Genetically humanized pigs exclusively expressing human insulin are generated through custom endonuclease-mediated seamless engineering

These authors contributed equally to this work. These authors contributed equally to this work. Dear Editor, Type 1 diabetes (T1D) is a lifelong (chronic) disease and a major health problem throughout the world. This disease can be treated by either insulin injection or islet transplantation. Islet transplantation is considered as a better treatment for T1D patients, because islets can produce and release insulin at the appropriate time, resulting in tight blood glucose control. However, islet transplantation is performed only for brittle T1D patients due to lack of sufficient donors: only 1 of 333 patients with insulin-dependent diabetes mellitus (IDDM) can obtain human islets (Frank et al., 2005). Porcine insulin has been applied to treat T1D. Recently, pig islets have also proved to be an attractive alternative resource for solving the donor shortage issue for islet transplantation. Many preclinical trials of porcine islet transplantation have been performed and have already achieved long-term survival of porcine islets in primate recipients (Cardona et al., 2006; Bottino et al., 2014). Previous efforts have been focusing on overcoming rejection of recipients to the donor organs (Bottino et al., 2014). Porcine insulin differs from human insulin by one amino acid (alanine in pigs and threonine in humans) at the carboxy-terminal of the B chain (i.e. position B30) (Sonnenberg and Berger, 1983) (Supplementary Figure S1). This single amino acid difference can induce human antibodies to act against porcine insulin. As a result, the effectiveness is decreased when porcine insulin is used for the long-term treatment of diabetic patients (Clark et al., 1982). For the same reason, the transferred pig islets may not function well in human patients for a long period. To address this issue, here we attempt to generate genetically humanized pigs that produce human insulin rather than porcine insulin. We edited porcine INS (pINS) gene in fibroblasts by using transcription activator-like effector nucleases (TALENs) or CRISPR/Cas9, combined with single-stranded oligonucleotides (ssODNs) as homology donors. By using somatic cell nuclear transfer (SCNT) technology, we successfully generated the genetically modified pigs exclusively expressing human insulin. Gene-editing animals have been generated by injecting the mRNA or proteins of custom endonuclease (such as zinc-finger nuclease, TALEN, and Cas9/gRNA) into one-cell stage embryos, which mostly resulted in mosaicism of the modification. In addition, the genotypes generated by embryo microinjection are variable and unpredictable (Crispo et al., 2015). One or two rounds of further breeding should be performed to obtain desired homozygotes with identical genotype and phenotype, which is particularly time and labor consuming for large animals with long gestation term and sex maturation age, such as pigs. To address this issue, our group has established an approach that employs gene-targeting somatic cells with SCNT, which could generate gene-editing animals with a single and identical mutation by one round of SCNT experiment. Recent reports have demonstrated that ssODNs, which are easily available, are more effective donors than traditional double-strand DNA homology for homology-directed repair (HDR), with the aid of custom endonucleases such as zinc-finger nuclease, TALENs, and CRISPR/Cas9 (Bedell et al., 2012; Yang et al., 2013). In addition, when ssODNs are used as HDR donors, selection marker gene is not necessary, which enables to create seamless site-specific mutations. The success of ssODN-mediated targeting strategy has been reported in cell lines or early stage embryos, but not yet in somatic cells, where homologous recombination with ssODNs is expected to be more difficult due to the limited proliferation competency. The length of ssODNs may affect the HDR efficiency in somatic cells. To optimize the length of ssODNs as donors, a HEK293 cell line with a fluorescence reporter was established, in which enhanced green fluorescent protein (EGFP) gene was mutated by deleting T at 456 site (ΔEGFP), and thus a stop codon (TATA>TA-A) was created to prevent the EGFP expression. Seven ssODNs with different lengths were used as donors, combined with either TALENs or Cas9/gRNA, to repair ΔEGFP (Supplementary Figure S2A). Flow cytometry analysis was performed to determine the repairing efficiency (Supplementary Figures S2B and S3). We found that the repairing efficiency increased when the length of ssODNs ranged from (0+0) nt to (25+25) nt and remarkably decreased when the length of ssODNs reached (30+30) nt. However, the repairing efficiency increased again when the length of ssODNs increased from (30+30) nt (Figure 1A, Supplementary Figure S3). It had been reported that when short ssODNs of 15−25 nt were used as donors, an alternative genome repair mechanism, i.e. microhomology-mediated end