Single-cell RNA sequencing for engineering and studying human islets

Abstract

The islets of Langerhans are complex tissues composed of several cell types that secrete hormones. Loss or dysfunction of the insulin-producing β cells leads to dysregulation of blood glucose levels, resulting in diabetes. A major goal in cellular engineering has been to generate β cells from stem cells for use in cell-based therapies. However, the presence of other cell types within these islets can mask important details about β cells when using population-level assays. Single-cell RNA sequencing have enabled transcriptional assessment of individual cells within mixed populations. These technologies allow for accurate assessment of specific cell types and subtypes of β cells. Studies investigating different stages of β cell maturity have led to several insights into understanding islet development and diabetes pathology. Here, we highlight the key findings from the use of single-cell RNA sequencing on stem cell-derived and primary human islet cells found in different maturation and diabetic states.

Introduction

Diabetes results from the deregulation of blood glucose levels caused by either a lack of insulin production by the pancreas or a defect in the way insulin is processed and used by the body. The key tissues responsible for maintaining normoglycemia are the islets of Langerhans found within the pancreas, which consist of β, α, δ, ε, and pancreatic polypeptide (PP) endocrine cells. β cells compose the largest population of these endocrine cell types in islets and are responsible for secreting insulin in a glucose-responsive manner. The destruction of β cells (type 1 diabetes, T1D) or impairment of β cells and insulin function (type 2 diabetes, T2D) leads to the inability of the body to control blood glucose levels and properly use glucose as an energy source. The progression of both T1D and T2D is complex, often involving other non-endocrine cell types. In T1D, for example, β cells are destroyed by infiltrating immune cells [1]. Deregulation of other surrounding cell types can influence β cell behavior because of the systematic network that governs islet performance [2]. Furthermore, the physical islet microenvironment can also affect β cells, leading to defects in islet composition and function [3].

Due to the central role of β cell function in diabetes, significant effort has been dedicated to engineering human stem cell-derived islets (SC-islets) containing stem cell-derived β cells (SC-β cells). Chemical signals in the cell culture media targeting signal transduction networks have successfully specified fate selection of differentiating human pluripotent stem cells to SC-islets [4, 5]. These cells have been further matured to be able to mount a dynamic insulin secretion response when exposed to changes in extracellular glucose [6, 7]. Most recently, control of the microenvironment during differentiation has further enhanced differentiation efficacy and the resulting function of SC-islets from multiple cell lines [8]. These cells are being explored for their utility in diabetes cell replacement therapy [9–11] and understanding diabetes pathology, for example, by studying SC-islets generated from patients [12–15]. These SC-islets remain immature compared to human islets from the adult body, however, indicating the need for further improvements to these differentiation strategies.

Studying islets and β cells to unravel mechanisms associated with their development is challenged by the heterogeneity of cell populations in the islet. For example, other non-β cell populations, including the other endocrine cells, endothelial cells, and immune cells, can vary in proportion from donor-to-donor as well as individual islets [16]. These cells mask important details about β cell maturity and function with conventional bulk assays, such as gene expression quantified with RT-PCR. Advancements in single-cell technologies, particularly RNA sequencing, have enabled more robust study of islet heterogeneity and gained considerable attention in recent years [17]. In this review, we discuss single-cell RNA sequencing studies that highlight important discoveries and insights in the understanding of β cell development and diabetes onset.

