The Coordinated KNR6–AGAP–ARF1 Complex Modulates Vegetative and Reproductive Traits by Participating in Vesicle Trafficking in Maize

The KERNEL NUMBER PER ROW6 (KNR6)-mediated phosphorylation of an adenosine diphosphate ribosylation factor (Arf) GTPase-activating protein (AGAP) forms a key regulatory module for the numbers of spikelets and kernels in the ear inflorescences of maize (Zea mays L.). However, the action mechanism of the KNR6–AGAP module remains poorly understood. Here, we characterized the AGAP-recruited complex and its roles in maize cellular physiology and agronomically important traits. AGAP and its two interacting Arf GTPase1 (ARF1) members preferentially localized to the Golgi apparatus. The loss-of-function AGAP mutant produced by CRISPR/Cas9 resulted in defective Golgi apparatus with thin and compact cisternae, together with delayed internalization and repressed vesicle agglomeration, leading to defective inflorescences and roots, and dwarfed plants with small leaves. The weak agap mutant was phenotypically similar to knr6, showing short ears with fewer kernels. AGAP interacted with KNR6, and a double mutant produced shorter inflorescence meristems and mature ears than the single agap and knr6 mutants. We hypothesized that the coordinated KNR6–AGAP–ARF1 complex modulates vegetative and reproductive traits by participating in vesicle trafficking in maize. Our findings provide a novel mechanistic insight into the regulation of inflorescence development, and ear length and kernel number, in maize.


Introduction
Maize is a very important grain and feed crop with the highest yield in the world. The grain yield of maize (Zea mays L.) is closely correlated with the number of kernels produced from the ear inflorescence. During normal development, the female inflorescence meristem (IM) gives rise to indeterminate spikelet pair meristems (SPMs). Each SPM differentiates into two determinate spikelet meristems (SMs), and their development is terminated by the formation of floral meristems (FMs). The lower florets produced from FMs are abortive, whereas the fertile upper florets develop into kernels after double fertilization [1][2][3][4]. The phytohormone auxin plays a critical role in the development of inflorescences and florets in maize. The loss of functions of auxin biosynthesis-related genes, such as SPARSE INFLORESCENCE1 [5] and VANISHING TASSEL2 [6], results in greatly reduced numbers of spikelets and florets, shown as sparse or barren inflorescences. In addition to the genes involved in auxin synthesis, those involved in auxin localization and signaling participate in the regulation of axillary meristem (AM) development [7]. Polar auxin transport is mediated by the auxin influx carrier AUXIN/LIKE AUXIN proteins

Sequence Extraction and Phylogenetic Analysis
To predict ARF proteins encoded in the maize genome, the raw ARF hidden Markov model (HMM) file, downloaded from the EMBL-EMI protein database, was used to query the B73 genomic database (Zea_mays.AGPv4.pep.all.fa; http://plants.ensembl.org/, accessed on 2020) using the HMM, then those predicted hits were aligned by ClustalW2 to produce a set of high-quality proteins with E-value < 1 × 10 −20 and an intact GT-Pase domain. The high-quality protein set was used to construct a maize-specific ARF HMM using hmmbuild in the HMMER v3 suite [39]. The maize-specific ARF HMM was used to query the B73 genomic database again, and the hits with E-value < 0.01 were referred to as maize ARF proteins. The 16 sequences were aligned using MUSCLE 3.8.31 (http://www.drive5.com/muscle/downloads.htm, accessed on 2020), and the phylogenetic trees were constructed using MEGA-X 10.1 with the maximum-likelihood method.

Vector Construction and Genetic Transformation
Two guide RNAs (gRNAs), GGCAAUAAGAUUGGUGAGGG and GCUGUGUA-GAAGGAAACCAC ( Figure S1), that target the coding region of KNR6 were designed using CRISPR-P 2.0 (http://crispr.hzau.edu.cn/CRISPR2/, accessed on 2017). Similarly, the gR-NAs GACGGAUUUGAGGCCCAACA and GUGGCUCUCCAGAUCCAAAA, targeting the AGAP gene, were also designed ( Figure S2). These gRNAs were then synthesized by the GeneCreate Company (Wuhan, China). The synthesized gRNAs were transformed independently into ZmUbi-hspCas9 through a recombination reaction using CloneExpress Multis (Vazyme Biotech, Nanjing, China). The KNR6-ZmUbi-hspCas9 and the AGAP-ZmUbi-hspCas9 constructs were transformed into Agrobacterium tumefaciens strain EHA101 and then were introduced independently into immature embryos of maize-inbred line KN5585 through Agrobacterium-mediated transformations [38] by the Wimibio Company (Changzhou, China). Independent T 0 transgenic plants were identified by PCR genotyping of the Bar gene, and the gene-specific editing was identified using gene-specific PCR and sequencing. The PCR primers used are listed in Table S1. KNR6-edited lines, knr6 cr1 and its respective non-transgenic line, KNR6 NT1 , as well as two AGAP-edited lines, agap cr1 and agap cr2 , and their respective non-transgenic lines, AGAP NT1 and AGAP NT2 , were developed.

