Reprogramming MHC specificity by CRISPR-Cas9-assisted cassette exchange

The development of programmable nucleases has enabled the application of new genome engineering strategies for cellular immunotherapy. While targeted nucleases have mostly been used to knock-out or knock-in genes in immune cells, the scarless exchange of entire immunogenomic alleles would be of great interest. In particular, reprogramming the polymorphic MHC locus could enable the creation of matched donors for allogeneic cellular transplantation. Here we show a proof-of-concept for reprogramming MHC-specificity by performing CRISPR-Cas9-assisted cassette exchange. Using murine antigen presenting cell lines (RAW264.7 macrophages), we demonstrate that the generation of Cas9-induced double-stranded breaks flanking the native MHC-I H2-Kd locus led to exchange of an orthogonal H2-Kb allele. MHC surface expression allowed for easy selection of reprogrammed cells by flow cytometry, thus obviating the need for additional selection markers. MHC-reprogrammed cells were fully functional as they could present H2-Kd-restricted peptide and activate cognate T cells. Finally, we investigated the role of various donor template formats on exchange efficiency, discovering that templates that underwent in situ linearization resulted in the highest MHC-reprogramming efficiency. These findings highlight a potential new approach for the correcting of MHC mismatches in cellular transplantation.

cleavage 14 , simply through the inclusion of several targeting guide RNA (gRNA) sequences, which enables HDR mechanisms to potentially replace long genomic regions with an exchange cassette 15 .
We rationalized that the Major histocompatibility complex (MHC) locus would serve as a relevant proof-of-concept to demonstrate the potential for nuclease-mediated genomic exchange of immune alleles. In immunity, the highly diverse MHC locus educates and activates adaptive immunity by presenting foreign peptides from invading pathogens to guide adaptive immune responses. Yet this protective response is often detrimental during transplantation as the host MHC complexes can present, and respond vigorously to, allogeneic peptides that are derived from the donor MHC molecules 16,17 . The precise matching of MHC alleles between donor and recipient is therefore critically important to ensure long-term survival of donated cells. However, many clinical procedures, such as hematopoietic stem cell (HSC) transplantation for malignant (e.g., leukemia, lymphoma) and non-malignant disorders (e.g., severe combined immunodeficiency disorder), suffer from a sparse availability of correctly matched donors 18 . Several ex vivo genome editing approaches have been explored to mitigate MHC mismatches and generate more compatible or even "universal" donor cells. Previous studies of MHC-gene editing have been simply related to knockout; for example, ZFNs have been used to knockout the conserved MHC subunit, beta-2 microglobulin 19 , or the MHC allele directly 20 , both of which resulted in an elimination of surface expression of the MHC protein. These MHC-knockout cells are promising as "off-the-shelf " therapies but activation of downstream adaptive immune responses is compromised by the deletion of MHC alleles. To date, there have been no previous reports of MHC allelic exchange or replacement. Therefore, instead of MHC knockout, a strategy for replacement of MHC alleles would be more beneficial for improving donor-recipient matching because full immunological functionality is maintained.
As a first step towards MHC reprogramming, we used CRISPR-Cas9 to mediate rapid and scarless exchange of entire MHC-I alleles at the native locus (H2-K1) in murine-derived antigen-presenting cell lines (RAW264.7 macrophages, CT26 fibroblasts). Multiplex targeting of Cas9 with two gRNAs was used to introduce DSBs flanking the MHC-I allele, enabling HDR-mediated replacement with a MHC donor cassette (provided as a double-stranded DNA template). Thus, CRISPR-Cas9 assisted cassette exchange (CACE) was able to effectively replace an endogenous MHC-I allele (H2-Kd) of ~3.4 kb with a highly similar orthogonal allele (H2-Kb). The altered cellular phenotype was validated by the surface expression of the new MHC alleles, thereby allowing detection and facile recovery of correctly modified cells by fluorescence-associated cell sorting (FACS). The MHC-reprogrammed cells displayed a phenotype similar to a wildtype H2-Kb+ antigen-presenting cell line (JAWSII dendritic cells), while also being able to present a H2-Kb-specific antigenic peptide and activate cognate CD8+ T cells. Additionally, we investigated the efficiency of HDR with different MHC donor construct designs such as linear, plasmid, minicircle, and self-linearizing minicircle templates. We found that self-linearizing minicircle templates were the most efficient format for MHC exchange. Our approach shows the feasibility of replacing large MHC alleles at the native locus and suggests future utility for the correcting of MHC mismatches in allogenic cellular transplantations.

