Genetic analysis of the barley variegation mutant, grandpa1.a

Background Providing the photosynthesis factory for plants, chloroplasts are critical for crop biomass and economic yield. However, chloroplast development is a complicated process, coordinated by the cross-communication between the nucleus and plastids, and the underlying biogenesis mechanism has not been fully revealed. Variegation mutants have provided ideal models to identify genes or factors involved in chloroplast development. Well-developed chloroplasts are present in the green tissue areas, while the white areas contain undifferentiated plastids that are deficient in chlorophyll. Unlike albino plants, variegation mutants survive to maturity and enable investigation into the signaling pathways underlying chloroplast biogenesis. The allelic variegated mutants in barley, grandpa 1 (gpa1), have long been identified but have not been genetically characterized. Results We characterized and genetically analyzed the grandpa1.a (gpa1.a) mutant. The chloroplast ultrastructure was evaluated using transmission electron microscopy (TEM), and it was confirmed that chloroplast biogenesis was disrupted in the white sections of gpa1.a. To determine the precise position of Gpa1, a high-resolution genetic map was constructed. Segregating individuals were genotyped with the barley 50 k iSelect SNP Array, and the linked SNPs were converted to PCR-based markers for genetic mapping. The Gpa1 gene was mapped to chromosome 2H within a gene cluster functionally related to photosynthesis or chloroplast differentiation. In the variegated gpa1.a mutant, we identified a large deletion in this gene cluster that eliminates a putative plastid terminal oxidase (PTOX). Conclusions Here we characterized and genetically mapped the gpa1.a mutation causing a variegation phenotype in barley. The PTOX-encoding gene in the delimited region is a promising candidate for Gpa1. Therefore, the present study provides a foundation for the cloning of Gpa1, which will elevate our understanding of the molecular mechanisms underlying chloroplast biogenesis, particularly in monocot plants. Supplementary Information The online version contains supplementary material available at 10.1186/s12870-021-02915-9.


