CRISPR/Cas9‐based genome editing of 14 lipid metabolic genes reveals a sporopollenin metabolon ZmPKSB‐ZmTKPR1‐1/‐2 required for pollen exine formation in maize

Summary Lipid biosynthesis and transport are essential for plant male reproduction. Compared with Arabidopsis and rice, relatively fewer maize lipid metabolic genic male‐sterility (GMS) genes have been identified, and the sporopollenin metabolon in maize anther remains unknown. Here, we identified two maize GMS genes, ZmTKPR1‐1 and ZmTKPR1‐2, by CRISPR/Cas9 mutagenesis of 14 lipid metabolic genes with anther stage‐specific expression patterns. Among them, tkpr1‐1/‐2 double mutants displayed complete male sterility with delayed tapetum degradation and abortive pollen. ZmTKPR1‐1 and ZmTKPR1‐2 encode tetraketide α‐pyrone reductases and have catalytic activities in reducing tetraketide α‐pyrone produced by ZmPKSB (polyketide synthase B). Several conserved catalytic sites (S128/130, Y164/166 and K168/170 in ZmTKPR1‐1/‐2) are essential for their enzymatic activities. Both ZmTKPR1‐1 and ZmTKPR1‐2 are directly activated by ZmMYB84, and their encoded proteins are localized in both the endoplasmic reticulum and nuclei. Based on protein structure prediction, molecular docking, site‐directed mutagenesis and biochemical assays, the sporopollenin biosynthetic metabolon ZmPKSB‐ZmTKPR1‐1/‐2 was identified to control pollen exine formation in maize anther. Although ZmTKPR1‐1/‐2 and ZmPKSB formed a protein complex, their mutants showed different, even opposite, defective phenotypes of anther cuticle and pollen exine. Our findings discover new maize GMS genes that can contribute to male‐sterility line‐assisted maize breeding and also provide new insights into the metabolon‐regulated sporopollenin biosynthesis in maize anther.


Introduction
In flowering plants, the anther cuticle and pollen exine are two crucial lipid layers for pollen grain formation (Wan et al., 2020).The former covers the anther outer surface and protects pollen development from biotic and abiotic stresses during anthesis (Yeats and Rose, 2013).The latter, the outermost layer of the pollen wall, plays an important role in pollen-stigma interaction and double fertilization in addition to protecting pollen grains after releasing them from anthers (Shi et al., 2015).Since lipids and their derivatives are the major chemical components of both layers, they often display defective structures when lipid biosynthesis or transport pathways are blocked.Plant mutants with defective anther cuticle or pollen exine usually display malesterile or partially sterile phynotypes (Wan et al., 2020).
Notably, most GMS genes in plants have been identified using the forward genetic strategy, which depends on the available GMS mutants, and the identification process is time-consuming.Recently, eight GMS genes encoding transcription factors (TFs) have been successfully identified in maize via CRISPR/Cas9 genome editing, demonstrating its effectiveness in the discovery of GMS genes and mutant creation (Jiang et al., 2021a).Based on cytological observation and landmark developmental events, maize anther development is divided into 14 stages (S1-S14), and the characteristics of each stage are described in detail (Wan et al., 2019).Pollen exine formation begins with the primexine appearance at S8b in maize.At S9, a thin exine composed of continuous tectum, foot layer and bacula is formed, then thickened apparently at S10, and completed at S11 and S12 (An et al., 2019).Anther cuticle appears at S10, becomes apparent at S11 and matures at S12 and S13 (An et al., 2019).Correspondingly, most maize lipid metabolic GMS genes for anther and pollen development have expression peaks during stages S8-S10.Therefore, selecting candidate lipid metabolic genes with expression patterns from S8 to S10 for CRISPR/Cas9 mutagenesis may efficiently discover new GMS genes in maize.
Sporopollenin, a complex biopolymer of lipidic monomers and phenylpropanoid, is the main constituent of pollen exine (Li et al., 2019;Shi et al., 2015;Xue et al., 2020).In Arabidopsis and rice, sporopollenin biosynthesis depends on a lipid metabolon consisting of acyl-CoA synthetase (ACOS), PKS and TKPR (Grienenberger et al., 2010;Lallemand et al., 2013;Yang et al., 2023).Mutations in any of these genes may result in complete or partial male fertility.These sporopollenin biosynthetic enzymes interact and use the products of upstream enzymes as substrates for sequential catalytic reactions to ensure metabolic efficiency (Lallemand et al., 2013;Yang et al., 2023).Although the molecular mechanisms of each component affecting pollen exine formation remain unknown, such a metabolon has also been reported in tobacco, moss, Hypericum perforatum and canola (Daku et al., 2016;Karppinen et al., 2008;Qin et al., 2016;Wang et al., 2013).This suggests that a conserved sporopollenin biosynthetic pathway is likely present in land plants.Recently, we have found that ZmPKSB is essential for normal male fertility in maize (Liu et al., 2022).However, whether the PKSB-TKPR sporopollenin metabolon exists in maize has not yet been uncovered.
Here, we found that ZmTKPR1-1 and ZmTKPR1-2 are GMS genes by screening and phenotyping 14 maize lipid metabolic gene mutants generated by CRISPR/Cas9.Among them, only tkpr1-1/-2 double mutants display complete male sterility with delayed tapetum degradation and defective pollen exine and anther cuticles.Both ZmTKPR1-1 and ZmTKPR1-2 are directly activated by TF ZmMYB84.ZmTKPR1-1 and ZmTKPR1-2 are duallocalized in the ER and nuclei.ZmPKSB, ZmTKPR1-1 and ZmTKPR1-2 interact with each other and form a complex, which constitutes a biosynthetic sporopollenin metabolon.The detailed functional differences between ZmTKPR1-1/-2 and ZmPKSB in lipid metabolism for anther cuticle and pollen exine formation are revealed.