Islet Heterogeneity

The islet contains a complex heterogenous network of several cell types that cooperatively work to maintain glucose homeostasis [18]. Population heterogeneity within the islet has been previously addressed with conventional single marker- and bulk population-based assays, including histology studies, flow cytometry, and mass cytometry [8, 19]. However, single-cell RNA sequencing has provided much better resolution of genetic information within each cell, allowing greater in-depth analysis of both currently identified and unidentified cell populations. While β α, δ, ε, and PP cells are the canonical endocrine cell types that make up pancreatic islets, other non-endocrine cells have been detected with this technology (Figure 1). For example, Maruno et al. used single-cell RNA sequencing to document various cell types found in the human islet [20]. In addition to the canonical endocrine cell types, endothelial cells, ductal cells, and mesenchymal cells were also detected. They also revealed additional marker genes for α cells (PLCE1, KLHL41 and FEV) and β cells (PFKFB2 and SIX2) that were not previously associated with these cell types. Segerstolpe et al. revealed 5 subtypes of β cells, 2 subtypes of α cells, and 2 subtypes of acinar cells in human islets that were unable to be resolved with previous techniques. The β cell subpopulations were defined by different combinatorial expressions of RBP4, FFAR4, ID1, ID2, and ID3 [21]. Furthermore, Baron et al. revealed two β cell populations differing by levels of several genes, including UCN3 and endoplasmic reticulum (ER) stress genes, such as HERPUD1, HSPA5, and DDIT3 [22]. The increased level in ER stress genes was associated with increased demand for insulin synthesis within β cells [23, 24]. The study also detected ε, exocrine, Schwan, immune, and ductal cell subpopulations in the islet [22]. In another study, Teo et al. used single-cell RNA sequencing to show that endocrine cells undergo de-differentiation into polyhormonal cells after isolation [25]. The robustness of this technology enabled polyhormonal cells, which express all hormone genes, to be distinguished from single hormonal endocrine cell populations.

Figure 1.

Summary of islet representative gene markers of cell types found in islets and SC-islets [20, 26].

SC-Islet Development and Maturation

Islet development has been studied using single-cell RNA sequencing tools to reveal new genes and pathways associated with β cell development. Because the pancreas undergoes large transcriptome changes during islet organogenesis [27], profiling these changes at the single-cell level can reveal important information for understanding endocrine cell fate, commitment, and maturation. SC-islets have been a convenient resource to study the development of pancreatic islets [5, 6, 8, 28, 29], and multiple reports have attempted to use single-cell RNA sequencing to explore heterogeneity and elucidate endocrine organogenesis [4, 5, 30]. In particular, Veres et al. sequenced cells that were undergoing in vitro SC-islet differentiation [26]. The study showed that the transcriptomic changes in stem cell differentiations recreated many aspects of pancreatic endocrine cell development in vivo. The study used single-cell RNA sequencing to resolve islet heterogeneity during development, sequencing cells from various stages of differentiation, including pancreatic progenitors, endocrine cell induction, and terminal SC-islet differentiation. SC-islets generated in this study produced key endocrine populations, including β, α, and δ cells along with both poly-hormonal and non-endocrine cells. The study also observed that about 20% of the cells resembled enterochromaffin cells (EC cells), which are not found in native human islets. While these cells express pancreatic markers, such as PDX1, CHGA, and NKX6–1, they also exhibit other EC cell markers, including LMX1A, ADREA2A, FEV, and TAC1. They also express genes associated with serotonin synthesis, a feature of native EC cells, including TPH1, DDC, and SLC19A1. Because of the overlap in genes expressed by endocrine and EC cells, this EC population was undetected until its transcriptome was revealed on a single cell level. The cause of EC cell generation in these in vitro differentiations remains unknown. However, because these cells carry pancreatic features, the fate of native EC cells may be closely related to pancreatic endocrine cells during development. In another study, Hogrebe et al. used single-cell RNA sequencing to reveal that the state of the actin cytoskeleton controlled pancreatic endocrine and exocrine fate during differentiation [8]. In particular, a depolymerized cytoskeleton of pancreatic progenitors favored endocrine induction while hyper-polymerization favored exocrine induction. In contrast, an intermediate level of cytoskeletal signaling facilitated development of NKX6–1+ pancreatic progenitors, which are necessary to generate functional SC-β cells. These cytoskeletal modulations could also be applied more broadly to influence differentiation to other endodermal cell fates, such as liver and intestine. These insights allowed for the development of an improved SC-islet differentiation protocol with enhanced function that worked more reproducibly across cell lines, highlighting the utility of single-cell RNA sequencing technologies for elucidating pathways involved in pancreatic development.