Materials and Methods
All plasmids and primers used in this work are described in Supplementary Table S1.

MHC-I locus sequencing.
Unless otherwise specified all primers were ordered from Integrated DNA Technologies (IDT). Genomic DNA was extracted from RAW264.7 and JAWSII cells (1 × 10 6 each) by the addition of 100 μ l QuickExtract solution (Epicentre) according to the manufacturers instructions. PCR using the Kapa2G Fast Readymix (Kapa Biosystems) and with primers p11 and p12 was performed to amplify ~4300 bp of DNA at the H2-K locus. Primers p13 and p14 were used to amplify a backbone fragment from pUC19 plasmid [New England Biolabs (NEB)]. Gibson cloning master mix (NEB) was used to combine the respective fragments to create the storage plasmids pUC19-H2-Kb and pUC19-H2-Kd, later used for Sanger sequencing. Additionally, ~1 kb regions 5′ and 3′ to the H2-Kd locus were amplified using primers p15-p16 and p17-p18, respectively and Sanger sequenced.
Determination of CRISPR cleavage sites at MHC locus. Typically, 2 × 10 6 RAW264.7 cells were recovered from 100 mm cell culture plates using Accutase solution (LT) and washed once with Opti-MEM media (LT). All spin steps were performed at 90 g for 5 min. Electroporation of RAW264.7 cells was performed with the Nucleofector 4D (program DS-136, Lonza). 5 μ g of each of the various px458 plasmids was diluted into a volume of 100 μ l SF buffer (V4XC-2024, Lonza) prior to electroporation. Following electroporation, the cells were plated in 100 mm culture dishes containing 10 ml of RPMI growth media for 24 h. After dissociation with Accutase solution, FACS (Influx, BD Biosciences) was performed to isolate Cas9 + cells based on 488 positive fluorescence (GFP), cells were returned to culture for 6 days. The cells were removed from the plate, washed once in PBS and genomic DNA was recovered from 1 × 10 6 cells using 100 μ l QuickExtract solution. Small fragments of DNA covering the putative cleavage sites were amplified from the genomic DNA using primers p1 and p2 for CRISPR guides #10 and #11, p3 and p4 for guides #12 and #13. For each primer set, control DNA was also amplified from wildtype RAW264.7 genomic DNA. CRISPR-Cas9 cleavage of the genome was determined using a Surveyor Mutation Detection Kit (IDT). Following hybridization of 600 ng of fragment DNA in the presence of 1 x Thermopol buffer (NEB), 1 μ l of surveyor enzyme was added in a 15 μ l final volume and digested for 40 min at 42 °C. All samples were run on 2% agarose gels (Sigma-Aldrich) for the detection of cleavage products. In order to test deletion of the MHC gene at the native locus, RAW264.7 cells were nucleofected as above with 2.5 μ g of each of the px458-G10 and px458-G13 plasmids. Primers flanking the H2-Kd locus (p1 and p4) were used for amplification with Kapa HiFi Hotstart 2x Readymix. PCR products were run on 1% agarose gels.
