Creating Class I MHC–Null Pigs Using Guide RNA and the Cas9 Endonuclease

Classical class I MHC genes synthesize polymorphic proteins that form a heterotrimeric complex with the β₂ microglobulin (β₂m) protein and short peptides (1, 2). The polymorphism arises from the existence of multiple class I genes each having many alleles. A groove, formed by the α1 and α2 domains of the protein, contains most of the variation between different class I polypeptides and has pockets that accommodate specific amino acids and determine peptide anchor positions (3–5). Variation in non-anchor residues allows a single class I MHC protein to present a diverse set of peptides.

Rodent models and in vitro assays have provided much of the mechanistic information regarding the role of class I molecules in Ag presentation. The heterotrimeric complex is scanned by Ag receptors on CD8⁺ T lymphocytes (6, 7). Some MHC-displayed peptides originating from infectious agents or cancers activate CD8 cells. Despite understanding these processes in great detail, identifying relevant Ags has been difficult. Variations in both the class I MHC protein and in other Ag presentation machinery can independently shape the peptide repertoire (8–10). In addition to their role in CD8⁺ T lymphocyte activation, class I MHC molecules also regulate the activity of other immune effectors such as NK cells. Understanding the nuances of how class I MHC regulates immunity in various species will benefit from the development of additional animal models.

To improve the ability to study class I MHC function in pigs, we have used the Cas9 nuclease and guide RNA (gRNA) to disrupt class I MHC genes, which are also known as swine leukocyte Ags (SLAs). Following somatic cell nuclear transfer (SCNT), we have created cloned animals lacking class I MHC protein expression. The development of class I MHC–deficient swine will aid the study of immunity to infectious diseases that cost the agricultural industry hundreds of millions of dollars annually (11). These novel pigs can also be used in studies of xenotransplantation in an effort to eliminate the human anti-pig immunity that currently prevents the use of pig tissues as replacements for failing human organs.

Fetal fibroblast cells from a cloned pig with known class I SLA alleles were used in this study (12). Fetal fibroblasts cultured in stem cell media (FFSCs) were resuspended and cultured in MEM-α (Invitrogen, Carlsbad, CA)/EGM-MV (Lonza, Basel, Switzerland) media supplemented with 10% FBS (HyClone, Logan UT), 10% horse serum (Invitrogen), 12 mM HEPES (Sigma-Aldrich, St. Louis, MO) and 1% penicillin/streptomycin (Life Technologies, Grand Island, NY), and cultured in collagen-I–coated plates (Becton Dickinson, Bedford, MA) at 38.5°C, 5% CO₂ and 10% O₂. The cells treated with SLA-specific gRNA and Cas9 contained a GGTA1 gene that had been previously inactivated (13). SLAs expressing control cells were derived from GGTA1-deficient animals. The genetic backgrounds of the control and experimental animals are very similar, as they were cloned from cells originating from a single donor.

Plasmid pX330-U6-Chimeric_BB-CBh-hSpCas9 (Addgene plasmid 42230, http://www.addgene.org/42230/) was used to clone the designed annealed oligonucleotides (Fig. 1E, Supplemental Fig. 1) to generate gRNA using the CRISPR-associated Cas9 nuclease system (14). One microgram of plasmid pX330 was digested with BbsI (New England Biolabs, Ipswich, MA) for 30 min at 37°C. Each pair of phosphorylated oligonucleotides was annealed using a Veriti themocycler (Applied Biosystems, Grand Island, NY) starting at 37°C for 30 min, followed by a step at 95°C for 5 min and then ramp down to 25°C at 5°C/min. Digested pX330 was ligated to the annealed par of oligonucleotides for 10 min at room temperature. Ligation reaction was used to transform TOP10 competent cells (Invitrogen), following the manufacturer’s protocol. The QIAprep kit (Qiagen, Valencia, CA) was used to isolate plasmid from 15 colonies per treatment. DNA clones were sequenced (DNA Sequencing Core Facility, Department of Biochemistry and Molecular Biology, Indiana University–Purdue University Indianapolis) and used to transfect porcine FFSCs.

