A method for the generation of human stem cell-derived alpha cells
Nature Communications volume 11, Article number: 2241 (2020) Cite this article
The generation of pancreatic cell types from renewable cell sources holds promise for cell replacement therapies for diabetes. Although most effort has focused on generating pancreatic beta cells, considerable evidence indicates that glucagon secreting alpha cells are critically involved in disease progression and proper glucose control. Here we report on the generation of stem cell-derived human pancreatic alpha (SC-alpha) cells from pluripotent stem cells via a transient pre-alpha cell intermediate. These pre-alpha cells exhibit a transcriptional profile similar to mature alpha cells and although they produce proinsulin protein, they do not secrete significant amounts of processed insulin. Compound screening identified a protein kinase c activator that promotes maturation of pre-alpha cells into SC-alpha cells. The resulting SC-alpha cells do not express insulin, share an ultrastructure similar to cadaveric alpha cells, express and secrete glucagon in response to glucose and some glucagon secretagogues, and elevate blood glucose upon transplantation in mice.
Although diabetes primarily involves beta cell dysfunction, there is mounting evidence that alpha cell defects play a role in disease etiology 1 , 2 . Patients with type 1 diabetes must cope with significant fluctuation in blood glucose levels including acute hypoglycemia 3 .
In the healthy pancreas, hormone-expressing endocrine cells function within the islets of Langerhans to precisely regulate blood glucose and energy metabolism. During hypoglycemia (low blood glucose), islet alpha cells secrete glucagon 4 which raises blood glucose levels by increasing glycogenolysis and gluconeogenesis in the liver 5 , 6 . Although alpha cells persist in diabetic islets, these alpha cells are often incapable of mounting an appropriate glucagon response, perhaps due to the absence of alpha cell−beta cell interactions 7 . Recent studies implicate dysfunction of alpha cells as a contributing factor in the elevated blood glucose levels observed in diabetic patients 4 , 7 .
Published methods to make pancreatic beta cells 8 , 9 , 10 , 11 all report a minor portion of alpha and delta (somatostatin secreting) cells. Additionally, there are several reports on the conversion of various cell types into alpha cells via transdifferentiation 12 , 13 , 14 . In 2011, Rezania et al. 15 reported a protocol for generating glucagon-positive cells that exhibited some glucagon secretion in vitro; however, upon transplantation of 1.9 million cells into mice, these cells had limited physiological effects. Despite these early efforts to generate glucagon-positive cells, the production of alpha cells has not been reproduced nor widely adopted by the field.
We previously reported the generation of functional pancreatic beta cells from human pluripotent stem cells using a six-step directed differentiation protocol 8 . That protocol generates beta cells as well as side populations including polyhormonal cells and nonendocrine cells 11 . The presence of polyhormonal cells in pancreatic differentiations have also been observed in other reported protocols 9 , 16 , 17 . Several reports described polyhormonal cells as having features of immature beta cells 9 , 17 , 18 , 19 ; however, beyond the expression of insulin protein, there has been little evidence to support the similarity of polyhormonal cells to beta cells, let alone their capacity to differentiate into beta cells. In contrast, others have reported that polyhormonal cells are present during development, contribute to alpha cells later in development, express several markers of alpha cells, and give rise to glucagon-expressing cells when transplanted 11 , 15 , 20 .
Reports from the literature have referred to INS+/GCG+ cells in turn as “bi-hormonal” 21 , “polyhormonal” 9 , 22 , 23 , 24 , “alpha-like cells” 17 , or “alpha cells” 25 . The nomenclature of the INS+/GCG+ cells produced in directed differentiation protocols has become equally confusing. In Veres et al. 11 , we performed extensive scRNA-seq studies on beta cell differentiations and observed a population of cells expressing many markers of alpha cells which we labeled as “alpha-like” cells irrespective of insulin expression. In this manuscript we refine this definition of “alpha-like” cells to distinguish between cells that have detectable insulin protein which we define as “pre-alpha” cells and those that do not express insulin, which we define as “SC-alpha cells”. We favor the term “pre-alpha” rather than “bi-hormonal” or “polyhormonal” because it more accurately depicts the similarity of these cells to alpha cells yet distinguishes them from more mature alpha cells that do not express insulin (SC-alpha).
