; Suppl Table 2). Gene-set enrichment analysis (GSEA) using Z-scores of transcripts found in >20% of cells highlighted pathways associated with elevated α-cell exocytosis at 1 mM glucose and lower exocytosis at 10 mM glucose (Fig 2C). Consistent with a role for metabolism in determining α-cell responsiveness, we found mitochondrial respiratory chain complex assembly as a positive correlate to exocytosis at low glucose and a negative correlate at high glucose (Fig 2C–D). A separate over-representation analysis (ORA) yielded similar results, highlighting the electron transport chain in the glucose-regulation of α-cell exocytosis (Suppl Fig 3A).

Figure 2. Patch-seq suggests roles for the mitochondrial respiratory complex in α-cell function, and endocrine development and cell fate in α-cell dysfunction.

A) Schematic diagram illustrating the isolation of 400 α-cells from 24 donors without diabetes (ND) and 7 donors with T2D assessed by patch-seq, some of which have been partly published previously (Camunas-Soler et al., 2020).

B) The distribution of exocytosis - transcript correlations from α-cells of ND donors at 1 and 10 mM glucose (see also Suppl Table 2).

C) Gene set enrichment analysis (GSEA) Gene ontology (GO): Biological Pathways using Z-scores as weighting across the transcriptome (using genes expressed in >20% of α-cells), separately at 1 and 10 mM glucose, reveals pathways linked to facilitation (positive values) or suppression (negative values) of exocytosis. FDR < 0.05 unless indicated otherwise, and absence of bars indicates no significant enrichment.

D) Leading-edge transcripts for mitochondrial respiratory chain complex assembly that flip correlation with exocytosis from positive to negative in α-cells of ND donors as glucose increases from 1 to 10 mM.

E) An endocrine system development pathway is enriched in ND α-cells with low exocytotic responses at 1 mM glucose.

F) Leading-edge genes underlying the signal in panel E, which includes transcripts involved in islet cell lineage and α-cell identity.

G) The distribution of exocytosis - transcript correlations α-cells from donors with T2D at 1 and 10 mM glucose (see also Suppl Table 3).

H) GSEA GO: Biological Pathways with Z-scores as weighting (using genes expressed in >20% of α-cells) reveals development and cell fate pathways linked to increased exocytosis in T2D α-cells at 10 mM glucose (FDR < 0.05 unless indicated otherwise).

I) Selected leading-edge transcripts that underlie the cell fate commitment (left) and pancreas development (right) pathways, including islet lineage and α-cell identity markers.

False discovery rate (FDR) for pathways identified by GSEA were < 0.05 unless indicated otherwise.

Ca²⁺ channel transcripts were not reduced, in α-cells of T2D donors compared with those from ND donors (Suppl Fig 3B–C). Hyperglycemia may induce α-cell mitochondrial dysfunction (Knudsen et al., 2019), and we saw a modest but significantly higher mitochondrial respiratory chain complex transcript expression in α-cells of T2D donors compared with α-cells from ND donors (Suppl Fig 3D,E). Intriguingly, transcripts associated with endocrine development appeared as anti-correlates of exocytosis at 1 mM glucose (Fig 2C,E–F) suggesting a role for cell differentiation state in the responsiveness of α-cells. In α-cells from donors with T2D, the effect of glucose on the distribution of transcriptome-wide correlations was reversed, such that high glucose associated with more positively correlated genes (Fig 2G; Suppl Table 3). GSEA of these correlations highlighted a role for cell development state in the inappropriately high α-cell exocytosis at 10 mM glucose in T2D (Fig 2H). Several leading-edge transcripts enriched in these pathways, including transcription factors important for pancreatic endocrine maturity like GATA6, PAX6, RFX6, and RFX3, overlap with those that correlated with inappropriately low exocytosis in α-cells of ND donors at 1 mM glucose (Fig 2F,I).

Glucose control of mouse α-cell exocytosis, and impaired ‘electrophysiological identity’ following high-fat feeding

Following an ~10-minute exposure to 1 mM glucose, depolarization of mouse islets with 20 mM KCl elicited a transient stimulation of glucagon release that was blunted in islets kept at 5 mM glucose (Fig 3A), consistent with the glucose-dependent suppression of α-cell exocytosis. These differences could not be explained by the indirect paracrine effects of insulin, as insulin secretion evoked by KCl was slightly higher at 1 mM glucose than at 5 mM glucose (Fig 3B). Similar to human α-cells, increasing glucose suppressed exocytosis in mouse α-cells (Fig 3C). Although the physiological impact of glucose on glucagon secretion involves key paracrine signals (Briant et al., 2016), the suppression of exocytosis by increasing glucose was similar when cells were seeded at 10% of normal density (Fig 3D) and required glucose metabolism since the non-metabolizable analog 2-deoxyglucose (2-DG) did not mimic the effect of glucose (Fig 3E). The glucose-suppression of Ca²⁺ entry and exocytosis could be prevented by rotenone, but not antimycin A (Fig 3F), suggesting a signal at or between mitochondrial respiratory chain complexes I and III. This could include reactive oxygen species/H₂O₂ produced by complex I via reverse electron transport (Onukwufor et al., 2019) or by complex III towards the cytosol (Muller et al., 2004). Indeed, direct intracellular dialysis of H₂O₂ suppressed α-cell Ca²⁺ influx and exocytosis in mouse (Fig 3F) and human (Suppl Fig 1E–F) α-cells.

Figure 3. Glucose-suppression of mouse α-cell exocytosis requires glucose metabolism and the mitochondrial respiratory chain.

A-B) Glucagon (A) and insulin (B) secretion from islets of 10–12 week old male C57bl6 mice as glucose remains at 5 mM (blue) or drops from 5 to 1 mM (black), and upon subsequent stimulation with 20 mM KCl. A control (grey) was maintained at 5 mM glucose (n = 3, 3, 3 mice).

C-D) Mouse α-cell exocytosis with increasing glucose (n = 32, 30, 14, 27 cells from 4 mice), and at very low (10% of normal) density (D; n = 6, 6, 13, 9 cells from 2 mice).

E) Effect of 2-Deoxy-D-glucose (2-DG) on glucose-regulation of α-cell exocytosis (n = 20, 20, 21, 46 cells from 4 mice).

F) Effect of the complex I inhibitor rotenone (0.5 μM) and the complex III inhibitor antimycin A (0.5 μM) on α-cell Ca²⁺ charge entry during a 500 ms depolarization from −70 to 0 mV, and exocytosis, at 1 and 10 mM glucose. Also the effect of direct intracellular dialysis of H₂O₂ (10 μM) via the patch pipette at 1 mM glucose (left n = 19, 15, 18, 17, 11, 19, 16 cells; and right n = 20, 19, 16, 26, 16, 26, 22 cells from 3 mice).

**- P < 0.01; and ***- P < 0.001 by one-way ANOVA (C,D,E) or two-way ANOVA (A,B,F), followed by Tukey post-test to compare points or groups with the 5 mM glucose points (A, B) or with the 1 mM glucose control group.