FFSCs were seeded in early passage (passage 2) onto six-well plates 24 h before transfection. Cells were harvested and counted and 1 × 10⁶ cells were resuspended in 800 μl fresh sterile electroporation buffer (75% cytosalt buffer: 120 mM KCl, 0.15 mM CaCl₂, 10 mM K₂HPO₄ [pH 7.6], 5 mM MgCl₂) and 25% Opti-MEM (Life Technologies). Cells were mixed with 2 μg plasmid DNA in 4-mm cuvettes. Transfection was performed using the Gene Pulser Xcell (Bio-Rad, Hercules, CA) following the manufacturer’s protocol for mammalian cells. Treated cells were seeded into six-well plates and grown until confluent. Cell screening was performed using a BD Accuri C6 flow cytometer (BD Biosciences, San Jose, CA), using mouse anti-pig SLA class I–FITC Ab (AbD Serotec, Raleigh, NC). Cells with low expression for SLA class I Ag were expanded and FACS was used at least twice with the FACSVantage SE at the Indiana University Flow Cytometry Resource Facility. Selected treatments were used for SCNT.

Genomic DNA was isolated from pig cells using the QIAamp DNA mini kit (Qiagen). RNA samples were isolated using the RNeasy Plus mini kit (Qiagen) following the manufacturer’s protocol. RNA quality and quantity were affirmed by Agilent bioanalyzer analysis (Core Facility of the Department of Biochemistry and Molecular Biology, School of Medicine, Indiana University–Purdue University Indianapolis). RNA samples were reverse transcribed using a OneStep RT-PCR kit (Qiagen). PCR products were purified and ligated into the pCR4-TOPO TA (Invitrogen). Transformed bacteria were plated onto Luria–Bertani agar containing 50 μg/ml kanamycin for clone selection. Plasmids were isolated using the QIAprep Spin Miniprep kit (Qiagen). Nucleotide sequences were performed by the Sanger method using custom sequencing service (Elim Biopharmaceuticals, Hayward, CA). Primers used in this study are shown in Supplemental Fig. 1.

For flow cytometry, porcine PBMCs were prepared using Ficoll-Paque Plus as described previously (15). PBMCs were stained with the following Abs: mouse anti-pig PerCP-Cy5.5 CD3, PE CD4, FITC CD8α, and mouse isotype control (BD Biosciences). Dead cells were excluded from analysis using fixable viability dye eFluor 660 (eBioscience, San Diego, CA). Analysis was performed using an Accuri C6 flow cytometer and CFlow software (Accuri, Ann Arbor, MI) and FlowJo software (Tree Star, Ashland, OR).

Primary kidney endothelial cells were isolated using 0.025% of collagenase type IV from Clostridium histolyticum (Sigma-Aldrich) and cultured for 3 d in RPMI 1640 medium supplemented with 10% FBS and 100 μg/ml endothelial cell growth supplement (BD Biosciences).

Fibroblasts were grown under the same conditions used to maintain fetal fibroblasts as described above.

SCNT was performed as described previously (16) using in vitro–matured oocytes (DeSoto Biosciences, St. Seymour, TN). Cumulus cells were removed from the oocytes by pipetting in 0.1% hyaluronidase. Only oocytes with normal morphology and a visible polar body were selected for SCNT. Oocytes were incubated in manipulation media (Ca-free NCSU-23 with 5% FBS) containing 5 μg/ml bisbenzimide and 7.5 μg/ml cytochalasin B for 15 min. Oocytes were enucleated by removing the first polar body plus metaphase II plate, and one cell was injected into each enucleated oocyte. Couples were fused and activated simultaneously by two DC pulses of 180 V for 50 μs (BTX cell electroporator, Harvard Apparatus, Holliston, MA) in 280 mM mannitol, 0.1 mM CaCl₂, and 0.05 mM MgCl₂. Activated embryos were placed back in NCSU-23 medium with 0.4% BSA and cultured at 38.5°C, 5% CO₂ in a humidified atmosphere for <1 h before being transferred into the recipient. Recipients were synchronized occidental gilts on their first day of estrus. Animals used in this study were under protocols approved by the Institutional Biosafety and Institutional Animal Care and Use Committee of Indiana University–Purdue University Indianapolis.