Results
Expression patterns of 14 lipid metabolic genes during maize anther development and their CRISPR/Cas9induced mutations Given that most lipid metabolic GMS genes for pollen exine formation are highly expressed during anther developmental stages S8-S10, we first selected 14 maize metabolic genes with expression peaks at these stages based on transcriptome analysis of anther RNA-Seq data in three maize inbred lines (B73, Zheng58 and M6007) and then verified their expression patterns by qRT-PCR analysis (Figure 1a,b; Table S1).Among them, Zm00001eb025490 exhibited high expression levels from S5 to S9-10, with three expression peaks at S6, S8b and S9, respectively.Zm00001eb223980 displayed high expression levels from S8b to S9, with a peak at S8b-9.Zm00001eb035410, Zm00001eb317040, Zm00001eb028240 and Zm00001eb294840 shared similar expression patterns, with the common peaks at S9. Zm00001eb185610 exhibited two expression peaks at S8b and S9, respectively.Zm00001eb147560 was highly expressed from S8b-9 to S9-10, with a peak at S9-10.Zm00001eb073500, Zm00001eb250600, Zm00001eb387330 and Zm00001eb026670 had similar expression patterns, with the common peaks at S9-10.Zm00001eb033930 was expressed at the late stages of anther development, with a peak at S10 (Figure 1a,b).Thus, the temporal expression patterns of these 14 genes covered the critical periods of exine formation in maize anther.
We then investigated their potential functions in maize anther and pollen development through CRISPR/Cas9-induced gene mutagenesis (Figure S1).Consequently, two of them, Zm00001eb035410 and Zm00001eb317040, were found to be essential for maize male fertility, and their double mutants, not single mutants, showed complete male sterility.Both genes are homologues and encode tetraketide a-pyrone reductases, ZmTKPR1-1 and ZmTKPR1-2, respectively (Figures 1c and 2).However, single mutants of the remaining 12 genes showed normal male fertility in maize, including 11 lipid biosynthetic genes (Zm00001eb025490, Zm00001eb223980, Zm00001eb147560, Zm00001eb185610, Zm00001eb028240, Zm00001eb278580, Zm00001eb073500, Zm00001eb294840, Zm00001eb250600, Zm00001eb387330 and Zm00001eb026670) and one lipid transport-related gene (Zm00001eb033930) (Figures 1c, S2a,b  and S3).These results indicate that CRISPR/Cas9-based genome editing combined with anther transcriptome analysis is an effective approach to discovering new GMS genes, especially for those depending on simultaneous mutations of mutilple homologous genes.
Scanning electron microscopy (SEM) showed no significant difference in anther outer and inner surfaces, and microspores between WT and tkpr1-1/-2 before stage S9 (Figure S4b).From S10 to S13, the outer and inner surfaces of tkpr1-1/-2 anther were smooth, and no three-dimensional knitting cuticle and Ubisch bodies appeared when compared with those of WT.Additionally, tkpr1-1/-2 microspores adhered to each other and displayed defective vacuolization and starch accumulation (Figure S4b).These results indicate that simultaneous mutations of ZmTKPR1-1/-2 result in the abnormal development of anther cuticle, Ubisch body and pollen grains in maize.

