The CRISPR-Cas12 family nucleases, particularly the Cas12i subtypes, are considered promising alternatives to Cas9 for genome editing in plants. We previously developed a new Cas12i variant, CasY7, which has been successfully applied in clinical trials; its performance in plants remains to be investigated. Initial testing in stable transgenic maize and rice showed that the codon-optimized CasY7 (pCasY7e1) achieved average editing efficiencies of 58.7% and 62.3% across five target sites, respectively, outperforming the typical Cpf1 (pCpf1) control that targets the same sites. To further enhance activity, we fused T5 exonuclease to CasY7 (pCasY7e2), which shifted mutation profiles toward larger deletions, and subsequently integrated an MS2 aptamer into the crRNA scaffold (pCasY7e3). The optimized pCasY7e3 system increased editing efficiencies to 87.7% in maize and 82.9% in rice-approximately 2.7-fold higher than pCpf1. We further demonstrated multiplexed editing in maize, generating biallelic dwarf mutants, and validated functionality in hexaploid wheat with editing efficiencies up to 58.8%. Overall, our comprehensive validation across 942 transgenic plants confirmed robust editing in maize, rice, and wheat, establishing CasY7 as a high-efficiency addition to the CRISPR toolkit.

Self-incompatibility (SI) is an important plant mechanism that prevents inbreeding depression by recognizing and rejecting self-pollen, thereby promoting outcrossing. However, SI can also act as a barrier in breeding programs, presenting significant challenges to breeders. Passion fruit (Passiflora edulis), a tropical fruit species of substantial economic importance, also serves as a valuable system for investigating SI mechanisms within the Passifloraceae. Nevertheless, the molecular basis of SI in passion fruit has not yet been elucidated. In this study, we investigated the SI system in yellow passion fruit (P. edulis f. flavicarpa) and employed transcriptomic analysis to examine the time-course transcriptional responses following different pollination treatments. Transcriptomic analysis revealed distinct gene expression dynamics under different pollination treatments: self-pollinated samples exhibited stronger and earlier transcriptional changes, whereas the number of differentially expressed genes (DEGs) in cross-pollinated samples was relatively lower. Numerous pathways previously associated with sporophytic self-incompatibility (SSI) were enriched in the stigma samples after self-pollination. Reactive oxygen species (ROS) are crucial signaling molecules involved in pollen germination and pollen tube growth during SI responses. Our results showed that ROS-related pathways were enriched in stigma tissues after self-pollination. In addition, oxidative stress-related responses were detected in the style shortly after self-pollination, suggesting that plastid-associated or general oxidative stress processes may also be involved, although the precise source of ROS requires further validation. FERONIA, ROP9, and ARC1 are key genes related to the SI system in Brassica. In the passion fruit SI response, the expression levels of these genes increased in the style, indicating a spatial expression pattern different from that reported in classical Brassicaceae SSI systems. Together with cytological observations showing that self-pollen rejection occurs at the stigma surface, our results suggest that yellow passion fruit may employ an SSI-like regulatory framework while exhibiting a lineage-specific spatial deployment of SI-related regulators. Overall, this study provides new transcriptomic insights into the SI mechanism of yellow passion fruit, establishes a molecular framework for understanding SI in P. edulis f. flavicarpa, and offers novel insights into the diversity of plant SI systems.

AI for scientific discovery is entering an agentic era, where protein-engineering systems are expected to prioritize future wet-lab experiments rather than merely fit static measurements. We introduce TadA-Bench, a million-variant wet-lab replay benchmark from 31 TadA directed-evolution rounds for future-round discovery toward agentic protein engineering. TadA-Bench preserves the campaign chronology and defines a fixed-data replay task: given earlier experimental rounds, models rank variants that appear only in later rounds. It provides aligned DNA, RNA, and protein views, and uses Seq2Graph, a graph-based label-unification pipeline, to reconcile noisy enrichment measurements into consistent cross-round activity labels. Random-split controls show strong interpolation, but future-round ranking and finite-budget candidate selection are much weaker. Controlled analyses suggest that evolutionary coverage is more informative than local data density, positioning TadA-Bench as a reproducible wet-lab replay substrate for future-round discovery toward agentic protein engineering; the data and code are released on Hugging Face and GitHub.

