RFX6 Maintains Gene Expression and Function of Adult Human Islet α-Cells

Dysfunction of pancreatic islet cells is a hallmark of diabetes. Previous studies have identified islet-enriched transcription factors (TFs) crucial for regulation and maintenance of islet cell subtypes, especially insulin-secreting β-cells and glucagon-secreting α-cells. One such TF is regulatory factor X-box binding 6 (RFX6), a conserved winged helix TF expressed in all pancreatic endocrine cells (1–8).

RFX6 is required for pancreatic islet cell development and insulin production in zebrafish, mice, and humans (3,5–8). Human RFX6 mutations can result in monogenic diseases, including Mitchell-Riley syndrome, neonatal diabetes, and a subset of maturity onset diabetes of the young (RFX6-MODY) (3,9–12). RFX6 expression is also dysregulated in human β- and α-cells in type 1 diabetes and type 2 diabetes (T2D) (13,14).

Despite higher expression in human α-cells than in β-cells, the function of RFX6 in α-cells remains poorly understood (15). Homozygous RFX6-null mice exhibit phenotypes consistent with human Mitchell-Riley syndrome and patients with RFX6-MODY, including reduction of islet cells expressing insulin, glucagon, somatostatin, or ghrelin. However, these mice die soon after birth, precluding studies of RFX6 function in mature α-cells (6). Heterozygous RFX6-null mice do not develop diabetes, despite the association of heterozygous RFX6 mutations with MODY in humans (6,12). Previous studies have investigated the effects of β-cell–specific Rfx6 knockout in mice, revealing dysregulated insulin secretion and calcium channel activity (3,5). However, analogous studies of selective α-cell RFX6 targeting have not been reported. Moreover, differences between mouse and human islets in cytoarchitecture, gene regulation, and function necessitate research in human models to comprehensively understand diabetes pathogenesis (16–18). While RFX6 function has been studied in the human β-cell line EndoC-βH2, human α-cell lines have not yet been generated (3). These challenges and knowledge gaps motivated our studies of RFX6 function in α-cells from primary human islets. Here we innovated an approach for selective lentiviral targeting of adult human α-cells to study the impact of acute RFX6 loss using transplantation-based assays, RNA-sequencing (RNA-seq), cleavage under targets and release using nuclease (CUT&RUN) genomic mapping, and patch clamp electrophysiology (2,4,19–21). These studies reveal that RFX6 is required in human α-cells to maintain gene regulation and hallmark functions.

See Supplementary Table 1 for all reagent catalog numbers.

Human Pancreas Analysis Program (HPAP-T2D) organ procurement and processing were performed as previously described (22). SmartSeq2 single-cell RNA-seq fastq files from 21 donors (10 T2D, 11 non-T2D) were obtained from PANC-DB. Alignment to the human genome (Genome Reference Consortium Human Build 38 with External RNA Controls Consortium sequences) was performed using STAR, and gene counts were determined using htseq-count (intersection-nonempty, Geographical Topic Phrases annotation) with Ensembl 97 release genes (23,24). Cells with <1,000 genes detected or <80,000 mapped reads were excluded. Gene expression was normalized to counts per million after removal of counts corresponding to External RNA Controls Consortium spike-ins and log-transformed after addition of a pseudocount. Data analyses, including normalization, scaling, and clustering, used Seurat v3.1.1 following the developer's manual (https://github.com/satijalab/seurat), unless otherwise noted (25,26).

Islets from previously healthy subjects were procured via the Integrated Islet Distribution Program, Alberta Disease Institute IsletCore, University of San Francisco Islet Core, Stanford Diabetes Center Islet Core, National Diabetes Research Institute, and International Institute for the Advancement of Medicine. See Supplementary Table 2 for donor information.

Commercial GIPZ human lentiviral RFX6 shRNA, scramble control, and RFX6-FLAG overexpression constructs were obtained from Dharmacon. Lentiviruses were produced by transient transfection of HEK293T cells with the vectors and packaging constructs. Supernatants were collected and concentrated using PEG-it and then stored at −80°C prior to transduction of primary human cells.

Pseudoislet generation was performed as previously described in detail (4).