Molecular characterization and transcriptional regulation of ZmTKPR1-1/-2
To investigate the evolutionary history of ZmTKPR1-1/-2, we conducted a protein sequence alignment with their orthologues from 11 different plant species.Results showed that the putative common NAD(P)H binding domain and NAD(P)-dependent epimerase/ dehydratase domain are conserved across these species (Figure S5).Phylogenetic analysis classifies these orthologues into three clades.Clade I contains ZmTKPR1-1, ZmTKPR1-2 and their monocot orthologues, including rice OsTKPR1 (Os09g0493500) (the closest paralogue of ZmTKPR1-2) (Xu et al., 2019) and Os08g0515900 (the closest paralogue of ZmTKPR1-1), while all the members of Clade II are from dicots, including Arabidopsis AtTKPR1 (Tang et al., 2009).AtTKPR2 and OsTKPR2 are distributed in Clade III, which is separated from Clade I and II branches, indicating that they have a far phylogenetic relationship with ZmTKPR1-1 and -2 (Figure 4a).Microsynteny analysis revealed that ZmTKPR1-1 and -2 are probably duplicated paralogues after evolutionary differentiation of monocots and dicots, as their orthologues in monocots rice, sorghum and wheat showed good syntenies with the AtTKPR1 single locus (Figure 4b).These results suggest that ZmTKPR1-1 and -2 and their orthologues were relatively conserved during the evolution of gramineous plants.
Subcellular localization analysis performed in maize protoplasts and tobacco leaves revealed that both ZmTKPR1-1 and -2 were dual-localized in the nuclei and ER (Figures 4c and S6a,b).qPCR analysis in different maize tissues showed that ZmTKPR1-1 and -2 transcripts were detected exclusively in anthers at specific developmental stages and both peaked at S9 (Figures 4d,e and S6c,d).Remarkably, no significant difference was detected in the expression levels of ZmTKPR1-1 and -2 between tkpr1-1/-2 and WT anthers, suggesting that the male-sterility phenotype of tkpr1-1/-2 results from their protein function defects.
Given that TF ZmMYB84 is required for tapetal PCD and pollen exine formation (Fang et al., 2022;Jiang et al., 2021a) and the expression pattern of ZmMYB84 overlaps with those of ZmTKPR1-1 and -2 from S8a to S9-10, with the same peak at S9 (Figures 4e and S6d), we speculate that ZmTKPR1-1 and -2 may be regulated by ZmMYB84.qPCR analysis revealed that ZmTKPR1-1 and -2 transcripts were almost undetectable in myb84 anther (Figures 4f and S6e).Thus, ZmMYB84 is necessary for the expression of ZmTKPR1-1 and -2.To further confirm whether ZmTKPR1-1 and -2 are the direct target genes of ZmMYB84, we conducted a transient dual-luciferase reporter (TDLR) assay and an electrophoretic mobility shift assay (EMSA) and found that ZmMYB84 bound to the promoters of ZmTKPR1-1 and -2 and activated their promoter activities (Figure 4g,h).Collectively, ZmMYB84 directly activates ZmTKPR1-1 and ZmTKPR1-2 expression.
ZmPKSB, ZmTKPR1-1 and ZmTKPR1-2 interact and form a multienzyme complex Growing evidence has proved that various proteins involved in lipid biosynthetic pathways can form dynamic protein-protein and protein-lipid interactomes or metabolons (Nakamura, 2017).Since ZmPKSB (Liu et al., 2022), ZmTKPR1-1 and -2 exhibit similar expression patterns during anther development and their encoded proteins are localized in the ER (Figures 4c-e and S6), we speculated that ZmPKSB, ZmTKPR1-1 and -2 may interact and form a complex to conduct their functions in maize anther development.
ZmPKSB, ZmTKPR1-1 and ZmTKPR1-2 enzymatic activities define a sporopollenin metabolon in maize In Arabidopsis and rice, PKS and TKPR are involved in a conserved sporopollenin metabolon (Lallemand et al., 2013;Yang et al., 2023).Mutations of both ZmPKSB and ZmTKPR1-1/-2 lead to male sterility, and ZmPKSB and ZmTKPR-1-1/-2 form a ª 2023 The Authors.Plant Biotechnology Journal published by Society for Experimental Biology and The Association of Applied Biologists and John Wiley & Sons Ltd., 22, 216-232 multienzyme complex, implying that a similar metabolon may also exist in maize anther.
ZmPKSB and ZmTKPR1-1/-2 exhibit different effects on pollen exine and anther cuticle formation Considering that ZmPKSB, ZmTKPR1-1 and -2 are the components of sporopollenin metabolon mainly for pollen exine formation in maize and the formation processes of pollen exine and anther cuticle share certain common pathways of lipid metabolism (Wan et al., 2020), we performed cytological comparison among WT, pksb and tkpr1-1/-2 mutants to determine whether ZmPKSB and ZmTKPR1-1/-2 have comparable effects on the formation of both lipid layers.As expected, both mutants exhibited a thinner pollen exine compared to WT (Figure 7a).Nevertheless, the pollen exine of tkpr1-1/-2 was obviously thinner than that of pksb.Additionally, anther cuticle showed opposite changes in both mutants.Compared with WT, anther cuticle was smooth in tkpr1-1/-2 anther, but denser in pksb anther (Figure 7b), suggesting that ZmPKSB and ZmTKPR1-1/-2 have different effects on pollen exine and anther cuticle formation.
Our previous research has shown that the expression of ZmPKSB depends on ZmMYB84 (Liu et al., 2022).Taken together with these results obtained here, we proposed a working model to elucidate the roles of the ZmPKSB-ZmTKPR1-1/-2 metabolon in maize anther tapetal cells (Figure 7g).ZmPKSB, ZmTKPR1-1 and -2 are activated by ZmMYB84, and the multienzyme complex ZmPKSB-ZmTKPR1-1/-2 forms a sporopollenin metabolon in anther tapetal cells.Then, sequential enzymatic reactions occur to generate reduced tetraketide a-pyrone by this metabolon.Subsequently, as crucial precursors of sporopollenin and cutin/ wax, lipid metabolic products are transported by ZmABCG26 and ZmABCG2a to form pollen exine and anther cuticle.In tkpr1-1/-2 anther, the reduced expression of ZmABCG26 blocks the transport of sporopollenin precursors, and the biosynthesis and transport of cutin/wax monomers are disturbed, likely due to the expression alterations of cutin/wax-related genes, ultimately resulting in a greatly thin pollen exine and smooth anther cuticle.