Low-glucose stimulation of glucagon secretion was enhanced in mice fed a high fat diet (HFD) for 12–14 weeks compared with age-matched controls (18–20 weeks; Fig 4A), similar to what we have shown previously (Kellard et al., 2020). There was no difference in insulin secretion at 5 and 1 mM glucose but the response to high-K⁺ was reduced in mice fed a HFD compared with controls (Fig 4B), possibly reflecting the disruption of a direct coupling between Ca²⁺ channels and insulin granules (Collins et al., 2010). As in humans, exocytosis at 1 mM glucose in α-cells from chow fed mice was inhibited by the P/Q-type Ca²⁺ channel blocker agatoxin, but not the L-type channel blocker isradipine (Fig 4C). This was reversed in α-cells from HFD mice where exocytosis is resistant to agatoxin but sensitive to isradipine (Fig 4C). Importantly, the P/Q-channel agonist GV-58, which delayed channel inactivation in α-cells from both the chow-fed and HFD mice (Fig 4D), rescued exocytosis following suppression by 5 mM glucose in α-cells from chow-fed, but not HFD-fed, mice, consistent with a loss of P/Q-type channel coupling to exocytosis (Fig 4E). Mouse α-cells are distinguished from β-cells by a characteristic right-shifted Na⁺ current half-inactivation, and these two cell types primarily express Scn3a and Scn9a, respectively (Zhang et al., 2014). Many α-cells from the HFD-fed mice, identified by glucagon immunostaining, showed a leftward shift in Na⁺ channel half-inactivation towards a ‘β-cell-like’ phenotype (Fig 4F; Suppl Fig 4A–C).

Figure 4. ‘β-cell like’ properties of α-cells from HFD-fed mice.

A-B) Glucagon (A) and insulin (B) secretion from islets of male C57/bl6N mice fed HFD for 10–12 weeks starting from 8 weeks of age (red) compared with age matched (18–20 week) chow fed controls (black). (n = 3, 3 mice).

C) Effect of the P/Q-type Ca²⁺ channel blocker agatoxin (100 nM) and the L-type Ca²⁺ channel blocker isradipine (10 μM) on total exocytosis from chow diet (CD) and HFD α-cells at 1 mM glucose (n = 34, 42, 34, 40, 29, 39 cells from 8 HFD and 8 CD mice).

D-E) Effect of the P/Q-channel activator GV-58 (10 μM) to delay Ca²⁺ current inactivation in both CD and HFD α-cells (D, n = 23, 36, 24, 23 cells from 3 CD and 3 HFD-fed mice) and on exocytosis (E) from CD (grey) and HFD (pink) α-cells at 1 and 5 mM glucose (n = 15, 16, 16, 21 cells from 3 CD mice; n = 14, 13, 15, 14 cells from 3 HFD-fed mice).

F) Steady-state voltage-dependent Na⁺ current inactivation curves (left) and individual half-inactivation voltages (right) from α-cells of chow fed and HFD mice (n = 36, 44 cells from 9 CD and 9 HFD mice, measured at 1 mM glucose). Half-inactivation voltages from fit curves are indicated.

G) From a separate set of CD and HFD α-cells assessed by patch-seq (Suppl Fig 4), mean expression (black) and % of cells expressing (blue) some α-cell identity transcription factors in CD and HFD α-cells.

H) Mean expression (black) and % of cells expressing (blue) the β-cell Na⁺ channel isoform, Scn9a, in HFD α-cells.

I) Correlation Z-scores of the negative shift in Na⁺ channel steady-state inactivation (at 1 mM glucose) in HFD α-cells with transcripts involved in α-cell lineage and identity.

*- P < 0.05; **- P < 0.01; and ***- P < 0.001 by the Student’s t-test (F), or by one-way ANOVA (D,E) or two-way ANOVA (A,B,C) followed by Tukey post-test to compare points or groups with the 5 mM glucose points (A,B), with the 1 mM glucose control group, or as indicated.

This switching of both Ca²⁺-channel-exocytosis coupling and Na⁺ channel half-inactivation towards ‘β-cell-like’ phenotypes upon HFD feeding is reminiscent of observations from genetically induced α- to β-cell trans-differentiation (Chakravarthy et al., 2017). However, the present changes occurred while glucagon expression was maintained since the α-cells were identified by glucagon immunostaining. We explored these findings in separate experiments where α-cells from chow-fed and HFD-fed mice were collected for patch-seq (Suppl Fig 4A) where we confirmed the negative shift in Na⁺ current inactivation (Suppl Fig 4B,C) and the switch from P/Q- to L-type Ca²⁺ channel dependence of exocytosis (Suppl Fig 4D). After HFD feeding some endocrine transcription factors were lower than in α-cells of chow-fed mice (Fig 4G; Suppl Table 4), although there was no clear change in exocytotic or Ca²⁺ channel transcripts (Suppl Fig 4E). The β-cell Na⁺ channel transcript Scn9a was higher (Fig 4H) in α-cells from HFD than chow-fed mice, and the HFD-induced shift in Na⁺ channel inactivation at 1 mM glucose correlated with expression of α-cell identity and islet lineage factors (Fig 4I; Suppl Table 5). Thus, after HFD feeding some α-cells adopted electrophysiological properties typically associated with β-cell function, and a transcriptional profile consistent with altered α-cell identity and/or maturation.

Electrophysiological fingerprint modeling links human α-cell behavior and cell phenotype

We next compiled data from islet cells of 67 donors collected at 1, 5 and 10 mM glucose, and identified either by scRNA-seq or immunostaining (Fig 5A). Unlike in mice, but similar to previous reports (Braun et al., 2008; Ramracheya et al., 2010), Na⁺ channel properties are similar between human α- and β-cells and could not be reliably used alone to distinguish these cell types (Fig 5B,C; Suppl Fig 4C). In ND α-cells, the peak Na⁺ current significantly correlated with genes known to impact α-cell activity (Fig 5D). Top positive correlates include genes involved in α-cell or islet function; including Na⁺ and Ca²⁺ channels (SCN3A, SCN3B, CACNA1A), α- and islet-cell lineage transcription factors (MAFB, FEV, NKX2.2, NEUROD1), G-protein coupled receptors (GIPR, SSTR1, FFAR1, GPR119), and secreted factors (LOXL4, WNT4, UCN3, IL16). Negative correlates were less abundant but included the vitamin D binding protein (GC) which is known to regulate α-cell Na⁺ currents (Viloria et al., 2020). GSEA for KEGG Pathways and GO: Biological Process terms revealed additional pathways enriched in ND α-cells with larger Na⁺ currents (Suppl Fig 5A,B), including the serotonin, GABA and glutamate pathways. Numerous cell surface receptor transcripts also appear as correlates to Na⁺ channel activity (Fig 5E, Suppl Fig 5C). We validated the ability of agonists for some of these to increase human α-cell Na⁺ currents (Fig 5F). Transcript correlates of peak Na⁺ current, and voltage-dependence of inactivation, from α-cells of ND and T2D donors are provided in Suppl Table 6.

Figure 5. Na⁺ current properties correlate with transcriptomic markers of α-cell function.

A) Schematic diagram illustrating an expanded dataset of electrically profiled human islet cells, and tSNE representations of α- and β-cells identified by immunostaining or sequencing along with the relative distribution of Na⁺ current half-inactivation values.

