T cell-specific inactivation of mouse CD2 by CRISPR/Cas9

The CRISPR/Cas9 system can be used to mutate target sequences by introduction of double-strand breaks followed by imprecise repair. To test its use for conditional gene editing we generated mice transgenic for CD4 promoter-driven Cas9 combined with guide RNA targeting CD2. We found that within CD4+ and CD8+ lymphocytes from lymph nodes and spleen 1% and 0.6% were not expressing CD2, respectively. T cells lacking CD2 carryied mutations, which confirmed that Cas9 driven by cell-type specific promoters can edit genes in the mouse and may thus allow targeted studies of gene function in vivo.

Genome modification in the mouse has become an essential research tool for addressing gene function in vivo. Since 1987, targeted mutagenesis in the mouse has been performed by manipulating embryonic stem (ES) cells in vitro 1 . Chimeric mice that were partially derived from these ES cells were used to find new, manipulated mouse lines. As an extension to this technology, recombinases such as Cre are used to avoid the use of resistance genes, to allow embryonic-lethal mutations and to make inducible, cell-type specific knockouts and knock-ins [2][3][4] . While, recombinase-mediated conditional gene targeting has drastically widened the scope of the technology, it still relies on ES cells for introduction of the recombinase targets such as loxP sites. Further, mouse lines carrying a loxP-containing allele need to be intercrossed with a Cre transgene to address target gene function. Altogether, the minimal time for generation of such conditional mutants is more than a year. Recently, a paradigm shift in germline mutagenesis has taken place through DNA editing by designer nucleases. Zinc finger nucleases, Tal effector nucleases, and the CRISPR/Cas9 system have facilitated the generation of gene-modified animals without the use of ES cells [5][6][7][8][9] . These designer nucleases can be tailored to introduce single-or double-strand breaks into target genes, which are repaired through cell-intrinsic repair mechanisms. Non-homologous end joining-mediated repair leads to insertions or deletions 10 and homologous recombination in the presence of a donor fragment facilitates introduction of defined mutations 11 . In the mouse, designer nuclease-mediated mutagenesis is achieved through injection of nuclease-encoding plasmids, RNA or protein into zygotes followed by screening of the founder animals 5,6 . Recently, the Clustered Regularly Interspaced Short Palindromic Repeats (CRISPR) in combination with CRISPR associated gene 9 (Cas9) system has surpassed the other designer nucleases in terms of efficiency and flexibility. It has proven to allow both insertions/deletions and specific targeted mutagenesis in the germline of the mouse 8,12-14 and in human cells 14,15 . The original CRISPR/Cas9 system from Streptococcus pyogenes 16 was modified for application in molecular biology and now relies on a single 102 bp long guide RNA (gRNA), of which 20 bp determine the target sequence 17 . Since, the exact choice of gRNA is subject to further restrictions, such as presence of the Protospacer Adjacent Motif (PAM) sequence (NGG) and uniqueness within the target genome, online tools have been developed for designing functionally competent gRNAs 18,19 . The CRSIPR/Cas9 system has been used to modify genes in the germline by transient expression in the zygote and was found to facilitate somatic mutations in the mouse when expression of the system was under control of the Cre/loxP system and gRNA was delivered through viruses or transfection 20 . Only recently, it was shown in mice that doxycycline-induced Cas9 allows conditional in vivo gene editing 21 . In Drosophila melanogaster, Cas9 was shown to enable efficient cell-specific gene editing after its placement under a tissue-specific promoter 22 . However, direct gRNA/Cas9-mediated conditional gene editing using cell-type specific promoters similar to conditional mutagenesis by Cre has been so far not reported for the mouse (Fig. 1a). Such an approach would allow single-step analysis of gene function within a cell lineage circumventing tedious gene targeting and crossing necessary for Cre/loxP-mediated conditional gene ablation.