Heterozygous familial hypercholesterolemia is a common genetic disorder characterized by lifelong elevation of serum low-density lipoprotein cholesterol (LDL-C) and premature atherosclerotic cardiovascular disease. YOLT-101 is an investigational in vivo gene therapy that uses adenine base-editing technology, delivered via GalNAc-modified lipid nanoparticles to inactivate PCSK9 and achieve sustained LDL-C reduction. Here we report interim results from an ongoing clinical trial evaluating primary (safety and tolerability) and secondary (lowering of PCSK9 and LDL-C levels) outcomes of a single intravenous dose of YOLT-101 in adults with heterozygous familial hypercholesterolemia and uncontrolled LDL-C. Six participants (three men and three women) received escalating doses of YOLT-101 (0.2, 0.4 or 0.6 mg kg-1). No grade ≥3 adverse events occurred. Transient and self-limited infusion-related reactions and elevations in liver enzymes were the most common adverse events. A single infusion of YOLT-101 induced dose-dependent and durable reductions in circulating PCSK9 and LDL-C, with sustained reductions of 74.4% and 52.3%, respectively, at 24 weeks in the 0.6 mg kg-1 cohort (n = 3), demonstrating promise for future clinical development. ClinicalTrials.gov registration: NCT06458010 .

Existing benchmarks for biological language models (BLMs) inadequately capture the challenges of real-world applications, often lacking realistic out-of-distribution (OOD) scenarios, evolutionary depth, and consistency in measurement. To address this, we introduce TadABench-1M, a new benchmark based on a wet-lab dataset of over one million variants of the therapeutically relevant TadA enzyme, purpose-built to embody these three essential attributes. Generated across 31 rounds of wet-lab evolution, it offers unparalleled evolutionary depth and naturally presents a stringent OOD challenge. To ensure measurement consistency across this extensive campaign, we developed Seq2Graph, a scalable graph-based algorithm that systematically unifies multi-batch experimental data. Our high-fidelity benchmark highlights a critical finding: while state-of-the-art BLMs excel on a standard random split of the data (Spearman’s ρ ≈ 0.8), they fail dramatically on a realistic temporal prediction task (ρ ≈ 0.1). This stark performance gap validates the importance of our benchmark’s design principles and suggests that evolutionary depth is critical for building models with realistic utility.

Sequence-specific gene knockdown technologies are crucial for fundamental research and therapeutic applications. RNA interference and CRISPR interference, while extensively utilized for gene expression manipulation, face limitations due to their ectopic or transient expression. In this study, we developed a generalizable and efficient method to downregulate gene expression in human 293T cells by introducing de novo upstream ATGs (uATGs) of genes using CRISPR-Cas9-mediated genome editing. Through CRISPR library screening, in-depth sequencing, and flow cytometry analysis, we validated that the introduction of uATGs served as an effective method to suppress protein expression. Our findings further revealed that this strategy can be tailored to diminish endogenous gene expression in tumor cells without affecting the mRNA transcription levels. Importantly, by introducing a uATG into the 5′, untranslated region (UTR) of the Uox gene, we successfully established a Uox-knockdown (KD) mouse model of hyperuricemia associated with metabolic disorders. This model demonstrated hyperuricemia, with serum uric acid levels that exceeded 400 µmol L−1, along with renal dysfunction, as indicated by elevated serum creatinine and blood urea nitrogen levels. Examination of the kidneys from 8-week-old Uox-KD mice revealed abnormal histopathological characteristics, including partial dilation of Bowman’s capsules and renal tubules, focal nephron collapse and necrosis, and lymphocytic infiltration. In addition, the mice exhibited lipid and glucose metabolism disorders, all while maintaining a normal lifespan. This spontaneous hyperuricemia model has potential as a valuable tool for long-term studies on hyperuricemia and gout. Taken together, we present an efficient approach for the constant suppression of specific gene expression in mammalian cells and the development of a Uox-KD mouse model of hyperuricemia via CRISPR-Cas9-mediated uATG introduction. This offers broad implications for fundamental research and therapeutic applications.

