Mutations in the C1 element of the insulin promoter lead to diabetic phenotypes in homozygous mice

Genome editing technologies such as CRISPR–Cas9 are widely used to establish causal associations between mutations and phenotypes. However, CRISPR–Cas9 is rarely used to analyze promoter regions. The insulin promoter region (approximately 1,000 bp) directs β cell-specific expression of insulin, which in vitro studies show is regulated by ubiquitous, as well as pancreatic, β cell-specific transcription factors. However, we are unaware of any confirmatory in vivo studies. Here, we used CRISPR–Cas9 technology to generate mice with mutations in the promoter regions of the insulin I (Ins1) and II (Ins2) genes. We generated 4 homozygous diabetic mice with 2 distinct mutations in the highly conserved C1 elements in each of the Ins1 and Ins2 promoters (3 deletions and 1 replacement in total). Remarkably, all mice with homozygous or heterozygous mutations in other loci were not diabetic. Thus, the C1 element in mice is required for Ins transcription in vivo.

T he promoter of the gene encoding insulin comprises sequences that are immediately upstream of the transcription initiation site [1][2][3][4][5] . The insulin gene is specifically expressed in pancreatic β cells [4][5][6] , and its promoter regulates the rate of transcription in response to physiological regulators 7 . In the insulin promoter, a region of~350 bp can direct β cell-specific expression of insulin, which is regulated by ubiquitous as well as pancreatic β cell-specific transcription factors 8 . In β cells, insulin gene (Ins) expression is controlled by multiple regulatory elements in the basal insulin promoter, which include the A, C and E elements, and three highly conserved enhancer elements in the human insulin promoter: A3 (bases -225 to -220) 9-12 , A2-C1 (bases -149 to -116) 13 and E1 (bases -111 to-102) 14,15 . Certain transcription factors that govern β cell differentiation (pancreatic and duodenal homeobox1; Pdx1, neuronal differentiation; Neu-roD and v-maf musculoaponeurotic fibrosarcoma oncogene family, protein A; MafA) interact directly with insulin promoter regulatory elements [9][10][11][12][13][14][15] . Most of these proteins bind to specific sites and form multiprotein transcriptional complexes situated on the insulin promoter 16,17 . The MafA and A2-binding factors cooperatively activate insulin gene expression 18 . However, these events were identified through in vitro studies, and we are unaware of comparable in vivo studies.
The use of clustered regularly interspaced short palindromic repeat (CRISPR)-Cas9 systems to manipulate mammalian genomes provides enormous opportunities for curing human diseases [19][20][21][22] . The nonhomologous end joining (NHEJ) mechanism leads to site-specific repair of the DNA break site, which causes different unpredictable insertions and deletions of various sizes. We took advantage of these mechanisms to develop mice in which the promoter regions of Ins1 and Ins2 were partially deleted. Although these features of the CRISPR-Cas9 system are disadvantageous for the generation of knockout mice with deletion of a single protein-encoding gene, the system is advantageous for evaluating the functions of promoter regions in vivo because multiple mice with different alterations of promoter sequences can be generated concurrently.
Here, we used CRISPR-Cas9 technology to generate mice with partial deletions of the Ins1 and Ins2 promoters, which showed that only homozygous mice with mutations in the highly conserved C1 elements of the Ins1 and Ins2 promoters developed diabetes, indicating the value of genome editing in studying the regulation of insulin synthesis in vivo.