Here we build upon our previous reports to develop a protocol for the generation of SC-alpha cells. The protocol robustly produces a transient pre-alpha cell intermediate, which has a transcriptional signature similar to alpha cells except that they express insulin in addition to glucagon. We identify a small molecule capable of driving pre-alpha cells to an alpha cell identity and demonstrate the ability to produce approximately 30% SC-alpha cells in vitro. Our results show that these SC-alpha cells have a transcriptional signature similar to human cadaveric alpha cells, are responsive to some glucagon secretagogues, and elicit a robust physiological response within 4 weeks of transplantation in mice. The SC-alpha cells generated can recapitulate central aspects of alpha cell biology and represent an efficient and scalable avenue to produce alpha cells for use in islet organoids for cell replacement therapy, drug screening, disease modeling, and may accelerate exploration of alpha cell biology.
Optimization of differentiation to generate pre-alpha cells
Our previously published protocol for generating stem cell-derived beta cells (SC-beta cells) produces a small but significant population of pre-alpha cells 8 , 11 . Figure 1a shows a typical result wherein 27% of the cells are glucose-responsive SC-beta cells and 9% are pre-alpha cells expressing both insulin and glucagon. Kelly et al. 16 suggested that these pre-alpha cells (referred to as polyhormonal in ref. 16 ) come from progenitors that fail to express NKX6.1. Thus, we sought to modify our beta cell protocol to prevent or reduce induction of NKX6.1 at stage 4. Using the HUES8 embryonic stem cell line, we observed that removal of KGF, SANT-1 and treatment with LDN only on day 2 of stage 3 results in a decrease of NKX6.1-positive cells 11 . In addition, we observed that treatment with LDN on day 1 of stage 4 and incubation with no factors for the remainder of stage 4 resulted in a significant population of Chromogranin A+/NKX6.1− cells. Together these protocol modifications (Fig. 1b ) resulted in a large fraction of pre-alpha cells as marked by the coexpression of insulin and glucagon proteins (Fig. 1c ). In the HUES8 cell line, this pre-alpha cell optimized protocol produces an average of 62.6 ± 2.3% insulin+ and glucagon+ coexpressing pre-alpha cells and a small percentage (80%) expressed both insulin and glucagon proteins at the beginning of the culture period. After 14 days, little change had occurred as >80% of the cells still expressed insulin and glucagon protein. After 21 days, a population of insulin-, glucagon+ cells appears with ~20% of the cells expressing only glucagon protein. This population of monohormonal SC-alpha cells that appeared at day 21, persists at day 28 of extended culture. These results indicate that upon extended culture in vitro, a fraction of pre-alpha cells reduce insulin protein expression and become SC-alpha cells.
PKC activation promotes maturation of SC-alpha cells
Given that pre-alpha cells can convert to SC-alpha cells in vivo and in vitro, we sought to identify signals that promote this conversion by performing a small-molecule screen. HUES8 embryonic stem cells were differentiated using the pre-alpha cell optimized protocol to the end of stage 5 (stage 6 day 1), arrayed into 384-well plates and incubated with a custom 43-member small-molecule library (Supplementary Table 3 ) where compounds were chosen for their activity in targeting signaling pathways. After 96 h of treatment (stage 6 day 5), cells were fixed and stained for insulin and glucagon (Fig. 3a ). High-content imaging was used to quantify the dispersed (2D) cell populations for the percentage of cells expressing each hormone individually and the percentage of pre-alpha cells marked by expression of both hormones (Fig. 3b, c ). The effect of each compound was evaluated in quintuplicate assays. The protein kinase c (PKC) activator phorbol 12,13-dibutyrate (PDBu) decreased the percentage of insulin protein-expressing cells and increased the percentage of glucagon-expressing cells compared to vehicle controls. To confirm that the effects of this compound were not unique to the planar assay format, PDBu was evaluated in the final stage of the 3D directed differentiation protocol. Over the course of a 28-day treatment, PDBu induced a significant population of pre-alpha cell clusters to reduce insulin protein expression compared to control (33 ± 6% vs. 16 ± 5%; n = 5; Fig. 3d and Supplementary Fig. 5 ). The resulting cell population contained significantly more alpha cells that did not coexpress the insulin protein (Fig. 3e ).