Integration of transcriptome, bioinformation and gene editing to discover new GMS genes in maize
In flowering plants, the anther cuticle is composed of lipidic polyester cutin and cuticular wax and pollen exine is mainly composed of sporopollenin and tryphine.All these materials are built by lipids or lipid derivatives, and many lipid metabolic GMS genes are indispensable for anther and pollen development (Wan et al., 2020).To rapidly enrich the genetic resources of lipid metabolic GMS genes in maize, we combined anther transcriptome analysis and CRISPR/Cas9 genome editing technology to generate mutants of 14 potential candidate genes (Figures 1 and  S1-S3).By phenotypic analysis, we discovered that ZmTKPR1-1 and ZmTKPR1-2 are new maize GMS genes, as simultaneous mutagenesis of both genes resulted in complete male sterility (Figure 2b).In contrast, the single homozygote mutants of 14 lipid metabolic genes either exhibited no male sterility or only slight male sterility.The fact that the mutant does not show an obvious phenotype is probably due to genetic robustness (El-Brolosy et al., 2019).Among several proposed mechanisms underlying genetic robustness, functional redundancy is considered to be a crucial one (Hu et al., 2023;Tautz, 1992).In this study, ZmTKPR1-1 and ZmTKPR1-2 are paralogous genes, and only their double mutants exhibited complete male sterility, reinforcing the functional redundancy mechanism.Further, no highly similar paralogues were found in the other 12 lipid genes (data not displayed in this paper).Rewiring of genetic networks (Barab asi and Oltvai, 2004) and genetic compensation (El-Brolosy et al., 2019;Rossi et al., 2015) are two other underlying mechanisms of genetic robustness, which may explain the male fertile phenotype of single mutants of the 12 lipid genes without paralogues.Therefore, integrating anther transcriptome analysis, functional redundancy analysis and gene editing is an effective strategy for discovering new male sterile genes.

Functional diversification of TKPR1 genes in anther and pollen development among different plants
ZmTKPR1-1 and ZmTKPR1-2, similar to their orthologues AtTKPR1 and OsTKPR1, were found to play conserved roles in controlling male fertility, supporting the functional conservation of lipid metabolic GMS genes among different plants (Gomez et al., 2015).Nevertheless, our results highlight the diversified and different functions of TKPR1 genes in controlling lipid metabolism and male fertility in different species.The singlegene mutants of ZmTKPR1-1 or ZmTKPR1-2 are male fertile in maize, whereas single-gene mutants of OsTKPR1 (the closest orthologue of ZmTKPR1-2) are completely male sterile in rice (Xu et al., 2019), and Os08g0515900, the closest orthologue to ZmTKPR1-1, may not be required for male fertility in rice.
In addition, AtTKPR1 and OsTKPR1 are only localized in the ER, while ZmTKPR1-1/-2 are localized in both the ER and nuclei (Figures 4c and S6a,b).The nuclear localizations of ZmTKPR1-1/-2 suggest that they may act similarly to TFs or regulators.To date, the multiple functions of several enzymes have been identified as transcriptional activators (Ke et al., 2020;Wang et al., 2023), but nuclear-localized TKPR has not been found in other plants.OsUGE1, a UDP-glucose epimerase, can act as a transcriptional activator to promote anther tapetal degradation in rice (Wang et al., 2023).Our results also revealed that tapetal PCD was delayed in tkpr1-1/-2 anther (Figure S4c).Whether there is a new regulatory mechanism for ZmTKPR1-1/-2 on maize reproductive development still needs further investigation.

Multiple approaches identify protein-protein interaction to define a sporopollenin metabolon in plant anther
Metabolons are transient enzyme-enzyme assemblies of sequential enzymes that mediate substrate channelling to enhance metabolic efficiency (Fernie et al., 2018;Srere, 1985).A prerequisite to defining such enzyme associations as metabolons is to prove their PPIs (Zhang and Fernie, 2021).Considering that the detection of such transient and dynamic enzyme complexes is more difficult than that of stable protein complexes, multiapproaches are required to test such PPIs, including biochemical experiments in vivo and in vitro and structure-based methods (Norris et al., 2022;Xu et al., 2023;Yu et al., 2023).In Arabidopsis, rice and Brassica napus, the PPIs in the sporopollenin metabolon consisting of ACOS-PKS-TKPR have been identified by Y2H, BiFC, CoIP, GST-pulldown and FLIM/FRET assays (Lallemand et al., 2013).However, structure-based PPI analysis has not been used in this metabolon.In this study, in addition to Y2H, BiFC and CoIP tests, we confirmed the pairwise interactions between ZmPKSB, ZmTKPR1-1 and ZmTKPR1-2 based on AlphaFold2 protein structure predictions, molecular docking and site-directed mutagenesis (Figures 5 and S7).Importantly, we found that seven and four docking sites are indispensable for ZmPKSB-ZmTKPR1-2 and ZmTKPR1-1-ZmTKPR1-2 interactions (Figure S7b,c), respectively.Different from that in maize (this study) and rice, BnTKPR1 is not involved in the sporopollenin metabolon in Brassica napus, and AtTKPR1 fails to form a homodimer in Arabidopsis (Lallemand et al., 2013;Qin et al., 2016).Therefore, our results reveal both conserved and diversified functions of components in the sporopollenin metabolon between dicots and monocots.
The ZmPKSB-ZmTKPR1-1/-2 metabolon is required for the biosynthesis of the reduced tetraketide a-pyrones for pollen exine formation Owing to its extreme chemical and physical recalcitrance, sporopollenin is the toughest material in the plant kingdom (Montgomery et al., 2016).For decades, the determination of sporopollenin components has become a challenging task.Recent genetic studies and analytical characterization of sporopollenin through solid-state NMR and targeted degradation techniques have found that polyhydroxylated a-pyrone subunits and hydroxylated aliphatic units are the curial components of sporopollenin (Grienenberger and Quilichini, 2021;Li et al., 2019;Mikhael et al., 2020).Our data indicate that ZmTKPR1-1/-2 has similar reduction activities to their orthologues, AtTKPR1 and OsTKPR1, and can produce reduced tetraketide a-pyrones (belong to polyhydroxylated a-pyrones), suggesting that the loss-of-function mutation of TKPR1-1/-2 leads to a lack of this core component of sporopollenin.Thus, tkpr1-1/-2 microspores exhibit very thin and completely disorganized exine without foot layer, baculua and tectum, which is consistent with the functions of ZmTKPR1-1/-2 in sporopollenin biosynthesis.Interestingly, while the pksb mutation reduces the expression of ZmTKPR1-1/-2, the pollen exine of pksb has a foot layer, baculua and tectum (Figures 5e, 7a,b and S7e).For this possible reason, we speculate that in the pksb anther, the reduced expression of ZmPKSA-1/-2 (paralogues of ZmPKSB) may still have polyketide synthase activities (Liu et al., 2022), which could produce a small amount of polyhydroxylated a-pyrones.In Arabidopsis, the nearly identical defective pollen exine of the pksa/ b double mutant and the tkpr1 mutant supports the above inference (Grienenberger et al., 2010).Further studies may confirm the hypothesis by creating a pksa/b double mutant in maize.

