Molecular Regulation of Maize Plant Height and Its Implications for Breeding Maize Tolerant to Enhanced Planting Density

Abstract

The continuous increase in maize yield has become increasingly dependent on high-density planting, rendering plant height (PH) and plant architecture core traits that balance light interception efficiency and lodging resistance. Although many PH regulators have been well characterized, we are still confused on how to handle these regulators for maize breeding. At the same time, biotechnological advances have provided us with tools to address this issue. This review presents perspectives on integrating molecular information underlying maize PH regulation with these biotechnological tools to improve high-density tolerance in maize. We first compile an inventory of genes associated with maize PH and classify them into hormone-related and non-hormonal pathways, and reveal that loss-of-function mutations in most of these genes exhibit prominent allelic characteristics, including extreme dwarfism and pleiotropy. We discuss the underlying mechanisms of pleiotropy, which is important for successful editing aimed at achieving moderate dwarfism. Then, we discuss current strategies for high-density tolerance improvement, including genome editing strategies, introgression of dominance genes, multi-gene stacking, and phenotypic prediction, and offer key insights for their practical implementation. We offer insights and perspectives for future high-density tolerance improvement, which requires cell-resolution gene regulatory networks, AI-enabled precision genome editing, and phenotypic prediction models incorporating all environmental factors influencing plant growth. This review aims to provide guidance for researchers and breeders in translating molecular knowledge into improved high-density tolerance, lodging resistance, and yield gain.

Keywords

maize / plant height / high-density tolerance / molecular mechanism / genome editing

1 INTRODUCTION

The global population is projected to reach nearly 10 billion by 2050, demanding a remarkable increase in food production (Assefa et al., 2018). As a global staple crop, maize is expected to contribute substantially to meeting this ever-growing demand. A key to unlocking future yield gains lies in optimizing plant architecture for higher planting densities, a principle that was central to the Green Revolution's success in wheat and rice (Peng et al., 1999; Sasaki et al., 2002). This principle has been pursued by the global maize research community (Duvick, 2005). Increasing density deteriorates within-canopy light quality and intensifies competition among plants, triggering shade-avoidance responses (SAR) that promote internode elongation and increase lodging risk (Wang et al., 2016). Plant height (PH) is a pivotal trait correlated with both canopy structure and mechanical stability. Accordingly, an important breeding goal is semi-dwarf, density-adapted plant architecture that increases yield potential at the unit area level.

Many PH genes have been cloned in maize, and most of them are related to gibberellin (GA), auxin, and brassinosteroid (BR). Examples include the GA biosynthesis genes anther ear1 (An1) (Bensen et al., 1995) and DWARF1 (D1) (Chen et al., 2014), brachytic2 (Br2), which modulates auxin transport (Multani et al., 2003), and the BR biosynthesis gene brassinosteroid-deficient dwarf1 (BRD1) (Makarevitch et al., 2012). However, the deployment of these genes is often limited by (i) unfavorable pleiotropy, (ii) genetic-background dependence, (iii) frequent recessive inheritance, and (iv) linkage with undesirable loci. This creates a fundamental challenge for molecular breeders: How to handle a catalog of PH genes to design high-density tolerant maize ideotypes? The conventional approach to address this issue is “from find to validate to deploy” based on one or a few genes. With the advancement of biotechnology, we are now moving to a new approach of “from design to stack to test”. This new approach requires us to: (i) deeply understand the gene regulatory networks (GRNs) that govern PH in concert with other agronomic traits, (ii) utilize genome editing to create tunable alleles that minimize negative pleiotropy, and (iii) integrate these modified alleles into multi-gene stacks whose performance in target environments can be predicted by advanced models.

In this review, we firstly retrospect the regulators of maize PH, and explore the mechanistic origins of pleiotropy. Then, we discuss the current strategies for developing semi-dwarf maize materials and varieties with enhanced high-density tolerance. Finally, we provide our opinions on how to accelerate the breeding process of density-tolerant semi-dwarf maize with the aid of biotechnological tools. The insights and perspectives synthesized in this review may serve as a valuable reference for maize geneticists and breeders alike.

2 Genetic Pathways Underlying Dwarfism in Maize

Let us begin with the current understanding of the regulatory mechanisms underlying PH in maize. PH is shaped by coordinated regulation of cell division, expansion, and differentiation. In maize, most cloned dwarfing genes fall into hormone-associated pathways, while some act through non-hormonal networks (Fig. 1; Table 1).

2.1 GA pathway

2.1.1 GA biosynthesis

GA biosynthesis starts with a cyclization step that converts geranylgeranyl diphosphate (GGPP) to ent-copalyl diphosphate (CPP), which is subsequently converted to ent-kaurene; ent-copalyl diphosphate synthase (CPS) catalyzes the first step (from GGPP to CPP), while ent-kaurene synthase (KS) catalyzes the second step (from CPP to ent-kaurene). Both CPS and KS belong to the diterpene synthase family, a group of cyclases that catalyze the formation of diterpenoid compounds. In maize, An1 encodes CPS, and loss-of-function An1 causes GA-responsive dwarfism accompanied by developmental abnormalities (Bensen et al., 1995). However, an1 mutants still accumulate ~20% of wild-type ent-kaurene, implying there are additional CPS/KS-like activities. The redundancy of CPS/KS-like activities is supported by the identification of specific KS family members in maize. Kaurene synthase-like 3 (ZmKSL3), mapped to the classic GA-deficient dwarf-5 (d5) region, directly links diterpene synthase to PH regulation (Fu et al., 2016).

The next stage in GA biosynthesis is the early oxidation process. The classic Dwarf3 (D3) gene encodes a cytochrome P450, and acts immediately downstream of the CPS/KS-mediated cyclization steps. It catalyzes the early oxidation of ent-kaurene to form ent-kaurenol, and subsequently ent-kaurenal, which are essential intermediates for the synthesis of bioactive GAs. Similar to the CPS/KS-mediated cyclization step, defects in this oxidation step exert a strong effect on PH but often have pleiotropic effects (Winkler and Helentjaris, 1995), as blocking this oxidation step prevents the formation of bioactive GAs.

Following the early oxidation step, the GA biosynthesis pathway proceeds to the late activation stage, catalyzed by GA20-oxidases (GA20ox) and GA3-oxidases (GA3ox). GA20ox catalyzes the sequential oxidation of C₂₀-GA precursors to form C₁₉-GA intermediates, while GA3ox (Chen et al., 2014) further oxidizes these intermediates to produce bioactive GAs that directly regulate PH and broader developmental processes (Yamaguchi, 2008). Targeted suppression of ZmGA20ox3 and ZmGA20ox5 in vegetative tissues produces short-stature plants with improved standability while maintaining reproductive GA levels (Paciorek et al., 2022). The GA20ox family in maize comprises multiple members with distinct expression patterns and functional roles. ZmGA20ox1, ZmGA20ox3, and ZmGA20ox5 are predominantly expressed in vegetative tissues, whereas ZmGA20ox2 and ZmGA20ox4 show higher expression in reproductive structures (Paciorek et al., 2022; Zhang et al., 2020). The multi-member structure also provides functional redundancy, buffering against complete loss-of-function mutations that would otherwise cause severe dwarfism.

2.1.2 GA catabolism and transport

In maize, GA2-oxidases (GA2ox) are key gibberellin deactivation enzymes that help maintain GA homeostasis by converting bioactive GAs and their immediate precursors into inactive forms. Dosage-based tests have shown that overexpression of a GA2ox7.3 gene in maize caused dwarf plants and shortened internodes (Deng et al., 2024). GA and its intermediates can move between organs and compartments (Binenbaum et al., 2018). Mechanistic support for transporter-mediated routing has been demonstrated in Arabidopsis, including NPF (Nitrate transporter 1/Peptide transporter family) transporters mediating GA translocation and vacuolar hormone transport (Binenbaum et al., 2018). In maize, functional characterization of transport components remains limited.

2.1.3 GA signaling

GA is perceived by soluble nuclear receptors of the GA insensitive dwarf 1 (GID1) family. In rice, bioactive GA binds GID1, and the GA–GID1 complex then promotes interaction with DELLA repressors, triggering DELLA ubiquitination and proteasomal degradation and thereby releasing GA-responsive growth programs (Ueguchi-Tanaka et al., 2005). However, maize GID1 alleles are less commonly reported in the public literature. DELLA proteins repress GA responses, and stabilized DELLAs lead to GA-insensitive dwarfism. The maize genome contains two DELLA proteins, DWARF8 (D8) and DWARF9 (D9). The classical d8 and d9 mutants were among the first dominant dwarfing genes characterized in maize, exhibiting moderate to severe GA-insensitive dwarfism (Harberd and Freeling, 1989). The transcription factor ZmNAC32 interacts with D8 and protects it from degradation by masking the ubiquitination site (Peng et al., 2026). This stabilization of D8 suppresses the expression of growth-related genes, thereby inhibiting internode elongation. Meanwhile, miR164e negatively regulates this process by cleaving ZmNAC32 transcripts, forming a “miR164e–NAC32–DELLA” regulatory module that precisely controls internode elongation and PH.