Precise transgene-free gene upregulation remains a challenge in crop biotechnology, as conventional enhancers often exceed CRISPR-mediated knock-in size constraints and face regulatory hurdles. Here we establish a foundational cross-species resource of compact transcriptional enhancers developed via STEM-seq, a high-throughput screening platform that systematically evaluated 81 475 genomic elements across maize, wheat, tomato, and soybean. This screen identified 6904 natural short transcriptional enhancers (STEs; 60-80 bp) exhibiting a broad range of activation efficiencies, with the most potent elements derived from wheat (up to 46.3-fold activation). Augmenting this resource, we developed BaseSearch, an AI-driven design framework, which computationally generated 5000 synthetic STE candidates and achieved a 9.1% success rate (11.4× higher than genome-wide screening). This set included ten ultra-potent enhancers outperforming natural counterparts by 2.27-fold (64.5-fold vs. 28.4-fold activation). Notably, the compact size of these STEs aligns with regulatory frameworks that favor endogenous sequence lengths, offering potential pathways for policy-compatible precision breeding. This integrated platform provides a substantial collection of functionally validated enhancers for crops, supplying the research community with immediately applicable elements for engineering agronomic traits while advancing the fundamental understanding of plant cis-regulation.

Plant core promoters (PCPs) are key genetic elements that control gene expression and have significant value for crop breeding and plant synthetic biology. Natural promoters (NPs) are constrained by their limited diversity and narrow activity range, and it remains unclear whether synthetic promoters (SPs) can transcend these natural constraints. Here, we present TargetGAN, a deep-learning framework trained on 76,851 NPs that integrates a generative adversarial network (GAN) with a pre-trained activity predictor to enable the de novo design of PCPs with user-defined activity. We used TargetGAN to generate 55,296 SPs and selected 5,250 for high-throughput functional validation using STARR-seq. Of these, 2,909 were successfully characterized, with a moderate correlation (Pearson correlation coefficient = 0.6435) between predicted and experimental activity. Surprisingly, 29 SPs exhibited ultra-high activity, exceeding the maximum activity of the tested NPs. Further orthogonal validation using luciferase reporter assays showed a strong positive correlation with STARR-seq measurements across a broad dynamic range. Notably, the most active synthetic candidate, SP1482, significantly outperformed the strongest tested NP, the UBI core promoter, achieving a 128-fold increase in expression relative to the 35S minimal promoter. Interpretable motif analysis suggested that ultra-high-activity promoter design can be achieved through the precise arrangement of strong activating motifs. These results demonstrate that TargetGAN is a robust and generalizable framework for the targeted generation of PCPs tailored to user-defined activity levels and will be a powerful tool both for precise gene regulation in plant systems and for overexpression analysis in genetic engineering and synthetic biology.

Ex vivo autologous haematopoietic stem cell (HSC) gene therapy provides a promising treatment option for haematological disorders. However, current methods involve complex processes and chemotherapeutic conditioning, leading to limited accessibility for treatment and major side effects. Here we develop antibody-free targeted lipid nanoparticles (LNPs) for mRNA delivery to HSCs in vivo, enabling efficient base editing of the γ-globin gene (HBG1/2) promoter target in human HSCs to reactivate fetal haemoglobin in derived erythroid cells. Delivery of ABE8e/sgRNA mRNA with optimized LNPs achieves efficient in vivo base editing of HBG1/2 in transfusion-dependent β-thalassaemia (TDT) patient-derived HSCs engrafted in immunodeficient NCG-X mice, showing restored globin chain balance in erythroid cells. Our research indicates that using LNPs for genome editor delivery achieves efficient editing of endogenous genes of human HSCs. This non-viral delivery system eliminates the need for collecting or mobilizing HSCs, providing a potent and one-time treatment potential for blood disorders such as sickle cell disease and TDT. A lipid nanoparticle is used to deliver a base editor in vivo to haematopoietic stem cells in mice.

Artificial intelligence (AI) has revolutionized the protein engineering process from multiple aspects, including representing protein information, generating protein designs, and evaluating protein properties. This review aims to introduce the recent progress of AI in protein research. We first introduce how AI models represent protein sequences, structures, and other properties. Further, the applications of generative models in protein design are introduced. The use of predictive models and smart agents in the evaluation process is then discussed, including high-precision protein property simulation and wet lab experimental design. Additionally, we discuss the future development of AI in protein research and the potential challenges it may encounter.