Creation of MHC-exchanged RAW264.7 cell line. Linear donor templates were amplified from a pMC.BESPX-ET1 plasmid stock using Kapa HiFi Hotstart 2x Readymix with primers p21 and p24. 25 cycles of amplification were performed with an annealing temperature of 64 °C and an extension time of 2 min. DNA was purified over Zymospin II columns, eluted in ddH 2 O, and concentrated to ~1 μ g/ml using a Concentrator 5301 (Eppendorf). 5 μ g of this plasmid was mixed with 5 μ g of the linear donor template and nucleofected into RAW264.7 cells. After 24 h of growth, Cas9 (GFP)-positive clones were isolated by FACS and cultured for an additional 6 days in 6-well plates containing 3 ml of complete media. The media was changed once after 3 days of growth. Cells were labeled with 1 μ M SIINFEKL peptide (eBioscience) in complete media for 30 mins followed by 1.25 μ g/ml anti-H2-Kd-PerCP-efluor 710 (Clone SF1-1.1.1, eBioscience), and 1.25 μ g/ml anti-H2-Kb-SIINFEKL-APC (Clone 25-D1.16, eBioscience, USA). FACS was used to isolate single-cell clones that were negative for H2-Kd expression and positive for high H2-Kb-SIINFEKL expression. Single-cells were grown in 96 well plates containing 100 μ l complete media. To confirm transcript expression of the new allele at the H2-K locus, mRNA was isolated from each of the single-cell sorted clones (C4, F4, F5, and G5) and parental RAW264.7 and JAWSII cells using 200 μ l of TRIzol ® reagent (LT). The mRNA was purified using a Direct-zol MiniPrep kit according to the manufacturers instructions (Zymo Research). First-strand cDNA was generated from mRNA with Maxima reverse transcriptase (LT) and used for subsequent PCRs. Allele specific primers that bridge the exon 3 -exon 4 junction for both the H2-Kb allele (p7 and p8) and the H2-Kd allele (p5 and p6) 25 were used with Kapa HiFi Hotstart 2x Readymix to amplify small DNA fragments for detection by agarose gel and Sanger sequencing.

Genomic characterization of MHC exchanged RAW264.7 cell lines.
To determine whether integration of the H2-Kb allele occurred at the correct locus, PCR was performed on genomic DNA with primers p9 and p29, which lie outside of the 5′ homology arm and within exon 3 of H2-Kb, respectively. Evaluation of the presence of remaining H2-Kd was also analyzed by PCR using primers p5 and p6. Finally deletion at the locus was evaluated by PCR as above using p1 and p4. For sequencing of the entire H2-K locus, fragments of ~8 kb in size were PCR-amplified from the genome with primers p37 and p38 using LongAmp Taq DNA polymerase (NEB). For each PCR, the reaction was split between a minimum of 5 separate tubes and then pooled for subsequent steps. Gibson cloning was used to clone these fragments into pUC19 storage plasmids (amplified with p39 and p40), and Sanger sequenced performed using primers p41, p42, p43, p44, p45 and p46.
T cell activation assay. Approximately, 2.5 × 10 5 JAWSII and 1.25 × 10 5 RAW264.7 cells were cultured in 96-well plates in 200 μ l of complete media supplemented with 10 μ g/ml E.coli 0111:B4 LPS and 10 ng/ml IL-4 cytokine. After overnight culture, samples (except for controls) were incubated with 1 μ M SIINFEKL peptide in complete media for 1 h at 37 °C and then irradiated with 20 Gy in a Cellrad irradiator (Faxitron, USA). 1 × 10 5 B3Z hybridoma cells were added to each well and the plate was spun briefly at 250 g to settle the cells. The plate was incubated overnight at 37 °C with 5% CO 2 before the cells were transferred to 96-well U-bottomed plates and spun at 2000 rpm for 2 min. The cells were washed twice with PBS and lysed with 100 μ l lysis buffer containing PBS with 0.5% Nonidet P40 (Sigma-Aldrich) and 0.15 mM Chlorophenol Red-ß-D-galactopyranoside (Sigma-Aldrich) 26 . The assay was incubated for a further 6 h at room temperature with absorbance readings at 570 nm recorded every 10 min in an Infinite Pro M200 plate reader (Tecan). B3Z cells cultured alone were included as a control.