Background
As the characteristic organelle in plant cells, the chloroplast is critical for plant photosynthesis and has a significant impact on biomass and economic yield. An increasing number of studies have revealed that chloroplasts also make important contributions to plant immunity through the synthesis of secondary metabolites and defense phytohormones, such as reactive oxygen species, nitric oxide, jasmonic acid, and salicylic acid [1,2]. To restrict pathogen infection, chloroplasts can navigate to the penetration site and directly suppress host cell invasion [3]. Therefore, the understanding of chloroplast biogenesis is necessary to meet the increasing food demand under rising population pressure.
Chloroplasts are differentiated from undeveloped plastids, which contain undifferentiated vesicles and lack stacked thylakoids (grana), the mounting-platform for chlorophyll. Chloroplast biogenesis is highly complex, being orchestrated by anterograde (nucleus to chloroplast) and retrograde (chloroplast to nucleus) signaling [4,5]. More than 95% of the ∼3000 proteins found in chloroplasts are encoded by nuclear genes and imported into chloroplast following synthesis in the cytosol, suggesting that chloroplast development is predominantly controlled by the nuclear genome [6]. Therefore, identification and functional characterization of such nuclear genes is important to understand the regulatory mechanisms underlying chloroplast biosynthesis. Variegation mutants have provided ideal models to identify genes or factors involved in chloroplast development [7]. Welldeveloped chloroplasts are present in the green tissue areas, while the white areas contain undifferentiated plastids that are deficient in chlorophyll. Unlike albino plants, variegation mutants survive to set seed and enable investigation of cross-communication between the nucleus and plastids.
Two representative Arabidopsis variegation mutants, immutans (im) and variegated 2 (var2), have been characterized and provided fundamental perspectives to the understanding of chloroplast biogenesis in plants [8][9][10][11][12]. The im mutant is caused by loss-of-function of a nuclear-encoded plastid terminal oxidase (PTOX), normally present in the thylakoid membranes. This plastoquinol oxidase also has homology to the alternative oxidase (AOX) in mitochondria [8,9]. The VAR2 gene, also known as Filamentous temperature-sensitive H2 (FtsH2), encodes a chloroplast-targeting ATP-dependent zinc metalloprotease homologous to the E. coli FtsH [13]. Both IM and VAR2 play roles in photoprotection and in regulation of redox state of the photosynthetic electron transport chain [8,10,14,15]. Lack of these proteins resulted in photodamaged/photooxidized plastids under high light, particularly in the white sections. Moreover, FtsH-mediated proteolysis is involved in retrograde signaling activated by ROS [16].
The green sections containing competent chloroplasts in variegation mutants may indicate the existence of compensatory mechanisms to escape the defect of mutation. Suppressor screening in the im and var2 backgrounds has identified a few second-site mutations restoring the variegation phenotype. Suppressors of im include a thylakoid membrane protein (Chlororespiratory reduction 2-2, Crr2-2) and a plant combinatorial and modular protein (PCMP) family member (Proton Gradient Regulation 5, PGR5) [17]. The PCMP family is closely related to pentatricopeptide repeat (PPR) proteins functioning in the editing and maturation of organellar RNA. Both Crr2-2 and PGR5 are also required for alternative electron transport pathways that alleviate photodamage during chloroplast biogenesis and photosynthesis [18]. Most of the identified var2 suppressors are involved in chloroplast translation or rRNA processing and editing, such as a chloroplastlocalized pseudouridine synthase (Suppressor of Variega-tion1, SVR1), a ClpR1 subunit of the chloroplast ClpP/R protease (SVR2), a chloroplast translation initiation factor 2 (Fu-gaeri1, FUG1), a chloroplast translation elongation factor EF-G (Snow Cotyledon 1, SCO1), and a PPR protein (SVR7) [19]. Although variegated mutants and suppressor screening enabled the cloning of many genes involved in chloroplast development, there are still major gaps in the knowledge of chloroplast biogenesis and the variegation mechanism.
Due to the difference in chloroplast development between monocotyledonous and dicotyledonous plants, variegation mutants of monocots exhibit striping phenotype with alternating white and green bands on the leaf. Barley (Hordeum vulgare subsp. vulgare) is the fourth most important cereal crop which is used as feed grain, human food, and raw material for the malting and brewing industry. It is also a valuable model monocot for plant genomics research. Chemical-and radiation-mediated mutagenesis has created abundant genetic material for barley improvement and genomic research (http://www.nordgen.org/). Introgression of characterized mutations from various sources into the common background, Bowman, has generated a series of near-isogenic lines (NILs) with single mutated alleles, providing a powerful tool for rapid gene identification [20]. Several variegated mutants have been identified but few have been genetically characterized [21], including the allelic mutants of grandpa (gpa or gp) identified in 1940s [22][23][24].
An allelism test and linkage study with morphological markers anchored the Gpa gene onto the long arm of chromosome 2 (2HL) [20,25]. However, the genetic control of the gpa mutants has not been investigated. The gpa1.a allele was caused by a spontaneous mutation in cultivar Lyallpur (GSHO 519), and it was introduced into the Bowman background with introgression [20]. Using BW397, a NIL of Bowman carrying the gpa1.a mutation, we characterized and finely mapped the Gpa1 gene in the present study. It was revealed that chloroplast biogenesis is defective at the white stripes in the mutant, and the Gpa1 gene is located in a gene cluster functionally related to photosynthesis or chloroplast differentiation. One gene in the delimited region codes for a putative PTOX homologous to IM of Arabidopsis, and we found that a large deletion occurring to this gene totally disrupted its function in BW397. Therefore, this PTOX-encoding gene was considered as a promising candidate for Gpa1. The high-resolution genetic map provided here lays the foundation for the cloning of this gene, which will further our understanding of molecular mechanisms underlying chloroplast biosynthesis.