Beyond their role as repressors of GA signaling, DELLA proteins also function as central hubs that integrate multiple pathways. DELLAs physically interact with jasmonate–ZIM–domain (JAZ) proteins to coordinate growth with defense responses (Qi et al., 2014), repress auxin signaling by interacting with auxin response factor 6 (ARF6) (Ben-Targem et al., 2021), inhibit BR responses by binding to the transcription factor BZR1 (Li et al., 2012), integrate ABA signaling by upregulating the ABA transporter AIT1.1 (Shohat et al., 2020), and bind to phytochrome-interacting factors (PIFs) such as PIF3 and PIF4 to couple GA signaling with light perception (Li et al., 2016). In maize, the JA co-receptor F-box protein COI1 can directly trigger proteasome-dependent degradation of DELLA proteins (Feiz et al., 2024). In the coi1 quadruple mutant, both D8 and D9 accumulate to high levels, leading to shortened internodes and reduced photosynthetic efficiency. These findings highlight DELLAs as valuable precise editing targets for optimizing plant architecture and stress tolerance in maize breeding.

2.2 BR pathway

The classical BR-deficient dwarf mutant in maize, nana plant1 (na1), is caused by disruption of a DET2-like (de-etiolated 2) steroid 5α-reductase acting in early BR biosynthesis, displaying reduced stature and altered leaf morphology. Na1 supplies upstream substrates and metabolic flux (Hartwig et al., 2011). BR-deficient dwarf1 (BRD1) encodes a BR C-6 oxidase and catalyzes the terminal oxidation and activation step of late BR biosynthesis, converting intermediates into bioactive BRs. The null allele brd1-m1 causes extreme dwarfism with stacked leaves and severe reproductive defects (Makarevitch et al., 2012). Brassinosteroid (BR)-mediated regulation of plant architecture is also evidenced by the UPA1 and UPA2 loci, which jointly control maize leaf angle (Tian et al., 2019). UPA2 involves a polymorphism upstream of ZmRAVL1, which encodes a B3-domain transcription factor, and influences endogenous brassinosteroid content and leaf angle by regulating BRD1 (underlies UPA1). The beneficial UPA2 allele from teosinte boosts yield under dense planting. The lac1 mutant displays a "smart canopy" with upright upper leaves and flat lower leaves (Tian et al., 2024). LAC1 encodes a brassinosteroid C-22 hydroxylase responsible for BR biosynthesis, and its disruption reduces BR specifically in upper leaves via a RAVL1-mediated shade response. Field trials show that lac1 increases yield at high densities, with further gains when combined with the teosinte UPA2 allele.

Arabidopsis studies established the core BR signaling module that BR is perceived by the plasma-membrane receptor kinase BRI1 (BR insensitive1), which functions with co-receptors (often from the BAK1/SERK family). BR signaling inhibits the GSK3-like (glycogen synthase kinase3) kinase BIN2 (BR insensitive2), in part via BSU1-type (BRI1 suppressor1) phosphatases, allowing the accumulation and activation of BZR1/BZR2 ( BR resistant1/2) transcription factors (Wang et al., 2006). In maize, RNAi knockdown of ZmBRI1 homologs produces BR-insensitive phenotype characterized by shortened internodes, dark-green and upright/twisted leaves, reduced auricle formation, and altered patterns of cell division and elongation (Kir et al., 2015), and maize BIN2-like GSK3 kinases (ZmSK2: SHAGGY-like kinase 2) show strong developmental defects when perturbed (Wang et al., 2022). GRAS42 (GRAS transcription factor 42) is also a signaling component in maize, and GRAS42 loss results in semi-dwarf plants with shorter/wider leaves and a more upright leaf angle (Kaur et al., 2024), consistent with partial attenuation of BR signaling rather than complete BR deficiency.

2.3 Auxin pathway

The IPyA (indole-3-pyruvic acid) pathway is a two-step auxin biosynthetic route that involves two conserved enzyme families. Tryptophan aminotransferases (TAA1/TAR) enzymes convert tryptophan (Trp) into IPyA, and YUCCA flavin monooxygenases convert IPyA into auxin. In maize, vanishing tassel2 (VT2) encodes a grass-specific TAA/TAR-type aminotransferase, and vt2 mutants show insufficient endogenous auxin synthesis, and strong developmental defects, including reduced internode development and short stature (Phillips et al., 2011).

The brachytic2 (Br2) gene encodes a ATP-binding cassette subfamily B transporter, and functions in intercalary meristems and nodal regions to export auxin into long-distance auxin streams. Br2 loss reduces auxin export/transport, shortens lower internodes, and reduces PH (Multani et al., 2003). Interestingly, allelic variations at Br2 in teosinte reduce internode elongation under competitive conditions (Wills et al., 2013), suggesting that Br2 alleles from wild relatives could be mined for semi-dwarf variants. PIN proteins establish auxin efflux polarity and thus auxin gradients. ZmPIN1a and ZmPIN1b influence maize architecture via polar auxin transport (Carraro et al., 2006). Perturbations in PIN1-dependent auxin maxima formation and polar flux organization can affect plant growth and PH (Carraro et al., 2006; Gallavotti et al., 2008).

Auxin levels are buffered by conjugation. Stress response analysis of the ZmGH3 (gretchen hagen3) family supports a role for ZmGH3-mediated conjugation in reducing free auxin pools and modulating growth and stress responses (Feng et al., 2015). Mechanistically, ZmGH3 proteins function as ATP-dependent auxin–amido synthetases that conjugate auxin to amino acids (e.g., auxin-Asp/auxin-Glu), thereby lowering the free auxin pool and reshaping auxin homeostasis and downstream growth programs (Staswick et al., 2005).

Auxin signaling relies on TIR1/AFB (Transport inhibitor response 1/auxin signaling F-box) F-box auxin co-receptors that promote ubiquitin-mediated degradation of Aux/IAA repressors, thereby releasing ARF (auxin response factor) transcription factors to activate auxin-responsive growth genes. In maize, many components of this nuclear signaling module are conserved, and pathway activity intersects nutrient and stress responses (Wang et al., 2023a). Using a synthetic nuclear auxin-response circuit, a study showed that all 16 maize Aux/IAA repressors can be degraded after auxin treatment, and that how fast they disappear depends on which receptor and which Aux/IAA are paired. The maize auxin receptor ZmAFB2/3b1 made the circuit respond to lower auxin doses (i.e., higher sensitivity) than circuits built with Arabidopsis receptors (Ramos Báez et al. 2020). Together, these results indicate that natural or engineered changes within the TIR1/AFB–Aux/IAA–ARF core module can tune the strength and timing of auxin-driven transcription, which matters for plant architecture.

2.4 Other hormone pathways that regulate maize PH

2.4.1 Strigolactone (SL) pathway

SLs are carotenoid-derived hormones best known for suppressing axillary branching, but in maize SL deficiency also affects stem thickness, internode elongation, and adventitious root development. Knockout of the SL biosynthetic gene ZmCCD8 (carotenoid cleavage dioxygenase8) caused pleiotropic architecture changes including reduced internode elongation and markedly reduced stem diameter, and these phenotypes can be rescued by GR24 (a synthetic strigolactone analog) (Guan et al., 2012).

In maize, SL perception is mainly mediated by ZmD14A/B, which encode α/β-hydrolase–type strigolactone receptors. Upon binding SL, ZmD14 proteins undergo conformational changes and form a signaling complex with the SCF F-box protein, which then promotes ubiquitin–proteasome–mediated degradation of downstream repressors such as ZmD53 and SMXL proteins. ZmD53 is a central SL signaling repressor that interacts with ZmD14A/B in a GR24-dependent manner, and a gain-of-function ZmD53 form produces strong SL-insensitive phenotypes (Liu et al., 2021a).

2.4.2 Ethylene pathway

Ethylene signaling can reduce cell elongation and often promotes radial thickening and lignification, potentially lowering PH while strengthening stalks. Consistent with this, transcriptomic studies of maize dwarfs frequently enrich ethylene-related regulators (Gao et al., 2024). At the gene-to-phenotype level, evidence is strongest for ZmACS7 encoding aminocyclopropane-1-carboxylic acid synthase 7, and corresponding to the semidwarf3 locus. A C-terminal change increases ZmACS7 protein stability, elevates ACC/ethylene accumulation, suppresses internode cell elongation, and ultimately reduces PH (Li et al., 2020a).

