Applying CRISPR Screen in Diabetes Research

The goal of studying genetics is to elucidate the function of genes in certain biological processes and diseases. The reverse genetic approach is hypothesis driven and knowledge based: perturbing genes of interest and characterizing the consequent biological phenotype, hence, a “genotype-to-phenotype” approach. By contrast, the forward genetic approach is a “phenotype-to-genotype” approach, which starts with a phenotype of interest and, through high-throughput genetic screening, identifies certain gene perturbations. The forward genetic screen is commonly carried out in model systems, including yeast, nematodes, flies, and zebrafish, and has uncovered many mechanisms and pathways in signaling and organism development (1–4). Until recently, it was technically difficult to perform forward genetic screens in mammalian systems because altering the phenotype often required full loss of function, which was time-consuming and costly to do on a large scale. However, in the last decade, forward genetic screens have been revolutionized by the use of RNA interference (RNAi) and CRISPR (Clustered Regularly Interspaced Short Palindromic Repeats) technology. The CRISPR system in particular, in combination with next-generation sequencing (NGS), has made forward genetic screening feasible, efficient, and prevalent in the study of mammalian systems. Since the introduction of the technology, numerous studies have been carried out, leading to discoveries in a variety of research fields, particularly cancer biology, virology, immunology, and basic cell biology (5–8). However, it has not been widely adopted in the area of diabetes research, partially because diabetes research often involves cross talk between multiple organs, tissues, or cell types, which makes it challenging to use a CRISPR screen. Nevertheless, one can begin to answer many questions in diabetes research by reducing the investigation to the study of a single cell type, and using CRISPR screens can provide initial clues toward mechanisms within the disease, which can be further studied on a systematic level. Here we provide a methodology review on the CRISPR screen, summarize how it revolutionized research in various fields, and give perspective on how it can facilitate diabetes research.

The CRISPR/Cas9 system has rapidly become a popular and effective tool in many areas of biological research. The technology was recently recognized by the Nobel Prize in Chemistry, awarded in 2020 to Emmanuelle Charpentier and Jennifer A. Doudna (9). CRISPR technology has been widely used in gene editing by inducing mutations and activating or suppressing expression of genes (10). The action of the CRISPR system is mediated by a Cas9/guide RNA (gRNA) complex. The SpCas9 protein (originating from Streptococcus pyogenes and the most widely used and well-characterized Cas9 homolog) is guided to a target genome locus by a single guide RNA (sgRNA), a sequence of ∼20 base pairs, directly upstream of a PAM (protospacer adjacent motif) sequence (5′-NGG-3′) (11). There, the protein induces a double-stranded DNA break (12–14). One of the cell’s responses to this damage is nonhomologous end joining, an error-prone repair mechanism that can cause an indel (insertion or deletion) mutation, resulting in a frameshift or premature stop codon and rendering the corresponding protein inactive (15) (Fig. 1A).

The generation of a deactivated Cas9 protein (dCas9) has further broadened the possibilities of CRISPR application beyond gene knockout. Engineered dCas9 lacks its endonuclease activity but still retains its binding capability to the gRNA. The gRNA is designed to target the promoter region of a gene, and when the dCas9 is fused with a transcriptional activator or suppressor, the complex will activate or interfere with the expression of the target gene (16,17). CRISPR activation (CRISPRa) is achieved by fusing the dCas9 to an activator domain, most commonly VP64 (a tetramer of VP16), which recruits transcription machinery to the genome locus that dCas9/gRNA binds (17,18). The level of gene activation using just dCas9-VP64 is relatively low, so researchers have developed additional activators (e.g., SAM, VPR, and SunTag) that enhance dCas9-VP64 to further increase gene expression level (19–22) (Fig. 1B). On the other end, CRISPR interference (CRISPRi) uses a dCas9 fused with the transcriptional repressor domain Krüppel-associated box (KRAB), which recruits corepressor proteins, induces hetero-chromatin formation, and achieves gene-specific expression silencing (Fig. 1C) (23). It should be noted that the phenotypic effect of using CRISPRi is often not as robust as that of a CRISPR knockout, which can pose an additional challenge in CRISPR screens that require high sensitivity. However, in certain instances CRISPRi can be beneficial, such as the study of essential genes where full gene knockout would kill the cells. It can also be a useful tool in pharmacological discovery, as the CRISPRi system would more closely mimic the effect of a drug than a complete knockout (24).