Phenotype characterization of the gpa1 mutant
Normally, the seedlings of gpa1 have an albino first leaf and display chlorosis at the tip of the second leaf (Fig. 1a). A striped pattern then develops with expansion of the second leaf ( Fig. 1a and b). Striping occasionally occurs on the third and subsequent leaves. Consistent with the visual difference, chemical analysis also indicated that levels of chlorophyll a, b and total in striped leaves of BW397 were significantly lower than that of WT Bowman plants (Fig. 1c). An adult BW397 plant produces albino or striped flag leaf, spike, awns and even anthers ( Fig. 1d and Additional Fig. 1), and the mutant is much shorter than WT (Fig. 1d). The mutants display reduced fertility with an average of 4 seeds/spike compared to 18 seeds/spike in WT (Fig. 1e). Seed size of BW397, measured as 100-seed weight, is only 3/4 that of the WT's ( Fig. 1f). In addition to the striped leaf phenotype, the gpa1 mutation causes a systemic effect on barley growth and development.
To confirm the striped phenotype is associated with defects in chloroplast biogenesis, we compared the chloroplast ultrastructure of Bowman and BW397 using transmission electron microscopy (TEM) (Fig. 2). As expected, numerous normal chloroplasts with well-organized thylakoids are developed in the WT cells ( Fig. 2a and b). Thylakoids are arranged in grana stacks, which are connected by stroma lamellae (Fig. 2c). On the contrary, BW397 cells in the albino sections contain few undeveloped plastids (Fig.  2d). Lamellae, acting like the skeleton in a normal chloroplast, is not well developed in the mutant (Fig. 2d, e, and f). Stacked grana and thylakoid membranes are also missing in these undifferentiated plastids ( Fig. 2e and f). Therefore, chloroplast biogenesis in the white stripes of BW397 is abolished by the gpa1 mutation.

Genetic mapping of Gpa1
A total of 510 F 2 plants derived from the cross between Bowman and BW397 were used for genetic mapping of Gpa1. Of those, 133 were striped, while the remaining 377 F 2 plants exhibited normal green leaf color. The segregation ratio of striped/green fits 1:3 (χ 2 = 0.316, df = 1, and P = 0.57), suggesting that the gpa1 mutation is monofactorial recessive. As an immediate strategy for gene localization and marker discovery, SNP array analysis was used to genotype 48 F 2 plants (24 striped and 24 green) together with the parental lines. As the Gpa1 gene was anchored to 2H [20], we focused on the SNPs  Table 1).
Initial genotype analysis of 48 F 2 progeny revealed that the Gpa1 gene was located in~5.56 Mb region flanked by two array SNPs, JHI-Hv50k-2016-139,629 and JHI-Hv50k-2016-142,540 (Additional Table 1). To increase the resolution of mapping, we enriched this region with SSR markers previously co-located to the stripped phenotype [20], known SNPs on 2H consensus map [26], and specific 50 k markers within this region ( Table 1). The analysis of an additional 104 F 2 progeny genotyped with these markers further delimited the Gpa1 gene between M4 and M6 (Fig. 3a). When we enlarged the segregating population to 510 F 2 plants, the Gpa1 region was narrowed down to a 410-kb region by M19 and M8 (Fig. 3b), where one co-segregating SNP, M20, was identified.

Physical localization of Gpa1
Fifteen putative protein-coding genes were identified in the fine-mapping interval according to the reference genome assembly [27] (Table 2) Blue copper protein function as an electron shuffler in electron transfer reactions, such as biological nitrogen fixation, respiration and photosynthesis. Structure analysis indicated the putative blue copper proteins, G7, G9, G10, and G12, in the Gpa1 region contain a domain identified in plastocyanin, the long-range electron carrier between photosystems II and I [28]. Alternative oxidase is involved in the regulation of redox state of the electron transport chain in organelles [8,9,18]. Particularly, the putative coding product of G14 shares high homology with IM (AT4G22260), the plastid terminal oxidase in Arabidopsis [8,9]. Therefore, G14 was named HvPTOX hereafter. For the G11 gene, two different products, one PPR protein and one DNA gyrase, were predicted in the sense and antisense strand, respectively (Additional Fig. 2). The putative introns of the DNA gyrase gene contain coding exons in the reverse complementary strand for the predicted PRR  Fig. 2). The EST match for the putative DNA gyrase was identified (FD525137), but we did not find ESTs for the predicted PPR from the available databases. DNA gyrase or topoisomerase has been linked to regulation of DNA replication and transcription during chloroplast biogenesis [29]. Although lacking EST matches, the predicted PPR protein is highly homologous to SVR7, one of the suppressors of the Arabidopsis var2 mutation [19]. Therefore, members of this functionally related gene cluster were selected for further analysis.