Different effects of ZmPKSB and ZmTKPR1-1/-2 on lipid metabolism for anther cuticle formation
The anther cuticle is a crucial lipid layer covering the outer surface of the anther epidermis.Although previous studies have shown that anther cuticle formation and pollen exine formation share certain common lipid biosynthetic pathways (Ariizumi and Toriyama, 2011;Shi et al., 2015;Wan et al., 2020), there is still a large gap in the mechanism underlying anther cuticle formation compared with pollen exine formation.In particular, different lipid genes involved in the same biosynthesis pathway but with different roles in anther cuticle formation have rarely been systematically compared.
ZmPKSB and ZmTKPR1-1/-2 can form a protein complex that increases their sequential catalytic activities (Figures 5 and 6), but their mutants exhibit opposite defective cytological phenotypes of the anther cuticle (Figure 7b).The possible reasons are as follows.
x-OH fatty acids (HFAs) have been shown to be the common substrates for the biosynthesis of anther cutin monomers and pollen sporopollenin precursors (Djukanovic et al., 2013;Dobritsa et al., 2009;Li et al., 2010;Yi et al., 2010).Thus, excess x-OH fatty acyl-CoAs may be used to generate more xHFA for anther cuticle formation when pollen exine formation is blocked in pksb anthers (Liu et al., 2022).The contents of cutin monomers C16 xHFA and C18: 1 xHFA in pksb anthers were significantly higher than those in tkpr1-1 and WT anthers (Figure 7c), further supporting the above speculation.Finally, lipid metabolism in the anthers of flowering plants is a very complex biological process, and its underlying mechanism remains unclear (Wan et al., 2020).The lipid products produced by these lipid metabolic GMS genes have a myriad of diverse functions.Instead of simple single lines downward, the catalytic processes of these enzymes are actually intertwined and eventually form pollen walls, anther cuticles and membrane structures of subcellular organelles, which requires further investigation.

Plant materials, growth conditions and phenotype characterization
Maize inbred lines B73, Zheng 58 and M6007 and hybrid Hi II were used in this study.B73 and M6007, corresponding to the WT lines of maize GMS mutants lob30 and ms7-6007, respectively, were originally obtained from the Maize Genetics Cooperation Stock Center (http://maizecoop.cropsci.uiuc.edu).Zheng58 and Hi II are maintained in our laboratory.All plants were grown in the experimental stations of the University of Sciences and Technology Beijing (USTB) in Beijing and Sanya, excepting T 0 transgenic plants that were grown in a greenhouse under long-day conditions (16 h/8 h (day/night) at 26 °C/22 °C).A Canon EOS 700D digital camera and an SZX2-ILLB stereomicroscope (Olympus, Japan) were used to photograph the images Metabolon ZmPKSB-ZmTKPR1-1/-2 for maize exine formation 227 of tassels and anthers, respectively.Pollen grains were stained with a 1% I 2 -KI solution and captured using an Olympus SZ51 microscope (Olympus, Japan) (Zhang et al., 2018a).