Results
Construction of CRISPR-Cas9 expression vectors. Despite the evolutionarily conserved function of insulin, the promoter and transcribed sequences of Ins are not well conserved between species 23,24 (Fig. 1a, Supplementary Fig. 1). However, our detailed analysis of published mammalian promoter sequences revealed that most of the critical promoter sequence elements are well conserved 25 , particularly bases -151 to -103 of the mouse Ins1 and Ins2 promoters and bases -149 to -102 of the human insulin promoter (Fig. 1a, Supplementary Fig. 1). These sequences comprise the GG2-A2, C1 and E1 elements (Fig. 1a, Supplementary  Fig. 1), which was previously known as the rat insulin promoter element 3 (RIPE3), to which cell-specific and ubiquitous factors bind 13 . For example, NeuroD binds the E1 element to form a heterodimer with the ubiquitously expressed bHLH factor E47. The C1 binding factor, which regulates pancreatic β cell-specific and glucose-regulated transcription of the insulin gene, is a member of the large family of Maf transcription factors [16][17][18][23][24][25][26][27][28][29] . The C1 element shares an overlapping DNA-binding region with the insulin enhancer element A2, and A2.2, a β cell-specific activator, binds to the overlapping A2 element 30 . Pdx1 binds to the GG2 element (bases -145 to -141 of the human insulin promoter 31 ) ( Supplementary Fig. 1).
We observed that bases -170 to -147 and -145 to -104 of the promoters of mouse Ins1 and Ins2 are identical and that bases -149 to -147 (AGG) and bases -114 to -112 (TGG) serve as protospacer adjacent motif (PAM) sequences in the guide RNA (gRNA) of the CRISPR-Cas9 system [19][20][21] . Therefore, we designed two gRNAs to target the promoter regions of mouse Ins1 and Ins2 (Fig. 1a). We used the pX330 vectors described in the "Methods" to express these gRNAs and Cas9 in mouse zygotes 32 .
Generation of mice with insulin promoter mutations. To generate mice with insulin promoter mutations, pX330-1 st /2 nd gRNA vectors were comicroinjected into the pronuclei of 310 one-cell-stage embryos obtained from C57BL/6J mice. Morphological abnormalities were undetectable in 252 (81.3%) of the embryos immediately after microinjection, which were consecutively transferred into the oviducts of pseudopregnant recipient the Institute for Cancer Research (ICR) mice to produce 86 neonates. Sequence analysis revealed that 23 mice had partial deletions and one replacement in either the Ins1 or Ins2 promoter or in both promoters and that 63 mice had normal or mosaic patterns. Mosaicism may be explained by mutations that occurred after the first or later embryonic divisions. Furthermore, Cas9 PCR products were not detected in 23 mice with deletions in either the Ins1 or Ins2 promoter or in both.
Among the 23 mice, we detected 5 and 6 types of mutations in the C1 elements or in the sequences bordering the C1 elements of the Ins1 and Ins2 promoters, respectively (Fig. 1b). To evaluate the importance of these promoter regions in vivo, the 11 mice with these mutations were crossed with wild-type C57BL/6J mice. Forty-one heterozygous mice with deletions in the Ins1 promoter or Ins2 promoter or in both were not diabetic (Supplementary Figs. 2, 3, Supplementary Table 1). The F1 mice were intercrossed to obtain homozygous F2 mice that had deletions in only the Ins1 promoter or the Ins2 promoter. These mice were not diabetic (Fig. 1d).
Thirty homozygous mice with mutations in both the Ins1 and Ins2 promoters were obtained by crossing the F3 and F4 mice (Supplementary Table 2). Homozygous mice with ≤3-base deletions in the Ins1 promoter and any deletion in the Ins2 promoter were not diabetic (Fig. 1e). Homozygous mice with any deletion in the Ins1 promoter and ≤3-base deletions in the Ins2 promoter were not diabetic (Fig. 1e). The most significant outcome of these studies was that the four types of mice with homozygous genotypes (1C 6 2C 6 , 1C m 2C 6 , 1C 6 2C 3 , and 1C m 2C 3 mice) with >3-base deletions in both the Ins1 and Ins2 promoters were diabetic (Fig. 1f).
Quantitative RT-PCR analysis of mRNAs extracted from the pancreatic islets of the 4 types of mice (fasted) with diabetes showed that the mRNA levels of Ins1 and Ins2 were decreased compared with those of normal mice (fasted) (Fig. 2a, b). There were no significant differences in the levels of the Pdx1, NeuroD, and MafA mRNAs ( Supplementary Fig. 4). Moreover, insulin promoter activity was significantly decreased for the two types of Ins1 and Ins2 promoters ( Fig. 2c-f). MafA, Pdx1, and NeuroD synergistically activated the wild-type promoters. Mutations in C1 result in a sharp decrease in synergistic activation by MafA, Pdx1, and NeuroD. In particular, MafA did not induce the activation of promoters with C1 mutations (Fig. 2d, f).

Discussion
Here, we used the CRISPR-Cas9 system to evaluate the functions of the promoter regions of Ins1 and Ins2 in vivo. We generated mice with 5 and 6 distinct deletions in the Ins1 and Ins2 promoters, respectively (Fig. 1b). Homozygous mice with deletions of ≤3 bases in the Ins1 promoter with any deletion in the Ins2 promoter (1C 1 2any, 1 2 2any, and 1 1 2any) or Ins1 promoter and with deletions ≤3 bases of the Ins2 promoter (1any2 3 , 1any2C 1 , 1any2 2 , and 1any2 1 ) were not diabetic (Fig. 1e, Supplementary Table 2). The 1-3 deleted bases were located between the C1 and E1 elements or within the sequences bordering the C1 element (Fig. 1b). Therefore, deletions of these 1-3 bases in the Ins1/2 promoters did not affect insulin promoter activity. The other homozygous mice with deletions in both the Ins1 and Ins2 promoters were diabetic (Figs. [3][4][5][6]. These data suggest that these sequences are important for insulin promoter activity in vivo. In vitro studies show that the C1 elements of the Ins1 and Ins2 promoters are required for Ins expression 13 . Furthermore, these sequences contain cell-specific and ubiquitous motifs that confer enhancer activity 13 (Fig. 1d). It has been reported that mice with single homozygous null mutations of Ins1 or Ins2 were not diabetic because of compensatory responses, and there was a dramatic increase in Ins1 transcripts in the Ins2 −/− mice 33 . Our data also showed that compensatory transcription of a functional insulin gene in homozygous mice with a deletion in the Ins1 or Ins2 promoter ( Supplementary Fig. 3c, d) did not lead to diabetes.
In conclusion, we used the CRISPR-Cas9 system to analyze the insulin promoter by taking advantage of the site-specific repair mediated by NHEJ at the DNA break site, which leads to different unpredictable insertions or deletions of various sizes. Our findings show, to the best of our knowledge and for the first time, that the C1 elements of the insulin promoter in mice are required for its activity in vivo. More importantly, the CRISPR-Cas9 technique will provide a tool to generate knockout mice that can be used to evaluate promoter regions.