Islets were dissociated as with whole pseudoislet generation. Prior to lentiviral transfection, α-cells were isolated via the EasySep Release Human PE Positive Selection kit using a phycoerythrin (PE) anti-human CD26 antibody, following the manufacturer’s protocol. CD26⁺ cells were transduced with 1 × 10⁹ viral units/mL lentivirus and cultured separately from untransduced CD26^neg cells for 4 days in an ultra-low attachment 96-well plate. Cells were then dissociated and seeded into a 24-well AggreWell400 plate at ∼2 × 10⁵ cells/well, following the manufacturer’s protocol. MagSep pseudoislets were cultured overnight before further analysis.

RNA extraction and quantitative RT-PCR (qRT-PCR) were performed as previously described in detail (2,4). TaqMan probes are listed in Supplementary Table 1.

MagSep pseudoislets shipped overnight from Stanford were dissociated into single cells and cultured in 35-mm dishes in DMEM with 10% FBS and 1% penicillin-streptomycin for up to 2 days. Only cells positive for green fluorescent protein (GFP) were patched to confirm knockdown, while cell identity was confirmed by post hoc immunostaining for glucagon. Electrophysiological measurements of voltage-dependent exocytosis were performed as previously described in detail (27).

For measurement of calcium channel activity using previously described methods (27), barium was used as a charge carrier. The extracellular solution contained (in mmol/L): 105 NaCl, 5.6 KCl, 20 BaCl₂, 1.2 MgCl₂, 10 HEPES, and 5 tetraethylammonium-Cl (pH 7.4 with NaOH). The intracellular solution contained (in mmol/L): 130 CsCl, 5 tetraethylammonium-Cl, 1 MgCl₂, 10 HEPES, 10 EGTA, 4 MgATP, and 1 Li guanosine-5′-triphosphate (pH 7.15 with CsOH). Currents were measured 30 s after obtaining the whole-cell configuration by 100-ms depolarizations from −50 mV with 10 mV increases up to 60 mV. Currents were normalized to the cell size (pico amperes/pico farad [pA/pF]). Data were analyzed using FitMaster (SmartEphys HEKA) and Prism (GraphPad Software).

Batches of 25 pseudoislets (Supplementary Table 2) were used for in vitro secretion assays, as previously described (2,4). Pseudoislets were equilibrated at 7 mmol/L glucose (glucagon secretion) or 2.8 mmol/L glucose (insulin secretion), then incubated at 7 mmol/L and 1 mmol/L glucose (glucagon secretion) or 2.8 mmol/L, 16.7 mmol/L, and 16.7 mmol/L glucose plus 3-isobutyl-1-methylxanthine (insulin secretion) for 60 min each. Supernatants collected at the end of each incubation and pseudoislet lysates were quantified by glucagon or human insulin ELISA kits. Results are presented as a percentage of total glucagon or insulin content. Assays were conducted in RPMI 1640 media supplemented with 1% FBS and the above glucose concentrations.

We performed experiments using sex- and age-matched mice. Sex was not considered a factor in the statistical analysis of the data. GKO-NSG mice were generated and raised in the Stanford Animal Facility and genotyped as previously described (21).

Human pseudoislets (n = 1,000) (Supplementary Table 2) suspended in Matrigel were transplanted into the left renal capsular space of host mice, as previously described (2,4). Transplant recipients were 8-week-old GKO-NSG mice (sex- and age-matched within experiments) and were anesthetized using ketamine/xylazine.

Mice received intraperitoneal insulin injections at 0.5 units insulin/kg body weight at least 1 month after transplantation to allow for engraftment and detection of glucagon. Tail vein blood samples were collected at 0, 15, 30, 45, and 60 min after injection. Serum glucagon levels were measured by glucagon ELISA.

