Dear Editor,
The emergence of CRISPR/Cas9-based genome editing technology has energized many applied biotechnology fields, including livestock breeding (
Zhao et al., 2019). Given that the most naturally occurring beneficial variants in farm animals are regulatory alleles (
Georges et al., 2019), practical livestock editing should mostly involve knocking-in of the natural variants, or of designed elements to regulate key genes. However, earlier CRISPR/Cas9-dependent strategies for installation of knock-in alleles showed limitations of low efficiencies and purities (
Anzalone et al., 2020).
An important breakthrough toward precise genome editing was marked by the development of prime editor (PE), which presented a highly innovative design for the installation of small-sized edits (
Anzalone et al., 2019). The canonical PE protein features a fusion of Cas9 nickase (nCas9, H840A) moiety and a reverse transcriptase (RTase) domain. A prime editing guide RNA (pegRNA) is comprised of a conventional sgRNA module (for initially directing a nick at the target site) and an additional 3′ sequence extension (for subsequently programming synthesis of edits at the nick by reverse transcription (RT)). The resultant intermediate 3′-flap structure may be subsequently resolved by the cellular repair pathways for the eventual installation of edits (
Anzalone et al., 2019). Empowered by such a double-strand DNA break (DSB)-independent and RT-dependent mode of action, PE has exhibited high editing versatility and genomic safety in diverse application contexts (
Chen and Liu, 2023). On the other hand, PE’s editing efficiencies have often been sub-optimal and inconsistent, presenting a technical challenge to inspire continuous enhancement efforts (
Chen and Liu, 2023). An important sub-class of such efforts centered on the development of nuclease PE (PEn, based on the nuclease-active Cas9, Fig. S1) that could increase the rates of desirable edits at some sites (
Li et al., 2023). To upgrade PEn toward more efficient installation of precise edits, we recently adopted an accessory protein regulator of DSB repair [a ubiquitin variant, also called i53 (
Canny et al., 2018)] to establish a high-activity uPEn platform (
Li et al., 2023). The potent activities of uPEn presented exciting opportunities, despite that it remained liable for a fraction of DSB-derived editing impurities (
Li et al., 2023). We envisioned that the cutting-edge PE/PEn tools could be suited for the installation of defined, beneficial genome edits in livestock.
PPARG (encoding PPARγ, a lipogenic transcription factor) represents an attractive target for gain-of-function editing, due to its association with intramuscular fat-dependent meat quality traits (
Liu et al., 2020). For proof-of-principle, the present study would aim to precisely edit
PPARG in Hu sheep, a major Chinese indigenous sheep breed with a high reproduction rate and commendable meat production (
Yue, 1996).
We first adapted the canonical PE protein component into a very recent “PEmax” architecture (with several elements of optimization over the initial PE (
Chen and Liu, 2023)). Furthermore, several split forms of the PEmax were constructed similarly as in a previous study (
Liu et al., 2022) (Fig. S2A). The basic split design featured separate nCas9 and RTase parts. In another variant split form, the non-essential RNase-H domain was further removed from the RTase [ΔRH] (
Chen and Liu, 2023). In parallel, a GCN4 tag (1×) and the corresponding single-chain fragment variable (scFv) binder were respectively placed on the nCas9 and RTase (with or
w/
o the RH) moieties to potentiate their interactions (
Liu et al., 2022). The activities of various PEmax forms were tested in human HEK293T cells for disabling a premature TAG stop codon within a reporter construct (Fig. S2A). Note that all pegRNA constructs in the present study were assembled with an enhancement design of 3′ xrRNA-joining, in accordance with our previous work (
Zhang et al., 2022). The results of reporter editing showed that the full-length PEmax and the split/tagged form with the RTaseΔRH (PEmax-SPLΔRH-tag) were overall the top performers, followed by the SPL-tag and SPLΔRH forms with slightly lower activities (Fig. S2B and S2C), while the modest differences in activity patterns of different PE forms were apparently not attributed to their expression levels (determined by immunoblotting of nCas9, Fig. S2D). We also determined the efficiencies of different PEmax formats for editing three previously tested genomic loci in HEK293T cells (
Zhang et al., 2022). Herein, the locus-dependent activities of the four different split PE forms were comparable with each other, while they were modestly lower than those of the full-length form (Fig. S2E). These results suggested the possibility of exploiting the editors in either the full-length or split format for later testing.