Preparation of MHC linear and minicircle donor templates. A MHC linear donor template was
amplified from pMC.BESPX-ET1 plasmid DNA using Kapa HiFi Hotstart 2x Readymix as above. To obtain minicircle DNA, single colonies from pMC.BESPX-ET1 and pMC.BESPX-ET2 were selected from fresh plates and grown for 2 h at 30 °C in LB medium (BD Biosciences) with Kanamycin. 200 ml of TB medium (Affymetrix) containing Kanamycin was inoculated with 2E-4 O.D 600 equivalents and grown for up to 20 h at 30 °C. If the pH did not exceed 8, and the optical density was between 4 and 6, the culture was induced with 200 ml LB modified to include 40 mM NaOH and 0.02% Arabinose and grown for a further 5.5 h. Midiprep columns (Zymo) were used to purify the minicircles and the concentration was normalized after concentration measurement with NanoDrop (Thermo Fisher). 1 μ g samples of each minicircle, pre-and post-induction, were digested with FastDigest EcoRI (LT) and run on a 1% agarose gel to confirm expression.
Creation of MHC exchanged CT26 cell line. A Nucleofector 4D (Lonza, USA) was used to electroporate 2.5 μ g of each of the px458 CRISPR guide plasmids px458-G10 and px458-G13, along with 5 μ g of pMC-BESPX-ET2 minicircle product, in a volume of 100 μ l SE buffer into CT26 cells with program CM-150. After 24 h of growth, Cas9 (GFP)-positive clones were isolated by FACS and plated in 6-well plates containing 3 ml of complete media for an additional 6 days. Bulk cells were harvested using QuickExtract solution for PCR analysis of integration with primers p9 and p10.
Statistical analysis. For pairwise comparison between experimental groups in the B3Z assay, student T-tests were performed using the R software package. In each case a Bonferroni correction was applied to account for multiple comparisons. For comparison of donor template designs an ANOVA test was applied, taking into account batch effects, and once again accounting for multiple comparisons with a Bonferroni correction.

Results
Characterization of CRISPR-Cas9 target sites in the MHC locus. Murine RAW264.7 macrophages, a well-characterized antigen-presenting cell line derived from the Balb/c mouse strain, offer a suitable model system to evaluate reprogramming in the highly polymorphic MHC locus. Our overall strategy was based on using CRISPR-Cas9 to generate two DSBs in the native MHC-I H2-Kd gene while also providing a replacement donor template containing an orthogonal MHC-I allele of H2-Kb (derived from the C57BL/6 mouse strain) and homology regions flanking either side of the DSBs (Fig. 1a). HDR would then facilitate the exchange of the > 3 kb MHC alleles. No selection marker (i.e., fluorescence reporter gene or antibiotic resistance gene) is required to be incorporated into the HDR donor because cells surface express MHC, thus fluorescence labeling with H2-Kd-and H2-Kb-specific antibodies can be used for selection. In order to find gRNA sites in the MHC locus of RAW264.7 macrophages, we first amplified and Sanger sequenced the H2-Kd gene and corresponding flanking regions of up to 1 kb. The exons of the H2-Kd allele share 91% nucleotide (nt.) identity with those of the H2-Kb allele from the C57BL/6 genome, with the most diversity located in the peptide binding regions encoded in exons 2 and 3 ( Supplementary Fig. S1a). The CRISPR design tool (http://crispr.mit.edu 23 ) was used to identify gRNA sites (compatible with S. Pyogenes Cas9 and its protospacer adjacent motif (PAM, 5′ -NGG)) in the sequenced H2-Kd gene and the C57BL/6 genome was used as a reference for off-target events (a reference BALB/c genome was not available). Due to the inherent high degree of similarity between MHC genes, finding unique gRNA sites in this locus is especially challenging. We identified a selection of potentially H2-Kd-exclusive gRNAs sites within the exons at the 5′ and 3′ portion of the gene (Fig. 1b). To minimize the size of the required donor template, we limited the search for gRNA sites to the region between exons 2 and 7. Thus, we were able to exclude exon 1 as a target as it encodes the MHC signal peptide that is cleaved upon expression, and exon 8, as it shares 100% nt. identity between the two alleles. Each gRNA was cloned into the CRISPR-Cas9 plasmid pX458 (pSpCas9(BB)-2A-GFP), where Cas9 and green fluorescence protein (GFP) are expressed under the control of the same promoter, but as two separate proteins due to the self-cleaving T2A peptide 22 . The resulting pX458 plasmids were individually electroporated (nucleofection protocol) into wildtype RAW264.7 cells; at ~24 h, cells expressing Cas9 (via 2A-GFP) were isolated by FACS and expanded. The activity of Cas9 at each gRNA site was determined by measuring NHEJ via Surveyor nuclease assays 27 . We observed strong Cas9-induced cleavage (with the expected fragment sizes) at each of the gRNA sites tested (Fig. 1b,c). Based on their strong NHEJ activity, gRNA10 and gRNA13 were evaluated for multiplexed targeting. After simultaneous electroporation with pX458-G10 and -G13, deletion of the genomic region in between these two gRNA was verified by PCR (Fig. 1d).