Pseudoislets were dispersed into single cells and stained with LIVE/DEAD Fixable Near-IR Dead Cell Stain Kit. Cells were then labeled with the following conjugated antibodies: Alexa Fluor 647-ectonucleoside triphosphate diphosphohydrolase-3 (AF647-NTPDase3; 1:50), PE-CD26 (1:100), PE-Texas Red-CD45 (1:100), and PE-Cy7-HPi2 (1:100). Incubations were conducted at 4°C for 30 min. Labeled cells were sorted on a special order five-laser FACS Aria II using a 100-µm nozzle with appropriate controls and doublet removal. Sorted cells were collected into low retention tubes containing cell staining buffer supplemented with Ribolock RNase inhibitor. Cytometry data were analyzed using FlowJo 10.8 software.

RNA from ∼5,000 sorted live α- or β-cells having RNA integrity number >7 were used for RNA-seq library construction. SMART-seq v4 Ultra Low input RNA kit amplified cDNA, and Nextera XT DNA Library Preparation kit generated RNA-seq libraries. Barcoded libraries were multiplexed and sequenced as paired-end 150-base pair reads on the Illumina NovaSeq 6000 platform. Sixteen libraries were generated from four different donors for αRFX6 knockdown (αRFX6^kd), βRFX6^kd, and respective controls.

RNA-seq analysis used FastQC v0.11.9 for quality control. All libraries had >30 million reads and Trim Galore! v0.6.5 trimmed barcodes. STAR v2.7.9a aligned reads to the human genome (Genome Reference Consortium human build 38) (23). RSEM v1.3.1 quantified transcripts per million (TPM) (28). DESeq2 identified differentially expressed genes (P-adjusted cutoff: 0.1) (29). Gene set enrichment analysis used DAVID v2002q3 and g:Profiler e106_eg_p16_65fcd96 (30,31).

Extracted nuclei from 500,000 dispersed RFX6-FLAG pseudoislet cells per condition were processed using CUTANA ChIC/CUT&RUN protocol v3.1 (10-min RT incubation with proteins pAG-MNase, 2-h nutation at 4°C with 100 mmol/L CaCl₂). Libraries were prepared with the CUTANA CUT&RUN Library Prep Kit, according to manual instructions. Sequencing on the NovaSeq 6000 platform generated paired-end 150-base pair reads (Novogene; >25 million reads). The CUT&RUNtools pipeline (32) performed read trimming (Trimmomatic), alignment (Bowtie2), and peak calling (HOMER’s macs2.narrow outputs) (33–35). Genome browser tracks, peak calling, and motif identification (using “makeUCSCfile,” “findPeaks,” and “findMotifs” commands, respectively) were performed. Motif enrichment P values were generated by HOMER, and gene annotation was generated with the GREAT algorithm using the default parameters (36).

Figure legends provide information on the number of replicates, standard deviation, and statistical analyses for qRT-PCR, glucose-stimulated insulin secretion (GSIS)/glucose-suppression of glucagon secretion (GSGS), intraperitoneal insulin tolerance test (IP-ITT), and area under the curve. GraphPad Prism v9.1 and R v4.1.1 were used for graphing and statistical analyses. ComplexHeatmap in R was used to generate heat maps. Browser tracks were generated with the University of California Santa Cruz genome browser (37). Biorender.com was used for method graphics.

All studies were conducted with deidentified human islets and are therefore not considered as human subject research by Stanford University Institutional Review Board guidelines. All animal experiments and methods were approved by the Stanford Administrative Panel on Laboratory Animal Care (APLAC 29985).

The data sets generated and analyzed during the current study were uploaded to the National Center for Biotechnology Information’s Gene Expression Omnibus (GSE248854) (38). Our studies used GKO-NSG mice from The Jackson Laboratory (Research Resource Identifier: International Mouse Strain Resource [IMSR]_JAX: 029819).

