Clustered regularly interspaced short palindromic repeats (CRISPR)-Cas systems can be engineered as programmable transcription factors to either activate (CRISPRa) or inhibit transcription. Apomixis is extremely valuable for the seed industry in breeding clonal seeds with pure genetic backgrounds. We report here a CRISPR/dCas9-based toolkit equipped with dCas9-VP64 and MS2-p65-HSF1 effectors that may specifically target genes with high activation capability. We explored the application of in vivo CRISPRa targeting of maize BABY BOOM2 (ZmBBM2), acting as a fertilization checkpoint, as a means to engineer parthenogenesis. We detected ZmBBM2 transcripts only in egg cells but not in other maternal gametic cells. Activation of ZmBBM2 in egg cells in vivo caused maternal cell-autonomous parthenogenesis to produce haploid seeds. Our work provides a highly specific gene-activation CRISPRa technology for target cells and verifies its application for parthenogenesis induction in maize.

Summary: The integration of haploid induction and genome editing, termed HI‐Edit/IMGE, is a promising tool for generating targeted mutations for crop breeding. However, the technical components and stacking suitable for the maize seed industry have yet to be fully characterised and tested. Here, we developed and assessed three HI‐Edit/IMGE maize lines: Edit^Wx, Edit^Sh, and Edit^Wx&Sh, using the haploid inducer CHOI3 and lines engineered using the CRISPR‐Cas9 system targeting the Waxy1 (Wx1) and Shrunken2 (Sh2) genes. We meticulously characterised the HI‐Edit/IMGE systems, focusing on copy numbers and the mutant alleles mtl and dmp, which facilitate haploid induction. Using B73 and six other parental lines of major commercial varieties as recipients, HI‐Edit/IMGE demonstrated maternal haploid induction efficiencies ranging from 8.55% to 20.89% and targeted mutation rates between 0.38% and 1.46%. Comprehensive assessment verified the haploid identification, target gene editing accuracy, genome background integrity, and related agronomic traits. Notably, Edit^Wx&Sh successfully combined distinct CRISPR‐Cas9 systems to induce multiple desired mutations, highlighting the potential of HI‐Edit/IMGE in accelerating the integration of edited traits into commercial maize varieties. Our findings underscore the importance of meticulous Cas9 copy number characterisation and highlight potential challenges related to somatic chimerism. We also validated the performance of single‐cross haploids derived using the HI‐Edit/IMGE process. Our results confirm the industrial applicability of generating targeted mutations through pollination and provide critical insights for further optimising this technology.

Base editing technologies can improve crops, but their efficiency in maize remains suboptimal. This study attempts to overcome these limitations by examining optimized cytosine and adenine base editors (CBEs and ABEs), namely evoAPOBEC1, evoFERNY, evoCDA1, TadA8.20, and TadA8e, for precise genome editing in transient and stable expression maize cells. Employing a seed fluorescence reporter (SFR) system for rapid screening of BE transformants and transgene‐free progenies, we enhanced editing efficiencies and heritability. Notably, TadA8.20 and evoCDA1 attained multiplexed editing efficiencies of up to 100.0% and 79.0% at the tested loci, respectively, with some homozygous and bi‐allelic mutants exceeding 72.4% and 73.7%. Precise editing of ZmACC1/2 (acetyl‐CoA carboxylase) improved herbicide resistance, with ZmACC2 mutants displaying improved performance. This study advances crop genetic engineering by facilitating robust, multi‐locus modifications without altered agronomic performance, enhancing herbicide tolerance in maize. The successful utilization of these BE is a significant step forward in agricultural biotechnology and precision breeding.

The Manipulated Genic Male Sterile Maintainer (MGM) system, a next-generation hybrid seed technology, enables efficient production of sortable seeds from genic male sterile (GMS) lines. However, implementing robust MGM systems in commercial maize inbred lines requires stable transformation, a genotype-specific and laborious process. This study aimed to integrate MGM technology into the commercial maize inbred line Z372, developing both GMS and MGM lines. We utilized the MGM line ZC01-3A-7, which contains the MS26ΔE5 editor T-DNA and MGM T-DNA, previously established in the highly transformable ZC01 recipient plants. Through a combination of crossing and backcrossing with Z372, we targeted the fertility gene Ms26 within the Z372 genome for mutation using the in vivo CRISPR/Cas9 activity within the MS26ΔE5 editor T-DNA construct. This approach facilitated precise editing of the Ms26 locus, minimizing linkage drag associated with the Ms26 mutation. Whole-genome SNP analysis achieved a 98.74% recovery rate for GMS and 96.32% for MGM in the BC2F2 generation. Importantly, the Z372-GMS line with the ms26ΔE5 mutation is non-transgenic, avoiding linkage drag and demonstrating production readiness. This study represents a significant advancement in maize breeding, enabling the rapid generation of GMS and MGM lines for efficient hybrid seed production.

The convergence of genome editing (GE) and artificial intelligence (AI) is shifting crop improvement from empirical optimization toward predictive design. In this Perspective, we propose that GE and AI are linked by a reciprocal innovation cycle. AI improves GE by enabling more accurate guide RNA design, editing outcome and off-target prediction, and data-driven development of next-generation editing systems. In turn, GE provides a powerful experimental platform to validate AI predictions, dissect causal genotype-to-phenotype relationships, and generate high-resolution perturbation datasets for iterative model refinement. This bidirectional interplay is beginning to accelerate the engineering of complex agronomic traits, including trait stacking, metabolic rewiring, climate resilience, stress tolerance, nutritional enhancement, and de novo domestication. Here, we argue that this synergy is emerging most clearly in predictive editing design and iterative model validation, although robust breeding-scale evidence remains limited. We also discuss emerging opportunities for AI-guided closed-loop editing platforms and generative biological design, together with key bottlenecks, including limited training data, biological complexity across scales, model interpretability, and heterogeneous regulatory landscapes. We argue that integrating AI with GE can establish a practical framework for predictive crop design and enable more efficient and sustainable crop engineering.