Plasmid construction, maize transformation and genotyping
For the mutagenesis of 14 maize lipid metabolic the CRISPR/ Cas9 plasmids were constructed using the pBUE411 vector as a backbone (Xing et al., 2014).The CRISPR-P 2.0 (http://crispr.hzau.edu.cn/CRISPR2/) was used to choose specific gRNAs targeting the coding sequences of 14 maize lipid metabolic genes, and off-target analysis of gRNAs was carried out on the website (http://www.rgenome.net/cas-offinder/).A CRISPR/Cas9 plasmid contains two gRNAs.For assembly of the two gRNAs, the PCR fragment was amplified based on pCBC-MT1T2 with the primer pair specific for each gene (Table S3), and the purified PCR fragment was cloned into a BsaI-digested pBUE411 vector.
The CRISPR/Cas9 recombinant plasmids were used for Agrobacterium tumefaciens-mediated transformation into maize (Hi-II) following the previously published protocols (Frame et al., 2002).The positive transformants were selected by PCR amplification using primers Bar-F and Bar-R.The genotyping of transgenic progenies in T 0 , F 1 and F 2 generations was performed as described previously (Jiang et al., 2021a).All primers used in this study are listed in Table S3.

Cytological analysis and microscopy
For transverse section, SEM and TEM analyses, the fresh anthers of WT and tkpr1-1/-2 from stages S5 to S13 were immersed in FAA solution (Coolabor, China) or 3.5% glutaraldehyde solution overnight according to the description of a previous study (An et al., 2020), respectively, and the images were photographed using a BX-53 microscope (Olympus, Japan), a HITACHI S-3400N scanning electron microscope (Hitachi Japan) and a HITACHI H-7500 transmission electron microscope (Hitachi Japan), respectively.

TUNEL assay
The anthers of WT and tkpr1-1/-2 at different stages were collected and prepared to perform paraffin sections.The selected paraffin sections were dewaxed in xylene and dehydrated in a gradient ethanol series.DNA fragmentation in tapetum was detected by TUNEL assay using a TUNEL kit (DeadEndTM Fluorometric TUNEL System, Promega) according to the manufacturer's instructions.Signals from anther samples were photographed under a fluorescence confocal scanner microscope (TCS-SP8, Leica).

Phylogenetic and microsynteny analysis
The 29 orthologues of ZmTKPR1-1 and ZmTKPR1-2 in 11 plants were obtained from the EnsemblPlants website (https://plants.ensembl.org).The alignment of amino acid sequences was performed with DNAMAN8 software.The phylogenetic tree was constructed using MEGA11 software with the neighbourjoining method (Tamura et al., 2021).
For the microsynteny analysis, we first identified the neighbouring genes of ZmTKPR1-1 and ZmTKPR1-2 in the B73 maize reference genome (AGPv5).Additionally, we obtained the flanking genes of ZmTKPR1-1 and ZmTKPR1-2 orthologues in Arabidopsis, rice, sorghum and wheat from the Phytozome V12 database (https:// phytozome.jgi.doe.gov/pz/portal.html).To determine the microsynteny of ZmTKPR1-1 and ZmTKPR1-2, we conducted multiple sequence alignments around the flanking genes of ZmTKPR1-1 and ZmTKPR1-2, as well as their orthologues in three monocot species and one eudicot species, using the MCScan software.

Subcellular localization of ZmTKPR1-1 and ZmTKPR1-2
The full-length coding sequences (CDS) of ZmTKPR1-1 and ZmTKPR1-2 were amplified by PCR, and the resulting fragments were inserted into the expression vectors pUC19 and pJG185 using the infusion method.The recombinant pUC19 vectors were co-transformed into maize protoplasts with HDEL-mCherry (an endoplasmic reticulum (ER) marker) and GHD7-mCherry (a nuclear marker) (Nelson et al., 2007).The recombinant pJG185 vectors were transiently expressed in tobacco leaves with ZmMs30-YFP, an ER marker (An et al., 2019).DAPI staining was used as a nuclear marker.The GFP-, YFP-, or mCherry-tagged fluorescent signals were detected using a fluorescence confocal scanner microscope (TCS-SP8, Leica).

Quantitative real-time PCR (qPCR) analysis
Total RNA was extracted using TRIzol reagent (Invitrogen), and DNase I (Promega) was used to eliminate genomic DNA contamination.Subsequently, cDNA synthesis was performed using 59 All-In-One RT MasterMix (ABM, Canada) according to the manufacturer's instructions.Quantitative real-time PCR (qRT-PCR) was carried out on a QuantStudio 5 Real-Time PCR system (ABI) using TB GreenTM Premix EX TagTM (TakaRa, Japan) and the corresponding primer set (Table S3).ZmCyanase (Zm00001d032736) and ZmUbi2 (Zm00001d05383) were employed as internal control genes.Each sample was replicated three times biologically, with each replicate performed in triplicate technically.The 2 ÀDDCt method was used to analyse the amplification data, and the results are presented as means AE standard deviation (SD).

Transient dual-luciferase assay
The cloning of the promoter regions of ZmTKPR1-1 and ZmTKPR1-2 (2000 bp and 2224 bp upstream of the ATG start codon respectively) into the pEASY-LUC vector using homologous recombination yielded the reporter constructs proZmTKPR1-1: LUC and proZmTKPR1-2: LUC.Additionally, the CDS region of ZmMYB84 (Zm00001d025664) was inserted into the pRTBD vector, generating the effector construct 35S-Ω: ZmMYB84.Maize protoplast transformation was then performed, and the relative LUC activity was measured as previously described (Hou et al., 2023).