Fig. 3: Screen to identify compounds that promote alpha cell identity.
Small molecules targeting known pathways (43 compounds) were incubated with pre-alpha cells in quintuplicate for 96 h. a Schematic of screening approach. Primary screening results showing b the percentage of cells expressing both insulin and glucagon and c the percentage of cells expressing only glucagon. The data are presented as mean ± SEM (n = 5 biologically independent samples). The PKC activator PDBu significantly reduced the percentage of pre-alpha cells and increased the percentage of cells expressing glucagon. d Representative flow cytometry results of pre-alpha cells at stage 6 day 1, and stage 6 day 28 control (no treatment), or 28 days treatment with the PKC activator PDBu. e Immunofluorescence of clusters treated with and without PDBu for 28 days. Scale bar = 50 μm. f Additional PKC activators evaluated for their effect on pre-alpha cells (stage 6 day 1). All PKC activators increased the percentage of SC-alpha cells. The data are presented as mean ± SEM, significance calculated using an ordinary one-way ANOVA with Dunnett multiple comparison test (n = 3 biologically independent samples). g Stage 6 of protocol converts pre-alpha cells into SC-alpha cells.
Full size image
To further explore the specificity of PKC activation, a small library of known PKC activators was evaluated. This structurally diverse set of compounds have all been reported to have PKC activating abilities. We added pre-alpha cell clusters (stage 6 day 1) to six-well plates and treated with either vehicle or each PKC activator for 28 days and assessed the percentage of monohormonal, glucagon-expressing alpha cells (Fig. 3f and Supplementary Fig. 6 ). While the control condition resulted in only 13.7% SC-alpha cells, treatment with each PKC activator resulted in a significant increase in the percentage of SC-alpha cells, though none more so than PDBu. To evaluate the stability of the shift from pre-alpha cells to SC-alpha cells, we withdrew PDBu from cells for 7 days and evaluated the percentage of monohormonal glucagon-expressing SC-alpha cells (Supplementary Fig. 7 ). Withdrawal of PDBu did not significantly affect the percentage of SC-alpha cells in our population. These results demonstrate that activation of PKC stably reduces the expression of insulin in the INS+/GCG+ population thereby accelerating the conversion process of pre-alpha cells to SC-alpha cells in vitro (Fig. 3g ).
As previously described, differentiation of hPSCs to beta cells results in a side population of pre-alpha cells. To this end, we evaluated the effect of PKC activation on the differentiation of SC-beta cells. Treatment of stage 6 SC-beta cells with the PKC activator PDBu did not have a significant effect on the number of beta cells in the population (Supplementary Fig. 8a ). However, when treated with PDBu, the pre-alpha side population in the differentiation was reduced and the percentage of SC-alpha cells was increased (Supplementary Fig. 8b ). Because previous reports pointed to variability when using different cell lines 30 , the robustness of this SC-alpha protocol was evaluated in the 1016 iPS cell line. Differentiated 1016 cells generate a similar percentage of SC-alpha cells (Supplementary Fig. 9 ) showing that the protocol is able to direct another stem cell line to SC-alpha cells using this protocol. To establish the performance of our SC-alpha cell protocol in comparison to previously published reports, we performed a head-to-head comparison where SC-alpha cells were generated with our protocol or the Rezania protocol 15 in order to assess protocol efficiency and the functional performance of the resulting cell populations (Supplementary Figs. 1 , 10 , 11 , 12 , 13 ). Using the HUES8 cell line, the percentage of SC-alpha cells produced was improved more than eightfold over previously reported protocols and demonstrates improved functional responses to glucose.