We next considered to design the modification at
PPARG. The
PPARG locus encodes two overlapping PPARγ isoforms with distinct promoters and translational start sites (i.e., broadly expressed γ1 and adipose-specific γ2 isoforms) (
Hu et al., 2022) (Fig. S3A). Therefore, editing at γ2-specific portion of
PPARG would selectively impact the adipocytes, and conceivably avoid a global effect. Interestingly, a closer examination of the sequences revealed that the Kozak motifs preceding mouse
Pparg (γ2) and sheep
PPARG (γ2) start codons deviate substantially from the consensus sequence of “GCCACC” (
Kozak, 1987) (Fig. 1A). Thus, a knock-in allele with a favorable Kozak motif preceding the start codon of
PPARG (γ2, mouse, and sheep) could potentially specify higher PPARγ2 expression. To this end, an initial trial was designed for precise Kozak motif replacement in mouse cells. Note that the suitable break position is located relatively distant (30 bp) from the Kozak sequence at the mouse target (an issue similarly present at the sheep target) (Fig. S3B), which is likely to undermine PE efficiencies (
Anzalone et al., 2019). Indeed, although the guide RNA target site was susceptible to cleavage by Cas9/sgRNA (Fig. S4A), we found that PEmax in combination with the pegRNA for replacement editing (“pegRNA-M-replc”, Table S1) failed to induce desirable sequence changes in mouse N2a cells (Fig. S4B and S4C).
Alternatively, based on the positions of available “NGG” protospacer adjacent motifs (PAMs), we next considered insertion (not replacement) of an optimal Kozak directly upstream of ATG, which could be programmed via small sequence insertion after the + 1 or + 2 position in reference to the
↓TATGG or
↓TTATGG DNA break positions (Fig. 1B). A similar design could also be applied for the sheep site (see later section, Table S1). Moreover, with a mouse
Pparg (γ2) upstream sequence-dependent reporter assay, we validated that the GCCACC-inserted construct drove higher reporter expression (Fig. S5A–C). Besides rational allele design to accommodate prime editing, we also chose to exploit our recently developed high-activity uPEn platform (
Li et al., 2023). According to our earlier reporter- and genomic loci-editing characterizations on the formats of PE (Fig. S2), either the full-length uPEn or its split/tagged form with the RTaseΔRH (uPEn-SPLΔRH-tag) was applied in further experiments (Fig. S6A). Different combinations of two uPEn formats and two pegRNAs (respectively, corresponding to cleavages at
↓TATGG and
↓TTATGG [pegRNA-M-Ins#1 and #2], see Fig. 1B) were co-transfected into mouse N2a cells. The Sanger sequencing and next-generation sequencing (NGS, primes listed in Table S2) results clearly indicated the occurrence of Kozak motif insertion in all experimental groups (10%–17% in all NGS reads), despite the concomitant presence of editing impurities (Figs. 1C, 1D and S6B). With both the full-length and SPLΔRH-tag uPEn, pegRNA-M-Ins#1 showed a trend of being relatively more active between the two pegRNAs for introducing accurate edits and indels (Fig. 1D).
Given the positive results in cultured cells, we next proceeded to test the knock-in of the consensus Kozak motif at Pparg (γ2) in mouse zygotes. The uPEn-SPLΔRH-tag format was exploited. The mRNAs for Cas9-GCN4 and scFv-RTaseΔRH, and the pegRNA component (pegRNA-M-Ins#1) were, respectively, prepared by in vitro transcription (IVT, see Table S3 for primer information). The C57/BL6JGpt 1-cell stage embryos (100) were microinjected with the editor mix (97.5 ng/μL of Cas9-GCN4 and 52.5 ng/μL of scFv-RTaseΔRH mRNA, and 50 ng/μL of pegRNA) in the cytoplasm, and were subsequently transferred to foster mothers. Out of 49 mice born, four showed clear modifications at the target site, upon initial genotyping of the toe clips (Fig. S7A). Targeted NGS of samples from the clearly modified mice assigned two F0 mice (#161 and #184) as founders with accurate knock-in modifications (desirable allele representing ~14% and 33% of total reads, respectively) (Figs. 1E and S7B). The #184 mouse was shown to also bear a high rate of indel allele (46%), and only a low percentage of WT allele (18.9%). The other two F0 mice presented either an imprecisely inserted allele (#158, ~46%) or a non-specific indel (#183, ~25%) (Figs. 1E and S7B). The edited F0 mice were subsequently bred with the WT mice to produce the F1 generation cohorts, in which all mutant allele types demonstrated evident germline transmissions (Figs. 1F and S7C). To understand the apparent lack of transmission of the WT allele by the #184 mouse (in > 15 progenies), we harvested the tissues from testis and several other internal organs from the aged F0 for subsequent NGS analyses. The result suggested that cells bearing a WT allele were comparatively under-represented in the germ cell compartment (Fig. S7D), indicated by a group-low, 4% allele frequency in the testis sample (8.4%–27% in other tissues).