Methods
Generation of mice with insulin promoter mutations. To construct two Cas9-single-guide RNA (sgRNA) expression vectors, deoxyoligonucleotides ( Fig. 1a: 1st gRNA and 2nd gRNA) were inserted into pX330 vectors 34 (Addgene, Watertown, MA). Female C57BL/6J mice were injected with pregnant mare serum gonadotropin and human CG (hCG) at 48-h intervals and were then mated with male C57BL/6J mice. The fertilized one-cell embryos were collected from the oviducts. Subsequently, 5 ng/μL pX330-1st/2nd gRNA vectors were injected into the pronuclei of these one-cell embryos, which were transferred into pseudopregnant Institute of Cancer Research (ICR) mice. F0 mice were genotyped to detect the presence of mutations in the Ins1/2 promoters. F0 mice were checked for the Cas9 transgene and for off-target effects. F0 mice were mated with C57BL/6N mice to obtain the F1 offspring.
The mutant mice and littermate wild-type mice were maintained on a C57BL/6 background. Mice were housed under a 12:12-h light/dark cycle in a room with controlled temperature and humidity. Food and water were provided without restrictions. The Institutional Animal Care and Use Committee of the University of the Ryukyus and the University of Tsukuba approved the animal studies.
Construction of insulin promoter-luciferase plasmids. Wild-type or mutated insulin promoter cDNAs containing~1000 bp of the 5′-flanking sequences of the Ins1/2 promoter regions were amplified using PCR with appropriate linkercontaining primers, which were then used to replace the promoter region of the Rat Ins2 promoter-reporter (luciferase) plasmid 35 using a ligation kit (TaKaRa, Tokyo, Japan).
Quantitative RT-PCR. Total RNA was extracted from the isolated islets using an RNeasy Mini Kit (Qiagen, Tokyo, Japan). After quantifying the RNA using spectrophotometry, 2.5 µg of RNA was heated at 85°C for 3 min and then reversetranscribed in a 25-µl reaction mixture containing 200 units of Superscript II RNase H-RT (Thermo Fisher Scientific), 50 ng random hexamers (Thermo Fisher Scientific), 160 µmol/l dNTP and 10 nmol/l dithiothreitol. The reaction mixture was incubated for 10 min at 25°C, 60 min at 42°C and 10 min at 95°C. The mRNAs were quantitated using the TaqMan real-time PCR system according to the manufacturer's instructions (Applied Biosystems, Foster City, CA, USA). PCR was performed for 40 cycles, and the reactions were incubated for 2 min at 50°C and 10 min at 95°C for the initial steps. In each cycle, denaturation was performed for 15 s at 95°C, and annealing/extension was performed for 1 min at 60°C. PCR was performed in 20 µl of solution using cDNAs synthesized from 1.11 ng of total RNA. The amount of mRNA was normalized by dividing the amount of the mRNA of interest by that of Gapdh mRNA. Primers specific for mouse Ins1, Ins2, Pdx1, NeuroD, MafA, and Gapdh were purchased as Assays-on-Demand Gene Expression Products (Applied Biosystems).
ELISA. The mice were fasted overnight. C-peptide was measured with a C-peptide ELISA kit (FUJIFILM Wako Shibayagi, Gunma, Japan).
Intraperitoneal glucose tolerance testing. Intraperitoneal glucose tolerance testing was performed on 6-and 12-week-old mice. The mice were fasted overnight, after which glucose (2.0 g/kg body weight) was intraperitoneally injected. The blood glucose levels were measured before injection and at 5, 30, 60, 90, and 120 min after injection.
Statistics and reproducibility. Error bars indicate standard error of at least triplicate measurements. Student's t test was used to compare two samples from independent groups using Microsoft Excel. Repeated measures ANOVA was performed to compare data among groups. The differences between the groups were considered to be significant when p < 0.05.