Nucleotide sequences were analyzed using the MacVector with Assembler 12.7 (MacVector Apex, Cary, NC). Overlapping forward and reverse sequence fragments assembled the complete coding sequence of each allele. Multiple sequence alignment was created for each locus. Sequences were confirmed by using the Immune Polymorphism Database–MHC SLA sequence database (http://www.ebi.ac.uk/ipd/mhc/sla/).

As depicted in Fig. 1A, the class I region consists of three classical MHC genes (SLA-1, -2, and -3). This genomic location also contains three pseudogenes (SLA-4, -5, and -9) and two poorly characterized class I–like genes (SLA-11 and -12) (17–20). The alleles of SLA-1, -2, and -3 present in the cells and animals used in this study are shown in Fig. 1B. Three SLA-1 alleles presumably exist as a result of gene duplication in one of the haplotypes (21, 22).

Classical class I MHC proteins are membrane proteins having three extracellular domains, a transmembrane segment, and a short cytoplasmic tail, and they are encoded by an eight-exon gene (Fig. 1C, 1D). Exons two and three are highly polymorphic whereas the remaining exons have limited variability (20, 23). Exon four encodes the extracellular α3 domain, which is critical for assembly of the class I molecule with β₂m and its transport to the cell surface (24, 25).

The Cas9 endonuclease is now commonly used to inactivate genes (14, 26–29). This is accomplished by designing gRNAs that recruit Cas9 to target sequences. Cleavage initiates repair mechanisms that are imperfect, leading to the creation of insertions and deletions that disrupt gene activity. Three separate gRNAs were designed that specified unique locations in exon 4 (Fig. 1D, 1E). All three targeted sequences are present in the alleles of SLA-1, -2, and -3 that are relevant to this study. Comparison with sequences from the National Center for Biotechnology Information indicates the targeted nucleotides are also present in SLA -4, -5, -9, -11, and -12 (pseudo)genes, although the consequences of gRNA treatment on these genes were not examined.

Fibroblasts expressing classical class I SLA molecules (Fig. 2A, untreated) were transfected with gRNA. Following gRNA and Cas9 treatment, expression of SLA-1, -2, and -3 was evaluated with an mAb that binds all three MHC proteins. Two rounds of flow cytometry sorting produced an SLA⁻ population of cells (Fig. 2A). The isotype control peak in sort 2 is difficult to see because of overlap with the class I SLA histogram. All three gRNAs successfully disrupted class I SLA genes when used alone and in combination (Fig. 2B).

Class I SLA⁻ cells, isolated from cultures treated with Cas9 and gRNA A+C, were used for SCNT (Table I). Two-hundred eleven cloned embryos were transferred to two recipients, of which one became pregnant. Fetuses were collected at day 32 of gestation. Fibroblast cells from well-formed fetuses 2 and 3 were grown in culture and retested for phenotypic expression of class I SLA by flow cytometry (Fig. 3). Although cells from fetus 3 appeared to express a small quantity of MHC protein, fetus 2 cells remained negative and were recloned, producing two other pregnancies. One sow spontaneously aborted at day 45 and the other produced three clonal piglets (Table I). One animal was sacrificed for cell collection at 1 wk of age. At the time of publication, the other two have been healthy for 7 mo. Renal cells from piglet 1, PBMCs from piglets 2 and 3, and fibroblasts from all three animals were negative for class I SLA cell surface expression (Fig. 4A).

cDNA from fetus 2 and piglets 1 and 2 were evaluated by allele-specific PCR to determine why class I MHC proteins were absent (Fig. 4B). PCR of cDNA from engineered animals failed to amplify SLA-2*0202, SLA-3*0402, and SLA-5*0502 transcripts. Apparently full-length transcripts representing SLA-1*1201 and SLA-2*1001 were obtained. These modified animals also contained multiple sizes of SLA-1*0702 and SLA-1*1301 transcripts. The parental unmodified cells were also analyzed for comparison and yielded primarily full-length transcripts (Fig. 4B). The lengths of the full-length transcripts were: SLA-1*0702 (1109 bp), SLA-1*1201 (1109 bp), SLA-1*1301 (1147 bp), SLA-2*0202 (1118 bp), SLA-2*1001 (1118 bp), SLA-3*0402 (1106 bp), and SLA-3*0502 (1109 bp).