B-C) The distribution of Na⁺ current amplitudes (B) and voltage-dependence of half-inactivation values (C) of β-cells (light blue) and α-cells (pink) (see also Suppl Fig 4C).

D) Selected significant positive (green) and negative (red) transcript correlates of peak Na⁺ current from α-cells of ND donors mapped to a proposed scheme of glucose-regulation of glucagon exocytosis (see also Suppl Table 6).

E) Correlation of α-cell peak Na⁺ current and selected transmembrane signaling receptor transcripts (see also Suppl Fig 5).

F) In α-cells of 6 donors with no diabetes (ND), Na⁺ currents measured with receptor agonists (colors matching receptors shown in panel E) upon depolarization from −70 to −10 mV. Peak current is shown at right: control (n = 53 cells); 0.5 μg/ml SLIT-2-N (n = 17 cells); 10 μM prostaglandin E₂ (PGE2, n = 17 cells); 0.5 μM lysophosphatidic acid (LPA, n = 14 cells); 0.2 μg/ml α-latratoxin (n = 24 cells); 10 mM L-glutamic acid (n = 21 cells); 100 nM glucose-dependent insulinotropic polypeptide (GIP, n = 7 cells); 200 nM somatostatin (SST, n = 24 cells); 5 μM epinephrine (n = 22 cells).

*- P < 0.05 and ***- P < 0.001 by one-way ANOVA, followed by the Benjamini and Hochburg post-test to method to compare groups controlling for false discovery rate.

Although α-cell lineage markers correlate with Na⁺ currents, the overlap in Na⁺ current properties between human α- and β-cells make it impossible to use these alone to interrogate shifts in human α-cell phenotypes. We therefore developed ‘electrophysiological fingerprinting’ classifier models (Fig 6A) that integrate Na⁺ current, Ca²⁺ current and exocytosis measures (Fig 6B) to characterize human α-cell ‘functional phenotypes’, similar to the logistic regression (Briant et al., 2017) and random forests (Camunas-Soler et al., 2020) models we used previously. In those models however, cell size was the major predictor of α-cell identity. Here we generated three independent models for cell type classification: Optimizable Ensemble classifiers that include (Model 1) or exclude (Model 2) cell size as an independent variable, and an Extreme Gradient Boosting model that also excludes cell size, but with additional restrictions to donor age, donor BMI and organ cold ischemic time (CIT) applied to the training data (Model 3). These models distinguished ND α- and β-cells well (Fig 6A) and, unlike some underlying parameters, were unaffected by potential confounders, such as donor sex, BMI, CIT, cell culture time and glucose concentration (Fig 6B).

Figure 6. Electrophysiological fingerprints define a loss of ‘functional identity’ in T2D α-cells.

A) Classifier models trained on islet cell electrical properties of α- and β-cells from donors with no diabetes (ND) using Optimizable Ensemble or Extreme Gradient Boosting (XGBoost) approaches, identify cell types with high accuracy regardless of the inclusion or exclusion of cell size from training data. 80% of data was used for training. 20% of data was reserved for validation and generation of confusion matrices. tSNE plots show cell types determined by immunostaining or sequencing (left) and assigned α_probability scores (right).

B) Ordinary least squares multiple regression of electrophysiological properties of α-cells from ND donors, Model scoring, and donor/isolation variables.

C) Calculated α_probability (and β_probability) values from Model 3 applied to cells collected for patch-seq, without a priori knowledge of cell type and correlation with canonical β-cell (light blue) and α-cell (pink) markers.

D) β_probability and α_probability values derived from all three models, of all β- and α-cells from ND or T2D donors.

E) Volcano plot of transcript correlations with Model 3 α_probability values (slope/standard deviation) in α-cells from donors with T2D. A negative correlation indicates transcripts associated with reduced α_probability.

F) Significantly enriched GO: Biological Pathways from Gene Set Enrichment Analysis (GSEA) performed using α_probability - transcript correlation slopes of α-cells from donors with T2D in all three Models as weighting.

***- P < 0.001 and **** P < 0.0001 as indicated within models using non-parametric Kruskal-Wallis test (D) followed by Dunn’s post-test to correct for multiple comparisons. Pink/red points in E indicate significance at P < 0.05. False discovery rate (FDR) for pathways identified by GSEA in F were <0.05 unless indicated otherwise.

Model output between 0 and 1 can be considered the probability an electrophysiological profile matches that of an α-cell, and we called this α_probability (with β_probability = 1-α_probability). Assignment of an α_probability to all ND α- and β-cells for which we had patch-seq data, without a priori knowledge of cell type, showed the effective separation of cell types and correlation with canonical markers (Fig 6C). All three models showed a significantly reduced β_probability and α_probability, in β- and α-cells respectively, from donors with T2D compared with cells from ND donors (Fig 6D). The expression of numerous transcripts correlated with a ‘loss of electrophysiological identity’ in T2D α-cells within one or more of the models, including lineage markers such as ISL1, NEUROD1, FEV, and RFX6 (Fig 6E; Suppl Table 7). GSEA using α_probability correlation slopes as weighting for transcripts expressed in >20% of cells highlighted GO: Biological Process terms related to mitochondrial respiratory chain complex assembly in all three models (Fig 6F), underscoring a link between increased respiratory chain complex expression and α-cell dysfunction.

Impaired functional identity and exocytosis in T2D α-cells enriched in lineage and immaturity markers

Human α-cells exist in states of variable chromatin accessibility at the GCG promoter and sites enriched with transcription factor motifs characterizing endocrine lineage and development (Chiou et al., 2021). Markers of α-cell lineage are heterogenous within our dataset, with evidence for increased expression in T2D (Fig 7A; Suppl Fig 6A,B). Intriguingly, this also included ARX which we confirmed at the protein level, along with MAFB and glucagon itself (Fig 7B). ARX-enriched (ARX^hi) cells expressed consistently higher levels of progenitor transcripts like NEUROD1, FEV, GATA6, and others (Suppl Fig 6B). We find no difference in α_probability scores between ND α-cells either enriched or deficient in these (Fig 7C; Suppl Fig 6). In T2D however, an impaired electrical fingerprint occurred selectively in α-cells with higher levels of ARX, NEUROD1, ISL1, PAX6, GATA6, FEV, NKX2–2, RFX6 and MAFB (Fig 7C; Suppl Fig 6). Accordingly, ARX and these other markers did not correlate with exocytosis in α-cells of ND donors but did correlate significantly with impaired exocytosis in T2D (Fig 7D). In separate scRNA-seq data (Avrahami et al., 2020) we confirmed the up-regulation of a juvenile α-cell profile (Arda et al., 2016), tissue development genes and a β-cell-like profile selectively in ARX^hi α-cells in T2D compared with α-cells from ND donors (Fig 7E; Suppl Fig 7). Finally, while the α-cells with low ISL1, NKX2–2, NEUROD1 and ARX from donors with T2D had normal exocytosis at 1 mM glucose, α-cells enriched in these transcripts showed impaired exocytotic function in T2D (Fig 7F).