RFX6 is expressed in all endocrine islet cells and most abundantly in α-cells, as determined from single-cell RNA-seq (Research Design and Methods) (Fig. 1A). These data demonstrate that the proportion of islet α- and β-cells expressing RFX6 is also reduced from organ donors with T2D (Fig. 1B). To investigate RFX6 function in human islet cells, we used lentiviral vectors coexpressing a GFP reporter to achieve shRNA-mediated suppression of RFX6 (hereafter, RFX6^kd); a nontargeting lentivirus was used as a control (hereafter, “control”) (Fig. 1C, Supplementary Fig. 1, and Supplementary Table 2). After dispersion, primary islet cells were infected with lentivirus, then reaggregated to form pseudoislets (hereafter, whole RFX6^kd pseudoislets), allowing subsequent molecular, cellular, and functional analyses (Fig. 1D) (2,4). We assessed in vitro GSGS in whole RFX6^kd pseudoislets 5 days after lentiviral transduction. GSGS data were normalized to total pseudoislet glucagon content. Whole RFX6^kd pseudoislets did not exhibit dysregulated glucagon secretion compared with controls, despite clear suppression of RFX6 mRNA and protein (Fig. 1E and Supplementary Fig. 1A and B).

Multiple factors may contribute to the observed lack of glucagon secretion phenotype, including the possibility that suppression of RFX6 expression in multiple islet cell types could mask α-cell phenotypes through nonautonomous mechanisms. To test this possibility, we developed a method for α-cell–selective gene targeting. An enriched primary human α-cell population was obtained by labeling dispersed islet cells with an antibody recognizing CD26, a known α-cell–specific surface protein (39,40). PE-conjugated α-CD26⁺ cells were isolated from CD26^neg non–α-islet cells using magnetic beads (MagSep), then infected with lentivirus expressing shRNA targeting RFX6 and a GFP transgene or with control lentivirus (Research Design and Methods) (Fig. 1C). We cultured infected CD26⁺ cells separately from CD26^neg cells for 4 days to allow for shRNA suppression of RFX6 mRNA, then reaggregated CD26⁺ and CD26^neg cells for functional studies (hereafter, MagSep RFX6^kd pseudoislets; Research Design and Methods) (Fig. 1F).

Five days after lentiviral infection, we used qRT-PCR and immunostaining to verify that the CD26⁺ fraction was enriched for α-cells (expressing glucagon [GCG]) and depleted of β-cells (expressing insulin [INS]) or δ-cells (expressing somatostatin [SST]) (Fig. 1G and Supplementary Fig. 2). In CD26⁺ cell fractions assessed by qRT-PCR, there was a marked reduction in RFX6 mRNA; by contrast, the CD26^neg fraction had no change in RFX6 mRNA (Fig. 1H). Thus, we suppressed RFX6 expression selectively in CD26⁺ cells. Next, we assessed imputed targets of RFX6 in these cells by qRT-PCR. Compared with control CD26⁺ cells, we observed a marked reduction in mRNA levels of CACNA1A, SLC5A1, RFX3, and CACNB2, but not of GCG, CACNA1C, ABCC8, or GCK (Fig. 1I). Static batch assays with MagSep RFX6^kd pseudoislets revealed a modest but significant reduction of glucagon secretion at a low glucose concentration (1 mmol/L glucose) compared with MagSep pseudoislets infected with control lentivirus (Fig. 2A). By contrast, batch assays did not reveal differences in insulin secretion by MagSep RFX6^kd pseudoislets (Fig. 2B). Thus, development of selective gene targeting in human islet α-cells revealed a requirement for RFX6 in maintaining α-cell gene regulation and function.

To assess α-cell functional defects after RFX6 suppression, we performed patch clamp electrophysiology on GFP⁺ cells isolated from control and MagSep RFX6^kd pseudoislets (Fig. 2C). GFP⁺ cells were confirmed to be GCG⁺ by immunostaining (27). In line with our GSGS data, MagSep RFX6^kd α-cells had significantly reduced exocytosis to a series of membrane depolarizations compared with control α-cells (Fig. 2D and E). Average exocytotic differences between RFX6^kd and control cells were larger after the first depolarization event, compared with subsequent depolarization events, suggesting an effect on docked or “readily releasable” glucagon granules (Fig. 2F and G). In prior studies in β-cells, loss of RFX6 led to perturbations of calcium channel activity (3). However, we found no significant differences in calcium channel activity between RFX6^kd and control α-cells (Fig. 2H). These findings complement prior studies of RFX6 in β-cells and suggest that RFX6-dependent regulation of exocytosis may differ in α- and β-cells (3,5).