The high efficiency of generating a gene mutation via CRISPR and the flexibility of activating or suppressing gene expression with CRISPRa/i are the foundation of the high-throughput CRISPR forward genetic screen. CRISPR screens are primarily performed in a pooled fashion, due to the ease with which a single mutation can be identified using NGS of the corresponding gRNA region. Arrayed CRISPR screens have been done on occasion but rarely on a genome-wide scale, as it is time-consuming and costly. Thus, the focus of this discussion will be on pooled CRISPR screening.

Pooled CRISPR screens can be classified based on the type and size of the library chosen. A large collection of knockout, activation, or inhibition CRISPR libraries, either genome-wide libraries or smaller, more focused, subpool libraries, for both the human and mouse genome, is readily available on Addgene, a nonprofit global repository of plasmids dedicated to academic research. Coverage of these libraries varies between a few hundred to >200,000 total gRNAs and 3–20 gRNAs per gene to maximize the chance of identifying screening hits. Libraries also include nontargeting guides, which are used as internal negative controls to evaluate noise and normalize read counts in a screen (25).

The overall workflow of a standard pooled CRISPR screen is illustrated in Fig. 2. The majority of the CRISPR libraries use lentivirus for cell transduction. There are a few mammalian CRISPR subpool libraries that use adeno-associated virus or retrovirus for delivery (26,27). These delivery methods enable direct in vivo CRISPR screens (adeno-associated virus) or screening of cells that cannot be transduced with lentivirus (retrovirus). For the lentivirus-based CRISPR libraries, the plasmid libraries first need to be amplified (Fig. 2A) and then are used to produce a large-scale lentivirus library (25,28). Many CRISPR knockout libraries are one-vector systems where each plasmid encodes both SpCas9 and an sgRNA. This type of library can be used to transduce any cell of interest without additional gene engineering. In some cases, a CRISPR library contains only gRNAs without SpCas9 or dCas9 fusion protein. In this case, researchers first need to generate a cell line that carries an SpCas9 or a dCas9 fusion protein transgene or use primary cells isolated from SpCas9 transgenic animals. Next, cells are transduced with the lentiviral library at a low multiplicity of infection (0.3–0.6), the ratio of viral agents to cell number, to minimize the likelihood of infecting a single cell with multiple gRNAs, confounding the results of the screen (Fig. 2B). The transduced cells are then selected and noninfected cells are removed from the culture (Fig. 2C), e.g., with antibiotic selection.

The next and most critical step for the CRISPR screen is to select or separate the mutated cells based on the phenotype of interest (Fig. 2D). This phenotype may be resistance to a certain drug treatment or condition, accelerated or arrested cell growth, increased or decreased gene expression or reporter activity, etc. Once cells are selected or separated, their genomic DNA (gDNA) is purified and a sequencing library is generated by amplifying the gRNA region using PCR (Fig. 2E and F). PCR primers typically contain the adapter sequence for NGS and barcodes that can be assigned to each gDNA sample, so amplified samples can be pooled and sequenced together. The cells before selection should be included as a baseline control, and the initial postamplification CRISPR library may also be included to account for bias during library amplification.

The output of NGS is a raw count of every gRNA in each sample for the researcher to determine which gRNAs are enriched and/or depleted (Fig. 2G). The strategy to identify significant hits will vary by screen based on the conditions, such as library size, selection protocol, and sensitivity (29,30). In addition, since all libraries contain multiple gRNAs per gene, analyzing the gRNA distribution on a gene level provides more confidence for the contribution of a particular gene toward a phenotype and can help account for potential off-target activity of a gRNA (11).

Bioinformatics tools such as RNAi Gene Enrichment Ranking (RIGER) (31) or Model-based Analysis of Genome-wide CRISPR/Cas9 Knockout (MAGeCK) (32–34) can be used to identify significantly enriched or depleted hits in these large data sets, and gene set enrichment analysis of significant hits may provide insight into overall pathways affecting the phenotype (35).

Before designing and executing a pooled CRISPR screen, there are a few important things that need to be considered, including the type of cells, the size and type of the CRISPR library, and the assays and strategies of separating cells (summarized in Fig. 3).

Ideally, a single cell type should be used in a pooled CRISPR screen, as using a mixture containing multiple cell types would introduce confounding into the separation or selection step. Pooled CRISPR screens require a large number of cells, as it is essential that sufficient coverage of the gRNA in the library is achieved to ensure saturation of the screen, prevent bias, and account for stochastic noise. The genome-wide CRISPR libraries contain a range of ∼60,000 to ∼200,000 gRNAs, and a minimum of 500 cells per gRNA is recommended throughout, as well as two to four replicates per sample (25). Hence, hundreds of millions of cells will typically be needed to initiate the CRISPR screen, so the cells of choice should be either a cell line that can be easily expanded or a type of primary cell that can be obtained and purified in a large quantity.