Plant Materials and Phenotypic Identification
All of the gene-edited lines and their respective non-transgenic lines were phenotyped during spring 2020 in Wuhan (30 • N, 114 • E), China; knr6 cr1 was crossed to agap cr1 and self-crossed during winter 2019 in Sanya (18.34 • N, 109.62 • E), China. Wild-type, two singlemutant and double-mutant individuals were selected from the segregating F 2 population of 607 plants by genotyping KNR6 and AGAP. Individuals of the four genotypes were phenotyped during spring 2020 in Wuhan. The inflorescence meristem lengths (µm) were measured under a scanning electron microscope. Botanical characteristics, including plant height (cm), ear height (cm), and tassel length (cm), were measured during the adult period. The ear length (cm), kernel number per row (KNR), and kernel number per ear were determined on air-dried ears.

Luciferase (Luc) Complementation Image Assay
The full-length coding sequences (CDSs) of AGAP and the ARF1 family genes were cloned into 35S::CLuc and 35S::NLuc independently using the recombinant enzyme combinations KpnI/SalI and BamHI/SalI, respectively. The constructs were introduced into A. tumefaciens strain GV3101 (Weidibo, Shanghai, China), and resulting strains harboring ARF1s-NLuc and AGAP-CLuc were coinfiltrated into Nicotiana benthamiana leaves. After culturing in the lysogeny broth medium supplemented with 100 µg mL −1 Kanamycin and 50 µg mL −1 Rifampicin until the optical density measured at 600 nm reached 0.6, the Agrobacterium pellet was collected and homogenized in suspension solution (10 mM MgCl 2 , 10 mM 4-Morpholineethanesulfonic acid (MES) at pH = 5.6, and 100 µM acetosy-ringone) for 2 h in the dark. The N. benthamiana plants were grown under greenhouse conditions with a 14 h/10 h light/dark photocycle at 23 • C. Leaves of N. benthamiana (5 to 6 weeks old) were inoculated by filtration through a 1 mL syringe with the A. tumefaciens strain carrying the appropriate plasmid construct [40]. Two days after inoculation, 1 mM luciferin was sprayed onto the inoculated leaves. The sprayed leaves were then maintained in the dark for 6 min to quench the fluorescence. A low-light cooled CCD imaging apparatus (Carestream Health, Rochester, NY, USA) was used to capture the luciferase image.

Subcellular Localization
The full-length CDSs of AGAP, ARF1.1, and ARF1.2 were cloned into pS1304-mCherry using SpeI digestion. The Golgi maker ST (AT2G03760) in Arabidopsis thaliana was cloned into pM999-GFP using XbaI. Plasmids were purified using NucleoBond Xtra Midi (Macherey Nagel, Berlin, Germany). The leaf tissues of 10-to 12-day-old maize line B73 grown in a dark incubator at 28 • C were cut into small squares (5 to 10 mm 2 ) with a new razor blade and incubated with 50 mL of enzyme solution (0.25% Macerozyme (Yakult Honsha Co., Ltd., Tokyo, Japan) R-10, 1.0% Cellulase (Yakult Honsha Co., Ltd.) R-10, 400 mM mannitol, 8 mM CaCl 2 , and 5 mM Mes-KOH, pH 5.6). The enzyme solution with leaf tissues was evacuated at~30 kPa for 20 min and then with gentle agitation (30 to 50 rpm) for 5 h at 25 • C. After incubation, the protoplast suspension was filtered through 100 µm mesh and protoplasts were collected by centrifugation at 100 g for 2 min. The pelleted protoplasts were resuspended in 5 to 10 mL of W5 solution (154 mM NaCl, 125 mM CaCl 2 , 5 mM KCl, 5 mM glucose, and 1.5 mM Mes-KOH, pH 5.6), and centrifuged for 2 min at 100 g. The intact protoplasts at the interface were transferred to a new Falcon tube containing 20 mL of W5 solution. The protoplasts were pelleted again by centrifugation at 100 g for 2 min and resuspended in 20 mL of W5 solution. The protoplasts were incubated on ice for 30 min. The protoplasts were pelleted again at 100 g for 2 min and resuspended in MMG solution (400 mM mannitol, 15 mM MgCl 2 , and 5 mM Mes-KOH, pH 5.6) at a density of 5 × 10 6 protoplasts/mL.
To transform DNA into protoplasts, plasmid DNA (10 to 20 µg total at a concentration of 2 mg/mL) was added to 100 µL of protoplast suspension followed by 110 µL of PEG solution (400 mM mannitol, 100 mM Ca(NO 3 ) 2 , and 40% polyethylene glycol 4000). The mixture was mixed gently and incubated for 15 min at room temperature. After incubation, the mixture was diluted with 440 µL of W5 solution. Protoplasts were recovered by centrifugation at 100 g for 2 min, resuspended in 1 mL of WI solution (500 mM mannitol, 20 mM KCl, and 5 mM Mes-KOH, pH 5.6), and incubated at 25 • C in the dark for 12-16 h. Expression of protein was monitored after transformation. Fluorescent signals were visualized using a FV1200 laser scanning confocal fluorescence microscope (Olympus, Tokyo, Japan) with the 488 nm laser line for GFP and 552 nm laser line for mCherry. Data were then processed using Radial Profile Plot in ImageJ 1.53j (Wayne Rasband and contributors, National Institutes of Health, USA)