Single cell RNA sequencing has also been used to demonstrate that β cells found in adults have transcriptome profiles that are different from β cells of newborns and SC-islets [31]. For example, both β cells from newborns and stem cell-derived sources have been shown to lack or have low expression of mature β cell genes such as MAFA and UCN3 [32, 33]. The maturation marker MAFA, in particular, has been shown to be positively associated with increased INS transcripts in β cells [34]. Since in vitro-generated β cells have yet to recapitulate these important aspects of mature adult β cells, however, the study of β cell maturation has been limited to donor islets [6, 28]. The mechanism of juvenile islet development into mature adult islets has not yet been elucidated, and it remains unclear why these functional characteristics develop over time. Early single-cell RNA sequencing studies demonstrated that β cells in juvenile islets exhibit less expression of β cell gene signatures [35]. Interestingly, many α cell gene signatures are expressed in β cells from juvenile islets. These data suggest that endocrine populations are collectively more similar during early development and are committed to their fates upon maturation. In another study, β cells from islets of older donors had molecular signatures and genetic profiles that were more similar to juvenile islets rather than from middle aged adult islets [36]. How this reversion to the juvenile β cell state influences function and if key maturation marker expression remains is unclear. Furthermore, old endocrine cells were found to have increased transcriptional noise, demonstrating endocrine cell fate drift from its pre-committed state [37]. For example, the β cell population demonstrated increased expression of GCG transcript, while α cells had increased INS transcript. This transformation was also associated with increased oxidative, cellular, and metabolic stress genes as well as a decrease in proliferation rate [38].

The Islet Transcriptome in Diabetes

T2D results from the deregulation of blood glucose levels that is not associated with autoimmune β cell death. The cause of T2D varies, often involving either insulin resistance or dysfunction of the β cells themselves [39]. As a result, the cellular cause of T2D is heterogenous, influencing the severity of diabetes. Nevertheless, transcriptome studies across patients investigating T2D associated with β cell dysfunction can provide important insights into the pathogenesis of this disease. Islets from T2D patients have been sequenced for comparison with healthy islet controls in several studies. In particular, a study by Xin et al. highlighted key differentially expressed genes between endocrine cell types found in T2D patients [40]. T2D did not affect islet composition, as the proportion of endocrine cell types were similar to healthy islet samples. Differential gene expression analysis, however, revealed over 200 genes detected in T2D. Interestingly, most of these differentially expressed genes in individual endocrine cell types were not previously associated with β cell identity or β cell function, with many being associated with non-islet cell growth. Genes that were upregulated in T2D within β cells included PAXBP1-AS1, APOL4, RPS3AP18, FXYD3, and GLS2, while downregulated genes included LAMB1, ZNF397, IGFBPL1, and COTL1. Notably, diabetes or β cell associated genes that were upregulated in T2D included only SYT13 while G6PC2, FFAR4, GLRA1 and SLC2A2 were downregulated. In the α cell population, G6PC2 was the only gene previously associated with diabetes that was significantly downregulated. Wang et al. also sequenced islet cells from T2D patients and found that the β cells had profiles more similar to juvenile β cells than adult β cells [35]. This pattern was similarly observed with α cells. The β cells in T2D patients also showed significant reductions in INS gene expression [21]. FXYD2, which encodes a subunit of the Na⁺/K⁺-ATPase, was downregulated in T2D. GPD2, which encodes for NADH associated with mitochondrial metabolism, and LEPROTL1, which encodes for the leptin receptor associated endospanin-2, were upregulated in T2D β cells. In α cells, RGS4, which inhibits insulin release in mice, was upregulated. Gene set enrichment analysis (GSEA) of T2D islets demonstrated significant downregulation of gene signatures associated with mitochondrial energy metabolism and protein synthesis. Meanwhile, gene signatures, including those for apoptosis, diabetic nephropathy and cytokine signaling, were upregulated [21]. These changes in gene expression and gene signatures detected with single cell RNA sequencing strategies across diabetic phenotypes are summarized in Table 1.