A researcher designing a CRISPR screen also must decide what kind of CRISPR gRNA library is most appropriate for their study. The most popular genome-wide CRISPR knockout libraries include genome-scale CRISPR-Cas9 knockout (GeCKO) (human and mouse) (36) and Brie (mouse) and Brunello (human) (12), and the most popular activation CRISPR libraries are the synergistic activation mediator (SAM) libraries (human and mouse) (21,25). At times, smaller-scale libraries that contain a specific category of genes can be beneficial for a more focused screen or screens with a relatively low cell number. We summarize all currently available CRISPR libraries from Addgene in Supplementary Table 1. In addition, a few premade CRISPR libraries and customized library synthesis services are available via commercial sources. The majority of the CRISPR libraries target human or mouse cells, so if the researcher wishes to analyze a different animal model, a custom CRISPR library needs to be synthesized or generated from mRNA using cells from the chosen species (37).

Any CRISPR screen requires a carefully designed assay and method of cell separation, which can vary widely in terms of phenotype, selection method, and complexity. The most common screens analyze cell viability and fitness by detecting gRNAs that are either enriched or depleted within a population after treatment. Enriched gRNAs induce a mutation that is beneficial for survival or proliferation in the given condition, while depleted gRNAs indicate mutations negative to cell health or growth (38). Detecting depleted gRNAs, also referred to as negative selection or dropout screening, requires greater coverage and may be more challenging. In addition to viability and fitness, any quantifiable trait can also be used as selection for a CRISPR screen. This is often done via FACS, where a fluorescent biomarker is used to separate cells and gRNA abundance is compared between the sorted groups (39,40). The screening options using FACS are limited to the available markers, such as antibodies or reporter cell lines, that are available. However, new and more complicated selection strategies using FACS are constantly being generated (41,42).

CRISPR screens can be used to search for essential genes in different cell types. By definition, the mutation of an essential gene leads to cell growth arrest or cell death. Finding essential genes can help with understanding of the basic machinery that is required for cell survival and proliferation, providing potential targets for anticancer therapy. If a gRNA targets an essential gene, it will slow down cell proliferation or kill the cell, so the abundance of this gRNA in the cell population will be depleted. For example, in the very first genome-wide CRISPR screen, Shalem et al. (43) screened the human GeCKO library on melanoma cell line A375 and embryonic stem cells HUES8, searching for essential genes, and found that gRNAs that target ribosome structural constituents were largely depleted. In another negative selection screen, Wang et al. (44) discovered a large group of uncharacterized genes involved in RNA processing to be essential for various human cancer cell lines. The same strategy can be applied to various cancer cell types, from B-cell lymphoma to pancreatic cancer, and the assay can be performed both in vitro or in vivo (45,46). In certain instances, this type of negative selection CRISPR screen was carried out with anticancer drug treatment, with the aim of looking for novel genes that could be targeted in synergy with known drugs. One of the examples is the identification of PRMT5 as synthetic lethality combinatorial target with gemcitabine in pancreatic cancer cells (47).

On the opposite end, positive selection screening is more common and has been widely used to study the mechanism of drug resistance and host factors for various viruses and identify tumor suppressors. The rationale of drug resistance screens is that, in the presence of a drug that would normally kill a cell, a gRNA-induced gene mutation that protects the cell from drug treatment would be enriched in the surviving cells after treatment. The target genes of the enriched gRNAs are then considered positive regulators for the drug-induced cell death or growth arrest. For example, in a study of acute lymphoblastic leukemia, Autry et al. (48) discovered that mutation of a previously unidentified gene, CELSR2, confers glucocorticoid resistance. Similarly, another CRISPR drug resistance screen found that BEND3 knockout confers resistance to antileukemia drug TAK-243 (49). Using this same strategy, CRISPR screening can also help explore the mechanism of resistance to immune cell killing. In a genome-wide CRISPR screen, Pan et al. (50) used T cellmediated killing of the B16F10 melanoma cell as a screening system and discovered that the Jak/Stat pathway, antigen-presentation pathway, and MAPK pathway are essential for T cell–mediated killing. Furthermore, they analyzed gRNAs that were depleted in the screen and showed that components of the chromatin remodeling complex sensitized the melanoma cells to cytotoxic T cell–mediated killing. In a similar study, Zhuang et al. (51) used CRISPR screening to find gene mutations that either protected or sensitized natural killer cell–mediated killing of leukemia cells. These are two examples of a positive selection and negative selection screen being achieved in the same CRISPR screen experiment.