Figure 7. A role for maturation state in α-cell dysfunction in T2D.

A) Heterogenous expression of transcript makers for islet cell lineage and α-cell maturity in 980 patch-seq α-cells.

B) ARX and MAFB protein expression in α-cells at the protein level in situ by immunostaining. Violin plots show the relative levels of GCG, ARX and MAFB expressed in ARX^low and ARX^hi α-cells.

C) α_probability values from the three separate classifier models in ND and T2D α-cells separated by high and low expression of ARX (see also Suppl Fig 6).

D) Correlation of exocytosis in ND and T2D α-cells with α-cell lineage and identity markers. Bars to the left of the centerline indicate correlation with low exocytosis at the given glucose concentration.

E) In a separate human islet single-cell dataset (Avrahami et al., 2020), expression of a juvenile α-cell gene-set in ND and T2D α-cells separated by low and high ARX expression. The heatmap displays relative expression levels as median log2 FPKM values.

F) Total exocytosis in ND and T2D α-cells at 1 mM glucose separated by low and high expression of ISL1, NEUROD1, NKX2-2, and ARX.

*- P < 0.05, **- P < 0.01, ***- P < 0.001 and **** P < 0.0001 two-way ANOVA followed by two-stage step-up method for estimation of FDR (C,D,F) or Tukey post-test (B).

DISCUSSION

Glucagon secretion is under the control of metabolic, paracrine, hormonal and neuronal signals (El et al., 2020). Much debate has centered around the question of whether glucose-suppression of glucagon secretion is mediated via intrinsic, paracrine or autonomic mechanisms although it seems likely that α-cells (like β-cells) adeptly integrate multiple signals for precise physiologic control of glucagon. A role for direct glucose-sensing is supported by the ability of glucose to modulate α-cell ATP-sensitive K⁺ channels (Zhang et al., 2013), and the impact of α-cell glucokinase manipulation on in vitro cell function and in vivo glucagon secretion and glucose homeostasis (Moede et al., 2020; Bahl et al., 2021; Basco et al., 2018). Here, we show that glucose suppresses human and mouse α-cell exocytosis, consistent with a recent report where exocytosis was measured by live-cell imaging (Omar-Hmeadi et al., 2020). A role for α-cell metabolism is supported by the effect of the non-metabolizable glucose analog 2-DG and the correlation of responses with mitochondrial respiratory complex assembly transcripts, particularly those of complex 1. Indeed, the ability of rotenone to block glucose-suppression of α-cell Ca²⁺ currents and exocytosis suggests a signaling role for reactive oxygen species and H₂O₂ generated by complex I. Consistent with this, intracellular application of H₂O₂ mimics the effect of glucose while reduced glutathione blocks the effect of glucose, and external H₂O₂ inhibits low glucose-stimulated glucagon secretion (Allister et al., 2013).

The study by Omar-Hmeadi et al. (2020) also show a U-shaped response of isolated α-cells to glucose, suggesting that glucose may increase α-cell exocytosis under some conditions. We find some hints of this in our own data, including in α-cells of donors with T2D. In ND α-cells this may depend on the culture conditions. Following prolonged low-glucose preincubation α-cells are larger, possibly due to continued glucagon granule fusion. Subsequent exocytotic responses are smaller when the α-cells are maintained at 1 mM glucose, while acute glucose-stimulation now increases exocytosis. This is reminiscent of the glucose-dependent increase in glucagon secretion from purified α-cells (Olsen et al., 2005). Thus, glucose acutely suppress P/Q-type Ca²⁺ channels via a complex I-dependent mechanism to limit depolarization induced α-cell exocytosis, while at the same time facilitating glucagon granule priming or docking to maintain the availability of granules. Altogether these processes will ‘tune’ α-cell responsiveness to myriad paracrine, endocrine, and neuronal inputs.

The P/Q-type Ca²⁺ current seen at low glucose in α-cells of ND donors appears absent in T2D, but channel expression is not reduced. Therefore, channel regulation appears disrupted. It seems possible that altered mitochondrial function in α-cells in T2D could drive dysregulation of P/Q-type Ca²⁺ channel activity, and indeed mitochondrial dysfunction has been demonstrated in α-cells from hyperglycemic mice (Knudsen et al., 2019) and mitochondrial morphology is altered in α-cells from western-diet-fed mice (Grubelnik et al., 2020). Genes involved in glucose metabolism appear downregulated in α-cells from donors with type 1 diabetes (Brissova et al., 2018) although these may exhibit a more extreme phenotype than in T2D, with down-regulation of several channels, exocytotic transcripts, and α-cell identity markers. We do find an up-regulation, particularly in ARX^hi T2D α-cells, of mitochondrial respiratory chain complex assembly transcripts and a clear correlation of this with the impaired α-cell phenotype by fingerprint modelling, however a demonstration of altered mitochondrial respiratory function and/or reactive oxygen species generation in α-cells from donors with T2D remains to be shown.

The coupling of glucagon exocytosis to P/Q-type Ca²⁺ channels, which we confirm at low glucose, has been demonstrated previously (Marinis et al., 2010; Ramracheya et al., 2010) and is similar to the direct coupling of insulin granule exocytosis to the activation of L-type Ca²⁺ channels (Wiser et al., 1999). These and other electrophysiological properties are used to distinguish α- and β-cells from rodents. Most commonly, mouse islet cell types are identified by a combination of cell size (α-cells are smaller) and distinct properties of Na⁺ current inactivation. In a model of genetically induced α-to-β trans-differentiation in mice, we reported a clear shift in electrophysiological phenotype consistent with the attainment of β-cell properties (Chakravarthy et al., 2017). Intriguingly, following high-fat feeding mouse α-cells undergo a negative-shift in Na⁺ current inactivation and convert from P/Q- to L-type Ca²⁺ channel dependence of exocytosis. While we provide some evidence for impaired α-cell identity, this is not associated with a clear trans-differentiation as β-cell markers are not increased and cells maintain positive immunostaining for glucagon. The shift in Na⁺ channel inactivation towards a ‘β-cell like’ phenotype, which could be related to changes in membrane composition (Godazgar et al., 2018), correlates with the expression of important α-cell lineage and identity transcription factors, suggesting that the change in electrical phenotype occurs more readily in cells with higher levels of these markers.

A subset of human α-cells, even from donors without diabetes, exist in an immature state and may suffer a further loss of mature identity in T2D (Avrahami et al., 2020), perhaps related to their greater epigenetic plasticity (Bramswig et al., 2013) and distinct states of chromatin accessibility (Chiou et al., 2021). We find that α-cells from ND donors with inappropriately low exocytosis are enriched in transcripts and pathways associated with endocrine development (FOXO1, PAX6, RFX6 and others). In T2D we see no obvious loss of α-cell transcription factors, or up-regulation of β-cell-defining transcription factors indicative of trans-differentiation per se. We do however find heterogeneity in many lineage markers, as reported previously by us (Camunas-Soler et al., 2020; Drigo et al., 2019) and others (Li et al., 2016). In our dataset, most of these show a small but significant increase in T2D. To assess a shift in electrophysiological phenotype of these human α-cells we could not use Na⁺ current inactivation alone as this feature overlaps with measurements from human β-cells. We therefore modified machine learning approaches that we previously used to improve the identification of mouse (Briant et al., 2017) and human (Camunas-Soler et al., 2020) islet cells. We used exocytosis, Na⁺ current, and Ca²⁺ properties as training data for three separate models with similar results. Two of these excluded cell size, to solely identify shifts in membrane ‘activity’, and all models accurately assigned probability values for α- and β-cells irrespective of ambient glucose, culture times and other important donor and isolation-related parameters. Correlation of α_probability values with transcriptomic data therefore emphasized pathways linked to a ‘loss of functional phenotype’ in T2D and suggest that α-cells with the most altered electrical phenotypes in T2D are those with higher expression of pancreatic endocrine lineage markers.