Transmission Electron Microscopic Observation of Golgi Apparatus
For transmission electron microscopy, 2-3 cm seedling roots were cut into 1-2 mm segments and fixed in 4% glutaraldehyde in 0.05 M sodium phosphate buffer overnight at 4 • C. After fixation, samples were stained with 1% osmium tetroxide, dehydrated through an ethanol series, and embedded in Spurr's resin (London Resin Company, London, UK). Sections (60-70 nm) were cut with a diamond knife, mounted on copper mesh grids, and stained with 4% uranyl acetate followed by Reynolds lead citrate. Sections were then examined using an H-7650 transmission electron microscope (Hitachi, Tokyo, Japan) at 100 kV.

FM4-64 Internalization Assay
The FM4-64 internalization assay was carried out as described by Fan et al. (2013) [41] and Wang et al. (2020) [42]. The agap cr and AGAP NT seedlings were incubated in 5 µM FM4-64 (Invitrogen, Carlsbad, CA, USA) for 10 min at room temperature. Then, the root cells that had been dyed with FM4-64 were transferred into 10 µM BFA. The roots were handsectioned, and the FM4-64 internalization was monitored using a FV1200 laser scanning confocal fluorescence microscope (Olympus) with a 552 nm laser line. The number of FM4-64 labeled puncta internalized per cell and the BFA bodies' size were determined using ImageJ 1.51K.

AGAP Gene Shows Pleiotropy in Vegetative and Reproductive Traits
A maize AGAP is phosphorylated by KNR6 to modulate the length of inflorescence meristems and, in turn, the lengths of mature ears and number of kernels borne per ear [38]. The studied AGAP contains 385 amino acid (aa) residuals with a conserved Arf_GAP domain (AGD) from the 50th to 164th aa and a C2 domain, having five Ca 2+ binding pockets, from the 226th to 372nd aa ( Figure S3A). It is highly homologous with SORBI_3009G121200 in Sorghum bicolor, SEVIR_3G267200v2 in Setaria viridis, and AT3G07940 in A. thaliana ( Figure S3B,C). To further reveal the roles of the KNR6-AGAP module in the ear inflorescence, two loss-of-function agap mutants (agap cr1 and agap cr2 ) were created using CRISPR/Cas9; agap cr1 has a 450-bp deletion between targets 1 and 2, and agap cr2 has a 1-bp deletion near the protospacer-adjacent motif of target 1. The deduced proteins translated from these two types of edited AGAP may lose both the AGD and the C2 domains to produce truncated proteins ( Figure S2A,B). Phenotypic observations revealed dramatic changes between the agap cr1 and its wild-type sibling (AGAP NT1 ). Specifically, agap cr1 produced a shorter plant (94.45 ± 8.76 cm, n = 4) and smaller ear (44.28 ± 6.75 cm) than AGAP NT1 (155.50 ± 7.12 cm and 65 ± 8.83 cm, respectively), and compact and asymmetrical internodes ( Figure 1A,E,G,H). The growth of the agap cr1 ear inflorescence was strongly suppressed, resulting in stunted ears ( Figure 1B,C). The agap cr1 tassel was clawlike, although the tassel length difference between agap cr1 (22.25 ± 3.72 cm) and wild-type (27.15 ± 2.83 cm) plants was not statistically significant (p = 0.08; Figure 1D). The agap cr1 leaves, at 39.86 ± 2.61 cm long and 12.13 ± 0.66 cm wide, were shorter but wider than wild-type leaves (69.82 ± 5.34 cm long and 9.41 ± 0.37 cm wide, Figure 1F,I,J). In the agap cr2 plants, the ear inflorescence and ear traits showed slight but statistically significant changes, with reduced ear lengths and kernel numbers, whereas the agronomically important traits, including plant and leaf architecture, did not significantly differ between agap cr2 and its non-transgenic sibling (AGAP NT2 ) ( Figure S4A-K). Thus, the loss of both AGD and C2 domains in AGAP produced strong effects on vegetative and reproductive traits. Although both mutants were predicted to translate a truncated and non-functional AGAP, the phenotype of the agap cr1 was quite different from that of the agap cr2 .