2.4.3 ABA pathway

ABA pathway genes regulate PH through three coupled layers: ABA biosynthesis, signal transduction, and crosstalk with other hormones. At the biosynthetic level, the VP14 gene encodes a 9-cis-epoxycarotenoid dioxygenase, and genetic work demonstrates that vp14 mutants are ABA-deficient, establishing VP14 as a key ABA biosynthetic regulator in maize (Tan et al., 1997). Endogenous ABA accumulation helps sustain primary root elongation while limiting shoot elongation, and this process directly affects plant stature (Saab et al., 1992; Spollen et al., 2000).

At the signaling level, genetic complementation experiments reveal a core ABA signaling network involving ZmPYLs (receptors), clade A ZmPP2Cs (negative regulators), and ZmSnRK2s (positive kinases) (Wang et al., 2018), and specific receptor–phosphatase modules (ZmPYL3–ZmPP2C) were supported by biochemical validation in maize (Wang et al., 2014). These modules control ABA sensitivity thresholds, which influence cell expansion, meristem activity, and internode growth. ABA may participate in the regulation of vegetative growth in an antagonistic manner with GA (Yue et al., 2021). For instance, drought stress elevates ABA levels, which can suppress the expression of GA biosynthesis genes like GA20ox, leading to reduced bioactive GA and subsequently shorter internodes (Liu et al., 2024). Collectively, these findings indicate that ABA pathway genes can indirectly shape PH or growth by modulating the trade-off between growth and stress response.

2.5 Non-hormonal regulatory pathways

2.5.1 Meristem homeostasis

A major non-hormonal determinant of PH is the size and activity of stem-cell pools in the shoot apical meristem (SAM). CLAVATA–WUSCHEL (CLV–WUS) pathway is well known as the central regulatory pathway coordinating stem cell proliferation with differentiation in Arabidopsis. Among the core components, CLV1 is a leucine-rich repeat receptor kinase, CLV2 is a receptor-like protein, CLV3 is a small secreted peptide ligand, and WUS encodes a homeodomain transcription factor. In maize, CLV-like signaling acts as a brake on meristem proliferation, thereby influencing organ initiation and plant stature. TD1 (thick tassel dwarf1) encodes a maize ortholog of CLV1; td1 mutants show increased meristem size and altered inflorescence traits, and importantly are slightly shorter than normal siblings (Bommert et al., 2005). FEA2 (fascinated ear2) encodes a CLV2-like protein that restricts meristem proliferation; fea2 mutants have enlarged meristems and fasciated ears, demonstrating CLV pathway conservation in maize (Taguchi-Shiobara et al., 2001). Weak or hypomorphic alleles in these genes can result in a modest increase in meristem size, leading to a slight increase in kernel row number without causing fasciation (Bommert et al., 2013). Concurrently, they may cause a minor reduction in PH due to altered internode patterning. Precise modulation of meristem homeostasis, perhaps through promoter editing to subtly reduce gene expression (Liu et al., 2021b), can simultaneously improve yield potential (more kernel rows) and lodging resistance (slightly shorter plants), a highly desirable combination for high-density ideotypes.

2.5.2 Light signaling and shade-avoidance networks

Under high density, SAR typically triggers internode elongation, and reduces yield stability. HY5 is a key transcription factor in light signaling, and DBB2 (dimerization partner of BBX proteins2) can physically bind HY5 and thereby inhibit HY5-driven transcription of the GA-catabolism gene ZmGA2ox4. Since ZmGA2ox4 deactivates GA, lowering ZmGA2ox4 expression would be expected to reduce GA breakdown, helping sustain higher GA levels and thus promote SAR-induced internode elongation (Wang et al., 2025). DBB2 also interacts with PIF4, a bHLH transcription factor, to enhance the transcriptional activation of cell elongation-related genes (Wang et al., 2025). Recent work has revealed additional layers of light–BR crosstalk relevant to high-density planting. In the lac1 smart canopy mutant, phytochrome A (PHYA) photoreceptors accumulate under shade conditions and directly interact with the RAVL1 transcription factor, promoting its degradation via the 26S proteasome (Tian et al., 2024). Since RAVL1 activates LAC1 expression and BR biosynthesis, this PHYA-mediated degradation reduces BR biosynthesis specifically in upper leaves, thereby decreasing upper leaf angle under dense stands. This mechanism provides a direct molecular link between light perception and BR-mediated architectural optimization, explaining how plants dynamically adjust leaf angles in response to crowding.

2.5.3 Epigenetic regulation: miRNAs and DNA methylation

Beyond direct genetic and hormonal controls, epigenetic mechanisms such as microRNAs (miRNAs) and DNA methylation add a fresh layer of PH determination. Several miRNAs have been implicated in the post-transcriptional regulation of genes governing maize architecture. For instance, the miR156–SPL module controls vegetative phase change and internode elongation, with miR156 overexpression delaying reproductive transition and altering plant stature (Chuck et al., 2007), while miR159 modulates the transcription of GAMYB-like transcription factors that mediate gibberellin signaling (Aya et al., 2009; Yu and Wang, 2020). Meanwhile, the miR164e–NAC32 module modulates PH by post-translationally regulating DELLA protein stability; miR164e overexpression promotes internode elongation, whereas NAC32 overexpression reduces PH (Peng et al., 2026).

DNA methylation regulates maize PH through multiple mechanisms. In the PAR1 epiallele, intronic cacta transposon methylation controls gene expression: hypermethylation maintains normal PH, while hypomethylation reduces it (Li et al., 2025). Similarly, differential intronic DNA methylation of ZmLSM2 modulates RNA processing and maturation within vascular stem cells, influencing stem mechanical strength and lodging resistance (Zhang et al., 2026). DNA methylation affects transcription factor binding at cis-elements, and such cis-regulatory variation explains ~72% of heritable phenotypic variation (Engelhorn et al., 2025). This collectively highlights that our understanding of the functional roles of DNA methylation remains in its infancy.

2.5.4 Cell wall and lignin pathways

Lignin biosynthesis mutants provide direct evidence that cell wall pathways alter stalk properties and sometimes growth allocation. The classic maize brown midrib3 (BM3) mutation occurs in the gene encoding caffeic acid O-methyltransferase (COMT), altering lignin composition and digestibility (Vignols et al., 1995). Beyond its role in silage quality, BM3 exemplifies a broader principle: modifications to lignin synthesis reshape the trade-off among PH, stalk stiffness, and biomass quality.

3 The molecular basis of pleiotropy in dwarfing genes

The successful deployment of dwarfing genes is often hindered by pleiotropy, the phenomenon where a single gene or locus influences multiple unrelated traits. Understanding its molecular basis is key to designing better breeding strategies by excluding the causal genes or sequences based on its mechanism. The pleiotropic effects observed in maize dwarfing mutants can be broadly categorized into several mechanistic classes.

3.1 Pathway-level crosstalk

Genes in hormone pathways are often pleiotropic because the hormones usually regulate diverse processes beyond internode elongation (Fig. 2–A). For example, severe mutations in An1 not only dwarf the plant but also affect floral development and fertility because GA is required for these processes (Bensen et al., 1995). DELLAs are well-established signaling hubs that integrate multiple hormone pathways (Daviere and Achard, 2016). Similarly, BR-deficient brd1 mutants exhibit extreme dwarfism alongside altered leaf morphology and reproductive defects, reflecting BR's role in cell division and differentiation (Makarevitch et al., 2012).

3.2 Spatiotemporal expression pattern

A classic example is the maize liguleless1 (LG1) gene, which encodes a squamosa promoter binding protein (SBP) domain transcription factor. LG1 is expressed at the boundary between the leaf blade and sheath, where it directs ligule and auricle formation (Zhong et al., 2025). However, LG1 is also expressed in spikelet primordia (Eveland et al., 2014). Mutations in LG1 affect both leaf morphology and tassel structure, demonstrating that its spatial expression domains account for its pleiotropic effects. The transcription factor bZIP29 regulates PH and endosperm filling, which can be explained by its spatiotemporal expression and a large number of target downstream genes (Yang et al., 2022; Zhang et al., 2025). These findings indicate that genes expressed in multiple tissues, or at distinct developmental stages inevitably influence diverse traits (Fig. 2–B).

3.3 Protein multifunctionality

Protein multifunctionality can arise from protein domain architecture, alternative splicing, post-translational modifications, or differential interaction partners. Many maize proteins contain multiple functional domains that mediate distinct activities. While loss-of-function of ZmAMP1 leads to a severe dwarf phenotype (Lv et al., 2014), a missense SNP mutation in ZmAMP1 underlies an early-flowering, semi-dwarf and lodging resistance phenotype (Wang et al., 2023b). The knotted1 (KN1) gene, a founding member of the KNOX homeobox transcription factor family, binds a cis-regulatory element of GA2ox1, and directly regulates GA catabolism (Bolduc and Hake, 2009). At the same time, KN1 exhibits pleiotropic effects on leaf morphology, meristem maintenance, and inflorescence architecture. The pleiotropy of KN1 arises from this dual functionality: its DNA-binding activity governs distinct transcriptional programs (Bolduc et al., 2012), while its protein-interaction capacity allows it to interact with other proteins (Tsuda et al., 2017). In maize, protein multifunctionality represents an important mechanism underlying pleiotropy (Fig. 2–C).