Glucagon secretion was modestly reduced after RFX6^kd in α-cells for 5 days in vitro, and we next investigated the durability of this glucagon secretion defect in vivo using a transplantation model (Figs. 2A and 3,A). Because mature glucagon is identical in mice and humans, we previously developed immunocompromised NOD scid IL2Rγ^null (NSG) mice harboring an in-frame deletion of Glucagon1-29 (GKO) to permit ELISA-based measures of glucagon output from transplanted human islets (hereafter, GKO-NSG) (21). Prior studies have shown that in GKO-NSG mice transplanted with human islets or pseudoislets, hypoglycemia elicits glucagon secretion (21). We transplanted 1) whole RFX6^kd and 2) MagSep RFX6^kd pseudoislets and their respective controls in GKO-NSG mice and assessed glucagon secretion at least 1 month after transplantation in response to hypoglycemia through IP-ITT (Fig. 3). All transplanted GKO-NSG mice experienced acute hypoglycemia upon insulin injection (Fig. 3B and E and Supplementary Figs. 3A and B and 4A and B). Mice transplanted with whole RFX6^kd or control pseudoislets had comparable glucagon excursions after IP-ITT (Fig. 3C and D and Supplementary Fig. 3C and D). By contrast, average glucagon secretion was significantly blunted in mice transplanted with MagSep RFX6^kd pseudoislets (Fig. 3F and G and Supplementary Fig. 4C and D), consistent with our in vitro GSGS findings (Fig. 2A). In summary, these data demonstrate that targeted RFX6^kd in human adult α-cells durably impaired glucagon secretion after transplantation in vivo.

To identify genetic mechanisms underlying phenotypes after RFX6^kd, we performed transcriptome analysis on FACS-purified CD26⁺ GFP⁺ α-cells (hereafter, αRFX6^kd cells: n = 4 independent donors) (Fig. 4A). qRT-PCR analysis confirmed depletion of RFX6 mRNA and enrichment of GCG mRNA in this α-cell–enriched subset. As expected, INS mRNA levels were also depleted (Supplementary Fig. 5A). We generated and sequenced bulk RNA-seq libraries of RFX6^kd and control α-cells and confirmed significant RFX6 suppression in the αRFX6^kd-cell subset (Fig. 4B). Principal component analysis revealed interdonor variability and separation of αRFX6^kd and control cells, consistent with previous studies (Supplementary Fig. 5B) (2). We used the DESeq2 algorithm to identify 915 differentially expressed genes (DEGs) after αRFX6^kd, with 558 showing reduced expression, and 357 showing increased expression (Fig. 4C and D and Supplementary Tables 3 and 4).

Gene ontology pathway analysis of downregulated genes suggested that αRFX6^kd impacted known regulators of endocrine function and identity. Prior studies have identified a subset of genes with human α-cell–enriched or α-cell–specific expression, including GPR119, GPR148, LOXL4, KLHL41, PLCE1, LDB2, ALDH1A1, PEMT, PTPRT, SPOCK3, and ARRDC4 (15,27,41–43), and αRFX6^kd resulted in reduced expression of these genes in α-cells (Fig. 4D). We also observed reduced expression of genes known to regulate islet development and fate (ISL1, IRF6, MAFB, LDHA, and JAK2), peptide hormone secretion (RFX6, GPR119, UCN3, SLC30A8, ISL1, RFX3, and GLUL), cell-cell signaling (UCN3, GPR119, TMEM64, FGF18, KCNA5, CACNA1A, NCAM1, SSTR2, EPCAM, and RAPGEF4), and vesicle-mediated transport (LGI3, SYTL2, BAIAP3, and EPG5) (Fig. 4E). Conversely, we observed increased expression of genes governing stress responses (HIF1A, DDR1, IFIH1, and IFIT3) and proliferation (ABCG1 and TRIB3) in RFX6^kd α-cells (Fig. 4F). We conclude that RFX6 is required in mature human α-cells to maintain regulation of genes governing α-cell identity and function.