Molecular characterization of pre-alpha and SC-alpha cells
To investigate the transcriptional changes that occur as cells transition from pre-alpha cells to SC-alpha cells, we performed single-cell RNAseq on pre-alpha cells and SC-alpha cells (Fig. 4 ). On a global level, the transcriptional profile of pre-alpha cells and SC-alpha cells are remarkably similar with insulin expression being a notable exception (Fig. 4a ). Analysis of 82 islet specific genes in comparison to human islet expression patterns further confirms the similarity of both pre-alpha cells and SC-alpha cells to primary human alpha cells (Fig. 4b ). These results suggest that the conversion of pre-alpha cells to SC-alpha cells represents a subtle maturation process rather than a cell fate change. Consistent with our global gene expression analysis, a KEGG and gene ontology pathway analysis demonstrates significant similarity between pre-alpha and SC-alpha cells (Fig. 4c, d and Supplementary Figs. 14 and 15 ) with the most significant pathways common between both cell types. These top pathways are consistent with the function of alpha cells and include protein export, protein processing in the ER, protein folding, and cellular signaling pathways. Despite these similarities, some differences were observed. Pre-alpha cells uniquely express genes enriched in the insulin secretion and metabolic stress pathways, while SC-alpha cells uniquely express genes enriched in the glucagon secretion and hormone action pathways.
Fig. 4: Molecular characterization of pre-alpha cells and SC-alpha cells.
a Scatterplot showing relative expression of pre-alpha (x-axis) and SC-Alpha (y-axis) transcript counts. Both axes are log2(cpm). Genes with fold change greater than 0.5 or less than −0.5, and p value < 0.01 as calculated using a Wilcoxon rank sum test, are highlighted in red. b Heatmap showing pre-alpha and SC-Alpha cells in comparison with human alpha, beta, and delta cells. Top 46 genes are alpha cell specific, middle 31 genes are beta cell specific, and bottom 5 genes are delta cell specific. Coloring is based on z-score from 1.5 as high (yellow) to −1.5 as low (purple). c Venn diagrams comparing KEGG pathway (left) and Gene Ontology-Biological Process (right) database overlap between the highest expressing 150 genes in pre-alpha and SC-Alpha cells. Significant terms defined as p value < 0.01 as calculated using an EASE Score (modified Fisher Exact test). d Heatmaps showing selected pathways from KEGG and Gene Ontology-Biological Process terms for pre-alpha and SC-Alpha cells. Columns show number of genes mapped to the pathway divided by number of possible genes in the pathway, in addition to its percentage. Scale is from –log(p value) of 2 (light red) to 10 (green) for KEGG and –log(p value) of 2 (light red) to 7 (green) for GO-BP. All pathways that were not significant as calculated using an EASE Score (modified Fisher Exact test), i.e. –log(p value) from 0 to 2, are labeled in red. Full heatmaps are shown in Supplementary Figs. 14 and 15 .
Full size image
Using pseudotime analysis, we established a sequence of transcriptional changes that occur as cells progress from pre-alpha to SC-alpha cells (Supplementary Fig. 16 ). The transition from pre-alpha to SC-alpha results in the sequential decrease of stress-related genes VTN and AFP followed by beta and delta cell genes INS, TSPAN1, GAL. Finally, alpha-cell-associated genes PLIN2, TXNIP, and GC are upregulated. To further examine the role of PKC activation on these transcriptional changes, we performed an additional scRNA-seq experiment to compare the transcriptional profile of SC-alpha cells generated with PDBu to untreated SC-alpha cells which resulted from a spontaneous conversion from pre-alpha to SC-alpha cells. (Supplementary Fig. 17a ). The tSNE plots from stage 6 day 28 cells generated with and without PDBu resulted in three major overlapping cell populations indicating significant similarity in their transcriptional profile. The predominant population expresses transcripts that are known markers of cadaveric alpha cells (ARX, IRX1, IRX2) 27 (Supplementary Fig. 17b ). Very few genes are differentially expressed between the SC-alpha generated with and without PDBu (Supplementary Fig. 17c and Supplementary Table 2 ). Of the genes that are differentially expressed, insulin transcript is downregulated 2.3-fold in cells treated with PDBu. Glucagon transcripts were not significantly affected by PDBu treatment.