3.4 Linkage drag

In breeding, when a favorable gene is introduced from a donor parent, tightly linked flanking regions carrying undesirable traits are co-inherited, creating the illusion of pleiotropy. For example, the rice Ghd8 gene affects grain protein content in the background of Ehd1 (wild-type allele), but is masked in the background of ehd1 (mutant allele) due to the epistasis of Ehd1×Ghd8 (Wei et al., 2024). A quantitative trait locus (QTL) cluster in a 5 Mb region around TB1 is associated with tiller number, ear row number, flowering time, and yield (Bouchet et al., 2017), indicating tight linkage of multiple trait-associated genes in this region. Understanding linkage drag as a cause of pleiotropy is crucial for overcoming constraints in maize breeding (Fig. 2–D).

4 Strategies for deploying dwarf genes in high-density maize breeding

The translation of our molecular knowledge into improved varieties requires a strategic approach to deployment. Breeders now have an advanced toolbox at their disposal, with strategies ranging from single-gene precision editing to multi-gene stacking and genome-wide prediction. The choice among these is not one-size-fits-all but depends on the breeding objective, the genetic architecture of the target trait, and the specific gene’s properties. This decision-making framework is illustrated in Fig. 3. For instance, if a known major gene like Br2 has negative linkage drag, precision editing is the preferred path. If a dominant, non-pleiotropic allele is desired, a forward-genetics cloning strategy might be initiated. For complex traits like yield stability under density, multi-gene stacking combined with phenotypic prediction becomes essential. This section critically evaluates each strategy, not as an isolated case study, but as an interrelated option within this broader decision-making context.

4.1 Precision editing of known genes to break negative effects and linkage drag

Genome editing changes breeding deployment mainly in three ways. First, it enables the creation of allelic series, so that breeders can tune the degree of PH reduction. Second, it helps avoid linkage drag by introducing desired variants (Fig. 3–A). Third, precision editing of multifunctional genes can create new alleles that retain only favorable effects, avoiding the adverse phenotypic consequences. Here, we introduce the editing of some promising genes as examples.

4.1.1 Br2

Since a complete knockout of Br2 by editing might mimic the strong dwarf phenotype. Editing by introducing a missense mutation in a non-critical domain, or by targeting the promoter to slightly reduce expression, can create weak or intermediate alleles. A previous study demonstrated that site-directed editing targeting the final exon of Br2 can generate a series of dwarf mutant alleles, facilitating the development of dwarf germplasm with tunable dwarfing degrees (Zhao et al., 2025). This tunability allows breeders to select the exact degree of dwarfing that maximizes lodging resistance without incurring a yield penalty. Direct generation of Br2 allelic editing variants in elite inbred lines can minimize background dependence and linkage drag. This approach can be combined with synergistic selection for yield-related and lodging resistance traits. Future deployment of Br2 editing may focus on combining domain-targeted edits with promoter modifications to achieve tissue- or stage-specific expression, restricting moderate dwarfing to early vegetative growth.

4.1.2 DELLA

d8 mutants behave as dominant GA-insensitive dwarfs (Harberd and Freeling, 1989), and influence developmental timing traits, highlighting both their breeding value and pleiotropy risk (Lawit et al., 2010). DELLA editing offers two advantages. First, dominant or semi-dominant effect can be achieved. Second, since DELLA proteins integrate GA signaling with other pathways, targeted editing of DELLA offers a feasible strategy to mitigate excessive elongation responses induced by dense planting and shade competition (Lawit et al., 2010). In a field evaluation study, hybrids carrying edited D8 were tested across 17 U.S. Corn Belt field locations from 2021 to 2022 at three planting densities (80000, 100000, 120000 plants ha^-1). Across these densities, edited hybrids showed ~34% lower PH and 32% lower ear height, while maintaining grain yield comparable to isogenic comparators comparators, and significantly reduced lodging (Weers et al., 2024). This study provides a strong proof-of-concept that DELLA editing can deliver “shorter and more standable” hybrids without a yield penalty.

4.1.3 GA20ox family

Targeted engineering of GA20ox family members, like ZmGA20ox3 and ZmGA20ox5, achieves moderate PH reduction while preserving normal reproductive development. The multi-member nature of the GA20ox family allows for this specificity, as suppressing a single isoform can avoid the global GA depletion caused by targeting a universally expressed biosynthesis gene like An1 or D3. CRISPR/Cas9-mediated editing of ZmGA20ox3 in maize generated semidwarf lines with concomitant reductions in endogenous GA levels (Zhang et al., 2020). Another notable example is a dominant miRNA-based suppression strategy that targets ZmGA20ox3 and ZmGA20ox5, producing a short-stature architecture in hybrid germplasm (Paciorek et al., 2022). Using multi-member gene families to create tunable phenotypic effects is a recurring theme in crop breeding.

4.1.4 Other genes for editing

Tuning the perception of light signals and their consequent plant architecture responses can prevent maize from over-elongating upon neighbor detection. Through CRISPR/Cas9-mediated knockout and transgenic overexpression lines, ZmPHYC1/ZmPHYC2 in maize have been demonstrated to regulate flowering time and PH, while conferring an attenuated SAR under shading conditions (Li et al., 2020b), indicating that genetic manipulation of ZmPHYCs can reduce PH and mitigate the SAR response in crowded planting environments.

While loss-of-function mutations in genes such as An1, D3 and Na1 result in extreme dwarfism, these genes remain potential targets for genome editing, as their partial loss-of-function or regulatory edits may generate graded reductions in PH. Moreover, promoter editing of CLE family genes has validated the feasibility of fine-tuning agronomic traits via targeted editing of regulatory regions (Liu et al., 2021b).

4.2 Cloning and deployment of dominant or semi-dominant dwarf genes

Placing a dominant or semi-dominant dwarf gene allele in a single parent can adjust hybrid performance, shorten breeding cycles, and simplify hybrid design. One of the best-studied dominant dwarf systems in maize is D8 (Harberd and Freeling, 1989). Another representative case is ZmRPH1 encoding microtubule-associated protein. Overexpression of ZmRPH1 reduced both PH and ear height by shortening internodes (Li et al., 2020c). This study evaluated hybrids derived from crossing a tester to ZmRPH1 overexpression lines and reported significantly reduced plant and ear height in hybrids, lower lodging rate, no obvious impact on flowering time and fertility, and in most cases, no significant reduction in yield traits, with density trials including 60000, 90000, and 120000 plants ha^-1.

Despite the above cases, de novo cloning of dominant or semi-dominant dwarf genes is still necessary for maize breeding because the current PH-reducing alleles are either too pleiotropic (Lawit et al., 2010), too recessive for efficient hybrid deployment (Paciorek et al., 2022), or too background-dependent (Fei et al., 2022) to serve as breeding-grade tools. Although dominance QTL for dwarfism are frequently detected in maize (Chen et al., 2018; Chen et al., 2025), it is still difficult and time-consuming to clone the genes underlying these QTL using conventional map-based cloning approach. GWAS (genome-wide association study) can simultaneously fine-map multiple candidate genes (Korte and Farlow, 2013), enabling high-throughput cloning of genes underlying maize dwarfism (Fig. 3–B). However, dwarfism caused by rare alleles is often difficult to detect with GWAS, and such alleles are more readily cloned using biparental or other designed mapping populations derived from a limited number of parental lines. QTG-seq (quantitative trait gene sequencing) was proposed as a rapid strategy to clone QTGs segregating in a biparental population within only four generations, by combining QTL partitioning with bulked-segregant whole-genome sequencing and an optimized candidate-gene identification algorithm (Zhang et al., 2019). Next-generation bulked segregant analysis (NG-BSA) can integrate biological big data, machine-learning models, high-throughput phenotyping, and large population sizes, enabling rapid and high-throughput cloning of QTGs in a streamlined workflow (Wang et al., 2023c).

Several issues should be considered when deploying dominant dwarf genes in commercial programs. First, dominance per se is not sufficient; the allele must exhibit stable expressivity across diverse environments and genetic backgrounds. Second, even dominant dwarf genes should be evaluated in hybrid combinations rather than inbred lines alone, because heterosis and epistasis can mask or modify dwarfing effects. Third, the genetic background should be considered: a major dominant dwarf gene provides the primary PH reduction, while polygenic modifiers compensate for yield penalties.