The positive selection CRISPR screen is also popular in the study of viral host factors. Viruses enter cells by binding to a receptor and propagate themselves using host factors, eventually killing the host cells. Therefore, in a CRISPR screen, any gRNAs that protect the host cells’ virus-induced cell death are those targeting the genes that are essential for the life cycle of the virus. These genes can potentially become important drug targets to fight against viral infection. This type of CRISPR screen has been applied to the studies of various viruses, such as Ebola (52), HIV (53), severe acute respiratory syndrome coronavirus 2 (54,55), and Zika (56), and both viral entry receptors and host factors that facilitate viral propagation were discovered.

Similar to use of cell death as a cell selection strategy, cell proliferation can also be used in a CRISPR screen. If a gRNA-induced gene mutation can either accelerate or arrest cell proliferation, the abundance of the gRNA will increase or decrease in the cell population, respectively. This strategy has been applied to the study of tumor suppression and cell senescence. For example, with a CRISPR screen Katigbak et al. (57) successfully discovered Sp3 and Phip as two rare tumor suppressors in Buikitt’s lymphoma, and CRISPR-induced mutations of these two new tumor suppressors promote lymphomagenesis in vivo. In another study, Wang et al. (58) used a genome-wide CRISPR screen to discover KAT7 as a driver of cellular senescence and that the mutation of KAT7 rejuvenates and promotes the proliferation of human mesenchymal precursor cells.

Selecting and separating cells based on a quantitative trait, most commonly through FACS sorting, is also a popular method for pooled CRISPR screens. To study an on-and-off quantitative trait on the transcription level, one can build a reporter cell line using a promoter of interest to drive the expression of either a fluorescent transgene or a membrane protein that can be labeled by fluorescence. For example, in a genome-wide CRISPRa screen for factors that drive neuronal fate and reprogramming, Liu et al. (59) created a reporter mouse embryonic stem cell line carrying a Tubb3 promoter–driven hCD8 transgene. After mouse embryonic stem cell differentiation, the Tubb3-positive neuronal cells were isolated by sorting hCD8⁺ cells using FACS where multiple genes were found to be able to induce neuronal differentiation. In another genome- wide CRISPR knockout screen, Panganiban et al. (40) used a reporter cell line that harbors a CHOP promoter–driven mCherry transgene and found that mutation of L3MBTL2 and MGA turned on mCherry expression under endoplasmic reticulum stress and promoted the cells toward apoptosis. It is also common to study a quantitative trait with an increased or decreased level by FACS sorting the top and bottom 5–10% of cells, comparing the distribution of gRNAs in the two groups, and finding enriched positive or negative regulators. To list a few examples, Parnas et al. (42) used genome-wide CRISPR screen to dissect the regulatory network in primary immune cells by comparing gRNA profiles in TNF^high versus TNF^low cells after LPS treatment. Loo et al. (60) conducted a CRISPR screen comparing Foxp3^high versus Foxp3^low regulatory T cells and found many key factors that control the Treg cell function. In a CRISPR screen by Hoshino et al. (61), mitochondria-specific GFP reporters and MitoTracker were used to quantify cellular mitochondria and, by comparing the gRNA profile in reporter^high and reporter^low cells, ADP/ATP translocase was found to drive mitophagy.

It is also possible to combine pooled CRISPR screening with the single-cell RNA sequencing (scRNA-seq) technique, which allows researchers to investigate phenotypes at the transcriptome level, and study gene functions in large number of cells in parallel. There are several systems developed for this purpose, including Perturb-seq (62,63), CRISPR droplet sequencing (CROP-seq) (64), and CRISP-Seq (65). These systems may be particularly useful in a screen using a heterogenous mixture of cells that consists of multiple cell types.

CRISPR-Cas9 technology has been used in the field of diabetes research on various fronts, mostly for precise gene editing purposes. Such examples includes the perturbation and study of type 2 diabetes risk genes in vitro (66), the creation of engineered brown-like adipocytes to improve glucose tolerance and insulin sensitivity (67), and CRISPR editing of stem cells to correct disease-causing mutations and prevent immune rejection in β-cell transplantation (reviewed in 68).