To evaluate the protein expression levels of insulin and glucagon, we performed Western blots on cell lysates from cells treated with PDBu (Supplementary Fig. 18 ). Cells treated with PDBu had a higher expression of glucagon protein compared to untreated cells. In untreated samples, glucagon expression is decreased. As we observed in our flow cytometry analysis (Fig. 3d ), the Western blot confirms that PDBu treatment decreases the level of insulin protein expression (Supplementary Fig. 18 ).
Structural and functional characterization of SC-alpha cells
SC-alpha cells express known markers of alpha cells including GCG, PDX1, IRX, PC2, but do not express beta cell-specific markers PC1 or NKX6-1 (Fig. 5a−j ). Electron microscopy reveals that the ultrastructure of SC-alpha cells resembles that of human primary alpha cells. In human islets, glucagon granules in alpha cells are dark and diffuse with round cores with an average size of 242 ± 8 nm (Fig. 5k ), which is smaller than the condensed insulin granules of the beta cell 31 . In SC-alpha cells, the secretory granules have a similar ultrastructure to alpha cells with an average size of 214 ± 12 nm (Fig. 5l ).
Fig. 5: Characterization of SC-alpha cells.
Pre-alpha (a, c, e, g and i) and SC-alpha cells (b, d, f, h, and j) stained for PC1 (a and b), PC2 (c and d), IRX1 (e and f), PDX1 (g and h), and NKX6.1 (i and j). Scale bar for a−j = 200 μm. The ultrastructure and granule morphology of human cadaveric alpha cells (k) and SC-alpha cells (l) as assessed by electron microscopy are similar. Arrows indicate granules; dashed line denotes the cell boundary. Inlay shows granule morphology. Scale bar for k and l = 500 nm. m SC-alpha cells secrete glucagon in response to glucose and are inhibited by somatostatin (SST). SC-alpha cell response to the glucagon secretagogue veratridine. Data are presented as mean ± SEM, significance calculated using a paired ratio Student’s t test (n = 3 biologically independent samples). n Representative electrophysiology recording of primary human alpha cell showing electrical activity in response to low (1 mM) and high (10 mM) glucose. Amplitude differences at low and high glucose are shown in the inlay (representative; n = 5 biologically independent cells). o Representative electrophysiology recording of SC-alpha cell showing electrical activity in response to low (1 mM) and high (11 mM) glucose. Amplitude differences at low and high glucose are shown in the inlay (representative; n = 30 biologically independent cells).
Full size image
To evaluate the ability of SC-alpha cells to respond to physiological conditions, we screened a number of glucagon secretion modulators (Fig. 5m ). Like mature human alpha cells 32 , 33 , 34 , 35 , 36 , glucagon secretion in SC-alpha cells is inhibited by glucose, an effect that is maximal at 7 mM glucose with higher glucose concentrations being less inhibitory (Fig. 5m ). This pattern of glucagon secretion in response to increasing glucose concentrations has been reported by others 37 . Glucagon secretion from SC-alpha cells is suppressed in the presence of exogenous somatostatin, normally released within the islets by the delta cells, unaffected by exogenous insulin and arginine and stimulated by veratridine, an activator of voltage-gated Na+ channels. Arginine evokes a transient stimulation of glucagon secretion ex vivo in mature alpha cells and time-resolved measurements may therefore be necessary to detect its stimulatory effect 38 .