4.3 Multi-gene stacking to achieve short stalks without decreasing yield

In contrast to a single major dwarf gene approach, modern breeding increasingly treats semi-dwarfing as a platform trait that must be combined with yield-protecting and density-adaptive traits (Fig. 3–C). This is because strong dwarfing leads to yield penalties, while density tolerance depends on more than just shorter plants. For instance, ZMM28 encodes a maize MADS-box transcription factor. Overexpression of ZMM28 driven by a constitutive promoter (event DP202216) confers enhanced growth rate, increased photosynthetic capacity, improved nitrogen uptake and nitrogen utilization efficiency, and assimilatory capacity. Consequently, this led to stable yield increases across multiple environments without altering PH or flowering time. The combination of D8 edits with DP202216 resulted in significant yield increase compared with D8 edits alone. This complementarity arises because ZMM28 boosts yield potential through enhanced source capacity, while D8 edits reduce lodging risk by lowering PH, together addressing two independent limitations to high-density productivity (Wu et al., 2019; Fernandez et al., 2022).

Another practical ideotype development is to combine semi-dwarf and suitable leaf angle and shapes, where the dwarf alleles reduce lodging risk and the leaf trait alleles improve light penetration and radiation use efficiency at high density. A recent study presents a comprehensive genomics-guided breeding strategy for developing ideotype maize optimized for high-density planting (Yao et al., 2026). Through a combination of map-based cloning, GWAS, expression analysis, the authors identified and validated eight genes, including ZmAIM1 and ZmSINAT4 for leaf angle, ZmCOL6 and ZmPDS10 for both PH and ear height, ZmKRP16 for PH, ZmOXS3 and ZmTAL1 for leaf width, and ZmA3A1 for leaf length. By pyramiding favorable alleles of the eight genes, along with other genome-wide loci associated with the target traits, into parental lines of Yufeng303, a very popular hybrid in Northern China. They developed four new hybrids that are better adapted to high-density planting, and achieved 4.1–9.2% higher plot yields than Yufeng303 under high-density conditions. This work serves as a template for genes/genomics-informed hybrid improvement in maize, effectively bridging the gap between gene discovery and breeding gains.

Multiplex editing expressing multiple gRNAs in one transformation dramatically increases the probability of generating useful combinatorial genotypes. A key platform for multiplex CRISPR editing in plants is the tRNA–gRNA (tRNA-guided RNA) processing system (Xie et al., 2015), in which a single Pol III transcript is designed as a tandem array of tRNA–gRNA units. After transcription, the plant’s endogenous tRNA maturation machinery (including RNase P and RNase Z) recognizes each tRNA segment and precisely cleaves at the tRNA boundaries, thereby releasing multiple functional gRNAs from one long precursor RNA. This design avoids building many separate U6:gRNA expression cassettes and provides a compact way to edit multiple genes in one transformation. In a maize implementation experiment (Qi et al., 2016), the authors optimized the architecture using maize glycine-tRNA as the processing element and demonstrated that the system worked robustly for multiplex mutagenesis, including expression cassettes containing up to four tRNA–gRNA units. Compared to single-gRNA strategies, the tRNA–gRNA system not only increased the number of target sites, but also enhanced mutagenesis efficiency.

4.4 Phenotypic prediction

As breeding shifts from single-gene manipulation to multi-allele stacks, classical marker-assisted selection becomes insufficient because many contributors are small-effect and background-dependent. Phenotypic prediction can be achieved through two methods: genomic selection (GS), which predicts breeding values using genome-wide markers, and machine learning (ML), which leverages multi-source data (integrating genotypic, high-throughput phenotypic, and environmental information) to enhance predictive accuracy (Fig. 3–D). Phenotypic prediction can (i) guide parent selection, (ii) predict which allele combinations will perform optimally in specific environments, and (iii) increase field testing efficiency by evaluating only hybrids with higher predicted trait values.

Density tolerance is a complex trait that emerges from interactions among three key factors: plant architecture (e.g., PH, leaf angle), stress physiology (e.g., shade avoidance, lodging resistance), and environmental conditions (e.g., light quality, water availability, nitrogen status). Accurate prediction therefore requires models that capture both polygenic inheritance and genotype-by-environment (G × E) interactions. Methods that build kernels combining genomic relationships with environmental covariates can boost predictive ability and improve genetic gain within a given cost (Montesinos-López et al., 2022). In tropical maize, environment data-based kernels and optimized training-set strategies have been proposed for traits including PH and ear height (Gevartosky et al., 2023). Conceptually, robust selection needs models that generalize across those variables and factors because field performance under crowding is highly dependent on environment factors.

4.5 Conceptual model

We propose a conceptual model for dwarf gene deployment (Fig. 3–E). Taking geneX in the GA20ox family as an example, loss-of-function or weak geneX alleles (from mutation, RNAi, genome editing, or natural variation) reduce bioactive GA accumulation in internodes, suppressing intercalary meristem cell division and parenchyma cell elongation. On the one hand, the cellular-level changes reduce final plant height and lower the gravity center of the plant, which is the main determinant of stem lodging resistance under high planting density. On the other hand, the reduced plant height optimizes canopy structure, improves light penetration into the lower canopy layers, and reduces mutual shading between adjacent individuals under high planting density. These two synergistic effects collectively enhance the standability and yield stability under high planting density. The favorable geneX allele can be stacked with the lac1 mutant allele. Plants carrying both alleles not only exhibit the favorable traits of reduced plant height and optimized canopy structure, but also retain the phenotype of the lac1 mutant—a more upright leaf posture and smaller leaf angle. These traits further optimize the spatial distribution of light interception and increase the effective photosynthetic leaf area at the population level. Stacking multiple genes (such as Zmm28, TD1, etc.) allows the integration of their combined effects into GS or ML models, enabling accurate prediction of field performance. This model connects molecular regulatory mechanisms to the breeding goal of enhancing high-density tolerance, offering an actionable framework for the rational design of high-density tolerant ideotypes.

How dwarf genes are deployed mainly depends on the specific context. Precision editing is preferable for directly improving existing elite lines by creating tunable alleles of known genes and breaking linkage drag. Deploying dominant or semi-dominant genes enables the introduction of novel variation from diverse germplasm. Multi-gene stacking is suitable for building well-designed ideotypes where genes in multiple pathways must be modulated or pyramided simultaneously. Finally, phenotypic prediction allows breeders to select the best combinations of alleles for specific environments. The choice among these strategies should be informed by clear breeding objectives, the germplasm resources at hand, and insights into the underlying molecular genetics.

5 PERSPECTIVES

5.1 Elucidating fine-scale gene regulatory networks

Density-tolerant semi-dwarf breeding is related to multiple complex traits, each of which is regulated by many genes. Therefore, the next wave of progress will rely on establishing fine-scale gene regulation networks (GRNs) that connect (i) hormone pathway, (ii) shade-avoidance and light signaling, (iii) cell wall and vascular development for stalk strength, and (iv) reproductive sink formation for ear and kernel development. The smart canopy concept exemplifies the need for tissue- and cell-specific GRN resolution. In the lac1 mutant, BR deficiency is not uniform but strongest in upper leaves due to differential regulation by PHYA–RAVL1 interactions (Tian et al., 2019). This spatial specificity cannot be predicted from bulk transcriptome analysis.

Recent advances in single-cell (scRNA-seq) and spatial transcriptomics enable high-resolution dissection of gene expression. Spatial transcriptome profiling of the developing ear and kernel links regulatory modules to reproductive sink traits that must remain robust under PH reduction (Wang et al., 2024; Fu et al., 2023). Single-cell and time-course meristem transcriptomes are revealing cell-level regulatory dynamics of inflorescence development and sex differentiation (Sun et al., 2024). Single-cell and spatial transcriptomics will be essential for mapping tissue-specific regulatory circuits and identifying nodes that can be tuned to optimize canopy architecture for high-density planting (Fig. 4–A). GRNs will become actionable when they can predict which genes or nodes can be tuned in internodes without sacrificing ear and kernel development, enabling tissue-specific stacking and rational selection of manipulation targets. The ideal targets should meet two criteria: (i) they function in specific tissues or developmental stages (e.g., internodes vs. ears), (ii) they have few paralogs or functional redundancy to minimize compensation effects.

5.2 AI-enabled precision in genome editing

Genome editing can already generate semi-dwarf alleles, but three bottlenecks remain: (i) generating edited populations and screening individual plants with desirable phenotypic traits requires a large number of editing events, (ii) editing may introduce off-targets and unintended changes, and (iii) combinatorial effects associated with multi-gene stacking should be evaluated in multi-environment trials (Jinek et al., 2012; Gehrke et al., 2018; Weers et al., 2024). These limitations can be addressed via AI-aided allele design.