We next identified DEGs in FACS-purified β-cells after RFX6^kd (βRFX6^kd). To isolate β-cells, we stained cells from transduced whole pseudoislets with an antibody specific for the human β-cell surface marker NTPDase3, then FACS-separated NTPDase3⁺ GFP⁺ cells (n = 4 independent donors) (Fig. 5A) (44). Principal component analysis revealed appropriate separation of RFX6^kd and control NTPDase3⁺ cells and INS mRNA enrichment (Supplementary Fig. 5C and D). Compared with controls, the βRFX6^kd-cell fraction also had significant RFX6 suppression (Fig. 5B). Knockdown of RFX6 in β-cells (63%) was slightly reduced compared with knockdown in donor-matched α-cells (77%) (Fig. 5C). DESeq2 analysis revealed 155 DEGs in βRFX6^kd, with 96 downregulated and 59 upregulated genes (Fig. 5D and Supplementary Tables 5 and 6).

We found that 45% (71 of 155) of βRFX6^kd DEGs overlapped with αRFX6^kd DEGs (Fig. 5E). This overlap included genes consistent with previous studies, including those encoding factors with well-characterized roles in peptide hormone secretion and ion channel function (CACNA1A, CACNA1C, SLC30A8, SLC5A1, KCNA5, and BAIAP3) and regulation of cell communication (NCAM1, TMEM64, STAT1, and IFI6) (Fig. 5F) (3). In addition, we observed 55% of DEGs (84 of 155) were altered after βRFX6^kd but not αRFX6^kd (Fig. 5G). This included ABCC8 and GCK, crucial regulators of β-cell stimulus-secretion coupling (45–47). Our finding corroborates prior work in mice showing that Abcc8 and Gck are regulated by Rfx6 in β-cells (5). Other DEGs identified in βRFX6^kd cells but not αRFX6^kd cells include genes encoding β-cell–enriched factors (C1orf127, IGSF1, and CAPN13), TFs (SOX9, STAT6, and FOS), regulators of hormone secretion (LRP5, CASR, CFTR, SPP1, and ROBO1), and cell adhesion (CXCR4, NEO1, CDON, ROBO1, PCDHA4, SPON1, and VCAM1). Thus, RFX6 regulates distinct molecular functions and biological processes in α- and β-cells, while also coordinately regulating an overlapping subset of genes with orthologous roles.

To identify islet cell genomic DNA bound by RFX6, we used CUT&RUN, a high-resolution chromatin profiling strategy (20). Since available antibodies for endogenous islet RFX6 did not meet standards for chromatin assays, we misexpressed RFX6 in human islet cells with a lentiviral construct encoding human RFX6 tagged with a FLAG immunoepitope (n = 3 independent donors) (Fig. 6A and Supplementary Fig. 6A–C). DNA bound by RFX6-FLAG protein was enriched with an anti-FLAG antibody and sequenced (Research Design and Methods). HOMER analysis was used, as previously described (2,35), to identify motifs bound by RFX6-FLAG. Peak centers of RFX6-FLAG sample libraries had enriched read densities compared with IgG controls, where there was minimal enrichment at the peak centers (Fig. 6B). The most enriched RFX6-FLAG–bound DNA-binding motif identified by HOMER was RFX6, demonstrating high specificity of the CUT&RUN assay (Fig. 6C). Other enriched motifs include sequences binding other RFX family members, including RFX5, and other TFs critical for islet cell identity and function like ISL1, NKX6.1, and MAFB (Fig. 6C).

RFX6-FLAG–bound genomic regions were analyzed using GREAT, which identified 10,610 associated genes (36). To prioritize candidate genes likely directly regulated by RFX6, we intersected the gene list generated by GREAT with the DEGs identified after transcriptome analysis of αRFX6^kd and βRFX6^kd. This data integration identified 715 total “overlapping genes” (Fig. 6D and Supplementary Table 7). This concordance between RFX6-FLAG–bound regions and DEGs after RFX6^kd provides evidence that our approach identified direct RFX6 targets.