AGAP Genetically Interacts with KNR6
AGAP physically interacts with KNR6 [38]. To determine the interaction in vivo, we created a KNR6 knock-out mutant (knr6 cr1 ) in which a 261-bp-encoding region between the second and third exons was deleted ( Figure S1). The ear inflorescences and mature ears of the knr6 cr1 showed reduced lengths, along with fewer kernels, relative to those of its non-transgenic sibling (Figure 2A-E). These phenotypes were similar to those of RNAi families reported by Jia et al. (2020) [38], indicating the key roles of KNR6 in the development of ear inflorescences and kernels. Next, the knr6 cr1 mutant was crossed to the weak agap cr2 mutant to develop an F2 segregation population. In this population, we identified the four genotypes, double-mutant knr6 cr1 /agap cr2 , two single mutants, knr6 cr1 /+ and +/agap cr2 , and the +/+ wild type, by genotyping. Compared with the wild-type individuals, each of the single mutants showed shorter ears with fewer kernels ( Figure 2F,G-I). Importantly, compared with the average ear lengths of 10.91 ± 1.19 cm and 10.91 ± 0.95 cm in the single mutants knr6 cr1 /+ and +/agap cr2 , respectively, the average ear length of the double mutant (knr6 cr1 /agap cr2 ) was 9.93 ± 0.76 cm (n = 14), indicating that double mutants had shorter ears than each of the single mutants ( Figure 2H). Similarly, the average kernel number per row (KNR) in the double mutant was 18.73 ± 1.49 (n = 13), which was fewer than in the single mutants knr6 cr1 /+ (20.53 ± 2.41, p = 0.0124) and +/agap cr2 (20.33 ± 2.19, p = 0.0276) ( Figure 2G). Thus, mutations at both genes enhanced the defective phenotypes resulting from the single gene mutation, suggesting that AGAP genetically interacts with KNR6 to influence ear length and KNR.

AGAP Genetically Interacts with KNR6
AGAP physically interacts with KNR6 [38]. To determine the interaction in vivo, we created a KNR6 knock-out mutant (knr6 cr1 ) in which a 261-bp-encoding region between the second and third exons was deleted ( Figure S1). The ear inflorescences and mature ears of the knr6 cr1 showed reduced lengths, along with fewer kernels, relative to those of its nontransgenic sibling (Figure 2A-E). These phenotypes were similar to those of RNAi families reported by Jia et al. (2020) [38], indicating the key roles of KNR6 in the development of ear inflorescences and kernels. Next, the knr6 cr1 mutant was crossed to the weak agap cr2 mutant to develop an F 2 segregation population. In this population, we identified the four genotypes, double-mutant knr6 cr1 /agap cr2 , two single mutants, knr6 cr1 /+ and +/agap cr2 , and the +/+ wild type, by genotyping. Compared with the wild-type individuals, each of the single mutants showed shorter ears with fewer kernels ( Figure 2F,G-I). Importantly, compared with the average ear lengths of 10.91 ± 1.19 cm and 10.91 ± 0.95 cm in the single mutants knr6 cr1 /+ and +/agap cr2 , respectively, the average ear length of the double mutant (knr6 cr1 /agap cr2 ) was 9.93 ± 0.76 cm (n = 14), indicating that double mutants had shorter ears than each of the single mutants ( Figure 2H). Similarly, the average kernel number per row (KNR) in the double mutant was 18.73 ± 1.49 (n = 13), which was fewer than in the single mutants knr6 cr1 /+ (20.53 ± 2.41, p = 0.0124) and +/agap cr2 (20.33 ± 2.19, p = 0.0276) ( Figure 2G). Thus, mutations at both genes enhanced the defective phenotypes resulting from the single gene mutation, suggesting that AGAP genetically interacts with KNR6 to influence ear length and KNR.