AI is accelerating precision allele design by enabling both protein-level engineering and expression-level optimization (Fig. 4–B). At the protein level, by modeling three-dimensional structures and protein interactions, researchers can predict how specific amino acid substitutions alter protein function. For example, AlphaFold 3-guided editing of D9 generated a mutant with reduced gibberellin sensitivity, achieving moderate dwarfism without yield penalties (Zheng et al., 2026). At the expression level, deep learning frameworks trained on epigenomic and expression data can predict how promoter variants affect transcription (Jaganathan et al., 2025), making it possible to design synthetic or edited promoters that drive expression in specific developmental contexts. Future breeding can integrate AI-driven protein engineering and expression-level optimization. By systematically evaluating whether target genes are amenable to protein-level modification or expression-level tuning, multiple favorable alleles can be efficiently pyramided for elite crop improvement. AI can simulate the combinatorial effects of pyramiding multiple edited genes or regulatory network nodes (such as GA20ox governing plant height and LAC1 controlling leaf angle) to predict field performance prior to material generation. This paradigm shifts breeding from trial-and-error gene stacking to rational, network-guided design.

5.3 Improving harvest index (HI)

A biological challenge for semi-dwarf, density-tolerant maize is avoiding trade-offs between ear sink strength and vegetative growth. Under high density, lodging can significantly reduce harvestable yield, so reducing plant and ear height can protect yield. However, if dwarfing influences reproductive development, the protected yield may still decline (Jin et al., 2023). Future progress will likely hinge on the coordinated manipulation of PH, lodging resistance, and ear development (Fig. 4–C).

Two practical trends support this direction. First, treating stalk strength and PH as equal targets without leading to yield reduction; transcriptomic studies are identifying pathways that can be stacked with PH alleles to reduce lodging (Xie et al., 2022). Second, nitrogen and density interact strongly with lodging and yield formation, so genetic gains must be validated under realistic N regimes. Field evidence indicates that optimized N management can improve culm mechanical strength and lodging resistance at high density (Ahmad et al., 2023). In summary, preferable multi-allele stacks will reduce plant and ear height and improve stalk biomechanics while maintaining yield performance, increasing HI under dense stands rather than simply shrinking the plant.

5.4 Phenotypic prediction by integrating high-throughput, continuous growth-related environmental factors

A key direction is incorporating real-time enviromics and high-throughput phenomics into prediction models, moving from “G only” or “G × E as a black box” toward mechanism-informed, data-rich prediction. Enviromics moves beyond simple location descriptors to a comprehensive characterization of the environment (Fig. 4–D). For instance, rather than just “Location A”, enviromics data would include real-time data of: (i) light quality and quantity: daily light integral and the ratio of red to far-red light (R:FR). (ii) thermal time: growing degree days. (iii) water stress indices: cumulative precipitation, evapotranspiration, and soil moisture. (iv) soil characteristics: texture, pH, organic matter, and nutrition availability (Ahmad et al., 2023). These high-dimensional enviromic data can be structured into environmental covariance matrices or environmental kernels (Resende et al., 2025), enabling ML models to capture complex, nonlinear relationships among environmental factors, genetic markers, and phenotypes.

The advancement of real-time, high-throughput phenotyping (HTP) technologies, including unmanned aerial vehicles (UAVs), ground-based robots, and spectral sensors, has reshaped data collection by generating continuous, time-resolved phenotypic measurements throughout the growing season (Song et al., 2021). Integrating these rich HTP data with genomic information can enhance prediction accuracy and accelerate breeding decisions. The key challenge is how to integrate phenomic data into predictive models. A common strategy is to reduce dimensionality using functional data analysis (Vieu, 2018), which summarizes temporal variation into a few scores that serve as input for prediction models. Another strategy is phenomic prediction, where HTP-derived traits are used as inputs in a standard GS model to directly predict target traits. This approach has been shown to achieve comparable or even higher prediction accuracy than genomic prediction for traits like grain yield (Zhu et al., 2022). When both phenomic and genomic predictions are combined via weighting, accuracy improves further (Jackson et al., 2023). Together, these advanced approaches align with the “genomics–phenomics–enviromics triangle” concept (Crossa et al., 2021), in which predictive accuracy depends on optimizing all three data types within an integrated framework.

Synthesizing these perspectives, we propose that a future maize breeding paradigm for density tolerance can integrate four elements: (i) GRN-guided target prioritization, (ii) AI-empowered precision editing to create allelic series, (iii) stacking favorable alleles, whether from natural variation or genome editing, to enhance HI, and (iv) phenomics–enviromics–genomics prediction to optimize selection across environments. The ultimate goal is ideotypes with moderate stature reduction, enhanced lodging resistance, optimized source-sink relationships, and stable high yield under dense stands. Achieving this will require sustained collaboration among molecular geneticists, breeders, physiologists, and computational biologists.

References

[1] Ahmad I , Batyrbek M , Ikram K . et al. Nitrogen management improves lodging resistance and production in maize (Zea mays L. ) at a high plant density. Journal of Integrative Agriculture, 2023, 22(2): 417–433

[2] Assefa Y , Carter P , Hinds M . et al. Analysis of long term study indicates both agronomic optimal plant density and increase maize yield per plant contributed to yield gain. Scientific Reports, 2018, 8(1): 4937

[3] Aya K , Ueguchi-Tanaka M , Kondo M . et al. Gibberellin modulates anther development in rice via the transcriptional regulation of GAMYB. The Plant Cell, 2009, 21(5): 1453–1472

[4] Bensen R J , Johal G S , Crane V C . et al. Cloning and characterization of the maize Anl gene. The Plant Cell, 1995, 7(1): 75–84

[5] Ben-Targem M , Ripper D , Bayer M . et al. Auxin and gibberellin signaling cross-talk promotes hypocotyl xylem expansion and cambium homeostasis. Journal of Experimental Botany, 2021, 72(10): 3647–3660

[6] Binenbaum J , Weinstain R , Shani E . Gibberellin localization and transport in plants. Trends in Plant Science, 2018, 23(5): 410–421

[7] Bolduc N , Hake S . The maize transcription factor KNOTTED1 directly regulates the gibberellin catabolism gene ga2ox1. The Plant Cell, 2009, 21(6): 1647–1658

[8] Bolduc N , Yilmaz A , Mejia-Guerra M K . et al. Unraveling the KNOTTED1 regulatory network in maize meristems. Genes & Development, 2012, 26(15): 1685–1690

[9] Bommert P , Lunde C , Nardmann J . et al. thick tassel dwarf1 encodes a putative maize ortholog of the Arabidopsis CLAVATA1 leucine-rich repeat receptor-like kinase. Development, 2005, 132(6): 1235–1245

[10] Bommert P , Nagasawa N S , Jackson D . Quantitative variation in maize kernel row number is controlled by the FASCIATED EAR2 locus. Nature Genetics, 2013, 45(3): 334–337

[11] Bouchet S , Bertin P , Presterl T . et al. Association mapping for phenology and plant architecture in maize shows higher power for developmental traits compared with growth influenced traits. Heredity, 2017, 118(3): 249–259

[12] Carraro N , Forestan C , Canova S . et al. ZmPIN1a andZmPIN1b encode two novel putative candidates for polar auxin transport and plant architecture determination of maize. Plant Physiology, 2006, 142(1): 254–264

[13] Chen C , Yang S , Zhang R . et al. qPH_1.01, a dominant QTL that contributes to plant height and ear height in maize. Plant Genome, 2025, 18(4): e70157

[14] Chen Q , Song J , Du W . et al. Identification and genetic mapping for rht-DM, a dominant dwarfing gene in mutant semi-dwarf maize using QTL-seq approach. Genes & Genomics, 2018, 40(10): 1091–1099

[15] Chen Y , Hou M , Liu L . et al. The maize DWARF1 encodes a gibberellin 3-oxidase and is dual localized to the nucleus and cytosol. Plant Physiology, 2014, 166(4): 2028–2039

[16] Chuck G , Cigan A M , Saeteurn K . et al. The heterochronic maize mutant Corngrass1 results from overexpression of a tandem microRNA. Nature Genetics, 2007, 39(4): 544–549

[17] Crossa J , Fritsche-Neto R , Montesinos-Lopez O A . et al. The modern plant breeding triangle: optimizing the use of genomics, phenomics, and enviromics data. Frontiers in Plant Science, 2021, 12: 651480

[18] Daviere J , Achard P . A pivotal role of DELLAs in regulating multiple hormone signals. Molecular Plant, 2016, 9(1): 10–20

[19] Deng S , Jiang S , Liu B . et al. ZmGDIα-hel counters the RBSDV-induced reduction of active gibberellins to alleviate maize rough dwarf virus disease. Nature Communications, 2024, 15(1): 7576

[20] Duvick D N . The contribution of breeding to yield advances in maize (Zea mays L. ). Advances in Agronomy, 2005, 86: 83–145

[21] Engelhorn J , Snodgrass S J , Kok A . et al. Genetic variation at transcription factor binding sites largely explains phenotypic heritability in maize. Nature Genetics, 2025, 57(9): 2313–2322

[22] Eveland A L , Goldshmidt A , Pautler M . et al. Regulatory modules controlling maize inflorescence architecture. Genome Research, 2014, 24(3): 431–443

[23] Fei J , Lu J , Jiang Q . et al. Maize plant architecture trait QTL mapping and candidate gene identification based on multiple environments and double populations. BMC Plant Biology, 2022, 22(1): 110