Further crossing of the heterozygous F1 mice (derived from the #161 founder) led to the delivery of homozygous knock-in progenies (Fig. S7E). A homogeneous population of adipocytes was prepared via differentiation of the stromal vascular fraction cells from the mice (
Fu et al., 2023). Comparisons between the control and differentiated cells regardless of the genotypes confirmed that PPARγ (both γ1 and γ2 isoforms) protein levels were markedly upregulated under the adipogenic condition (Fig. 1G). Importantly, the cells (differentiated) with Kozak knock-in at
Pparg (γ2) showed higher normalized protein expression for PPARγ2, but not PPARγ1, than their WT counterparts (Fig. S7F). This apparently led to a greater PPARγ2/γ1 ratio in the knock-in cells (Fig. 1H). In contrast, there were no genotype-dependent differences at the levels of
Pparg γ1/γ2 mRNAs (Fig. 1I). Although an immediate survey of common PPARγ target genes did not show evident differences between the WT and knock-in cells (Fig. S7G), future examinations on potential changes driven selectively by PPARγ2 are warranted (
Hu et al., 2022).
The above success in the mouse trial encouraged us to further apply uPEn to editing the Hu sheep zygotes. Programmed multi-locus edits in sheep would be a desirable strategy for balanced improvements of both meat production and meat quality traits (
Zhao et al., 2019). Accordingly, we considered applying a mix of sgRNA/pegRNA to simultaneously program knockout at
MSTN (
Zhao et al., 2019) and knock-in at
PPARG (
Liu et al., 2020) (Fig. 2A). To this end, a pegRNA was designed for Kozak motif insertion directly upstream of ATG at sheep
PPARG [γ2] locus (Fig. S8A), similar to the mouse site-targeting strategy (Fig. 1B). To ensure efficient knockout of
MSTN, two sgRNAs targeting the same essential exon (exon 2) were designed (Fig. S8B). The trial was also set to use different formats of uPEn (full-length and SPLΔRH-tag forms).
The pegRNA, sgRNAs, and the mRNAs for two uPEn formats were prepared by IVT. The fertilized eggs of Hu sheep were microinjected (in cytoplasm) with the RNA mix (100 ng/μL of uPEn (or 65 ng/μL of Cas9-GCN4 and 35 ng/μL of scFv-RTaseΔRH) mRNA, 50 ng/μL of pegRNA, and 25 ng/μL of mixed sgRNAs), followed by embryo transfer. Five months later, a total of 15 F0 lambs were born (Fig. 2B). Interestingly, one of the F0 lambs presented obvious macroglossia (enlarged tongue) (Fig. 2C), a phenotype likely indicative of muscle hyperplasia related to
MSTN editing. DNA samples from all newborns were subjected to targeted NGS at the
PPARG and
MSTN sites (Fig. S9A and S9B). For the uPEn group (
n = 7) and the uPEn-SPLΔRH-tag group (
n = 8), a major portion (4 and 4) of the newborns from each group presented accurate
PPARG knock-in alleles (allele frequencies > 1%) (Figs. 2D, 2E and S9A). In the corresponding positive newborns of the two groups (H1, H3, H4, H9 and H2, H5, H10, H11, respectively), the average frequencies of the knock-in alleles were 33% and 62%, respectively. Among these knock-in-positive founders, one (H2) also presented an evident level (~10%) of imperfect prime edit, and two others (H3 and H9) harbored undesirable indels at the
PPARG site (Figs. 2D and S9A), the latter apparently attributed to the Cas9 nuclease module in uPEn. Moreover, at the
MSTN locus, 3 newborns from the uPEn group and 6 from the uPEn-SPLΔRH-tag group were shown to bear knockout edits (1 more in the uPEn group featuring in-frame deletions) (Figs. 2D, 2E and S9B). In total, an impressive number of F0 newborns (together 6 out of 15) presented desirable two-locus knock-in/knockout modifications (2 from uPEn group: 1 unfortunately deceased after birth, and 4 from uPEn-SPLΔRH-tag group) (Fig. 2E). Although the present proof-of-principle experiment is not sufficient for quantitative comparisons between the full-length and the SPLΔRH-tag uPEn formats on zygote editing, our results have evidently demonstrated two active forms of uPEn that might accommodate different application needs. Besides the on-target editing assessments, several potential genomic off-target sites for the pegRNA (