Because the cDNA analyses did not explain the loss of every classical class I molecule, allele-specific PCR amplifications of cDNA were inserted into vectors. Multiple cDNA clones for each allele were sequenced in an effort to find inactivating insertions and deletions (Table II). These analyses were performed on untreated parental cells and on the modified cells isolated from fetus 2 and two SLA-deficient piglets. Most cDNA clones derived from the parental cells contained transcripts capable of encoding functional class I SLA proteins. A small number of cDNA clones indicated that some alleles (SLA-1*0702 and SLA-1*1201) expressed transcripts lacking 276 bases. In contrast, neither fetus 2 nor the cloned piglets contained full-length transcripts capable encoding functional class I SLA proteins. Transcripts lacking 276 bases became the predominant SLA-1*0702 species in piglets 1 and 2. In fetus 2, the SLA-1*0702 transcripts lacked either 276 or 4 bases (Supplemental Fig. 2). Neither of these transcript variants should encode functional protein because loss of 276 bases eliminates the α3 domain, and the 4-base deletion creates a frameshift mutation.

Although PCR analyses generated products for SLA-1*1201, SLA-1*1301, and SLA-2*1001, these were not confirmed in the sequence analyses. Instead, these cDNAs revealed multiple recombinant molecules (Table II). Alleles SLA-1*1301 and SLA-2*1001, SLA-1*1301 and SLA-1*0702, and SLA-2*1001 and SLA-12 recombined in close proximity to the location of the gRNA binding sites. These recombinants were incapable of encoding functional class 1 SLA molecules as a consequence of frameshifts arising from a 2-base deletion or 1-base insertion, or they lacked 276 bases of exon 4 (Supplemental Fig. 2).

Because class I MHC molecules are needed for the efficient development of mature T lymphocytes expressing CD3 and CD8 markers, a T cell subset analysis was performed (Fig. 5). PBMCs were gated for viability and expression of the CD3 protein (Fig. 5A) and then analyzed for expression of CD4 and CD8 (Fig. 5B). This analysis was performed on four separate blood draws from an SLA⁺ animal and on a total of five blood draws from two cloned SLA-deficient animals (Fig. 5C). The lymphocyte subsets in the control animal closely match what has been reported for other class I SLA–expressing pigs. The SLA-deficient animals showed a significant reduction in CD3⁺/CD8⁺ mature T lymphocytes and an increase in CD3⁺/CD4⁺ T cells.

In this study, we describe the creation of class I MHC–deficient pigs by simultaneously disrupting seven alleles of classical class I SLA genes. At the time of publication, the animals have been healthy and growing well for 7 mo. All cell types tested from these animals exhibit total loss of cell surface class I MHC proteins. Inactivation of a single gene, β₂m, also prevents expression of class I MHC heterotrimers at the cell surface. This disruption was not attempted because mice lacking β₂m lose the ability to regulate iron homeostasis (30–32). The high efficiency of the gRNA/Cas9 technology enabled the rapid production of animals deficient in class I MHC. The relative ease of this approach may enable the rapid production of many novel species lacking class I MHC gene activity.

Multigene families can be difficult to analyze because of functional redundancies. Simultaneous inactivation of all related genes minimizes these challenges by creating a null background that enables the study of one gene at a time. Research into the class I MHC gene family highlights these issues. The development of cell lines with defective expression of class I MHC gene products was crucial to various aspects of their biosynthesis and immune regulatory functions (6, 33–38). Animals lacking expression of class I MHC proteins have provided the opportunity to study the function of these molecules at the organismal level. This type of analysis is challenging in humans because of the the rare opportunity to identify and study naturally occurring defects in class I MHC biology (39). Mouse models have been far more productive in this regard because genetic engineering enabled the production of animals with a variety of gene defects preventing biosynthesis of class I MHC proteins (40–44). Manipulation of larger mammals using the genome engineering tools applied to murine models has been difficult because they reproduce slowly.