[24] Feiz L , Shyu C , Wu S . et al. COI1 F-box proteins regulate DELLA protein levels, growth, and photosynthetic efficiency in maize. The Plant Cell, 2024, 36(9): 3237–3259

[25] Feng S , Yue R , Tao S . et al. Genome-wide identification, expression analysis of auxin-responsive GH3 family genes in maize (Zea mays L. ) under abiotic stresses. Journal of Integrative Plant Biology, 2015, 57(9): 783–795

[26] Fernandez J A , Habben J E , Schussler J R . et al. Zmm28 transgenic maize increases both N uptake- and N utilization-efficiencies. Communications Biology, 2022, 5(1): 555

[27] Fu J , Ren F , Lu X . et al. A tandem array of ent -Kaurene synthases in maize with roles in gibberellin and more specialized metabolism. Plant Physiology, 2016, 170(2): 742–751

[28] Fu Y , Xiao W , Tian L . et al. Spatial transcriptomics uncover sucrose post-phloem transport during maize kernel development. Nature Communications, 2023, 14(1): 7191

[29] Gallavotti A , Yang Y , Schmidt R J . et al. The relationship between auxin transport and maize branching. Plant Physiology, 2008, 147(4): 1913–1923

[30] Gao Y , Ren H , Wang R . et al. Genome-wide identification of height-related genes using three maize dwarfs and RNA-Seq. Agronomy-Basel, 2024, 14(7): 1598

[31] Gehrke J M , Cervantes O , Clement M K . et al. An APOBEC3A-Cas9 base editor with minimized bystander and off-target activities. Nature Biotechnology, 2018, 36(10): 977–982

[32] Gevartosky R , Carvalho H F , Costa-Neto G . et al. Enviromic-based kernels may optimize resource allocation with multi-trait multi-environment genomic prediction for tropical maize. BMC Plant Biology, 2023, 23(1): 10

[33] Guan J C , Koch K E , Suzuki M . et al. Diverse roles of strigolactone signaling in maize architecture and the uncoupling of a branching-specific subnetwork. Plant Physiology, 2012, 160(3): 1303–1317

[34] Harberd N P , Freeling M . Genetics of dominant gibberellin-insensitive dwarfism in maize. Genetics, 1989, 121(4): 827–838

[35] Hartwig T , Chuck G S , Fujioka S . et al. Brassinosteroid control of sex determination in maize. Proceedings of the National Academy of Sciences of the United States of America, 2011, 108(49): 19814–19819

[36] Jackson R , Buntjer J B , Bentley A R . et al. Phenomic and genomic prediction of yield on multiple locations in winter wheat. Frontiers in Genetics, 2023, 14: 1164935

[37] Jaganathan K , Ersaro N , Novakovsky G . et al. Predicting expression-altering promoter mutations with deep learning. Science, 2025, 389(6760): eads7373

[38] Jin R , Li Z , Wang X . et al. Optimizing row spacing increases stalk lodging resistance by improving light distribution in dense maize populations. Agronomy-Basel, 2023, 13(2): 462

[39] Jinek M , Chylinski K , Fonfara I . et al. A programmable dual-RNA-guided DNA endonuclease in adaptive bacterial immunity. Science, 2012, 337(6096): 816–821

[40] Kaur A , Best N B , Hartwig T . et al. A maize semi-dwarf mutant reveals a GRAS transcription factor involved in brassinosteroid signaling. Plant Physiology, 2024, 195(4): 3072–3096

[41] Kir G , Ye H , Nelissen H . et al. RNA interference knockdown of BRASSINOSTEROID INSENSITIVE1 in maize reveals novel functions for brassinosteroid signaling in controlling plant architecture. Plant Physiology, 2015, 169(1): 826–839

[42] Korte A, Farlow A. 2013. The advantages and limitations of trait analysis with GWAS: a review. Plant Methods, 9(29).

[43] Lawit S J , Wych H M , Xu D . et al. Maize DELLA proteins dwarf plant8 and dwarf plant9 as modulators of plant development. Plant and Cell Physiology, 2010, 51(11): 1854–1868

[44] Li H , Wang L , Liu M . et al. Maize plant architecture is regulated by the ethylene biosynthetic gene ZmACS7. Plant Physiology, 2020a, 183(3): 1184–1199

[45] Li K, Yu R, Fan L, et al. 2016. DELLA-mediated PIF degradation contributes to coordination of light and gibberellin signalling in Arabidopsis. Nature Communications, 7(1): 11868.

[46] Li Q, Wang C, Jiang L, et al. 2012. An interaction between BZR1 and DELLAs mediates direct signaling crosstalk between brassinosteroids and gibberellins in Arabidopsis. Science Signaling, 5(244): ra72.

[47] Li Q , Wu G , Zhao Y . et al. CRISPR/Cas9‐mediated knockout and overexpression studies reveal a role of maize phytochrome C in regulating flowering time and plant height. Plant Biotechnology Journal, 2020b, 18(12): 2520–2532

[48] Li R, Xu Y, Xu Q, et al. 2025. An epiallele of a gene encoding a PfkB-type carbohydrate kinase affects plant architecture in maize. The Plant Cell, 37(1).

[49] Li W , Ge F , Qiang Z . et al. Maize ZmRPH1 encodes a microtubule‐associated protein that controls plant and ear height. Plant Biotechnology Journal, 2020c, 18(6): 1345–1347

[50] Liu L , Gallagher J , Arevalo E D . et al. Enhancing grain-yield-related traits by CRISPR–Cas9 promoter editing of maize CLE genes. Nature Plants, 2021b, 7(3): 287–294

[51] Liu Y , Chen Z , Zhang C . et al. Gene editing of ZmGA20ox3 improves plant architecture and drought tolerance in maize. Plant Cell Reports, 2024, 43(1): 18

[52] Liu Y , Wu G , Zhao Y . et al. DWARF53 interacts with transcription factors UB2/UB3/TSH4 to regulate maize tillering and tassel branching. Plant Physiology, 2021a, 187(2): 947–962

[53] Lv H , Zheng J , Wang T . et al. The maize d2003, a novel allele of VP8, is required for maize internode elongation. Plant Molecular Biology, 2014, 84(3): 243–257

[54] Makarevitch I , Thompson A , Muehlbauer G J . et al. Brd1 gene in maize encodes a brassinosteroid C-6 oxidase. PLoS One, 2012, 7(1): e30798

[55] Montesinos-López O A , Montesinos-López J C , Montesinos-López A . et al. Bayesian multitrait kernel methods improve multienvironment genome-based prediction. G3-Genes Genomes Genetics, 2022, 12(2): jkab406

[56] Multani D S , Briggs S P , Chamberlin M A . et al. Loss of an MDR transporter in compact stalks of maize br2 and sorghum dw3 mutants. Science, 2003, 302(5642): 81–84

[57] Paciorek T , Chiapelli B J , Wang J Y . et al. Targeted suppression of gibberellin biosynthetic genes ZmGA20ox3 and ZmGA20ox5 produces a short stature maize ideotype. Plant Biotechnology Journal, 2022, 20(6): 1140–1153

[58] Peng C , Zhao X , Xiao J . et al. The miR164e-NAC32 module orchestrates maize plant height via post-translational regulation of DELLA protein stability. Plant Communications, 2026, 7(2): 101670

[59] Peng J , Richards D E , Hartley N M . et al. 'Green revolution' genes encode mutant gibberellin response modulators. Nature, 1999, 400(6741): 256

[60] Phillips K A , Skirpan A L , Liu X . et al. vanishing tassel2 encodes a grass-specific tryptophan aminotransferase required for vegetative and reproductive development in maize. The Plant Cell, 2011, 23(2): 550–566

[61] Qi T , Huang H , Wu D . et al. Arabidopsis DELLA and JAZ proteins bind the WD-repeat/bHLH/MYB complex to modulate gibberellin and jasmonate signaling synergy. The Plant Cell, 2014, 26(3): 1118–1133

[62] Qi W , Zhu T , Tian Z . et al. High-efficiency CRISPR/Cas9 multiplex gene editing using the glycine tRNA-processing system-based strategy in maize. BMC Biotechnology, 2016, 16(1): 58

[63] Ramos Báez R , Buckley Y , Yu H . et al. A synthetic approach allows rapid characterization of the maize nuclear auxin response circuit. Plant Physiology, 2020, 182(4): 1713–1722

[64] Resende R T , Xavier A , Silva P I T . et al. GIS‐based G × E modeling of maize hybrids through enviromic markers engineering. The New Phytologist, 2025, 245(1): 102–116

[65] Saab I N , Sharp R E , Pritchard J . Effect of inhibition of abscisic Acid accumulation on the spatial distribution of elongation in the primary root and mesocotyl of maize at low water potentials. Plant Physiology, 1992, 99(1): 26–33