Reprogramming MHC specificity in RAW264.7 macrophages. To reprogram the MHC H2-Kd allele of RAW264.7 cells, a ~4 kb exchange template was constructed based on the H2-Kb allele (exons 2-7), which was derived from genomic DNA of the JAWSII cell line (originating from the C57BL/6 mouse strain). The H2-Kb region was flanked by left and right homology arms (~350 bp each) corresponding to RAW264.7 BALB/c genome (Fig. 1a). In order to prevent Cas9 cleavage of the exchange template, the PAM site was modified (NGG > NGA) or silent mutations were created within the gRNA recognition sequence (Supplementary Fig. S1b,c). The donor template, generated by PCR and thus in a linear format, was electroporated with pX458-G10 and pX458-G13 plasmids into RAW264.7 macrophages. After ~24 h, cells were isolated by FACS for Cas9 expression (via 2A-GFP) and expanded in culture. In order to detect cells that had undergone MHC replacement we used a Scientific RepoRts | 7:45775 | DOI: 10.1038/srep45775 flow cytometry-based assay that relied on differential monoclonal antibody labeling of H2-Kb and H2-Kd ( Supplementary Fig. S1d). To further improve the specificity of detection for H2-Kb expression, we used a monoclonal antibody that had specificity for the H2-Kb allele only when bound to a cognate peptide derived from ovalbumin (OVA 257 -264 or SIINFEKL) 27 . This peptide bound selectively and with high affinity to the H2-Kb MHC allele but not the H2-Kd allele (Supplementary Fig. S1d). In RAW264.7 cells that received pX458-G10 and -G13 plasmids alone (without donor template), we observed substantial knockout of the endogenous MHC gene, as 34% of cells no longer surface expressed H2-Kd (Fig. 2a). When the exchange template was included, an additional small population of cells (0.48%) appeared which were positive for H2-Kb-SIINFEKL and negative for H2-Kd. Single-cell clones were sorted from this population and expanded for further analysis.
Flow cytometry analysis on several of the single-cell colonies revealed high expression levels of the new H2-Kb protein with no detectable expression of native H2-Kd. The reprogrammed RAW264.7 cells expressed H2-Kb at similar levels to that of JAWSII dendritic cells (Fig. 2b and Supplementary Fig. S1e). To further confirm expression of the newly introduced H2-Kb allele, total mRNA was isolated and used to generate cDNA from each of the single-cell derived clones. The resulting cDNA was used in a PCR reaction with primers flanking exons 3 and 4 of the H2-Kb and H2-Kd genes, respectively (Fig. 2c). PCR products of 153 bp should only be generated if successful splicing of exons 3 and 4 occurs. All reprogrammed cell lines showed robust expression of H2-Kb transcripts, while residual H2-Kd transcripts were not detected (Fig. 2d). Sanger sequencing of these PCR products verified correct allelic expression and showed the correct splicing junction sequence for each clone, matching that of wildtype H2-Kd expression in JAWSII cells (Fig. 2e).