joining, would be employed by cells and resulted in sequence deletion and chromosome translocation (McVey and Lee, 2008). Therefore, ssODNs with the length < (25+25) nt are not suitable for a precise DNA repair. Conversely, when the length of ssODN is >35 nt, classical HDR will take place, which can accurately restore DNA sequence at a double-strand break (Supplementary Figure S2C–E). Therefore, we chose ssODN with lengths of (40+40) and (45+45) nt that show an optimum repairing efficiency as HDR donors to replace GCC with ACG in the pINS gene (Figure 1A). The efficiency of HDR mediated by ssODNs with TALENs was comparable with that of ssODNs and Cas9/gRNA. Thus, both TALENs and Cas9/gRNA were applied to modify the pINS gene (Figure 1B). Two pairs of TALENs and one gRNA targeting the exon II of pINS were designed and assembled, and their activities were evaluated by single-stranded annealing (SSA) assay (Li et al., 2014). The cleavage activity of pINS-TALEN2 is higher than that of pINS-TALEN1 and comparable with that of pINS-gRNA (Supplementary Figure S4). Thus, pINS-TALEN2 and pINS-gRNA were used to target pINS in fibroblasts. ssODNGCC>ACG with lengths of 88 nt (for TALENs) and 99 nt (for Cas9/gRNA), containing 41 nt right arm and 44 nt left arm, respectively, were used as HDR donors (Figure 1B). We introduced GCC>ACG mutation in the ssODN, and as a result, A>T amino acid substitution was induced at position B30 of porcine insulin. This mutation also generated a MluI restriction enzyme site, which helps to identify the genotype of targeted clones. For Cas9/gRNA, one silent mutation (GCG>GCA) on the PAM site was introduced to avoid repetitive digestion (Figure 1B). ssODN-mediated HDR with TALENs or Cas9/gRNA seamlessly edits pINS to generate cloned pigs expressing human insulin. (A) Optimization of ssODNs for HDR with TALENs or Cas9/gRNA. Seven ssODNs designed and synthesized for different lengths (31, 41, 51, 61, 71, 81, and 91 nt) were separately electroporated, alone or with TALENs or Cas9/gRNA, into reporter cells. Bar graph shows Tdtomato+/EGFP+ cell numbers in total 50000 cells quantitated by flow cytometry after 48 h of transfection (n = 3, error bar = SD). (B) Schematic diagram for precise editing of pINS gene by ssODN-mediated HDR with TALENs or Cas9/gRNA. Two pairs of TALENs and one gRNA were designed and assembled to target the Exon II of pINS (green highlight). Red letters indicate the substitution target site in pINS locus and corresponding mismatched nucleotides in the ssODNs that generate MluI restriction enzyme site (Italic). (C) Identification of selected colonies by PCR–MluI digestion. The red arrowhead marks the DNA band indicating HDR-mediated point mutation digested by MluI restriction enzyme. (D) Sanger-sequencing of targeted colonies identified by PCR and MluI digestion. The red box shows monoallelic and biallelic GCC>ACG substitutions in pINS locus. (E and F) Morphologically normal piglets were cloned from cell colonies targeted by ssODN-mediated HDR with TALENs (E) or Cas9/gRNA (F). All cloned piglets were genotyped by PCR–MluI digestion and Sanger-sequencing (Supplementary Figure S5). (G and H) RT-PCR (G) and Sanger-sequencing (H) detected the pINS mRNA-coding sequence of the cloned piglets. The red box shows monoallelic and biallelic GCC>ACG substitutions in pINS mRNA-coding sequence. (I and J) Insulin protein from the pancreas of the cloned piglets was purified with native-PAGE (I and Supplementary Figure S7) and analyzed by LC-MS (J and Supplementary Figures S9, S10). Human (black) and porcine (red) insulin standards were used as controls. Insulin proteins extracted from heterozygous (orange) and homozygous (blue) pigs were analysis. Porcine fetal fibroblast (pFF) cells isolated from E35d embryos of Bama mini-pigs were electroporated with circular plasmids of pINS-TALEN2 or Cas9/gRNA accompanied with ssODNs. For transient drug selection, a neomycin expression cassette was inserted into the TALENs or Cas9 expression vector. After selection with G418 (1 mg/ml for 10–14 days), the surviving cell colonies were expanded and initially identified through PCR and MluI digestion. Non-homologous end joining was observed in colonies generated by either pINS-TALEN2 (72/300, 24%) or Cas9/gRNA (32/90, 35.6%) (Supplementary Tables S1, S4, and S5). Among the 300 colonies generated by TALENs, six (6/300, 2%) were identified with one allele precise substitution and no colony was found with biallelic point mutations (Figure 1C). In colonies generated through the Cas9/gRNA method, a higher editing rate of 5.6% (5/90) was obtained. Interestingly, all of the five cell colonies were biallelically targeted (Figure 1C). Sequencing results further confirmed that all of the colonies were correctly targeted with humanized insulin sequence (Figure 1D, Supplementary Tables S4 and S5). The correctly targeted colonies with one allelic or biallelic mutation were used as donor cells for nuclear transfer. A total of 2613 cloned embryos were generated and transferred