To test whether gRNA/Cas9 could drive cell-type specific mutagenesis in the mouse, we placed Cas9 under control of a CD4 promoter (CD4dsCas9), thus initiating the expression in T lymphocytes during their maturation in the thymus (Fig. 1b) 23 . In a second construct, we placed a gRNA expression cassette under control of a U6 promoter (U6gRNA(CD2)). The gRNA was directed against a target sequence in exon 2 of CD2 (Fig. 1b), an easily detectable surface marker found on all T cells. In addition, we chose this target because its deficiency does not influence cell viability 24 and does not lead to changes in the cell's phenotype, that could impair this proof of concept study. The two constructs were co-injected into the pronuclei of oocytes of FVB/N mice and offspring were screened for presence of the transgenes. Two transgenic founders carrying both constructs were identified (Fig. 1c). We also obtained two founders carrying solely the Cas9 and two carrying solely the gRNA construct (data not shown). Further analysis of peripheral blood by flow cytometry revealed that in one of the double-transgenic founders a fraction of CD4 + (0.4%) and CD8 + (0.5%) cells lacked expression of CD2 (Fig. 2). No such populations were found in wildtype controls (data not shown), the other double-transgenic (Suppl. Fig. 1) or Cas9 and gRNA single-transgenic founders (Fig. 2). Furthermore, we analyzed pooled lymph nodes and spleen cells of two double transgenic F1 mice as well as of two wildtype littermates to verify the initial blood analysis. Here we observed that 1% of CD4 + and 0.6% of CD8 + T cells were not expressing CD2 anymore (Fig. 3). A transgene copy number analysis by real-time PCR was performed for this transgenic line and indicated numbers of integrated Cas9 being 26 and 41 and of the gRNA of 12 and 6/7 (Fig. 4a). In addition, CD4-Cas9 mRNA expression was found solely in T cells and thymocytes but not B cells of double transgenic (dTG) mice or wildtype thymocytes (Fig. 4b). U6-driven gRNA(CD2.0) expression was detected in total spleen and thymus of dTGs but not in WT controls (Fig. 4c). To confirm that the CD2 locus was edited by gRNA/Cas9, we single cell-sorted T cells lacking CD2 and sequenced a fragment around the target site after single cell PCR amplification and cloning (Fig. 5a). We observed that eight out of the nine sequenced CD2 − CD4 + T cells carried a mutation in the amplicon (Fig. 5b). In CD2 − CD8 + T cells two of the three amplicons were wildtype and one mutated (Fig. 5b). In all amplicons, we only found a single sequence, either mutated or wildtype. This may be either result of a technical bias leading to amplification of only one allele or alternatively an outcome of repair of the second allele by homologous recombination using an already mutated donor allele. In all 17 amplicons from CD2 + T cells as well as nine amplicons from CD19 + B cells, we obtained solely unmutated sequences (Fig. 5b). As an additional control, we performed single cell PCR on FVB/N wildtype CD4 + and CD8 + T cells and sequenced the PCR products, again yielding solely wildtype sequences (Fig. 5b). Sequence analysis of the mutations found within the amplicons from  CD2 − T cells revealed deletions spanning up to 19 bp and insertions of one to four base pairs, five were found as out of frame and one in frame (Fig. 5c). Two mutations were repeatedly found (Fig. 5c).
Even though we observed CD2 − T cells in a gRNA/Cas9 double-transgenic situation, we were surprised that the number of these cells was extremely low. This may have been a specific result of this particular founder due to expression variegation or may constitute a general problem of the approach. We therefore analyzed the T cell compartments in our different transgenic mice in greater detail. The ratio of CD4 + TCRβ + and CD8 + TCRβ + T to CD19 + B cells was found to be slightly increased in double-transgenic (8 (CD4 + TCRβ + ) and 2.2 (CD8 + TCRβ + )) versus Cas9 single (4.8 (CD4 + TCRβ + ) and 1.4 (CD8 + TCRβ + )) and gRNA transgenic animals (4.9 (CD4 + TCRβ + ) and 1.8 (CD8 + TCRβ + )), thus ruling out a general counter-selection of CD4 + and CD8 + T cells. Since expression of nucleases such as Cre recombinase were reported to lead to apoptosis in situations of high proliferative activity 25 and also for Cas9 toxic in vivo effects were described 22 , we analyzed CD4 + and CD8 + T cells for cell death (Suppl. Fig. 2a,b). We obtained no evidence of an increased number of dead cells in lymphocytes of the gRNA/Cas9 double-transgenic situation compared to Cas9 or gRNA single transgenic mice as well as wildtype mice. Thus, in our model we cannot find evidence for genotoxicity of gRNA and Cas9. Another important point regarding the use of CRISPR/Cas9 for in vivo and ex vivo use are off-target effects. Whereas some groups report high specificity of CRISPR/Cas9 26,27 others show high frequency of off-target effects 28,29 . Several groups found that off-target effects can be reduced by the use of optimized gRNAs 28,30,31 . In our case off-target effects were not analyzed because our on-target frequency of modification was already low 32 .