Reprogrammed MHC cell lines activate T cells. We next confirmed that the newly expressing MHC
H2-Kb alleles in reprogrammed RAW264.7 cells were capable of providing functional immune activity. To upregulate MHC expression and ensure a mature cell phenotype, LPS and IL-4 cytokines were added to cultures for 24 hours. Flow cytometry analysis revealed a moderate upregulation (~2-fold) of the maturation markers CD80 and CD86, while there was a nearly 2-fold upregulation in H2-Kb expression with LPS (10 μ g/ml) ( Supplementary Fig. S2a). The addition of IL-4 (10 ng/ml) had a less pronounced but detectable increase in the expression of these markers. The matured cells were incubated with the SIINFEKL peptide (H2-Kb-specific) and X-ray irradiated to prevent proliferation. The irradiated RAW264.7 macrophages were then co-cultured with a CD8 + T cell hybridoma reporter cell line (B3Z) (Fig. 3a). B3Z T cells express a TCR that is specific for the MHC-peptide complex of H2-Kb-SIINFEKL, they have also been engineered to secrete β -galactosidase upon TCR engagement 21 . All of the reprogrammed RAW264.7 clones were able to induce strong β -galactosidase expression in B3Z T cells, which was comparable to the positive control of JAWSII dendritic cells (Fig. 3b and Supplementary Fig. S2b). Activation of B3Z T cells was completely dependent on the presence of SIINFEKL peptide in the co-culture. The parental RAW264.7 macrophage cells promoted very little β -galactosidase expression from B3Z cells (comparable to B3Z cells alone) indicating that the H2-Kd allele cannot serve to activate these T cells even with SIINFEKL present.

Genotypic characterization of MHC-reprogrammed cells. While phenotypic analysis is able to show
positive expression of H2-Kb and the lack of expression of the wildtype H2-Kd protein in reprogrammed macrophages, it is insufficient to determine whether: (i) MHC exchange occurred on both chromosomes (biallelic) or (ii) MHC exchange occurred on a single chromosome (monoallelic) in combination with NHEJ-induced knockout on the other chromosome. To address this, PCR assays on genomic DNA were designed to evaluate the various possibilities of modification in the MHC locus (Fig. 4a). In all single-cell lines tested, the newly introduced H2-Kb allele was detected by PCR (with primers p9/p10), but not in the parental RAW264.7 cell line, indicating that at least one of the alleles had been exchanged (Fig. 4b). We next looked for the presence of wildtype H2-Kd alleles, only a single colony (F5) showed a positive PCR product (p5/p6) that was comparable to the RAW264.7 positive control (Fig. 4c). This suggests cell lines C4, F4, and G5 all possessed only the new H2-Kb allele. This result correlates with the lower phenotypic H2-Kb expression observed by flow cytometry for clone F5 as compared to the other isolated clones (Supplementary Fig. S2b). However, the lack of H2-Kd allele present at the endogenous locus could either be due to exchange of the second allele or deletion at this locus. To control for this a PCR assay (p1/p4) was used to detect for H2-Kd deletion (Fig. 4d). Cells receiving only px458-G10 and px458-G13 were used as a positive control for genomic deletion. We found no bands indicating deletion in any of the cell lines tested. From each of the cell lines, we used a split-pool PCR approach to clone the inserted H2-Kb gene into a sequencing plasmid (Fig. 4e and Supplementary Fig. S3a). Each single-cell colony (C4, F4, F5, G5) had three of their resulting bacterial colonies Sanger sequenced. We observed multiple mutations that differed from the input donor template in all cell lines (Ø), and in some cases the mutations were present in all three colonies sequenced although these appeared not to have altered the H2-Kb phenotype. Because split-pool PCR was used for amplification, it is very unlikely that mutations arose during cloning into the sequencing vector. Sequencing errors are unlikely to be generated at the same location in all three colonies. Thus, mutations were most likely generated during the PCR amplification of the linear donor template despite the use of a high fidelity polymerase (KapaHiFi). Recent studies using high-throughput sequencing on PCR amplicons have shown that reproducible polymerase hotspot errors are more common than previously believed, even with high-fidelity polymerases 28,29 . Therefore, the generation of long donor templates by PCR may result in erroneous variants, which would present a major problem for future therapeutic applications.