Heterozygous familial hypercholesterolemia (HeFH) is a genetic disorder characterized by persistently elevated low-density lipoprotein cholesterol (LDL-C) levels, leading to an increased risk of early-onset atherosclerosis cardiovascular diseases (ASCVD). YOLT-101, an in vivo base-editing therapeutic agent delivered via GalNAc-modified lipid nanoparticles, is designed to achieve permanent inactivation of proprotein convertase subtilisin/kexin type 9 ( PCSK9 ), enabling sustained LDL-C reduction. This trial enrolled participants with heterozygous genetic mutations in the low-density lipoprotein receptor (LDLR), and LDL-C levels of ≥2.6 mmol/L (without ASCVD) or ≥1.8 mmol/L (with ASCVD) despite receiving moderate- or high-intensity statin therapy. Eligible patients received a single intravenous infusion of YOLT-101 at ascending doses (0.2, 0.4, and 0.6 mg/kg). We report interim results from an ongoing clinical trial evaluating the safety, tolerability, pharmacodynamics, and efficacy of YOLT-101. Six participants were enrolled (median age, 48 years, range, 34-62) in the study. The most common adverse events (AEs) were transient infusion-related reactions (83.3%) and elevations in alanine/aspartate aminotransferase (50%). No study withdrawals or AEs of grade 3 or higher occurred. PCSK9 and LDL-C levels decreased in a dose-dependent manner following YOLT-101 administration. In the 0.6mg/kg group (n=3), mean PCSK9 levels decreased by 55.9% at week 1 and by 75.8% and 72.5% after 1 and 4 months, respectively; corresponding LDL-C reductions were 33.2%, 48.9%, and 50.4%, respectively. A single infusion of YOLT-101 at 0.6 mg/kg was well tolerated and led to sustained PCSK9 and LDL-C reduction, demonstrating promise for future clinical development. (Funded by YolTech Therapeutics; Registration Number: NCT06458010 )

Base editing allows for the precise modification of genetic information, providing new avenues for treating diseases1. The adenine base editor ABE8e is currently the most efficient and widely used tool for adenine base editing. ABE8e was developed through multiple rounds of directed evolution of Escherichia coli tRNA adenine deaminase, including phage-assisted continuous evolution (PACE)2,3. While PACE is highly effective, it is a complex system that poses challenges for implementation4. Despite its high efficiency, ABE8e is associated with limitations such as relatively higher bystander editing effects and elevated off-target activity5, which need to be addressed to further enhance its precision and safety. Here, we developed a novel method for the de novo discovery of evolved ABE components, particularly adenine deaminases. This process involves identifying candidate proteins through AI-based structural prediction and clustering, followed by the enhancement of deaminase editing activity through screening libraries created by sequential amino acid saturation mutagenesis. This evolutionary strategy simplifies the approach by employing saturation mutagenesis libraries tailored to specific segments, thereby enabling exploration of an expanded sequence space and increasing the likelihood of discovering adenine deaminases with superior capabilities. The newly developed hpABE5.20 here demonstrates a more refined editing window, reduced DNA off-target effects that are both sgRNA-dependent and -independent, and minimized RNA off-target activity, while maintaining robust editing efficiency relative to ABE8e. Furthermore, hpABE5.20 has been successfully applied for precise and effective therapeutic adenine base editing in cellular disease models, humanized mice, and non-human primates. ### Competing Interest Statement H.Z., C.L., and Z.W. have filed and been granted patent applications related to this work. Z.W., Y.L. and Y.W. are co-founders of YolTech Therapeutics, Shanghai, China. The other authors declare no competing interests. This work was supported by the National Key R&D Program of China 2023YFC3403401 & 2024YFA1803301 (Y.W.), the Shanghai Agricultural Science and Technology Innovation Program K2023001 (Y.L.), the National Natural Science Foundation of China 82270125 (Y.W.), 32300667 (J.L.), 32371535 (S.C.), 82450107(D.W.), the project of Shanghai Municipal Science and Technology Commission 23HC1400400 (Y.W.), the National Program for Support of Top-Notch Young Professionals (Y.W.).

The spatiotemporal knockdown of genes through genome editing heralds a new frontier in molecular breeding,yet it remains largely unexplored.Recognizing the intricate regulatory networks of endogenous microRNAs(miRNAs),we posited that integration of specific miRNA target sequences into the 3'untrans-lated region(UTR)of a gene could construct artificial miRNA-dependent regulatory circuits,facilitating pre-cise spatiotemporal gene suppression.To test this hypothesis,we selected three endogenous miRNAs with unique expression profiles by analyzing rice miRNA expression profiles.Results from both transient assays and stably-edited rice plants confirmed that in-locus incorporation of miRNA targets into the 3'UTR of target genes can substantially reduce their expression in a spatiotemporal manner.Using GID1 as a target gene,we found that knockin of the miR156a target led to a remarkable 97％constitutive reduc-tion;knockin of the tissue-specifically expressed miR396c target significantly reduced its expression in shoots alone;and knock-in of the long-day-induced miR528 target triggered a dramatic and temporal decrease of 95％specifically under such light exposure.These findings underscore the viability of miRNA-mediated,in-locus knockdown(MiRKD)as a convenient approach for crop breeding,leveraging miRNA expression traits and genome editing for conditional gene suppression.