Direct targets of RFX6 imputed by our analysis included ion channel subunits (CACNA1A, CACNA1C, KCNA5, and KCTD12), transmembrane transporters and receptors (SLC30A8, SLC5A1, GPR119, and ABCC8), enzymes (PCSK1, PCSK2, and GCK), and TFs (RFX6 and PAX6) previously shown to govern islet function (Fig. 6D). A subset of genes expressed in human α-cells (but not expressed in islet α-cells from other species such as mice) were identified as putative direct targets of RFX6, including GPR148, LOXL4, PLCE1, SPOCK3, and ARRDC4 (Fig. 6D and Supplementary Table 7) (15,27,41–43). Likewise, our analysis revealed direct interaction of RFX6 with a subset of genes expressed in human β-cells (but not in β-cells from mice) (43) such as ROBO1 and CASR (Fig. 6D and Supplementary Table 7). Genomic regions of presumptive RFX6 target genes in α-cells, such as GPR148 and LOXL4, bound by RFX6-FLAG also corresponded with regions of transcriptionally active chromatin, as identified by previous assay for transposase-accessible chromatin-seq studies (Fig. 6E and F) (48–50). These findings support the view that RFX6 directly activates genes encoding factors essential for mature α-cell function in human islets.

Work here fills significant knowledge gaps regarding RFX6 roles in mature adult pancreatic islet α-cells. Using a novel cell-selective lentiviral shRNA targeting approach to investigate functions of RFX6 in primary human adult islet cells, we have revealed a requirement for RFX6 in maintaining adult human α-cell gene regulation and glucagon secretion. By combining transcriptome analysis with CUT&RUN studies, we identified direct targets of RFX6, including genes crucial in endocrine cell fate and hormone secretion in addition to previously unrecognized RFX6-dependent targets. These findings elucidate a novel requirement for RFX6 in regulating mature human α-cell function.

Dysregulated RFX6 expression has been noted in islet cells from subjects with diabetes, and RFX6 loss was found to impair β-cell function in animal models and immortalized human β-cell lines (3,5–7,14). Far less is known about RFX6 functions in human α-cells compared with β-cells (13,27). Our in vitro studies, in vivo transplantation analysis, and patch clamp studies using MagSep RFX6^kd pseudoislets revealed a marked reduction of glucose-dependent glucagon secretion and impaired α-cell exocytosis. Physiological studies of glucagon secretion are challenged by unavoidable variation between individual human donors (Supplementary Fig. 3B and D); thus, our conclusions are based on multiple samplings and the average of our studies (e.g., Fig. 3F). Transcriptomic and chromatin analyses identified downstream targets of RFX6 known to regulate exocytosis and glucagon output, including known ion channels, transport proteins, and vesicular fusion proteins. Reduced human α-cell glucagon output has not been previously linked to targeted RFX6 loss, although a recent study of type 1 diabetes islets correlated reduced α-cell expression of RFX6 (and multiple other factors) with impaired glucagon secretion (13). Unexpectedly, we did not detect impairment of α-cell glucagon secretion without selective RFX6 suppression, suggesting that compensatory effects from RFX6 loss in β-cells (Fig. 5E) or other non–α-islet cells may offset autonomous effects of RFX6 loss in α-cells, although further studies are needed to explore this possibility.

To identify direct genetic targets of RFX6 in human islet cells, we used CUT&RUN after misexpression of FLAG-tagged RFX6. While caution should be exercised in interpreting results with a tagged RFX6 due to possible excessive binding and false positive associations with target genes, we limited calling of RFX6 direct target genes by integrating the CUT&RUN analysis with α- and β-cell DEG analysis after RFX6 suppression. This approach enhanced the reliability of identifying direct genetic targets of RFX6 imputed by CUT&RUN. Among these targets were genes previously associated with RFX6-dependent regulation in β-cells, including ABCC8 and GCK (3,5). We also observed distinct α- and β-cell DEGs. For example, while expression of α-cell ABCC8 or GCK after RFX6^kd remained unchanged, expression of genes regulating vesicle-mediated transport, nutrient processing, exocytosis, and ion channel activity were changed in α-cells, but not in β-cells, after RFX6^kd. These data indicate differential regulation of α- and β-cells by RFX6. For example, our findings with patch clamp suggest that RFX6 loss impaired exocytotic machinery but not calcium channel activity (Fig. 2H) (3,5).