Selection of the Gpa1 candidate
The four blue copper proteins within this region share at least 75% sequence identity, and the coding products of G9 and G12 vary by only one amino acid (aa) substitution. We speculated that these blue copper proteins may function redundantly, and mutation on one gene may not result in apparent phenotype change. Moreover, the AOX encoded by G13 is incomplete, and was eliminated from further analysis. Using Bowman, BW397, Lyallpur and its isogenic mutant GSHO 519, we focused on identifying sequence polymorphism between the gene alleles in G11 and G14 (HvPTOX). Although a few SNPs were detected between the Bowman and Morex alleles of G11, Bowman, BW397, GSHO 519 and Lyallpur share identical genomic sequences including the coding region. This suggested that G11 might not be one of the candidates for Gpa1.
Gene prediction and EST matches (DK626738, DK619131, BF626913, RUS39D06w and HB15J15r) showed that HvPTOX contains 10 coding exons and 9 introns (Fig. 4a). The full-length coding region was successfully amplified in Bowman and Lyallpur, but   Table 2). We only obtained the 3′ region of the coding sequence in the mutants (Fig. 4b, Additional Table 2). To capture the full length of HvPTOX, we conducted FPNI-PCR to acquire the coding sequence at the 5′ region. Sequencing of the product derived from FPNI-PCR indicated that the first three and almost half of the fourth exon were missing in the mutated allele. However, the sequence proximal to the fourth exon of HvPTOX in BW397 cannot be aligned to the assembled reference genome, and it is not homologous to any known protein coding sequences or transposable elements.
To determine if the gene mutation was caused by insertion or deletion, we designed more markers at the potential promoter and far-upstream regions (2.5 kb upstream of the start codon) of HvPTOX. Our result indicated that those markers were all missing in mutants (Fig. 4c). In line with the deletion of the promoter, RT-PCR confirmed that HvPTOX was not expressed in mutants (Fig. 4d). Its expression in mutants could not be detected even with 30 PCR cycles, in contrast to the expression abundance indicated by the 25 cycles in wild types (Fig. 4d). Quantitative real-time PCR (qRT-PCR) also revealed that the expression of gpa1 is significantly suppressed in mutants (Additional Fig. 5). Although the size of the deletion was unknown, HvPTOX structure, expression and thus function was totally disrupted in mutants. Therefore, the HvPTOX gene (G4/HORVU.MOREX.r2.2HG0177270) was selected as a strong candidate. The identity of Gpa1 still needs to be further verified in homologous or heterologous systems.