In this study we define an impaired α-cell phenotype in T2D that is largely restricted to a subgroup of cells expressing higher levels of markers that define not only α-cell maturity, but pancreatic endocrine lineage. At first look we would have expected that high levels of ARX should be indicative of mature α-cell function, but this is misleading given a strong overlap of ARX^hi cells with transcripts more traditionally associated with an endocrine progenitor state such as NEUROD1, ISL1, FEV and others. These cells indeed appear more ‘plastic’ in their electrophysiological responses, which may predispose them to an altered electrical phenotype in T2D. We confirm in a separate dataset the ‘de-repression’ of immature gene-sets associated with a ‘juvenile’ α-cell phenotype and with tissue development restricted to ARX^hi α-cells in T2D. Thus, we demonstrate that all α-cells are not equally impacted by disease and that a sub-set of α-cells defined by their maturation state may be key drivers of impaired glucagon responses in T2D.

Limitations of the study

While tempting to link impaired exocytosis in α-cells from donors with T2D to an impaired responsiveness to hypoglycemia in vivo, this must be considered in the context of in situ α-cell function which will be impacted by the local environment and by paracrine or hormonal signals. We studied single isolated α-cells here, and although we demonstrate a similar heterogeneity in ARX/MAFB protein expression in situ, α-cell function may be different within the intact pancreas and would likely be impacted by architectural changes in the disease state. Encouragingly though, we find clear differences between ND and T2D α-cells that persist in vitro, and among the novel findings of this study we also find wellestablished regulators of α-cell function (such as GIPR). Additionally, both α-cell function and transcript expression are likely dynamic and impacted by metabolic status or culture conditions. Indeed, pre-culture at low glucose can alter α-cell exocytotic function. Our approach to ‘electrophysiological fingerprinting’ addresses this in part since it is unaffected by glucose, time in single-cell culture (up to 3 days), or various donor-related parameters. Nonetheless, the relevance of possible dynamic shifts in α-cell phenotype in T2D remains unclear. We do not know, for example, if exocytotic function would be restored if T2D α-cells transition from an ARX^hi to ARX^lo state, or whether such a transition itself is impaired.

While α-cells enriched in lineage factors appear less mature and maintain some plasticity, whether T2D induces a true reversal of maturation or a more general phenotypic drift could still be questioned. While an enrichment of numerous gene-sets including β-cell genes in the ARX^hi cells in T2D could suggest the latter, we note that those are also all enriched in true juvenile α-cells (Avrahami et al., 2020). The exact links between these changes and dysregulated glucagon secretion remains somewhat speculative, particularly since we see no decrease in Ca²⁺ channel transcripts to explain the reduced Ca²⁺ currents. Interestingly, mitochondrial respiratory complex assembly transcripts are increased in T2D α-cells, most notably in the α-cells with impaired function and enriched for ARX. While altered mitochondrial function could drive P/Q-type Ca²⁺ channel inhibition and impaired exocytosis, and we provide some evidence linking the respiratory chain to α-cell dysfunction in T2D, we do not know if mitochondrial respiration is altered in T2D α-cells. This will require assessment of mitochondrial function and oxygen consumption in the ARX^hi sub-set of α-cells from donors with T2D as a future priority.

Finally, we should be careful when directly comparing the rodent and human studies. One clear difference we find is that human α-cells in T2D show impaired (low) exocytosis at 1 mM glucose and a modest increase with higher glucose, while the mouse HFD α-cells instead show altered Na⁺ channel inactivation and exocytosis-Ca²⁺-channel coupling. The exact reasons for these differences are unclear, but perhaps related to obvious differences between humans and the mouse model (disease/HFD duration, degree of dysglycemia, age, etc.). Nonetheless, in both the mice and human cells, we find evidence to suggest that α-cell dysfunction is linked to cell maturation state.

STAR METHODS

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Patrick MacDonald (pmacdonald@ualberta.ca).

Materials availability

This study did not generate new unique reagents.

• Raw sequencing reads are available in the NCBI Gene Expression Omnibus (GEO) and Sequence Read Archive (SRA) under accession numbers GSE124742 and GSE164875.

• The code and scripts generated during this study, as well as preprocessed datasets, are available at https://github.com/jcamunas/patchseq.

• Data S1 represents an Excel file containing the values that were used to create all the graphs in the paper. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Human islets

In most cases, human islets were from our in-house human islet isolation and distribution program (www.isletcore.ca) (Lyon et al., 2019). Some human islets were provided by the Clinical Islet Transplant Program at the University of Alberta, by Dr. Rita Bottino at the Alleghany Health Network (Pennsylvania, US), or from the Human Pancreas Analysis Program (Kaestner et al., 2019). Details of donors with no diabetes (ND) or type 2 diabetes (T2D) used in this study are shown in Supplementary Table 1. T2D was determined either by reporting of previous clinical diagnosis at the time of organ procurement, or by assessment of %HbA1c > 6.5 in a few cases that were considered as previously undiagnosed T2D. Human islets and dispersed cells were cultured in DMEM (ThermoFisher, #11885) with 10% FBS (ThermoFisher, #12483020) and 100 U/ml penicillin/streptomycin (ThermoFisher, #15070063) at 37°C and 5% CO₂. In our previous study (Camunas-Soler et al., 2020) we saw no obvious effect of sex on patch-seq data, or when analyses were corrected for sex as a co-variate. In the present study we find that electrophysiological fingerprint modelling is not impacted by donor sex (Fig 6B). All donors provided written informed consent for research. Human tissue studies were approved by the Human Research Ethics Board of the University of Alberta (Pro00013094, Pro00001754).

Mouse islets

Mouse islets were isolated from chow-fed (5L0D Picolab Laboratory Rodent Diet, #3005659–220) male C57bl/6N mice (Charles River, #CRL:027; RRID:IMSR_CRL:027) at 10–12 weeks of age or following 1012 weeks of high fat diet (60% of calories from fat; VWR Bio-Serv, #CA89067–471) starting from 8 weeks of age by collagenase digestion and hand-picking (Smith et al., 2020). Mouse islets and dispersed cells were cultured in RPMI (ThermoFisher, #11875) with 10% FBS and 100 U/ml penicillin/streptomycin at 37°C and 5% CO₂. Animals were housed at 20–24°C with a 12h:12h light:dark cycle and daily health checks. Studies were performed in accordance with institutional guidelines and were approved by the Animal Care and Use Committee at the University of Alberta (AUP00000291).