Electrophoretic mobility shift assay (EMSA)
The full-length coding sequence (CDS) of ZmMYB84 was fused with the carboxyl terminal of maltose-binding protein (MBP) and cloned into the vector pMCSG7.Biotin-labelled ZmTKPR1-1 and ZmTKPR1-2 promoter probes were generated by annealing primer pairs ZmTKPR1-1bio-F/ZmTKPR1-1bio-R and ZmTKPR1-2bio-F/ZmTKPR1-2bio-R, respectively.The fusion vector transformation, protein purification and EMSA were carried out following the methods described in a previous study (Hou et al., 2023).Excess core motif-mutated probes were used as competitors to test the specificity of the DNA-protein interaction.

Y2H assay
The matchmaker GAL4 Two-Hybrid System (Clontech) was employed to investigate PPIs.The full-length coding sequences of ZmPKSB, ZmTKPR1-1 and ZmTKPR1-2 were cloned into pGADT7 and pGBKT7 vectors.Amino acid substitution vectors, including pGBKT7-ZmTKPR1-1-9M, pGBKT7-ZmTKPR1-2-7M and pGADT7-ZmTKPR1-1-4M, were constructed using fusion PCR.Co-transformation of a pGADT7/prey plasmid and a pGBKT7/bait plasmid was performed in yeast strain AH109 and verified on selective media according to the manufacturer's instructions.Negative control vectors, pGADT7-T7 and pGBKT7-lam, were used, as well as a positive control with pGADT7-T7 and pGBKT7-53.Subsequently, vector co-transformation into yeast strains and yeast screening were conducted following the methods described in a previous study (An et al., 2020).

BiFC assay
To investigate PPIs, we employed the bimolecular fluorescence complementation (BiFC) assay, following the methods described in a previous study (Walter et al., 2004).The full-length coding sequences (CDSs) of ZmPKSB, ZmTKPR1-1 and ZmTKPR1-2 were fused with the N-terminal (1-155 amino acids) and C-terminal (156-239 amino acids) fragments of yellow fluorescent protein (YFP).These fusion constructs were then cloned into the pUC19-35S vector using homologous recombination (Li et al., 2005).HDEL-mCherry, an ER marker, was used in conjunction with the paired recombinant BiFC plasmids.The maize protoplasts were co-transformed with the BiFC plasmids and the ER marker and incubated in the dark at 28 °C for 12-16 h.YFP fluorescence was subsequently visualized using a confocal laser-scanning microscope (Leica TCS SP8) with excitation at 514 nm and emission detection at 525-565 nm.

Protein purification, enzymatic analysis and LC-MS/MS analysis
The full-length CDSs of ZmTKPR1-1 and its four mutations (À4aa, S128A, Y164A and K168A), as well as ZmTKPR1-2 and its four mutations (À1aa, S130A, Y166A and K170A), were cloned into the pMCSG7 vector, respectively.The vector construction and protein purification procedures followed the methods described in a previous study (Zhang et al., 2021).For enzyme activity assays, 10 lg recombinant protein of ZmPKSB was firstly incubated with C16/18-CoA (0.1 mM) in a reaction buffer (10 mM MgCl 2 , 5 mM ATP, 2.5 mM DDT and 0.1 mM malonyl CoA in 60 mM sodium phosphate buffer) for 30 min at 30 °C to generate triketide and tetraketide.Subsequently, 10 lg ZmTKPR1-1 or 10 lg ZmTKPR1-2 and 1 mM NADPH were added and then incubated for another 1 h at 30 °C to generate reduced tetraketide.To measure the enzyme activities of variants, ZmPKSB and C16/18-CoA were incubated in the reaction buffer at 30 °C for 30 min to generate triketides and tetraketides.After that, the products were equally divided into eight parts, and the recombinant proteins of variants were added to analyse their enzyme activities, respectively.All reactions were stopped with the addition of 20 lL 1 M HCl.Each final product was extracted with 800 lL of ethyl acetate, and the extract was then evaporated overnight in a vacuum.
The experiments utilized an Acquity UPLC system (Waters) connected to a Quattro Premier XE triple quadrupole MS system (Waters Micromass).Reaction products were placed into an Acquity UPLC BEH C18 column (2.1 9 100 mm, 1.7 lm) along with a precolumn (2.1 9 5 mm, 1.7 lm), and then separated through an increasing acetonitrile gradient in water with 0.1% formic acid.The flow rate was set at 0.45 mL/min with gradient conditions being 50%-100% acetonitrile for 25 min, sustained at 100% for 2 min, reduced to 50% for 1 min and then the column was balanced at 50% acetonitrile for 4 min.Full-scan, selected ion recording, daughter scan and multiple reaction monitoring modes were implemented for the analysis (Zhu et al., 2017).