Comparison of HDR efficiency with various MHC exchange donor templates. To avoid
PCR-based generation of donor templates, we investigated the use of alternative formats. Minicircles are a plasmid system that allows for selective removal of vector components, they rely on recombination of integrase AttB and AttP sites to remove bacterial gene elements prior to transfection in mammalian cells 24 . When compared to donor templates generated with standard plasmids, minicircles provide a smaller and less immunogenic format. We designed several minicircle donor templates and compared their exchange efficiency to our previously used PCR-generated exchange template (Fig. 5a and Supplementary Fig. S3b,c). Previous studies have suggested that linearized templates are more efficient for HDR integration when compared to circular templates 30,31 . Therefore, we created a "self-linearizing" minicircle exchange template, as it possessed the same gRNA sites (G10 and G13) present in the MHC locus, and thus this template would be linearized in situ (after transfection) by Cas9. We also included minicircles that were pre-linearized via an EcoRI restriction site. Equimolar quantities of each donor template were nucleofected into RAW264.7 cells along with the pX458-G10 and G-13 plasmids, and as before Cas9-expressing cells (via 2A-GFP) were sorted and expanded ( Supplementary Fig. S3d,e). Flow cytometry analysis was used to determine the fraction of MHC-reprogrammed cells based on H2-Kb-positive/H2-Kd-negative expression. After six-independent experiments, the self-linearizing minicircle template showed the highest fold improvement in MHC exchange compared to linear template (7.9 fold on average) (Fig. 5b). No significant differences in HDR efficiency were observed for the other donor templates. PCR analysis was performed on the bulk population and revealed correct integration in the MHC locus for all samples (Supplementary Fig. S3f). To confirm the reprogramming efficiency of the self-linearizing minicircle template, we performed exchange on a second H2-Kd-expressing cell line, CT26 fibroblasts. We had previously found this cell line to be recalcitrant to MHC exchange with linear PCR templates. However, when using the self-linearizing minicircle, we observed an easily detectable level of MHC exchange by flow cytometry (Fig. 5c), which was confirmed by genomic PCR (Fig. 5d). This level of exchange was comparable to that seen with RAW264.7 macrophages, suggesting that this approach is applicable to other cell types.

Discussion
Here we have demonstrated a proof-of-concept for reprogramming MHC-specificity in antigen-presenting cell lines using CRISPR/Cas-9 cassette exchange. In contrast to previous studies that simply knocked out MHC genes, we sought to improve MHC matching while retaining full immune functionality through the scarless replacement of MHC genes. The multiplexed capacity of Cas9 enabled the use of two gRNAs to excise the native MHC-I H2-Kd allele in RAW264.7 macrophages and replace it with a new H2-Kb allele via HDR. The surface expression of MHC proteins meant we could forgo the use of selection markers, thus MHC exchange was performed in a scarless manner, as we easily isolated cells displaying the new H2-Kb phenotype by FACS. Furthermore, the functionality of MHC-reprogrammed cells was verified through peptide antigen presentation and activation of cognate T cells. We subsequently investigated the design of the DNA donor templates and found the efficiency of exchange was significantly increased by using a self-linearizing minicircle template containing gRNA sites matching those targeted in the native H2-Kd locus. The newly designed donor template was even successful in facilitating MHC exchange in CT26 fibroblast cells, which we found previously to be resistant to exchange with linear template donors.