METHOD DETAILS

Patch-clamp recordings

Hand-picked islets were dissociated to single cells using StemPro accutase (ThermoFisher, #A1110501). Dispersed islet cells were cultured for 1–3 days, after which media was changed to bath solution containing (in mM): 118 NaCl, 20 TEA, 5.6 KCl, 1.2 MgCl₂, 2.6 CaCl₂, 5 HEPES, and with glucose as indicated (pH adjusted to 7.4 with NaOH) in a heated chamber (32–35 °C). Modulators of Ca²⁺ channels (isradipine, Sigma-Aldrich, # I6658; agatoxin, Alomone Labs, #STA-500; GV-58, Alomone Labs, #G-140) or mitochondrial respiratory chain inhibitors (antimycin A, Sigma-Aldrich, #A8674; rotenone, Sigma-Aldrich, #R8875) were added to the bath, and pH adjusted, as indicated in figure legends. For whole-cell patch-clamping, fire polished thin wall borosilicate pipettes coated with Sylgard (3–5 MOhm) contained intracellular solution (in mM): 125 Cs-glutamate, 10 CsCl, 10 NaCl, 1 MgCl₂, 0.05 EGTA, 5 HEPES, 0.1 cAMP and 3 MgATP (pH adjusted to 7.15 with CsOH). Electrophysiological measurements were collected using a HEKA EPC10 amplifier and PatchMaster Software (SmartEphys HEKA) within 5 minutes of break-in as described previously (Camunas-Soler et al., 2020). Quality control was assessed stability of the seal (>10 GOhm) and access resistance. Cells were identified by post-hoc immunostaining for insulin with a rabbit anti-insulin primary antibody (Santa Cruz; #SC-9168; RRID:AB_2126540) and goat anti-rabbit Alexa Fluor 488 secondary (ThermoFisher, #A-11076; RRID:AB_141930), and with a guinea pig anti-glucagon primary antibody (Sigma-Aldrich, #G2654; RRID:AB_259852) and goat antiguinea pig Alexa Fluor 594 secondary (ThermoFisher, #A-11076; RRID:AB_141930); or following collection for single-cell RNA sequencing analysis (Camunas-Soler et al., 2020).

For assessment of Ca²⁺ channel activity using Ba²⁺ as a charge carrier (Fig 1C,F) bath solution contained (in mM): 100 NaCl, 20 BaCl₂, 5 CsCl, 1 MgCl₂, 10 HEPES, with 0.2 μM tetrodotoxin and glucose as indicated (pH 7.4 with NaOH); and intracellular solution was (in mM): 140 Cs-glutamate, 1 MgCl₂, 20 TEA, 5 EGTA, 20 HEPES, 3 Mg-ATP (pH 7.15 with CsOH). For validation of receptor agonist effects on Na⁺ currents (Fig 5F), bath solution contained (in mM): 118 NaCl, 20 TEA, 5.6 KCl, 1.2 MgCl₂, 5 HEPES, and 5 mM glucose (pH 7.4 with NaOH); and intracellular solution was (in mM): 125 Cs-glutamate, 10 CsCl, 10 NaCl, 1 MgCl₂, 0.05 EGTA, 5 HEPES (pH 7.15 with CsOH). Compounds used were: lysophosphatidic acid (LPA; Sigma-Aldrich, #L7260), α-latrotoxin (Enzo Life Sciences, #ALX-630027-C040), Slit guidance ligand 2 (SLIT-2-N; Sigma-Aldrich, #SRP3155), prostaglandin E₂ (PGE2; Tocris, #2296), somatostatin (SST; Sigma-Aldrich, #S9129), glucose-dependent insulinotropic polypeptide (GIP;Eurogentec, #AS-65568), L-glutamic acid (Sigma-Aldrich, #G1251), and epinephrine (Sigma-Aldrich, #E4375). These were added in bath solution, at the concentrations indicated in the figure legend, and pH was re-adjusted if needed. Na⁺ current was activated by a 50 ms depolarization from −70 to −10 mV.

Pancreas patch-seq

Our protocol for pancreas patch-seq is outlined in a recent paper (Camunas-Soler et al., 2020). In brief, following patch-clamp cells were collected using a separate wide-bore collection pipette (0.2–0.5 MOhm) filled with lysis buffer (10% Triton, Sigma-Aldrich, #93443; ribonuclease inhibitor 1:40, Clontech, #2313A), ERCC RNA spike-in mix (1:600000; ThermoFisher, #4456740), 10 mM dNTP, and 100 μM dT (3’-AAGCAGTGGTATCAACGCAGAGTACTTTTTTTTTTTTTTTTTTTTTTTTTTTTTTVN-5’) and then transferred to PCR tubes and stored at −80°C. We generated cDNA and sequencing libraries using an adaption of the SmartSeq-2 protocol for patch-seq plates (Camunas-Soler et al., 2020; Picelli et al.,2014). Libraries were generated from the amplified cDNA by tagmentation with Tn5 and sequenced in the NovaSeq platform (Illumina) using paired-end reads (100 bp) to an average depth of 1 million reads per cell. Sequencing reads were aligned to the human genome (GRCh38 genome with supplementary ERCC sequences) using STAR (Dobin et al., 2013), and gene counts determined using htseq-count (intersection-nonempty) using a GTF annotation with Ensembl 89 release genes (Anders et al., 2015). Gene expression was normalized to counts per million (cpm) after removal of counts corresponding to ERCC spike-ins and transformed to log2 values after addition of a pseudocount. This was followed by QC filtering of the sequenced cells (Camunas-Soler et al., 2020) to remove low-quality cells and potential doublets. Raw sequencing reads are available in the NCBI Gene Expression Omnibus (GEO) and Sequence Read Archive (SRA) under accession numbers GSE124742 and GSE164875. Correlation between transcript expression and electrophysiology was as outlined previously (Camunas-Soler et al., 2020). Visualization as t-distributed stochastic neighbor embedding (tSNE) plots was with the Qlucore Omics Explorer v3.6. Gene-set enrichment analysis (GSEA) was performed with of Z-scores or slopes for transcripts found in >20% of cells using the WEB-based Gene SeT AnaLysis Toolkit (webgestalt.org) and weighted set gene coverage to reduce redundancy in identified terms (Liao et al., 2019).

Hormone secretion measurements

Groups of 150–175 hand-picked mouse islets were perifused using a BioRep perifusion system (BioRep, Miami). Islets were pre-incubated in perfused KRB buffer containing (in mM): 140 NaCl, 3.6 KCl, 2.6 CaCl₂, 0.5 NaH₂PO₄, 0.5 MgSO₄, 5 HEPES, 2 NaHCO₃ and 0.5 mg/ml Essentially fatty acid free BSA (Sigma A6003) for 30 minutes, and then perifused in the same KRB with changes in glucose and KCl as indicated. Samples were collected at 90 −210 sec intervals and stored for assay of glucagon (U-PLEX Mouse Glucagon Assay, Meso Scale Diagnostics, #K1525YK) and insulin (STELLUX Rodent Insulin Chemiluminescence ELISA, ALPCO, #80-INSMR-CH10) at −20°C.