We also sought to evaluate the stimulatory capacity of SC-alpha cells using electrophysiology. Using a fluorescent reporter cell line that expresses mCherry under the endogenous glucagon promoter, we identified alpha cells for patch clamp experiments and compared the electrical activity of these cells to primary human alpha cells in response to changing glucose concentrations. Like primary alpha cells, SC-alpha cells generate spontaneous overshooting action potentials in the presence of 1 mM glucose (Fig. 5n, o ). In the SC-alpha cells, these action potentials reflect activation of voltage-gated Na+, Ca2+, and K+ channels, similar to what has been demonstrated in adult human alpha cells 37 . The effects of high glucose (10−11 mM) were also similar: slight depolarization of the interspike membrane potential, an increase in action potential frequency and a reduction in action potential height (Fig. 5n, o inlays). These effects of high glucose are similar to those reported in mouse alpha cells and indicate that glucagon secretion may be influenced by action potential firing in a amplitude- rather than frequency-dependent fashion 39 . In all, the results demonstrate that SC-alpha cells resemble human primary alpha cells in their transcriptional profile, glucagon granule morphology, electrophysiology and physiological response to several (but not all) glucagon secretion modulators in static incubations.
SC-alpha cells reduce hypoglycemia in transplantation models
We evaluated the potential utility of these cells to modulate physiology in vivo by transplanting SC-alpha cells. Using a continuous glucose monitor (CGM), we evaluated the interstitial glucose concentrations in animals at 5-min intervals during a period of fasting-induced hypoglycemia. Control animals (sham surgery) exhibited hypoglycemia after 6 h of fasting while mice transplanted with SC-alpha cells were protected from hypoglycemia (Fig. 6a ). In SC-alpha-transplanted animals, blood glucose levels begin to return to baseline before food was restored while control animals continued to decrease blood glucose until food was restored. Furthermore, when administered an i.p. bolus of exogenous insulin, SC-alpha-transplanted animals are protected from hypoglycemia (Fig. 6b ). To evaluate the effect of transplanted SC-alpha cells on long-term blood glucose values, we measured interstitial blood glucose continuously starting 4 weeks after transplantation and continuing until week 8. During this time, mice were maintained under normal housing conditions with 24-h light/dark cycles and feeding ad libitum. At the end of the 4-week observation period, CGM glucose readings for all mice in a cohort were averaged and represented as an average blood glucose value per 5-min interval throughout a standard 24-h period with the associated standard error (Fig. 6c ). As has previously been reported 40 , 41 , 42 , 43 , control mice exhibited a characteristic fluctuation pattern in blood glucose values over a standard 24-h period with elevated blood glucose levels during active/feeding periods (dark) and lower blood glucose levels during resting periods (light). In SC-alpha-transplanted mice, we observed a similar pattern of glucose fluctuations, although the average blood glucose concentrations were elevated compared to control mice (Fig. 6d ). Transplantation of SC-alpha cells does not perturb the normal circadian regulation of blood glucose concentrations in these animals but the presence of additional glucagon secreting cells raises the basal blood glucose concentrations in these animals. Further evaluation of circadian and ultradian patterns in blood glucose levels confirms that there is no significant shift in periodicity of these previously reported glucose rhythms (Supplementary Fig. 19 ).
Fig. 6: In vivo characterization of SC-alpha cell transplantation.
a Fasting-induced hypoglycemia in control (n = 9 animals) and SC-alpha transplanted mice (n = 9 animals). Continuous glucose monitoring for 24-h period at 5 weeks post transplant. Food was removed after 15 h and restored at 22 h. SC-alpha-cell-transplanted mice are protected from hypoglycemia. b Insulin tolerance test on control (n = 9 animals) and SC-alpha (n = 9 animals) transplanted mice at 6 weeks post transplant. SC-alpha-cell-transplanted mice are protected from hypoglycema. c Continuous glucose monitoring of mice for 4 weeks reveals that mice transplanted with SC-alpha cells (n = 10 animals) have elevated blood glucose compared to control animals (sham surgery, n = 10 animals). Graph represents the average daily blood glucose value at 5-min intervals for all animals in each treatment group. Each data point represents the average of 280 CGM readings (10 mice × 28 days). d In transplanted mice, SC-alpha cells reduce instances of hypoglycemia in mice. Average blood glucose reading by CGM over 28-day period (n = 10 animals for each group). e Percentage of time over 4 weeks spent in range (≥70 mg/dL) and hypoglycemic (