Targeted nucleases have simplified the alteration of mammalian genes. Zinc finger nucleases and transcription activator–like nucleases consist of the nucleolytic portion of the FokI enzyme covalently attached to a protein domain that can be modified to bind to unique DNA sequences (28, 45, 46). An RNA-based technology employs the Cas9 endonuclease and a gRNA to achieve site-specific activity (13, 26–29). The generation of cleaved DNA stimulates at least two cellular repair mechanisms, either of which can be used to complete the genome editing process: 1) nonhomolgous end joining of DNA fragments is imperfect and leads to random modifications such as point mutations, insertions, and deletions; and 2) exogenous DNA templates transfer specific genetic modifications to the genome in a process relying on homology-directed repair (28).

Although gRNA can direct Cas9-mediated cleavage of unintended genomic locations even in the presence of sequence mismatches (47–49), the results reported in the present study arise as a direct consequence of modifying class I MHC genes. We were able to eliminate class I MHC protein expression by treating cells with three unique gRNAs (Fig. 2). The only common target of the gRNAs is exon 4 of classical class I genes and class I–related pseudogenes in the MHC complex. Although it is likely that any of the gRNA/Cas9 treatments would have generated SLA deficient animals, cells treated simultaneously with gRNAs A and C were chosen for SCNT in an attempt to increase the frequency of larger deletions to simplify analyses. Determining the true effects of simultaneously using two gRNAs requires additional analyses.

The combination of high polymorphism and identity between each classical class I MHC gene makes sequence analysis of this region difficult. PCR amplification of cDNA from these loci introduces point mutations, recombinations, insertions, and deletions (50). Therefore, we evaluated numerous cDNA clones to ensure accurate sequence analyses of each allele in the unmodified parental cells and in gRNA/Cas9-treated samples. Most parental cell cDNAs were determined to be full length (Table II), and a few cDNAs lacking 276 bases of exon 4 were found for SLA-1*0702 and SLA-1*1201. No transcripts capable of encoding functional class I proteins were found in the SLA-deficient animals. SLA-2*0202, SLA-3*0402, or SLA-3*0502 were completely absent. The fetus and piglets all expressed three unique recombinant transcripts not found in the unmodified parental line. These transcripts do not encode functional protein because of frameshifts in two of the recombinants as well as the deletion of 276 bp from the third. The only transcript common to the unmodified parental cells and gRNA/Cas9-treated cells originated from SLA-1*0702, but a 276-base deletion in this transcript eliminates the α3 domain of the class I protein.

The reduced frequency of CD4⁻CD8⁺ T cells in the SLA-deficient animals mimics what has been seen in class I MHC–deficient mice, suggesting that many aspects of T cell development are common between the two species (38, 51). Additionally, the T cell subset analysis of CD4 and CD8 expression on peripheral lymphocytes found four populations (double-negative, double-positive, CD4⁺CD8⁻, and CD4⁻CD8⁺) similarly to what others have reported (52–54). The apparently normal levels of CD4⁺CD8⁺ double-positive cells in class I SLA–deficient animals indicates that engagement of classical class I MHC molecules with either the TCR or CD8 molecule is not essential for the production or maintenance of these cells. The presence of reduced levels of CD4⁻CD8⁺ T cells may provide some protection from infection through a mechanism that does not involve class I MHC. This could explain the animals’ apparent good health. However, the fact that the pseudorabies virus has evolved to inhibit presentation of antigenic peptides suggests that class I SLAs are key to viral control in pigs (55). Longer term evaluation will be needed to determine the susceptibility of these animals to infectious diseases and cancer.

The enhancements provided by targeted nuclease technologies have dramatically expanded the scope of genome engineering. The ease with which the gRNA/Cas9 system has allowed simultaneous disruption of seven class I MHC alleles in swine suggests this technology should be easily translated to other complex organisms. The ability to simultaneously disrupt several genes in an organism promises to enhance the studies of gene families with redundant functions and will enable the creation of polygenic disease models.

We thank the Methodist Research Institute and Laboratory Animal Research Center staff for taking care of the animals, as well as Dr. Chris Burlak for teaching us the FlowJo software.

A.J.T. is the founder and equity holder of Xeno-Bridge, LLC, an entity that has applied for patents with a goal to genetically engineer pigs for xenotransplantation. R.A.S. serves as a contractor and consultant to businesses that operate in the field of xenotransplantation. Parts of this study were funded by Lung Biotechnology, LLC. The other authors have no financial conflicts of interest.