Although CRISPR screens have been widely used in many research fields, there are only a few examples where researchers have applied CRISPR screens in diabetes-related research (overview in Supplementary Table 2). Diabetes involves multiple organs or cell types communicating with each other. For example, the heterogeneity within the pancreatic islets, the cross talk between various types of immune cells in type 1 diabetes, and the complicated regulation mechanism of blood glucose make it difficult to design and execute CRISPR screens. However, it is still possible to use a single cell type to study certain aspects of diabetes, which may shed light on the understanding of diabetes on a systematic level. Here we list a few such examples.

In the study of pancreatic β-cells, Wei et al. (69) derived β-like cells from iPS cells, with GFP reporter under an insulin promoter, and identified vitamin D receptor (VDR) as a modulator of inflammation and β-cell survival. Fang et al. (39) used intracellular insulin immunofluorescence as a quantitative readout and discovered that the cohesin loading complex and the NuA4/Tip60 histone acetyltransferase complex regulate insulin transcription and release. Recently in our laboratory we used a genome-wide CRISPR knockout screen to identify genes that regulate pancreatic β-cell protection against autoimmune killing in type 1 diabetes. In this study, the autoimmune destruction provided a strong live-death selection pressure on transplanted pancreatic β-cells, and we were able to discover several gene mutations, including RNLS, a human type 1 diabetes GWAS gene that reduces intrinsic stress in β-cells and shields the cells from the autoimmunity (70).

CRISPR screens can also facilitate the mechanism study of metabolism in diabetes research. For example, Gulbranson et al. (71) used HA-GLUT4-GFP as a reporter for the genome-wide CRISPR screen to investigate GLUT4 exocytosis, which led to the discovery of RABIF (Rab-interacting factor) as a positive regulator of insulin-induced GLUT4 exocytosis. A similar reporter and assay were used in a CRISPR screen by Wang et al. (72), who identified Exoc7/Exo70 as another critical regulator of exocytosis. A CRISPR screen was also applied in a study of cellular LDL uptake by Emmer et al. (73), in which a fluorescent LDL was used and LDL^high and LDL^low cells were isolated by FACS and analyzed, with several previously unrecognized genes found with putative roles in LDL uptake by functional interaction with LDLR.

We believe that there are many topics in diabetes-related research that can benefit from CRISPR screens and novel discoveries still yet to be made. For example, pancreatic β-cell death and proliferation have been important topics in diabetes research. Genome-wide screens can be applied to study the mechanism of high lipid–induced, high glucose–induced, or endoplasmic reticulum stress–induced β-cell death, which is believed to be essential for both type 1 and type 2 diabetes. One could use CRISPR screens to study human embryonic stem cells or human-induced pluripotent stem cell differentiation into pancreatic β-like cells, or transdifferentiation of other cell types into pancreatic β-cells, by using cell lineage reporters such as Insulin-GFP. The loss of Insulin-GFP could also be a readout for a potential CRISPR screen to study β-cell de-differentiation in type 2 diabetes. Studying pancreatic β-cell function with CRISPR screens is technically challenging because insulin secretion is not a cell-autonomous phenotype, but assays can be carefully designed to measure an intermediate step during insulin secretion. One possible strategy could be to quantify ATP-to-ADP ratio by fluorescent reporter (74), or pH-sensitive green fluorescent protein (pHluorin) fused to phogrin could be a surrogate readout for insulin secretion (75). CRISPR screens could be used to study key metabolite uptake by metabolic organs/tissues, for example, using a fluorescent glucose analog, such as 2-NBDG, to study glucose uptake in adipocytes, muscle, or liver cells. Similarly, fluorescent fatty acid analogs also exist for the study of fatty acid uptake. In the research of brown adipocyte differentiation and function, the expression level of UCP1 or a UCP1 fluorescent reporter could serve as readout for a CRISPR screen. In the area of diabetes complications study, chronic hyperglycemia–induced death or dysfunction in vascular endothelium, kidney podocytes, and retina epithelia cells can also be potentially modeled in vitro and be a great subject for which to use CRISPR screens.

In summary, there are still many diabetes-related research questions that can be explored with a CRISPR screen approach. The resources and tools are all readily available, but the researchers need to creatively design appropriate assays and cell selection strategies to make the screens successful. Investigators can learn a great deal from the CRISPR screen studies in other fields that we discussed in this article and apply similar ideas to their specific research. We are looking forward to seeing many more great discoveries in the diabetes research area using CRISPR genetic screens.

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

Author Contributions. P.Y. and N.M. reviewed the current literature and contributed to writing the manuscript. N.M. reviewed CRISPR technology and screen methodology and generated figures and tables. P.Y. reviewed assay strategies and provided future perspectives.