PPARG editing, see Tables S4 and S5 for sequence and primer information) and the sgRNA (corresponding to the more active sgRNA2 against
MSTN, see Fig. S8) were amplified using the samples from the edited founders, and those from three lambs presenting no on-target modifications (H7, H8, and H13). The comparisons of the NGS reads from edited and non-edited animals showed minimal off-target editing at the sites tested (Fig. S10), although further comprehensive analyses would be encouraged. In the meantime, we envision that future employment of high-fidelity Cas9 variants (reviewed in (
Anzalone et al., 2020)) in uPEn may formally improve its genetic safety.
Collectively, extending on our recent development of an upgraded PE nuclease (uPEn) (
Li et al., 2023), the present study has further demonstrated its effectiveness in installing small knock-in edits in mice, and on enabling multiplexed knock-in/knockout edits in Hu sheep zygotes. The validated improvement of PPARγ2 expression in cells from the Kozak element-edited mice has suggested our editing strategy as a widely applicable approach to program target gene expression in animals. Moreover, the observed high rate of simultaneous
PPARG/
MSTN knock-in/knockout edits in Hu sheep (despite certain evidence of stochastic, single-site editing) has highlighted uPEn’s potential for multi-target applications toward rapid and balanced livestock genetic improvements (
Georges et al., 2019). In our two-site-targeting experiment, the
MSTN-editing efficiencies by uPEn/sgRNA [10/15] duly served as a high-end benchmark to underscore the strong performances of uPEn/pegRNA-dependent Kozak insertion at
PPARG [8/15 achieved] despite the expected activity restraints on pegRNA-driven editing owing to mechanistic complexities. One other technical aspect for future application of uPEn is whether chemically synthesized/modified pegRNA or IVT pegRNA (with 3′ stabilizing RNA structure) would be the preferred format. As an effort for direct comparison, we requested commercial synthesis (with HPLC purification) of chemically modified pegRNA (cm-pegRNA, featuring 2′-OMe and phosphorothioate modifications and regular length) and of unmodified pegRNA with a 3′-xrRNA motif (equivalent to an IVT xr-pegRNA), against the mouse
Pparg site for insertion of Kozak motif (Fig. S11A). In two mouse cell lines (LLC and N2a cells), while co-transfection of the cm-pegRNA with plasmid-borne uPEn induced visibly greater levels of accurate edits (than the group with xr-pegRNA), it also undesirably caused higher rates of indels, which eventually translated into sometimes lower edit/indel ratios (Fig. S11B–D). These results suggested that the cm-pegRNA did not necessarily outperform xr-pegRNA (with uPEn) in term of refining the ratio of precise PE/impurities. With the additional consideration of the current high cost of obtaining long, chemically synthesized RNA, our results support the use of IVT, motif-stabilized pegRNA for routine uPEn applications, unless in a context where the absolute editing efficiency is prioritized over editing purity. On another note, while our work was in progress, we noticed a recent study that reported high-efficiency adaptation of nickase-based PE for single nucleotide substitutions in mouse embryos (PE-embyro) (
Kim-Yip et al., 2024). Intriguingly, compared to its potencies for point mutations, the PE-embryo method installed larger edits (3–8 bp insertions) with significantly lower efficiencies (
Kim-Yip et al., 2024). Therefore, it would be instrumental in future studies to systematically compare uPEn and PE-embryo tools for installing various edit types in animals. Overall, our work strongly supports uPEn as a broadly useful and readily applicable editing platform for livestock genetic improvements and beyond.
The Author(s) 2025. Published by Oxford University Press on behalf of Higher Education Press.
PDF
(8575KB)