AGAP Participates in Vesicle Trafficking
To determine the roles of AGAP, we first examined the transient expression of AGAP-GFP in tobacco (Nicotiana benthamiana L.) leaves. We found that AGAP-mCherry signals were enriched in the cytoplasm ( Figure 3A,B). Furthermore, to understand where the organelles AGAP-mCherry localized, Arabidopsis thaliana ST-GFP was used as a marker to label the Golgi compartments. The AGAP-mCherry signal overlapped the signals of Golgi marker, with the similar relative pixel intensity ( Figure 3C-E), inferring that AGAP localized on membranes of the Golgi apparatus. Moreover, through transmission electron microscope observations, we found that 72% of the Golgi apparatus (n = 11) in the agap cr cells exhibited abnormal architectural features, with thin (64%) or circularized (18%) structures, and compact Golgi cisternae ( Figure 3F-I), instead of the typically flat and linear Golgi cisternae in AGAP NT cells ( Figure 3F). The circularized Golgi apparatus had curved Golgi cisternae, and the TGN/EE was maintained in the Golgi cisternae (Figure 3H). The results indicate that AGAP is required for Golgi organization.

AGAP Participates in Vesicle Trafficking
To determine the roles of AGAP, we first examined the transient expression of AGAP-GFP in tobacco (Nicotiana benthamiana L.) leaves. We found that AGAP-mCherry signals were enriched in the cytoplasm ( Figure 3A,B). Furthermore, to understand where the organelles AGAP-mCherry localized, Arabidopsis thaliana ST-GFP was used as a marker to label the Golgi compartments. The AGAP-mCherry signal overlapped the signals of Golgi marker, with the similar relative pixel intensity ( Figure 3C-E), inferring that AGAP localized on membranes of the Golgi apparatus. Moreover, through transmission electron microscope observations, we found that 72% of the Golgi apparatus (n = 11) in the agap cr cells exhibited abnormal architectural features, with thin (64%) or circularized (18%) structures, and compact Golgi cisternae ( Figure 3F-I), instead of the typically flat and linear Golgi cisternae in AGAP NT cells ( Figure 3F). The circularized Golgi apparatus had curved Golgi cisternae, and the TGN/EE was maintained in the Golgi cisternae ( Figure 3H). The results indicate that AGAP is required for Golgi organization. Because the roles of the Golgi apparatus are to process, sort and transport proteins synthesized by the ER and then send them to specific target membranes or secrete them out of cells, we observed the organelle compartments in the cells of 2-3 cm roots of 3-dayold seedlings using FM4-64 and BFA treatments. The vesicles in cells were identified by FM4-64-labeled fluorescent puntca, and the average vesicle number per cell was counted every 30 min after the FM4-64 treatment. A significant difference in the vesicle number per cell was observed between wild-type and agap cr2 cells at 60 min after the FM4-64 treatment. The vesicle number per cell in agap cr2 cells was 6.704 ± 1.64 (n = 27), less than that in AGAP NT cells (8.556 ± 1.62, n = 27; F = 18.729, p < 0.001), at 60 min after the treatment. The average vesicle numbers slowly increased in agap cr2 cells during the treatment process, reaching 9.333 ± 2.61 (n = 27) and 9.704 ± 2.46 (n = 27) at 90 and 120 min, respectively, after treatment. However, the vesicle number and rate of increase in AGAP NT cells were much greater than in agap cr2 cells (F = 18.433, p < 0.001 at 90 min and F = 95.061, p < 0.001 at 120 min after treatment) ( Figure 4A-C), indicating that the internalization of FM4-64-labeled puntca in the agap cr2 cells was delayed compared with in AGAP NT cells. Because the roles of the Golgi apparatus are to process, sort and transport proteins synthesized by the ER and then send them to specific target membranes or secrete them out of cells, we observed the organelle compartments in the cells of 2-3 cm roots of 3-day-old seedlings using FM4-64 and BFA treatments. The vesicles in cells were identified by FM4-64-labeled fluorescent puntca, and the average vesicle number per cell was counted every 30 min after the FM4-64 treatment. A significant difference in the vesicle number per cell was observed between wild-type and agap cr2 cells at 60 min after the FM4-64 treatment. The vesicle number per cell in agap cr2 cells was 6.704 ± 1.64 (n = 27), less than that in AGAP NT cells (8.556 ± 1.62, n = 27; F = 18.729, p < 0.001), at 60 min after the treatment. The average vesicle numbers slowly increased in agap cr2 cells during the treatment process, reaching 9.333 ± 2.61 (n = 27) and 9.704 ± 2.46 (n = 27) at 90 and 120 min, respectively, after treatment. However, the vesicle number and rate of increase in AGAP NT cells were much greater than in agap cr2 cells (F = 18.433, p < 0.001 at 90 min and F = 95.061, p < 0.001 at 120 min after treatment) ( Figure 4A-C), indicating that the internalization of FM4-64-labeled puntca in the agap cr2 cells was delayed compared with in AGAP NT cells.
The fungal toxin BFA drives the rapid agglomeration of endomembrane compartments and the accumulation of membrane proteins into cellular structures, termed "BFA bodies" [43][44][45]. Therefore, we re-treated those FM4-64-treated cells with BFA and observed the states and sizes of BFA bodies in the root epidermal cells. The fungal toxin BFA drives the rapid agglomeration of endomembrane compart ments and the accumulation of membrane proteins into cellular structures, termed "BFA bodies" [43][44][45]. Therefore, we re-treated those FM4-64-treated cells with BFA and ob served the states and sizes of BFA bodies in the root epidermal cells.