Gene silencing is crucial in crop breeding for desired trait development. RNA interference (RNAi) has been used widely but is limited by ectopic expression of transgenes and genetic instability. Introducing an upstream start codon (uATG) into the 5 ' untranslated region (5 ' UTR) of a target gene may 'silence' the target gene by inhibiting protein translation from the primary start codon (pATG). Here, we report an efficient gene silencing method by introducing a tailor-designed uATG-containing element (ATGE) into the 5 ' UTR of genes in plants, occupying the original start site to act as a new pATG. Using base editing to introduce new uATGs failed to silence two of the tested three rice genes, indicating complex regulatory mechanisms. Precisely inserting an ATGE adjacent to pATG achieved significant target protein downregulation. Through extensive optimization, we demonstrated this strategy substantially and consistently downregulated target protein expression. By designing a bidirectional multifunctional ATGE4, we enabled tunable knockdown from 19% to 89% and observed expected phenotypes. Introducing ATGE into Waxy, which regulates starch synthesis, generated grains with lower amylose, revealing the value for crop breeding. Together, we have developed a programmable and robust method to knock down gene expression in plants, with potential for biological mechanism exploration and crop enhancement.

Gene upregulation through genome editing is important for plant research and breeding. Targeted insertion of short transcriptional enhancers (STEs) into gene promoters may offer a universal solution akin to transgene-mediated overexpression, while avoiding the drawbacks associated with transgenesis. Here, we introduce an "in-locus activation" technique in rice that leverages specifically screened STEs for refined, heritable, and multiplexed gene upregulation. To address the scarcity of potent enhancers, we developed a large-scale mining approach and discovered a suite of STEs capable of enhancing gene expression in rice protoplasts. The in-locus integration of these STEs into eight rice genes resulted in substantial transcriptional enhancements, with up to 869.1-fold increases in the edited plants. Employing a variety of STEs, we achieved delicate control of gene expression, enabling the fine-tuning of key phenotypic traits such as plant height. Our approach also enabled efficient multiplexed gene upregulation, with up to four genes simultaneously activated, significantly enhancing the nicotinamide mononucleotide (NMN) metabolic pathway. Importantly, heritability studies from the T0 to T3 generations confirmed the stable and heritable nature of STE-driven gene activation. Coupled with our STE-mining technique, in-locus activation holds great promise to make gene upregulation a major application of genome editing in plant research and breeding.

Understanding the behavior of endogenous proteins is crucial for functional genomics, yet their dynamic characterization in plants presents substantial challenges. Whereas mammalian studies have leveraged in locus tagging with the luminescent HiBiT peptide and genome editing for rapid quantification of native proteins, this approach remains unexplored in plants. Here, we introduce the in locus HiBiT tagging of rice proteins and demonstrate its feasibility in plants. We found that although traditional HiBiT blotting works in rice, it failed to detect two of the three tagged proteins, a result attributable to low luminescence activity in plants. To overcome this limitation, we engaged in extensive optimization, culminating in a new luciferin substrate coupled with a refined reaction protocol that enhanced luminescence up to 6.9 fold. This innovation led to the development of TagBIT (tagging with HiBiT), a robust method for high-sensitivity protein characterization in plants. Our application of TagBIT to seven rice genes illustrates its versatility on endogenous proteins, enabling antibody-free protein blotting, real-time protein quantification via luminescence, in situ visualization using a cross-breeding strategy, and effective immunoprecipitation for analysis of protein interactions. The heritable nature of this system, confirmed across T1 to T3 generations, positions TagBIT as a powerful tool for protein study in plant biology.