Discussion
Plant chloroplast biogenesis is important for biomass and economic yield. Variegation mutants provide an ideal model to understand the development of functional chloroplasts. The allelic mutants of gpa has long been identified, but their genetic basis underlying variegation remains unknown. In the present study, we characterized the chloroplast ultrastructure of the gpa1.a mutant and localized the corresponding locus. The lack of chlorophyll in the white stripes of the mutant is caused by the interruption of chloroplast biogenesis. The Gpa1 gene was delimited within a 410-kb region containing a cluster of functionally related genes. A genomic deletion in the PTOXencoding gene in mutants abolished its gene function, suggesting it is a promising candidate for Gpa1.
Many functionally related genes are distributed nonrandomly throughout the genome as functional clusters in eukaryotes [30]. In plants, many biosynthetic gene clusters for secondary metabolism have been identified [31]. Those clusters may have arose from recruitment of genes from elsewhere in the genome through duplication and neofunctionalization, but not by horizontal gene transfer from prokaryotes [32]. From an evolutionary perspective, the physical clustering of functionally related genes may facilitate coordinated gene expression and prevent the breakup of co-adapted alleles during recombination [32,33]. The cluster members in the Gpa1 region may target chloroplast for photosynthesis or chloroplast development. However, their actual functional roles require further investigation. Using CRISPR technology to knock out the clustered genes individually will possibly solve this puzzling question.  Table 2. Full-length of the coding region could not be obtained with the primer combination of F2 and R1 in mutants, but half of that at the 3′ was successfully amplified with primers F3 and R1 (b). The promoter (F4 and R2) and a region at the 2.5 kb upstream of the start codon (F5 and R3) were deleted altogether in mutants (c). RT-PCR analysis with 25, 28 and 30 PCR cycles failed to detect the expression of Hvptox in mutants. Actin was used as the internal control. The original gel images were included in Additional Figs. 3 and 4 Although G11 may not be Gpa1, it is interesting to discover if two different genes are derived from the same genomic fragment. Eukaryotic operon-like transcription has been observed, and one mRNA can be translated to several separate products in the cytoplasm [34,35]. However, there have been no reports for two independent genes harbored on reverse complementary strands. The lack of EST matches for the PRR-coding gene suggests that it may not be a true gene, but this conclusion needs to be confirmed with rapid amplification of cDNA ends (RACE).
The PTOX protein encoded by G14 shares high identity with IM in Arabidopsis. The variegation mutant im in Arabidopsis and the orthologous ghost mutant in tomato both display the loss of PTOX caused defect in chloroplast biosynthesis [9,36]. It was discovered that PTOX functions as a terminal oxidase in controlling the redox state of the plastoquinone (PQ) pool in developing and mature thylakoids [37,38]. PTOX is involved in transferring electrons from plastoquinol (PQH2) to molecular oxygen, forming water and PQ [38]. The role in regulation of the redox state of the photosynthetic apparatus makes PTOX crucial for a growing number of biochemical pathways, such as the desaturation reactions in carotenoid biosynthesis, chlororespiration, PSI cyclic electron flow, and photoprotection [37,39,40].
Biosynthesis of carotenoid is impaired in the im mutant due to the lack of phytoene desaturation (PDS), and the intermediate phytoene is accumulated in the white sections of leaves [8,9]. The white areas of im might accumulate white photodamaged plasmids due to the lack of photoprotective carotenoid [10]. It was also demonstrated that the redox state of the PQ pool may control chloroplast biogenesis as a potent initiator of retrograde signaling [37,41]. Under these scenarios, the predicted HvPTOX (HORVU.MOREX.r2.2HG0177270) in the Gpa1 region was designated a strong candidate. This series of gpa mutants were identified almost 80 years ago, and fine localization of Gpa1 in this study has provided candidates for functional validation, facilitating cloning of this long-elusive gene.

Conclusions
In summary, we have characterized and genetically mapped the gpa1.a mutation causing a variegation phenotype in barley. Our results indicate that chloroplast biogenesis is defective in white sections of the mutant, and the gpa1 mutation imposes a systemic effect on barley growth and development. The Gpa1 gene was mapped to chromosome 2H within a 410 kb region. In addition, we have shown that Gpa1 is harbored in a gene cluster functionally related to photosynthesis or the chloroplast. Further studies have indicated that a genomic deletion disrupts the expression and function of the PTOX-encoding gene, G14/HORVU.MOREX.r2.2HG0177270. Therefore, the present study paves the way for the cloning of Gpa1, which will improve our understanding of the molecular mechanisms underlying chloroplast biogenesis.

Plant materials
An F 2 segregating population derived from the cross between Bowman (wild type) and BW397 (the gpa1.a mutant) was used for genetic mapping of the Gpa1 gene. The gpa1.a mutation was donated by GSHO519 carrying the genetic background of Lyallpur. Seeds of Bowman, BW397, GSHO519 and Lyallpur were obtained from the USDA-ARS National Plant Germplasm System (NPGS). A total of 510 F 2 plants were used to generate a genetic map. All plants together with parents were grown in a greenhouse under a 16 h light/8 h dark photoperiod at 25°C. Phenotyping was conducted at the 1st leaf stage and repeated at the boot stage. Phenotype of F 2 recombinants delimiting the Gpa1 region was confirmed with 30 F 3 individuals.