In summary, our study reveals that RFX6 is required for maintenance of genetic regulation and function in mature adult human α-cells, particularly in preserving genetic pathways responsible for regulated hormone secretion. Many islet transcriptional regulators, such as RFX6, PDX1, and MAFB, are expressed in multiple islet cell types, highlighting the potential of cell type-specific genetic targeting in deciphering cell autonomous and nonautonomous functions of these TFs. These findings advance our understanding of RFX6-dependent mechanisms that maintain adult human α- and β-cell functions. Moreover, they emphasize the importance of further exploring human α-cell biology to enhance our understanding of diabetes and pancreatic cancer. Our innovative approaches for targeted genetic manipulation in human primary islet cells can be extended to investigate cell-specific functions of other factors relevant to these diseases.

Acknowledgments. The authors thank Drs. S. Park, R.L. Whitener, M.A. Miranda, S.H. Kim, M. Neukam, H. Peiris, and Mr. J. Lam in the Kim group, as well as Professors A. Gloyn, R. Nusse, and W. Talbot (Stanford) for their technical and conceptual guidance, advice, and encouragement. The authors thank organ donors and their families, and the Alberta Diabetes Institute IsletCore, Integrated Islet Distribution Program, the National Disease Research Interchange, and the International Institute for the Advancement of Medicine, the University of San Francisco Islet Core, and the Stanford Diabetes Center Islet Core. In Alberta, we thank the Human Organ Procurement and Exchange (HOPE) program and Trillium Gift of Life Network (TGLN), and J Lyon for his efforts in human islet isolation.

Funding. The IsletCore was supported by the Alberta Diabetes Institute, and the Integrated Islet Distribution Program was supported by National Institutes of Health (NIH), National Institute of Diabetes and Digestive and Kidney Diseases (UC4 DK098085). V.M.N.C. was supported by a NIH National Institute of General Medical Sciences T32 training grant (5T32GM007790). M.F.Q. was previously supported by the Stanford Medical Scholars, Howard Hughes Medical Institute (HHMI) Medical Research Fellows Programs, Stanford Medical Scientist Training Program, and a Knight-Hennessy Scholars award. K.T. was supported by a graduate research fellowship award from the National Science Foundation (DGF-114747). R.J.B. was supported by a postdoctoral fellowship from JDRF (3-PDF-2018-584-A-N). RNA-seq data were generated using instrumentation purchased with NIH funds (S10OD025212 and 1S10OD021763). This manuscript also used data acquired from the Human Pancreas Analysis Program (HPAP-Research Resource Identifier [RRID]: SCR_016202) Database (https://hpap.pmacs.upenn.edu), a Human Islet Research Network (RRID: SCR_014393) consortium supported by National Institute of Diabetes and Digestive and Kidney Diseases grants (UC4 DK112217, U01 DK123594, UC4 DK112232, and U01 DK123716). Work here was supported by NIH National Institute of Diabetes and Digestive and Kidney Diseases awards (R01 DK107507, R01 DK108817, and U01 DK123743 to S.K.K., and R01DK126482 to S.K.K. and P.E.M.), a JDRF Center of Excellence Award (to S.K.K.), and NIH National Institute of Diabetes and Digestive and Kidney Diseases P30 DK116074 (S.K.K.), which supports the Stanford Islet Research Core and Diabetes Genomics and Analysis Core. P.E.M. holds the Canada Research Chair (Tier I) in Islet Biology. S.K.K. holds the KM Mulberry Chair at Stanford.

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

Author Contributions. V.M.N.C., M.F.Q., K.T., A.B., R.J.B., X.G., Y.H., M.N., W.Z., and C.C. performed experiments. V.M.N.C., M.F.Q., K.T., R.J.B., P.E.M., and S.K.K. acquired the funding. V.M.N.C., K.T., and S.K.K. conceptualized the study. V.M.N.C. and S.K.K. wrote the manuscript with input from all coauthors. S.K.K. supervised the study. V.M.N.C. and S.K.K. are the guarantors of this work and, as such, had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Prior Presentation. Parts of this study were presented in abstract form and as an oral presentation at the 2021 Western Region Islet Study Group, Stevenson, WA, 3–5 November 2021.