into 11 surrogate mothers (Supplementary Table S2). Five surrogates were confirmed pregnant through ultrasound detection one month post-embryo transfer. Four of these surrogates developed to term and gave birth to seven cloned piglets (Figure 1E and F, Supplementary Table S2). Genomic DNA was extracted from the ear tissues of these piglets for genotyping by PCR and MluI digestion. Of the seven piglets, three were cloned from TALEN-targeted colonies and were heterozygous, while four were cloned from Cas9/gRNA-targeted colonies and were homozygous (Figure 1E and F). The sequencing result further confirmed the desired precise modification in pINS in the cloned piglets (Supplementary Figure S5). Furthermore, no integration of exogenous gene was found in the seven cloned piglets by PCR amplification (Supplementary Figure S6), indicating that seamless site-specific modification has been successfully achieved, which will minimize the safety concern for future application in biomedicine. A heterozygote (A206-1) and a homozygote (A210-5) were born weak and both died one day after birth. The remaining cloned piglets were healthy and grew normally. A 3-month-old heterozygous piglet (A206-2) and a 3-month-old homozygous piglet (A210-6) were sacrificed to retrieve the pancreas and to verify whether the mutant piglets could express human insulin. The whole mRNA was extracted from the pancreas of the two piglets. pINS mRNA was amplified by RT-PCR and sequencing (Figure 1G and H). As shown in Figure 1H, the heterozygote contained both porcine (GCC) and humanized (ACG) insulin mRNAs, while the homozygote merely transcribed the humanized insulin mRNA with ACG. The insulin protein was also extracted from the pancreas of the two piglets by acid-alcohol extraction and purified by native-polyacrylamide gel electrophoresis (native-PAGE) (Figure 1I and Supplementary Figure S7). The purified insulin was quantified using a high-resolution accurate mass spectrometer (MS) in accordance with the manufacturer's instructions. Liquid chromatograph-MS (LC-MS) data obtained using a Thermo Scientific™ Q Exactive™ mass spectrometer revealed that the heterozygote expressed two kinds of insulin with molecular weights of 5807 and 5777 Da, corresponding to human and porcine insulin, respectively, while the homozygote exclusively expressed human insulin with a molecular weight of 5807 Da (Figure 1J, Supplementary Figures S9 and S10). Insulin from heterozygous and homozygous piglets was then intraperitoneally injected at the same dose into mice to evaluate the hypoglycemic activity. As shown in Supplementary Figure S8, the humanized insulin from pigs could decrease mouse blood glucose concentration to the similar level as that by standard human and porcine insulin within 1 h after injection. Although TALENs and CRISPR/Cas9 have greatly advanced gene targeting efficiency in pigs, the risk of undesired off-target effects still remains. We computationally predicted off-target sites possibly created by TALENs or Cas9/gRNA in the mutant pig genomes. A total of 39 potential TALEN-off-target (T-OT) sites with <9 mismatches/gaps in the spacer region from 32 to 60 bp, as well as 16 Cas9-off-target (C-OT) sites with up to 3 mismatches, were examined by T7 endonuclease I (T7EI) assay and sequencing (Supplementary Figure S11, Tables S6 and S7). No mutations were found in the potential off-target sites in all of the seven cloned piglets. The size, morphology, and histological structure of pancreas from the dead homozygous piglet (A210-4) appeared normal, compared with those from the wild-type piglets (Supplementary Figure S12). Hematological assay demonstrated that all of the blood indexes were within the range from the wild-type pigs at the same age (Supplementary Table S3). In summary, technologically, this study is the first attempt to generate seamless-engineering large animals by combining ssODNs with TALENs or CRISPR/Cas9 through SCNT. The approach established here can be extended to precisely create desirable SNPs in other large animals associated with human diseases and agricultural breeding. From the application aspects, insulin from humanized pigs is expected to have better effectiveness for treatment of diabetic patient. In addition, the INS-humanized pigs will provide more desirable source of xenogeneic islets and overcome risks caused by insulin differences between pigs and humans. [Supplementary material is available at Journal of Molecular Cell Biology online. This work was supported by grants from the National Basic Research Program of China (973 programs) (2011CB944203, 2011CB944204), the National High-Tech R&D Program of China (863 Programs) (2014AA021602), the National Natural Science Foundation of China (31401271), the Key Deployment Project of the Chinese Academy of Sciences (KSZD-EW-Z-005-003-002), the Science and Technology Planning Project of Guangdong Province, China (2014B030301058, 2015A030310119), and Bureau of Science and Technology of Guangzhou Municipality (201505011111498).]