Immunostaining and single-cell protein analysis

Human paraffin embedded pancreas biopsies were sectioned to 3 μm and immunostained with antibodies against MAFB (Cell Signaling Technologies, #41019; RRID:AB_2799192), ARX (R&D Systems, # AF7068; RRID:AB_10973178) and GCG (Sigma-Aldrich, #G2654; RRID:AB_259852). The secondary antibodies were Cy2, Cy3, or Cy5 conjugated (1:500; Jackson ImmunoResearch, 111–225144, RRID:AB_2338021; 115-165-003; RRID:AB_2338680, 713-175-147, RRID:AB_2340730). Nuclear co-staining was conducted with DAPI Fluoromount G (Southern Biotech, #0100-20). Immunofluorescent mages were acquired on a Zeiss Axio Imager M2 widefield microscope with ApoTome. Images containing islet and exocrine cells were processed for nuclei segmentation, individual cell detection and cytosol border inference using Qupath software v0.2.3 (Bankhead et al., 2017). Parameters for nuclei detection: background radius 20px, median filter 5px, sigma 7px, minimum nuclei area 10px², maximum nuclei area 1000px², threshold of 2, nuclei were split by shape, and cell boundaries were determined by an expansion threshold of 12px with smoothing. Next, the relative expression levels of nuclear ARX, nuclear MAFB and cytosolic GCG levels for each detected cell were exported as .csv files and imported into Cytomap v1.4 for spatial analysis (Stoltzfus et al., 2020). In Qlucore Omics Explorer v3.6, we performed dimensionality reduction using tSNE of all imaged islet dataset using default parameters with 30 perplexity and 0.5 theta. Data was normalized for each dataset and the relative GCG, ARX and MAFB levels were displayed in the tSNE plot. Next, we created a gate to select the α-cell population in our dataset and the relative levels of ARX and MAFB were plotted. This information was used to determine the identity of ARX^hi and ARX^low cells and the relative expression levels of ARX, MAFB and GCG for each α-cell subpopulation was plotted.

Gene-set and pathway analysis of scRNA-seq

To characterize the expression program of ARX^hi and ARX^low α-cells in ND and T2D samples, we conducted gene set enrichment analysis (GSEA) on aggregated ARX^hi versus ARX^low α-cell transcriptomes from our recently published data (Avrahami et al., 2020) using a hypergeometric test by Genomica software (http://genomica.weizmann.ac.il/) considering gene sets with a P value <0.01 and an FDR <0.05 to be significantly enriched. Pathway analyses were performed with Gene Set Enrichment Analysis (GSEA) (Subramanian et al., 2005).

QUANTIFICATION AND STATISTICAL ANALYSIS

Electrophysiological Fingerprint Modelling

Multiple regression was carried out on ND α-cells using Ordinary Least Squared (OLS) Regression with Statsmodel v0.12.2 (Seabold and Perktold, 2010). Independent variables included age, sex, body mass index (BMI), HbA1c, cold ischemia time (CIT), culture time, and glucose concentration. Dependent variables included: cell size (pF), total exocytosis (fF/pF), early exocytosis (fF/pF), late exocytosis (fF/pF), peak Na⁺ current (pA/pF), Na⁺ half-Inactivation (mV), early Ca²⁺ current (pA/pF), late Ca²⁺ current (pA/pF), Ca²⁺ charge entry during an initial depolarization (pC/pF), exocytosis normalized to Ca²⁺ charge entry (fF/pC), reversal potential (mV), and α_probability from Models 1–3 (see below). Cells lacking data for a dependent variable were only dropped from that specific OLS analysis. Classification of cell type was conducted using the above electrophysiological measures as dependent variables from ND donors in an Optimizable Ensemble that either included (Model 1) or excluded (Model 2) cell size in MATLAB, or using Extreme Gradient Boosting (XGBoost v1.0.2) in a Python v3.7.11 framework that excluded cell size and reversal potential and restricted training data to 32 ND donors within an age of 20–70 years, a BMI of 18.5–30.3, and a pancreas CIT of ≤20 hours. Fine tuning was performed with a predetermined minimum accuracy of 80% for both α- and β-cells and training was performed with early stopping set to 50 iterations and utilized AUCPR as the evaluation metric. Models were trained on 80% of the α- and β-cells, with 20% reserved for testing and validation. Confusion matrices were generated using scikit-learn v0.24.2 (Pedregosa et al., 2012). We applied the models across our combined immunostaining and patch-seq database of ND and T2D cells and utilized the classifier’s predicted probability scores to assess fit to α-cell (αprobability = 1.0) and β-cell (αprobability = 0.0) models.

Statistical Analysis

Data are expressed as mean and standard error (line plots), mean and 10–90 percentile range (box and whisker plots), or as violin plots with median and quartiles indicated. When comparing two groups we used Student’s t-test or the non-parametric Mann-Whitney test to compare ranks. When comparing more than two groups we used one-way or two-way ANOVA followed by either the Tukey post-test or two-stage step-up method for estimation of false discovery rate (FDR); or alternatively the non-parametric Kruskal-Wallis test followed by Dunn’s post-test to correct for multiple comparisons. Statistical tests used are indicated in the figure legends. p-values less than 0.05 were considered as significant. For correlation of electrophysiology with transcript expression, Spearman tie-corrected correlations were computed for each gene and significance was tested by bootstrapping (1,000 iterations) as described in our previous work (Camunas-Soler et al., 2020). For ORA and GSEA, false discovery rate (FDR) of reported pathways was <0.05 or <0.1 as indicated in the figures using Bengamini-Hochberg (BH) correction within the WEB-based GEne SeT AnaLysis Toolkit (webgestalt.org). We did not perform power calculations prior to experiments, and experimenters were not blinded. Animals were assigned randomly to CD and HFD groups, and all human donors were accepted for islet isolations if they met the requirement for negative serology reports. No cells that passed standard electrophysiology and sequencing quality control (described above) were excluded from analysis.

Supplementary Material

9

Supplementary Table 1. Human donors included in the study, related to Figures 1, 2, 5, 6, and 7

1

Supplementary Table 2. Transcript correlations with total exocytosis from alpha-cells of donors with no diabetes (ND) at 1 mM and 10 mM glucose, related to Figure 2B

2

Supplementary Table 3. Transcript correlations with total exocytosis from alpha-cells of donors with type 2 diabetes (T2D) at 1 mM and 10 mM glucose, related to Figure 2G

3

Supplementary Table 4. Differential expression analysis of transcripts in mouse alpha-cells from chow-fed versus high-fat-fed mice, related to Figure 4

4

Supplementary Table 5. Transcript correlations with Na⁺ current voltage-dependence of half-inactivation of alpha-cells from Chow-fed and HFD-fed mice, related to Figure 4

5

Supplementary Table 6. Correlation of transcripts with peak Na⁺ current and Na⁺ current half-inactivation in human ND and T2D alpha-cells, related to Figure 5

6

Supplemental Table 7. Correlation of transcript expression with α_probability values from three different classifier models applied to ND and T2D alpha-cells, related to Figure 6

7

Data Supplement S1. Source data used to generate all graphs, related to Figures 1, 3–7 and Supplemental Figures 1, 2, 4 and 6.