With the widespread use of clustered regularly interspaced palindromic repeats (CRISPR)/CRISPR-associated nuclease (Cas) technologies in plants, large-scale genome editing is increasingly needed. Here, we developed a geminivirus-mediated surrogate system, called Wheat Dwarf Virus-Gate (WDV-surrogate), to facilitate high-throughput genome editing. WDV-Gate has two parts: one is the recipient callus from a transgenic rice line expressing Cas9 and a mutated hygromycin-resistant gene (HygM) for surrogate selection; the other is a WDV-based construct expressing two single guide RNAs (sgRNAs) targeting HygM and a gene of interest, respectively. We evaluated WDV-Gate on six rice loci by producing a total of 874 T-0 plants. Compared with the conventional method, the WDV-Gate system, which was characterized by a transient and high level of sgRNA expression, significantly increased editing frequency (66.8% vs. 90.1%), plantlet regeneration efficiency (2.31-fold increase), and numbers of homozygous-edited plants (36.3% vs. 70.7%). Large-scale editing using pooled sgRNAs targeting the SLR1 gene resulted in a high editing frequency of 94.4%, further demonstrating its feasibility. We also tested WDV-Gate on sequence knock-in for protein tagging. By co-delivering a chemically modified donor DNA with the WDV-Gate plasmid, 3xFLAG peptides were successfully fused to three loci with an efficiency of up to 13%. Thus, by combining transiently expressed sgRNAs and a surrogate selection system, WDV-Gate could be useful for high-throughput gene knock-out and sequence knock-in.