AGAP Physically Interacts with Two Members of the Arf GTPase 1 (ARF1) Protein Subfamily
The AGAPs act by binding to GTPases and contribute to the hydrolysis of GTP-bound ARFs [46,47]. In the maize B73 genome, 16 ZmARFs were annotated and divided into four clades ( Figure 5A). To identify the ZmARFs interacting with the studied AGAP, we isolated five out of six annotated members of the ZmARF1 subfamily by PCR and analyzed their interactions with the AGAP using luciferase complementation image assays. We found only two proteins, Zm00001d043113 (ARF1.1) and Zm00001d008295 (ARF1.2), which interacted with the AGAP (Figure 5B), whereas the remaining three members did not show AGAP-interacting signals, indicating that the AGAP proteins preferentially bind to ARF1.1 and ARF1.2. Additionally, the GTP-bound active form of ARF is recruited to the Golgi compartments [48]. Consequently, the subcellular localization showed that ARF1.1 and ARF1.2 predominantly localized on the Golgi apparatus ( Figure 5C,D). Bioinformatics predictions revealed that neither of the ZmARF1 proteins, nor the AGAP, have transmembrane domains; therefore, we hypothesize that these two ZmARF1 proteins and the AGAP protein may be recruited to the Golgi apparatus by Golgi-bound effectors. The AGAPs act by binding to GTPases and contribute to the hydrolysis of GTPbound ARFs [46,47]. In the maize B73 genome, 16 ZmARFs were annotated and divided into four clades ( Figure 5A). To identify the ZmARFs interacting with the studied AGAP, we isolated five out of six annotated members of the ZmARF1 subfamily by PCR and analyzed their interactions with the AGAP using luciferase complementation image assays. We found only two proteins, Zm00001d043113 (ARF1.1) and Zm00001d008295 (ARF1.2), which interacted with the AGAP (Figure 5B), whereas the remaining three members did not show AGAP-interacting signals, indicating that the AGAP proteins preferentially bind to ARF1.1 and ARF1.2. Additionally, the GTP-bound active form of ARF is recruited to the Golgi compartments [48]. Consequently, the subcellular localization showed that ARF1.1 and ARF1.2 predominantly localized on the Golgi apparatus ( Figure  5C,D). Bioinformatics predictions revealed that neither of the ZmARF1 proteins, nor the AGAP, have transmembrane domains; therefore, we hypothesize that these two ZmARF1 proteins and the AGAP protein may be recruited to the Golgi apparatus by Golgi-bound effectors.