8

Supplementary Figures. Supplementary Figure legends and Supplementary Figures 1–7

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Insulin Antibody (H-86)	Santa Cruz	SC-9168; RRID:AB_2126540
Anti-Glucagon Antibody, clone 13D11.33	EMD Millipore	MABN238; RRID:AB_433707
Goat anti-Guinea Pig IgG (H+L) Highly Cross-Adsorbed Secondary Antibody, Alexa Fluor 594	ThermoFisher Scientific	A-11076; RRID:AB_141930
Monoclonal Anti-Glucagon antibody produced in mouse	Sigma-Aldrich	G2654; RRID:AB_259852
Human ARX Affinity Purified Polyclonal Ab antibody	R&D Systems	AF7068; RRID:AB_10973178
MAFB Antibody	Cell Signaling Technology	41019; RRID:AB_2799192
Cy3-AffiniPure Goat Anti-Mouse IgG	Jackson ImmunoResear ch	115-165-003; RRID:AB_2338680
Cy2-AffiniPure Goat Anti-Rabbit IgG	Jackson ImmunoResear ch	111-225-144, RRID:AB_2338021
Cy5-AffiniPure Donkey Anti-Sheep IgG	Jackson ImmunoResear ch	713-175-147, RRID:AB_2340730
Biological Samples
Human pancreatic islets	Alberta Diabetes Institute IsletCore	See Suppl Table 1
Human pancreatic islets	University of Alberta Clinical Islet Laboratory	See Suppl Table 1
Human pancreatic islets	Allegheny Health Network	See Suppl Table 1
Human pancreatic islets	Human Pancreas Analysis Program	See Suppl Table 1
Chemicals, Peptides, and Recombinant Proteins
RPMI 1640 media	ThermoFisher Scientific	11875
low glucose DMEM media	ThermoFisher Scientific	11885
Fetal Bovine Serum - Canadian origin	ThermoFisher Scientific	12483020
Bovine Serum Albumin	Sigma-Aldrich	A6003
Penicillin-Streptomycin	ThermoFisher Scientific	15070063
StemPro Accutase Cell Dissociation Reagent	ThermoFisher Scientific	A1110501
chemicals/salts for patch-clamp and KRB buffers	Sigma-Aldrich	n/a
isradipine	Sigma-Aldrich	I6658
agatoxin	Alomone Labs	STA-500
GV-58	Alomone Labs	G-140
rotenone	Sigma-Aldrich	R8875
antimycin A	Sigma-Aldrich	A8674
lysophosphatidic acid	Sigma-Aldrich	L7260
alpha-latrotoxin	Enzo Life Sciences	ALX-630–027-C040
SLIT-2-N human	Sigma-Aldrich	SRP3155
prostaglandin E2	Tocris	2296
somatostatin	Sigma-Aldrich	S9129
GIP, rat	Eurogentec	AS-65568
L-glutamic acid	Sigma-Aldrich	G1251
epinephrine	Sigma-Aldrich	E4375
LIVE/DEAD Fixable near-IR dead cell dye	Life Technologies	L10119
ERCC spike-in control	ThermoFisher Scientific	4456740
5X All-In-One RT Master Mix	Applied Biological Materials	G486
10% Triton	Sigma-Aldrich	93443
Recombination RNase inhibitor	Clontech	2313A
SUPERase-In RNase Inhibitor	ThermoFisher Scientific	AM2696
10 mM dNTP	ThermoFisher Scientific	18427088
DAPI Fluoromount-G	SouthernBiotech	0100–20
Critical Commercial Assays
U-PLEX Mouse Glucagon Assay	Meso Scale Diagnostics	K1525YK
STELLUX Rodent Insulin Chemiluminescence ELISA	ALPCO	80-INSMR-CH10
KAPA HiFi HotStart ReadyMix	KAPA Biosystems	KK2601
Nextera XT	Illumina	FC-131–1096
Deposited Data
Single cell mRNA-seq data	this paper	GEO: GSE164875
Single cell mRNA-seq data	Camunas-Soler et al., Cell Metab, 2020	GEO: GSE124742
Single cell mRNA-seq data	Avrahami et al., 2020	GEO: GSE154126
Processed patch-seq datasets	Camunas-Soler et al., 2020	https://github.com/jcamunas/patchseq
Experimental Models: Organisms/Strains
mouse C57BL/6NCrl Inbred	Charles River Laboratories	CRL:027; RRID:IMSR_CRL:027
5L0D Picolab Laboratory Rodent Diet	LabDiet	3005659–220
Mouse Diet, High Fat Calories	VWR / Bio-Serv	CA89067–471
Oligonucleotides
SmartSeq2 OligodT: 5’–AAGCAGTGGTATCAACGCAGAGT ACT30VN-3’	Picilli et al., 2014	n/a
SmartSeq2 TSO: 5’-AAGCAGTGGTATCAACGCAGAGTA CATrGrG-3’	Picilli et al., 2014	n/a
SmartSeq2 ISPCR: 5’-AAGCAGTGGTATCAACGCAGAGT-3’	Picilli et al., 2014	n/a
Software and Algorithms
Qlucore Omics Explorer v3.6	Qlucore	www.qlucore.com
WEB-based Gene SeT AnaLysis Toolkit	Liao et al., 2019	webgestalt.org
Graphpad Prism v9.0.0	GraphPad	https://www.graphpad.com
Custom analysis software	Camunas-Soler et al., 2020	https://github.com/jcamunas/patchseq
PatchMaster 2×90.1	Smart Ephys HEKA	https://www.heka.com
FitMaster 2×90.1	Smart Ephys HEKA	https://www.heka.com
STAR	Dobin et al., 2013	https://github.com/alexdobin/STAR
HTSeq	Anders et al., 2015	https://github.com/simon-anders/htseq
MATLAB R2020a	Mathworks	https://uk.mathworks.com/products/matlab.html
Qupath v0.2.3	Bankhead et al., 2013	https://qupath.github.io
Cytomap v1.4	Stoltzfus et al., 2020	https://gitlab.com/gernerlab/cytomap
Genomica	Segal Lab of Computational Biology	http://genomica.weizmann.ac.il/
GSEA 4.1.0	Subramanian et al., 2005	http://software.broadinstitute.org/gsea/index.jsp
Python v3.7.11	Python Software Foundation	http://www.python.org
XGBoost v1.0.2	XGBoost	https://xgboost.readthedocs.io/en/latest/python/index.html
Scikit-learn v0.24.2	Pedregosa et al., 2012	https://scikit-learn.org/stable/
Statsmodel v0.12.2	Seabold et al., 2010	https://www.statsmodels.org/stable/index.html

Highlights:

• Glucose suppresses α-cell exocytosis by inhibiting P/Q-type Ca²⁺ currents

• Patch-seq links maturation, respiration, and receptor expression to α-cell function

• Dysfunction of α-cells associates with a ‘β-cell-like’ electrophysiologic signature

• Impaired exocytosis occurs in α-cells enriched for lineage and immaturity markers