Base editors, including cytosine and adenine base editors (CBE and ABE), are promising tools for precise genome modification. They enable the generation of single nucleotide variants in plants for research and crop improvement (Li et al., 2020; Manghwar et al., 2019; Ren et al., 2021; Xu et al., 2021; Zeng et al., 2022). However, existing base editors are still limited in the types of base conversions they can induce. Recently, a new base editor was constructed by fusing an engineered N-methylpurine DNA glycosylase (MPG) with ABE to create AYBE. This has achieved efficient A-to-T and A-to-C (A-to-Y, AYBE) transversions in mammalian cells and also timely assessed in rice to induce A-to-T (AKBE) (Li et al., 2023; Tong et al., 2023; Wu et al., 2023). However, the editing activity of AYBE remains unexplored in maize, and its editing efficiency leaves room for further optimization. Here, by fusing the adenine base editor with a codon-optimized N-methylpurine DNA glycosylase (MPG) and co-expressing the maize translesion synthesis DNA polymerase η (Polη), we developed an optimized AYBE base editor (ZmAYBEv3) for both A-to-T and A-to-C base conversions with high efficiency in maize and other monocots plants. First, the human-derived MPG (hMPG) was engineered (G163R, N169S, S198A, K202A, G203A, S206A and K210A) and codon-optimized for maize (MzMPG), then fused to the C-terminus of the maize ABE editor ZmABE8e to construct the initial AYBE editor ZmAYBEv1 (Figure 1a). Two sgRNAs (sgRNA1 and sgRNA2) targeting maize genes ZmGA20ox3 and ZmCT2 were designed. Hundreds of young embryos from the inbred maize variety KN5585 were transformed with Agrobacterium for evaluation. Approximately 50 regenerated shoots from each transformation were pooled and genotyped using the next-generation sequencing (NGS). As expected, only A-to-G substitutions were detected in samples edited with the conventional ZmABE8e, while A-to-Y conversions were found in ZmAYBEv1 edited samples (Figure 1b). For example, at the A8 site of sgRNA1, the A-to-T and A-to-C conversion frequencies were 3.86% and 0.53%, respectively, demonstrating the A-to-Y editing activity of ZmAYBEv1. We then tested it in maize plants. A total of 45 T0 plants were obtained and genotyped by NGS (Liu et al., 2019). The results showed seven T0 plants contained A-to-Y substitutions, further demonstrating ZmAYBEv1's editing capability in plantlet (Figure 1c; Table S1). However, the chimerism state of A-to-Y substitutions (calculated from the proportion of NGS reads, Li et al., 2023) was too low in most mutants. Usually, T0 plants with a chimerism>10% are required to ensure heritability. Thus, only one mutant could be identified as a valid A-to-T edited line, and no A-to-C editing lines were found. This revealed the need for further improvement of ZmAYBEv1. Polη is involved in the replication of damaged DNA and may improve base editing efficiency (Tong et al., 2023). Accordingly, human and maize Polη (hPolη and ZmPolη) were incorporated into ZmAYBEv1 to construct ZmAYBEv2 and ZmAYBEv3, respectively (Figure 1a). Quick tests in maize embryos showed a significant increase in A-to-T and A-to-C editing efficiencies when using ZmAYBEv3 (Figure 1b), indicating positive regulation of ZmPolη on AYBE. We think that different base conversion types between pAYBEv2 and pAYBEv3 might be caused by the different enzymatic activity of hPolη and ZmPolη (Figure S1). To assess them in transgenic plants, 39 and 52 T0 plants were generated using ZmAYBEv2 and ZmAYBEv3, respectively, targeting the same two genes above (Figure 1c). As expected, substantially more A-to-Y edited plants (7 out of 23 with chimerism>10%, the same hereinafter) were identified in ZmAYBEv3 edited lines for ZmGA20ox3. At sgRNA2 of ZmCT2, an uneditable site for ZmAYBEv1 or ZmAYBEv2, an A-to-T editing plant was successfully obtained using ZmAYBEv3. Moreover, three A-to-C edited lines were also identified. To confirm the editing results, we then resequenced the ZmAYBEv3-derived lines by Sanger sequencing and further confirmed these results (Figure 1d; Figure S2). Notably, we also found that homozygous lines could be generated in T0 plants. The homozygous A-to-T editing at the sgRNA1 (A8) of ZmGA20ox3 produced a premature stop codon (AAG to TAG), resulting in a semi-dwarf phenotype of maize, even in T0 generation (Figure 1e,f). These results indicate ZmAYBEv3 has the highest editing efficiency, capable of both A-to-T and A-to-C editing. To further confirm the versatility of ZmAYBEv3, we targeted three additional maize genes (ZmLW2, ZmABH2 and ZmLBD5) for editing. We regenerated 51 T0 maize plants and performed NGS genotyping. The results showed successful A-to-Y editing by ZmAYBEv3 at all three genes, with an average efficiency of 35.3% (18/51). Notably, the A-to-T editing frequency reached 45.5% at the ZmLBD5 locus. Given the known transferability of base editors across monocot species, we also tested ZmAYBEv3 in rice on three genes (OskTN80b, OsWaxy and OsTB1). NGS and Sanger sequencing showed ZmAYBEv3 could efficiently induce A-to-Y editing at the three rice genes with an average efficiency of 21.1% (12/57) (Figure 1c,d; Figure S3). In some locus, the A8 site within a sgRNA seems the best targeting nucleotide (Figure 1g). Together, these results further validate the editing activity of ZmAYBEv3 in both maize and other monocot species. Collectively, the incorporation of ZmPolη enhanced the A-to-Y editing efficiency of ZmAYBEv3. Across five target sites in 103 T0 maize plants, 50 plants had A-to-Y conversions (chimerism >1%), validating its capabilities. Notably, 26 plants (25.2%) showed potentially heritable edits (chimerism>10%). ZmAYBEv3 also enables the possibility of obtaining homozygous edits within the T0 generation. The high editing efficiencies achieved by ZmAYBEv3 in maize and rice highlight its usefulness as an alternative tool to supplement existing base and prime editors for functional studies and trait improvement in crops. Supported by the National Key R&D Program of China (No. 2021YFD1201300) and the National Natural Science Foundation of China (No. 32070396) to Y.L. We thank WIMI for assistance with maize transformation. D.Z. and Y.L. designed the research; D.Z, H.P., K.L., Y.Z., F.Z., L.Y., S.R., Q.D. and J.X. performed experiments; D.Z. and Y.L. wrote and revised the manuscript. The authors declare no competing interests. The data that supports the findings of this study are available in the supplementary material of this article. Data S1 Supplemental methods. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.

As a powerful genome editing technology, CRISPR/Cas is revolutionizing both fundamental research and crop breeding, and has now evolved into large-scale editing tools that are efficient, simple, and programmable. With such CRISPR screening technologies, the numbers of genome-edited crops are rapidly increasing. Here, we describe the general workflow of a CRISPR screen in plants, including the selection of appropriate editors, genome-wide guide RNA design, pooled library construction, massive transformation, and high-throughput genotyping. We also discuss applications for the screening of candidate genes, the optimization of spatiotemporal expression, the evolution of protein activities, and the establishment of genome-wide libraries of knockout mutant. After considering the current challenges and limitations, we finally envision a virus-mediated strategy to improve CRISPR screens.