The downstream immune functionality of reprogrammed H2-Kb + MHC cells was validated, as they were able to engage and activate CD8 + T cells (B3Z cell line) in a manner indistinguishable from JAWSII cells, which are wildtype for H2-Kb. For future in vivo applications, biallelic scarless exchange would be preferential to monoallelic exchange combined with a knockout at the second allele. This would avoid frameshifts, which can result in the production of new out-of-frame MHC protein by-products. We tested for the presence of the original H2-Kd allele, correct integration of the new H2-Kb allele, and deletion at the locus. By this procedure we could confirm biallelic exchange was present in a majority of clones tested (3/4). As expected, cells with biallelic exchange showed higher MHC surface expression than monoalleic cells, thus suggesting phenotypic expression levels may offer a simple strategy for selection of cells with biallelic reprogramming. Upon genomic sequencing of the new H2-Kb alleles, we noted the introduction of several point mutations including some coding for amino acid changes. A split-pool PCR approach for sequencing allowed us to conclude that these mutations likely arose from the earlier PCR steps used to generate the linear donor template. To minimize the chance of mutation, we decided to test alternate donor template designs based on the minicircle technology. An important consideration when comparing plasmid-based generation versus PCR-based generation of donor templates is that the fidelity of bacterial replication is reportedly as low as 5 × 10 −10 errors per bp, which is over 500-fold lower than the highest fidelity polymerases 32,33 . Furthermore, minicircles offer the advantage over plasmids in having their bacterial plasmid backbone deleted, which results in lower immunogenicity (via removal of inflammatory CpG motifs) and a smaller DNA product that can be more efficiently delivered into cells 24 . The self-cleaving minicircle donor template, which contained gRNA sites matching those used to cleave the H2-Kb allele, was the most efficient design for exchange. We propose the higher efficiency stems from kinetic reasons; the reduced intracellular processing of the minicircle compared to the linear template should allow for more donor template to be present when Cas9 expression peaks after nucelofection 34 .
Due to their critical importance in allogeneic cellular transplantation, the highly diverse alleles of the MHC locus provide a therapeutically relevant target for immunogenomic editing. Currently, intensive screening is often required in order to find suitable donors with MHC alleles that match recipients, as compatibility of multiple MHC alleles (8/8 MHC alleles matched; HLA-A, -B, -C, -DRB1) correlates strongly with transplant acceptance and long-term survival 35,36 . Closely related donors are generally preferred as there is a higher chance of allelic matching, although due to haplotype inheritance, even siblings will only have a 25% chance of being a perfect MHC match. Therefore approximately 50% of cellular transplantations are performed with imperfectly matched MHC donors (6/8 or 7/8 alleles matched), which often results in much higher transplantation rejection and lower survival rates 36 . Traditionally, mammalian cells have been very difficult to modify due to low rates of HDR and the reduction of proliferative capacity following delivery of the nuclease components. However, the efficiency of genome editing in these cells has dramatically improved in recent years via the development of greatly improved delivery systems for nucleic acids, most notably through the recent use of adenovirus vectors 10,37 . Another important consideration is the possibility of adverse events (e.g. oncogenic mutations) occurring from Cas9 off-target activity. Several strategies have been developed to reduce the off-target activity of programmable nucleases, such as the use of truncated Cas9 gRNAs, fusion of catalytically inactive Cas9 to Fok1 nucleases domains, single-stranded genomic cuts with Cas9 nickases, and engineered high-fidelity versions of Cas9 [38][39][40][41][42] . In addition, the recent discovery of other programmable nucleases such as CRISPR-Cpf1 may also lead to strategies that enable high efficiency HDR with low off-target activity 43 . Moreover, the use of Cas9 as a ribonucleoprotein, rather than encoded by a plasmid, is becoming more widespread as a method to further improve the fidelity of editing 44 . All of these strategies for efficiency improvement could be readily adapted and incorporated into the approach described in this work but careful evaluation of transcriptome and proteome changes induced by replacement of the MHC locus will be required to demonstrate the specificity of the method. Likewise, while immortalized cell lines are sufficient for demonstrating proof-of-concept, our exchange methodology will need to be transitioned into primary cells for future progression towards clinical reality (e.g MHC exchanged murine bone derived macrophages co-cultured with SIINFEKL specific OT-1 primary CD8+ T cells) and subsequently towards more difficult to edit progenitor cell types such as stem cells. Careful design of in vivo models will also be essential to demonstrate tolerance of the engineered cells in transplantation. We hope that one day ex vivo MHC engineering of cells for transplantation could be very important in eliminating mismatches between hosts and donors or even applied to MHC exchange in induced pluripotent stem cell lines in order to generate cell banks for rare MHC genotypes 45,46 .