Analysis of anther cutin, wax and internal lipid
Anthers at stage S13 from both the WT and tkpr1-1/-2 were collected and subjected to freeze-drying.The calculation of the anther surface and the determination of the content and component of anther wax, cutin and internal lipid were performed as described previously (An et al., 2020).Metabolon ZmPKSB-ZmTKPR1-1/-2 for maize exine formation 229 Figure S1 The physical maps and target-site information of CRISPR/Cas9 constructs for editing14 lipid metabolic genes.Figure S2 CRISPR/Cas9 mutagenesis and characterization of the derived six lipid metabolic gene mutants and the proportions of aborted pollen grains in the single-gene mutants of ZmTKPR1-1 and ZmTKPR1-2.Figure S3 CRISPR/Cas9 mutagenesis and characterization of the derived six metabolic gene mutants.Figure S4 Cytological observation and TUNEL assay of WT and tkpr1-1/-2 anthers.Figure S5 The amino acid sequence alignment of ZmTKPR1-1, ZmTKPR1-2 and their orthologues from 11 plant species.Figure S6 Subcellular localization of ZmTKPR1-1 and ZmTKPR1-2 in tobacco leaves and qPCR analysis of ZmTKPR1-1, ZmTKPR1-2 and ZmMYB84.Figure S7 Predicted three-dimensional structures of protein complexes for ZmPKSB, ZmTKPR1-1 and ZmTKPR1-2 using AlphaFold2.Figure S8 Analysis of anther cutin, wax and internal lipid contents in WT and tkpr1-1/-2 anthers at stage 13. Figure S9 Expression of cutin-and wax-related genes in WT and tkpr1-1/-2 anthers.Table S1 Transcriptional levels of 14 investigated genes during anther development based on RNA-seq analysis in three maize lines.Table S2 The detailed cutin, wax and internal lipid compositions in WT, pksb and tkpr1-1/-2 anthers.Table S3 Primers used in this study.

Figure 1
Figure 1 Expression patterns of 14 lipid metabolic genes during maize anther development and their CRISPR/Cas9-induced mutations.(a) Expression patterns of 14 maize lipid metabolic genes during anther developmental stages S5 to S12 based on anther transcriptome analysis in B73, Zheng58 and M6007 genetic backgrounds.(b) qRT-PCR analyses of the 14 maize lipid metabolic genes during B73 anther developmental stages S5 to S12.(c) Gene information, CRISPR/Cas9-induced mutation types and their corresponding male-sterility or -fertility phenotypes of the 14 investigated maize lipid metabolic genes.

Figure 5
Figure 5 Interactions of ZmPKSB, ZmTKPR1-1 and ZmTKPR1-2.(a) Protein interactions of ZmPKSB, ZmTKPR1-1 and ZmTKPR1-2 shown by the Y2H assay.(b) BiFC analysis demonstrates the interactions of ZmPKSB, ZmTKPR1-1 and ZmTKPR1-2 in maize protoplasts.Protein name-n and -c indicate nYFP and cYFP fusions, respectively.HDEl-mCherry was used as an ER marker.(c) Protein interactions shown by the Co-IP assay in maize protoplasts.The nYFP-FLAG and ZmbHLH51 were used as negative controls.(d) Protein interactions of amino acid substitution variants that cannot form hydrogen bonds are shown by the Y2H assay.TKPR1-1-9M, TKPR1-2-7M and TKPR1-1-4M indicate nine, seven and four amino acid substitution variants, respectively.(e) qPCR analysis of ZmPKSB in WT and tkpr1-1/-2 anthers and ZmTKPR1-1 and ZmTKPR1-2 in WT and pksb anthers.NS and *** indicate the significant levels of P > 0.05 and P < 0.001 determined by a two-tailed Student's t-test, respectively.For (a and d), T7/P53 and T7/Lam were used as positive and negative controls, respectively.SD-Trp-Leu indicates double dropout medium; SD-Trp-Leu-His-Ade indicates quadruple dropout medium.

Figure 7
Figure 7 Cytological and lipid comparison among WT, pksb and tkpr1-1/-2 anthers and ZmMYB84-regulated ZmPKSB-ZmTKPR1 metabolon affecting anther cuticle and pollen exine formation in maize.(a) TEM observation of pollen exine in WT, pksb and tkpr1-1/-2 at stage S12.(b) SEM observation of anther outer surface in WT, pksb and tkpr1-1/-2 at stage S13.(c) The amount of anther cutin in WT, pksb and tkpr1-1/-2 at stage S13.(d) The amount of anther wax in WT, pksb and tkpr1-1/-2 at stage S13.(e) qPCR analysis of five downregulated genes in tkpr1-1/-2 anthers from stages S8a to S12.(f) qPCR analysis of five upregulated genes in tkpr1-1/-2 anthers from stages S8a to S12.(g) A working model of the ZmMYB84-regulated ZmPKSB-ZmTKPR1 metabolon affecting anther cuticle and pollen exine formation in maize.Ba, baculua; F, foot layer; Te, tectum.For (c to f), *, ** and *** indicate the significant levels of P < 0.05, 0.01 and 0.001 determined by a two-tailed Student's t-test, respectively.For (c and d), the red or blue lines represent the increase or decrease of the cutin and wax monomers in mutant anthers, respectively.The thickness of the lines represents the change in magnitude in terms of increase or decrease.

ª
2023 The Authors.Plant Biotechnology Journal published by Society for Experimental Biology and The Association of Applied Biologists and John Wiley & Sons Ltd., 22, 216-232

ª
2023 The Authors.Plant Biotechnology Journal published by Society for Experimental Biology and The Association of Applied Biologists and John Wiley & Sons Ltd., 22, 216-232

ª
2023 The Authors.Plant Biotechnology Journal published by Society for Experimental Biology and The Association of Applied Biologists and John Wiley & Sons Ltd., 22, 216-232