Table 1.

Summary of transcript and signature changes in β cells of various conditions when compared to healthy controls.

β Cell State	Gene Identity	Signatures	References
Juvenile	↑ GCG, SST ↓ INS, MAFA, MAFB, UCN3	↑ Proliferation, α cell genes ↓ β cell genes	[34]
Aged	↑ GCG, SST ↓ INS	↑ Immature β cells, Oxidative stress, Cellular stress, Metabolic stress genes ↓ Proliferation	[36–38]
T2D	↑ PAXBP1-AS1, APOL4, RPS3AP18, FXYD3, GLS2, SYT13 ↓ INS, LAMB1, ZNF397, IGFBPL1, COTL1, G6PC2, FFAR4, GLRA1, SLC2A2	↑ Immature β cells, apoptosis, diabetic nephropathy, cytokine signaling ↓ Non-islet cell growth, mitochondrial energy metabolism, protein synthesis, mTORC1	[13, 21, 40]
T1D	N/A	N/A	N/A

T1D results from the destruction of pancreatic β cells caused by autoimmune attack, resulting in the loss of insulin production. T cells become autoreactive by recognizing β cell antigens as a threat, thus initiating an immune response against the host’s own tissue, though the complete mechanism leading to T1D is not yet fully elucidated [41]. However, studies have shown that deregulation of antigen presenting cells, particularly resident macrophages, contribute to the development of T1D [42]. Deregulation of antigen presenting cells can trigger this autoimmunity, causing the acquisition of antigen specific T cells that target pancreatic β cells. Because T1D is caused from a heterogenous network of many cell types that are not limited within the pancreas, single-cell RNA sequencing can provide a useful resource for resolving the mechanisms by which this autoimmunity progresses.

Utilizing single-cell RNA sequencing to detect β cell populations in T1D can be difficult, however, as β cells are mostly destroyed by infiltrating T cells and isolation of these islets is difficult [43]. Therefore, single-cell RNA sequencing studies of islet cells in T1D is extremely limited. Instead, studies are often focused on peripheral blood mononuclear cells (PBMCs) since they contain circulating immune cells responsible for T1D. Wang et al. performed a comprehensive single-cell RNA sequencing study of various patients, including those with T1D [35]. Islet purity in T1D donors was very low when compared with healthy controls. Less than 10% of the extracted islet population were β cells, while the remaining majority were ductal cells. With limited endocrine cell detection, comparative analysis of these populations is challenging, though several studies have succeeded in gaining valuable information from these datasets. For example, Cerosaletti et al. used single-cell RNA sequencing to explore minority immune cell populations in PBMCs of T1D patients [44]. Deep sequencing of all collected cells enabled detection of antigen specific CD4⁺ T cells. These clonotypes expanded during the progression of T1D. Th2 associated genes including GATA3, CCR4, and IRF4 were upregulated in expanded CD4⁺ T cells of T1D subjects. Meanwhile, IFN response genes, such as IFNG, CD69, and GBP5, were downregulated when compared to healthy control cells.

Some genetic disorders are associated with the onset of diabetes. Single-cell RNA have been used to study the effects of these variants, interrogating differences in genes and pathways that causes β cell dysfunction. Balboa et al. recently used SC-islets differentiated from a patient with an INS gene mutation to model neonatal diabetes [13]. Using CRISPR to correct the mutation, they compared SC-β cells generated with and without the mutation and found with single-cell RNA sequencing that the mutation elevated ER stress and activated the unfolded protein response (UPR) pathway, decreasing genes associated with function. Much of these defects were attributed to decreased mTORC1 signaling. Maxwell et al. investigated SC-islets derived from patients with Wolfram syndrome, a monogenic form of diabetes caused by a mutation in the WFS1 gene [45]. The study compared defected SC-β cells with CRISPR-corrected SC-β cells from a Wolfram Syndrome patient. Single-cell RNA sequencing analysis show that the mutation corrected cells have increased INS transcript and decreased ER stress genes including CHOP, MANF, and EIF2A and the mitochondrial stress gene TXNIP. These studies show the promise of utilizing single-cell technologies to evaluate β cells for autologous cell therapy.