Discussion
The ear inflorescence developmentally originates from the apical meristem of axillary branches after the reproductive transition. The ear inflorescence meristem successively differentiates into indeterminate SPMs, determinate spikelet meristems, and then terminates with FM production [49]. The ear inflorescence meristem axially generates a variable number of florets and, in turn, kernels, which are developed from florets after double fertilization. Therefore, the number of kernels arranged axially on the ear is agronomically referred to as KNR. Undoubtedly, KNR is closely correlated with the determinacy of SPMs during ear inflorescence development. A complex functional hierarchy of genes participates in the regulation of SPM determinacy in maize, including genes involved in the CLAVATA-WUSCHEL (CLV-WUS) negative feedback loop, phytohormone biosynthesis and signaling, and in microRNA-mediated post-transcriptional regulation [12,[50][51][52]. A recent study found that KNR6 is a pleiotropic quantitative trait locus for ear length and KNR, and its underlying gene encodes a serine/threonine protein kinase that phosphorylates an AGAP that controls GTP-binding protein activity levels [38]. Because of the phosphorylation roles of serine/threonine protein kinases on small GTPase-coupled receptors [36,53] and maize AGAP [38], and the functions of AGAP in the transport of auxin in rice and Arabidopsis [16,45,54,55], it has been proposed that KNR6 acts in auxin-dependent inflorescence development by mediating AGAP phosphorylation [38]. However, downstream pathways and biological processes regulated by phosphorylated AGAP, which will expand our knowledge of the molecular regulation of reproductive axillary meristem determinacy, are relatively unknown in maize.
The AGAP family of genes is involved in plant organ differentiation and formation [16,23,25]. In Arabidopsis, an AGAP, VAN3/AGD3 controls vein patterning by participating in the trans-Golgi network [23]. A class II AGAP protein, Arabidopsis RPA/AGD10, modulates root hair development by activating the ARF1 subfamily of proteins [25]. In rice, OsAGAP negatively controls root growth and development by modulating the auxin influx pathway [16]. In this study, we found that a weak mutant of maize AGAP alters ear inflorescence-related traits, producing shorter inflorescence meristem and ear length, together with fewer florets and kernels. The strong mutant of maize AGAP showed pleiotropic effects on vegetative and reproductive traits, producing dwarfed plants, asymmetric internodes, and dramatically repressed ear development. We speculated that the 1-bp deletion (agap cr2 ) at the DNA level may be partially repaired or bypassed during translation to encode a functional AGAP protein, resulting in a weak mutant phenotype, whereas the large deletion at the DNA level (agap cr1 ) is not significantly repaired, leading to a strong mutant phenotype. The genetic effects of the agap mutant on the ear traits were enhanced by the knr6 mutant, providing in vivo evidence for KNR6-AGAP protein interactions. These results demonstrate that the KNR6-AGAP complex is required for vegetative and reproductive development in maize.
Generally, AGAPs, such as AGD7 and NEV/AGD5 in Arabidopsis, localize on membranes of the Golgi apparatus, and they participate in membrane trafficking from the Golgi apparatus to ER [27,56]. As in several known AGAP mutants [28,57], the Golgi apparatus showed a defective structure, the internalization of FM4-64 labeled puntca was delayed, and the agglomeration of the typical BFA bodies was not observed in the agap cells, indicating that the involvement of AGAP is required for intact Golgi structures and functions in vesicle trafficking. In addition to AGAP, GTPases are other key factors involved in endocytosis and the Golgi apparatus-to-ER retrograde trafficking in plants [27,[58][59][60]. The Arf_GAP domain in AGAP proteins binds to ARFs and then catalyzes active GTP-bound ARFs into inactive GDP-bound ARFs, whereas GDP-bound ARFs are activated by Arf-GEF [60,61]. In Arabidopsis, VAN3/AGD3, AGD7, and RPA/AGD10 bind to ARF1 and then regulate ARF1 activity [23,25,27]. AGD8 and AGD9 also recruit GDP-bound ARF1 to Golgi apparatus [57]. We found that two maize ARF1 subfamily proteins physically interact with AGAP. In these cases, AGAP functions may be achieved by interactions with ARF1 proteins. Thus, the signaling mediated by the two ARF1 proteins requires the involvement of AGAP in maize. We also used the pBridge vector system to find that KNR6, AGAP, and ZmARFs could not form a ternary protein complex ( Figure S5), although the protein-protein interactions were detected between KNR6 and AGAP, AGAP, and ZmARFs.