Taken together, our data indicate that gRNA/Cas9 can be used to mutate target genes in specific cell lineages in the mouse. The frequency of such mutations in our system was low. We excluded that this was result of the actual gene deficiency by using a target whose absence was shown to have no effect over T cell survival 24 . Also, we did not find evidence that nuclease activity would increase cell death and thereby remove the edited cells from the population, which could explain the low frequency of mutations. One reason for inefficient abrogation of CD2 expression may, of course, be a feature of this particular transgenic founder. Defined and characterized Cas9-driver lines that are crossed to gRNA lines may overcome such problems in the future. To address this point we analyzed gRNA expression and Cas9 expression on mRNA and protein level. gRNA was expressed in both spleen and thymus. However, while we clearly detected Cas9 mRNA, we failed to see expression of the protein (0.25 ng Cas9 protein detection limit, data not shown). Thus, low protein expression levels in our particular founder line may contribute to the low efficiency of the system. To further address whether or not the efficiency of the used gRNA contributed to the observed low mutation frequencies we compared by in vitro digestion of a PCR product the used gRNA (gRNA(CD2.0) to three other gRNAs (gRNA(CD2.1,CD2.2,CD2.3)) targeting the same region (Fig. 6). We found that gRNA(CD2.1) showed almost no cutting and gRNA(CD2.2) complete digest. The gRNA(CD2.0) used for generating the transgenic mice presented with good but not complete cutting efficiency, along with gRNA(CD2.3) (Fig. 6). It thus appears unlikely that the low in vivo efficiency of CD4-Cas9/ gRNA(CD2.0) is sole result of suboptimal gRNA performance. Methylation of target loci as factor contributing to efficiency of Cas9-mediated genome editing has been also discussed and assessed, though with contradictory results 18,32 , but chromatin structure in general appears to influence Cas9 binding 32,33 . Interestingly, Cre-induced Cas9 expression from within the gt(ROSA)26Sor locus combined with viral gRNA expression also resulted in only infrequent gene editing in 0.1% to 1.8% of lung cells 20 while a recent report of doxycycline-regulated Cas9-mediated gene editing showed significantly higher frequencies of mutated cells 21 . gRNA library transduction experiments in EL4 cells, a mouse thymic cell line, was found to also lead to only 0.94-3.60% mutant cells when analyzing different surface markers 34 . Furthermore, we were not able to efficiently mutate CD2 in T cell transfection experiments (data not shown). While certainly the evidence is still anecdotal, T cells may be difficult to modify by CRISPR/Cas9. One way to increase the frequency of gene editing events leading to inactivation of gene functionality may be the use of multiple gRNAs towards the same target. Yet, even in such an approach only 2/3 of alleles will carry an out-of-frame mutation. CRISPR/Cas9-mediated in vivo targeted mutagenesis may thus work best when a critical region, such as an enzymatic site, is functionally compromised even through in-frame in/del mutations. Nevertheless, also in the here-described simple version conditional DNA editing is extremely useful. It allows rapid assessment of cell-type specific phenotypic changes in a time frame incomparable to classical conditional mutagenesis, especially when mutant cells can be easily identified for phenotypic analysis. Additionally, for the investigation of onco-and tumor suppressor genes the observed low editing frequency may prove to be advantageous. Taken together, this proof of concept study shows that direct conditional gene editing by gRNA/Cas9 transgenesis may extend the toolbox for genetic studies in the mouse.

Methods
Methods were carried out in accordance with the approved guidelines.