Perspectives and Future Direction

New single-cell RNA sequencing methodologies, both in sequencing and computational analysis, are actively being developed and released in recent years. With increasing demands for single-cell approaches, this technology is expected to become even more widely available. Importantly, cost reduction strategies can benefit the field, allowing multiplexing of samples from separate conditions for a single run of sequencing [46, 47]. The new development of single-cell ATAC sequencing can also be useful to address epigenetic questions and verify findings not answered with single-cell RNA sequencing [48]. However, current commercially available technologies are limited by the ability to sequence both information from the same cell or sample. Chen et al. recently developed a novel sequencing platform, single-nucleus chromatin accessibility and mRNA expression sequencing (SNARE-Seq), which enables for both chromatin and mRNA information to be sequenced from the same cell [49]. Observing the downstream targets of identified transcription factors within β cells will be of interest with this technology, particularly in development and maturation.

While there are an increasing number of single-cell RNA sequencing reports focused on reporting novel genes of various cell types and conditions, follow-up investigations to address key biological questions in β cells remain lacking. Integration and analysis of multiple data sets from different studies will be key to generating new insights into β cell development and maturation. Specifically, identifying targets for improving SC-islet differentiations and restoring islet health in diabetes will be of particular interest. Recent computational tools are becoming available to help support these metadata-based studies, allowing integration of multiple single-cell data sets [46, 50].

While single-cell RNA sequencing tools, such as library preparation platforms and computational packages, have made many advancements, important limitations of note persist. Cross comparative studies involving multiple datasets from different sources can be challenging across different apparatuses and platforms, such as 10X genomics and Drop-Seq [51]. The use of different sample preparation techniques and reagents to generate barcoded cDNA libraries may impose technical variations across different datasets. Moreover, lowly expressed genes may be misinterpreted when comparing datasets of different sequencing depth. Normal biological occurrences, including transcriptional bursting and cellular stress signals, may also introduce noise variations in datasets of low cell number. Inconsistent interpretations or conclusions can also be drawn depending on the type of computational tools or packages used for the analysis [52, 53]. Due to these limitations, major findings from such data sets should always be confirmed with follow-up validation studies, such as PCR, protein expression, or phenotypic studies. In addition, including biological replicates, such as multiple islet donors, batches of differentiation, or stem cell lines, would improve the robustness and validity of single-cell RNA sequencing results.

Single-cell technologies can be utilized in other forms of investigations, including disease modeling and therapeutic studies for diabetes. Because single-cell studies on T1D are limited due to the lack of endocrine cell number in T1D patients, for example, future studies can focus on developing disease models to study immune cell interactions with islet cells that mimic T1D pathogenesis. Single-cell technologies will accelerate this research because they can detect multiple cell populations in cultures, allowing researchers to more easily study the effects of each cell population on each other. Observing these early events prior to islet autoimmunity may provide insights to future immunotherapeutic targets and strategies. Furthermore, while islet interaction and co-culture studies have been done with endothelial cells, they would further benefit from these single-cell sequencing strategies by demonstrating how these interactions with endothelial cells influence the β cell transcriptome [28]. Investigating therapeutic responses may be another application of single-cell RNA sequencing to study islet health. With multiplexing technologies, multiple drugs or therapeutic targets may be screened and tested in parallel for sequencing. For example, modern immunotherapeutic strategies for T1D and β cell modulating drugs for T2D can be explored with single-cell RNA sequencing to comprehend cellular responses and outcomes [54, 55]. As this technology becomes more economical and widely available, single-cell RNA sequencing will not only be restricted to cataloguing cell types but will be utilized to generate novel insights into the pathogenesis of diabetes and mechanisms of islet health.