Using our results, we established a model of the KNR6-AGAP-ARF1 complexmediated regulation of vegetative and reproductive traits in maize through its participation in vesicle trafficking ( Figure 6). We propose that the serine/threonine-protein kinase KNR6 binds to and then phosphorylates the AGAP, and the phosphorylated AGAP interacts with two specific ARF1 proteins to form the KNR6-AGAP-ARF1 complex. The multi-factor complex is recruited by a specific set of effectors on the Golgi apparatus to hydrolyze GTP into GDP, while Arf-GEF catalyzes the displacement of pre-bound GDP, allowing its replacement with GTP [60,61]. The GTP-bound ARF1 is active, whereas the GDP-bound ARF1 is inactive. The inactive ARF1 is released from membranes after altering its conformation [62,63]. The small GTPase cycle is widely considered a switch mechanism in membrane trafficking [64]. In the agap cr1 cells, GTP-bound ARF1 cannot be released from the Golgi apparatus because the GTP-GDP nucleotide cycle is interrupted, which further represses the internalization and agglomeration of vesicles and damages Golgi compartments. Thus, the failure of the ARF1 cycle interrupts the organization of vesicles from the Golgi apparatus and, in turn, causes cytological, physiological, and/or developmental malfunction through a set of unknown pathways. A proposed downstream pathway is an auxin-mediated regulatory mechanism, which remains to be further investigated.
Cells 2021, 10, x FOR PEER REVIEW 12 of 16 that KNR6, AGAP, and ZmARFs could not form a ternary protein complex ( Figure S5), although the protein-protein interactions were detected between KNR6 and AGAP, AGAP, and ZmARFs. Using our results, we established a model of the KNR6-AGAP-ARF1 complex-mediated regulation of vegetative and reproductive traits in maize through its participation in vesicle trafficking ( Figure 6). We propose that the serine/threonine-protein kinase KNR6 binds to and then phosphorylates the AGAP, and the phosphorylated AGAP interacts with two specific ARF1 proteins to form the KNR6-AGAP-ARF1 complex. The multifactor complex is recruited by a specific set of effectors on the Golgi apparatus to hydrolyze GTP into GDP, while Arf-GEF catalyzes the displacement of pre-bound GDP, allowing its replacement with GTP [60,61]. The GTP-bound ARF1 is active, whereas the GDPbound ARF1 is inactive. The inactive ARF1 is released from membranes after altering its conformation [62,63]. The small GTPase cycle is widely considered a switch mechanism in membrane trafficking [64]. In the agap cr1 cells, GTP-bound ARF1 cannot be released from the Golgi apparatus because the GTP-GDP nucleotide cycle is interrupted, which further represses the internalization and agglomeration of vesicles and damages Golgi compartments. Thus, the failure of the ARF1 cycle interrupts the organization of vesicles from the Golgi apparatus and, in turn, causes cytological, physiological, and/or developmental malfunction through a set of unknown pathways. A proposed downstream pathway is an auxin-mediated regulatory mechanism, which remains to be further investigated. Figure 6. A potential working model of the KNR6-AGAP-ARF1 complex in maize. Under normal conditions, KNR6 binds to and phosphorylates AGAP NT , and the phosphorylated AGAP interacts with ARF1, leading to the release of vesicles from membranes. When treated with BFA, vesicles from exocytosis and ER-to-Golgi apparatus trafficking are disrupted, with the former gathering into TGN/EE and the latter forming BFA bodies. In agap cr cells, endocytosis and Golgi apparatusto-ER retrograde trafficking involving AGAP are deficient. The vesicle efflux from TGN/EE is greater during exocytosis than the protein influx into TGN/EE during endocytosis, leading to the smaller TGN/EE. The number of vesicles transported from the ER to Golgi apparatus is greater than the number of vesicles transported from the Golgi apparatus to ER, leading to larger Golgi cisternae. When treated with BFA, the BFA bodies that originate from endocytosis and Golgi apparatus-to-ER trafficking are inhibited; therefore, the agglomeration of BFA bodies is delayed compared with in AGAP NT cells. Figure 6. A potential working model of the KNR6-AGAP-ARF1 complex in maize. Under normal conditions, KNR6 binds to and phosphorylates AGAP NT , and the phosphorylated AGAP interacts with ARF1, leading to the release of vesicles from membranes. When treated with BFA, vesicles from exocytosis and ER-to-Golgi apparatus trafficking are disrupted, with the former gathering into TGN/EE and the latter forming BFA bodies. In agap cr cells, endocytosis and Golgi apparatus-to-ER retrograde trafficking involving AGAP are deficient. The vesicle efflux from TGN/EE is greater during exocytosis than the protein influx into TGN/EE during endocytosis, leading to the smaller TGN/EE. The number of vesicles transported from the ER to Golgi apparatus is greater than the number of vesicles transported from the Golgi apparatus to ER, leading to larger Golgi cisternae. When treated with BFA, the BFA bodies that originate from endocytosis and Golgi apparatus-to-ER trafficking are inhibited; therefore, the agglomeration of BFA bodies is delayed compared with in AGAP NT cells.

Supplementary Materials:
The following are available online at https://www.mdpi.com/article/10 .3390/cells10102601/s1. Table S1: Primers used in this study, Figure S1: The guide RNAs of KNR6 and editing types in KNR6 knockout lines, Figure S2: Guide RNAs of AGAP and editing types in AGAP knockout lines, Figure S3: Protein structure, phylogenetic tree, and conserved domain analyses of AGAP, Figure S4: Vegetative and reproductive traits of AGAP NT2 and agap cr2 , Figure S5: Two maize ARF1 proteins, KNR6 and AGAP could not form a ternary protein complex.