[66] Sasaki A , Ashikari M , Ueguchi-Tanaka M . et al. A mutant gibberellin-synthesis gene in rice. Nature, 2002, 416(6882): 701–702

[67] Shohat H , Illouz-Eliaz N , Kanno Y . et al. The tomato DELLA protein PROCERA promotes abscisic acid responses in guard cells by upregulating an abscisic acid transporter. Plant Physiology, 2020, 184(1): 518–528

[68] Song P , Wang J , Guo X . et al. High-throughput phenotyping: Breaking through the bottleneck in future crop breeding. The Crop Journal, 2021, 9(3): 633–645

[69] Spollen W G , LeNoble M E , Samuels T D . et al. Abscisic acid accumulation maintains maize primary root elongation at low water potentials by restricting ethylene production. Plant Physiology, 2000, 122(3): 967–976

[70] Staswick P E , Serban B , Rowe M . et al. Characterization of an Arabidopsis enzyme family that conjugates amino acids to indole-3-acetic acid. The Plant Cell, 2005, 17(2): 616–627

[71] Sun Y , Dong L , Kang L . et al. Progressive meristem and single-cell transcriptomes reveal the regulatory mechanisms underlying maize inflorescence development and sex differentiation. Molecular Plant, 2024, 17(7): 1019–1037

[72] Taguchi-Shiobara F , Yuan Z , Hake S . et al. The fasciated ear2 gene encodes a leucine-rich repeat receptor-like protein that regulates shoot meristem proliferation in maize. Genes & Development, 2001, 15(20): 2755–2766

[73] Tan B C , Schwartz S H , Zeevaart J A . et al. Genetic control of abscisic acid biosynthesis in maize. Proceedings of the National Academy of Sciences of the United States of America, 1997, 94(22): 12235–12240

[74] Tian J , Wang C , Chen F . et al. Maize smart-canopy architecture enhances yield at high densities. Nature, 2024, 632(8025): 576–584

[75] Tian J , Wang C , Xia J . et al. Teosinte ligule allele narrows plant architecture and enhances high-density maize yields. Science, 2019, 365(6454): 658–664

[76] Tsuda K , Abraham-Juarez M , Maeno A . et al. KNOTTED1 cofactors, BLH12 and BLH14, regulate internode patterning and vein anastomosis in maize. Plant Cell, 2017, 29(5): 1105–1118

[77] Ueguchi-Tanaka M , Ashikari M , Nakajima M . et al. GIBBERELLIN INSENSITIVE DWARF1 encodes a soluble receptor for gibberellin. Nature, 2005, 437(7059): 693–698

[78] Vieu P . On dimension reduction models for functional data. Statistics & Probability Letters, 2018, 136: 134–138

[79] Vignols F , Rigau J , Torres M A . et al. The brown midrib3 (bm3) mutation in maize occurs in the gene encoding caffeic acid O-methyltransferase. The Plant Cell, 1995, 7(4): 407–416

[80] Wang H , Wu G , Zhao B . et al. Regulatory modules controlling early shade avoidance response in maize seedlings. BMC Genomics, 2016, 17(1): 269

[81] Wang P , Yang Y , Li D . et al. Cloning of a new allele of ZmAMP1 and evaluation of its breeding value in hybrid maize. The Crop Journal, 2023b, 11(1): 157–165

[82] Wang X , Han L , Li J . et al. Next-generation bulked segregant analysis for Breeding 4.0. Cell Reports, 2023c, 42(9): 113039

[83] Wang X , Jiao Z , Zhang Y . et al. DBB2 regulates plant height and shade avoidance responses in maize. Journal of Integrative Plant Biology, 2025, 67(5): 1323–1338

[84] Wang Y , Fu F , Yu H . et al. Interaction network of core ABA signaling components in maize. Plant Molecular Biology, 2018, 96(3): 245–263

[85] Wang Y , Luo Y , Guo X . et al. A spatial transcriptome map of the developing maize ear. Nature Plants, 2024, 10(5): 815–827

[86] Wang Y , Xu J , Yu J . et al. Maize GSK3-like kinase ZmSK2 is involved in embryonic development. Plant Science, 2022, 318: 111221

[87] Wang Y , Yu H , Zhang Y . et al. Interaction between abscisic acid receptor PYL3 and protein phosphatase type 2C in response to ABA signaling in maize. Gene, 2014, 549(1): 179–185

[88] Wang Z , Wang Q , Chong K . et al. The brassinosteroid signal transduction pathway. Cell Research, 2006, 16(5): 427–434

[89] Wang Z , Wang Y , Du Q . et al. The auxin signaling pathway contributes to phosphorus-mediated zinc homeostasis in maize. BMC Plant Biology, 2023a, 23(1): 20

[90] Weers B , Phillips J , Niu X . et al. Genome edited zm-D8 reduced stature maize hybrids have improved climate resilience and competitive yields. Field Crops Research, 2024, 317: 109527

[91] Wei X , Chen M , Zhang Q . et al. Genomic investigation of 18, 421 lines reveals the genetic architecture of rice. Science, 2024, 385(6704): eadm8762

[92] Wills D M , Whipple C J , Takuno S . et al. From many, one: genetic control of prolificacy during maize domestication. PLoS Genetics, 2013, 9(6): e1003604

[93] Winkler R G , Helentjaris T . The maize Dwarf3 gene encodes a cytochrome P450-mediated early step in Gibberellin biosynthesis. The Plant Cell, 1995, 7(8): 1307–1317

[94] Wu J , Lawit S J , Weers B . et al. Overexpression of Zmm28 increases maize grain yield in the field. Proceedings of the National Academy of Sciences of the United States of America, 2019, 116(47): 23850–23858

[95] Xie K , Minkenberg B , Yang Y . Boosting CRISPR/Cas9 multiplex editing capability with the endogenous tRNA-processing system. Proceedings of the National Academy of Sciences of the United States of America, 2015, 112(11): 3570–3575

[96] Xie L , Wen D , Wu C . et al. Transcriptome analysis reveals the mechanism of internode development affecting maize stalk strength. BMC Plant Biology, 2022, 22(1): 49

[97] Yamaguchi S . Gibberellin metabolism and its regulation. Annual Review of Plant Biology, 2008, 59: 225–251

[98] Yang T , Wang H , Guo L . et al. ABA-induced phosphorylation of basic leucine zipper 29, ABSCISIC ACID INSENSITIVE 19, and Opaque2 by SnRK2.2 enhances gene transactivation for endosperm filling in maize. The Plant Cell, 2022, 34(5): 1933–1956

[99] Yao W , Ku L , Wang B . et al. Breeding ideotype maize with enhanced yield through genomics-guided pyramiding of favorable alleles. Nature Genetics, 2026, 58: 916–927

[100] Yu S , Wang J . The crosstalk between microRNAs and gibberellin signaling in plants. Plant and Cell Physiology, 2020, 61(11): 1880–1890

[101] Yue K , Lingling L , Xie J . et al. Synthesis and regulation of auxin and abscisic acid in maize. Plant Signaling & Behavior, 2021, 16(7): 1891756

[102] Zhang H , Wang X , Pan Q . et al. QTG-Seq accelerates QTL fine mapping through QTL partitioning and whole-genome sequencing of bulked segregant samples. Molecular Plant, 2019, 12(3): 426–437

[103] Zhang J , Gu R , Miao X . et al. GWAS-based population genetic analysis identifies bZIP29 as a heterotic gene in maize. Plant Communications, 2025, 6(5): 101289

[104] Zhang J , Zhang X , Chen R . et al. Generation of transgene-free semidwarf maize plants by gene editing of gibberellin-oxidase20-3 using CRISPR/Cas9. Frontiers in Plant Science, 2020, 11: 1048

[105] Zhang Y, Wang Z, Du J, et al. 2026. Stem microanatomical phenomic uncovers a potential role for ZmLSM2 in regulating maize stem bending strength. Journal of Integrative Plant Biology: 70140.

[106] Zhao B , Xia Z , Sun C . et al. CRISPR /Cas9‐mediated genomic editing of Brachytic2 creates semi‐dwarf mutant alleles for tailored maize breeding. Plant Biotechnology Journal, 2025, 23(4): 1133–1135

[107] Zheng H, Chen M, Liu P, et al. 2026. Precision-edited D9 allele breaks the trade-off between maize dwarfing and yield. Journal of Genetics and Genomics: in press.

[108] Zhong Z , Yao M , Cao Y . et al. LG1 promotes preligule band formation through directly activating ZmPIN1 genes in maize. Journal of Genetics and Genomics, 2025, 52(3): 356–366

[109] Zhu X , Maurer H P , Jenz M . et al. The performance of phenomic selection depends on the genetic architecture of the target trait. Theoretical and Applied Genetics, 2022, 135(2): 653–665

RIGHTS & PERMISSIONS

Molecular regulation of maize plant height and its implications for breeding maize tolerant to enhanced planting density