DNA extraction
DNA was extracted according to the CTAB protocol [42]. Around 100 mg leaf samples were collected from plants at the three-leaf stage and quantified using a NanoDrop spectrophotometer (NanoDrop 8000, Thermo Fisher Scientific) according to the manufacturer's instructions. The final concentration was adjusted to 100 ng/uL for PCR application.

SNP genotyping and marker development
Forty-eight F 2 s (24 each for wild type and stripped) and parental lines were first genotyped with a barley 50 k iSelect SNP Array [43]. Marker positions are based on the barley pseudo-molecule assembly of Morex V1 [44]. Genotype calling was performed with the de novo calling algorithm in GenomeStudio (Illumina). Clusters of polymorphic SNPs were inspected and manually adjusted if necessary. The linked SNPs were used to develop semithermal asymmetric reverse PCR (STARP) markers to genotype the F 2 population [45]. PCR was conducted in a 10 μl reaction volume with Taq DNA polymerase (New England Biolab) according to the manufacturer's instructions. Sequences of priming element-adjustable primer (PEA-primer) 1 and 2 are 5′-AGCTGGTT-SP9-GCAACAGGAACCAGCTATGAC-3′ and 5′-ACTGCT CAAGAG-SP9-GACGCAAGTGAGCAGTATGAC-3′, respectively [45]. The thermal amplification program followed a touchdown protocol as described previously [45,46]. Stained with GelRed™ nucleic acid stain (Milli-poreSigma), amplicons were analyzed with 6% polyacrylamide gel which was imaged using a Typhoon™ FLA 9500 variable mode laser scanner (GE Healthcare Life Sciences, Marlborough, MA). The markers used in the present study are listed in Table 1.

Physical mapping and gene prediction
The genome sequences of barley cv Morex v2 and Golden Promise v1 were used for marker localization and physical mapping [27,47,48]. Gene predication and annotation was conducted with the programs of FGEN ESH and Pfam 32.0, respectively [49,50]. Gene annotation was also confirmed with the BLASTP program.
Fluorescent qRT-PCR was performed with three replications using the 7500 Fast Real-Time PCR system (Applied Biosystems, Grand Island, NY, USA). Amplification was conducted in a 10 μl reaction volume containing 2 μL cDNA (equivalent to 50 ng of total RNA), 1.5 μL of each gene-specific primer (2.5 μM), and 5 μL of SYBR Select Master Mix (Applied Biosystems). The actin gene was used as the internal control for real-time analysis and was amplified with primers 5′-AACTATGTTC CCAGGTATCGC-3′ and 5′-GACTCGTCGTACTCAT CCTTG-3′. Primers 5′-GAAACTACCAGCTCCAGA GG-3′ and 5′-GCGTTTGACATGCCTTCATC-3′ were used to target Gpa1 gene expression. Amplification conditions used were the same as described previously [51].

Chlorophyll measurement
Chlorophyll contents were measured by spectrophotometric determination with 4 replications [52]. Briefly, 0.2 g fresh leaf tissue of wild type and mutant plants at booting stage was collected and soaked in 5 ml of 95% ethanol in the dark for 48 h. After 5 min of centrifugation at 3000 rpm, the residual plant debris was removed. Supernatant was measured with a spectrophotometer (Beckman DU 7400) at 663 nm for chlorophyll a and 645 nm for chlorophyll b.

Fusion primer and nested integrated PCR (FPNI-PCR)
FPNI-PCR was performed for chromosome walking to identify the unknown genomic region [53]. Gene-specific primers used for primary, secondary and tertiary PCR respectively are, GSP1, 5′-CTGCACTCAATAGGCAGG GTGT-3′; GSP2, 5′-ACCGAGTCGCAACCAGCCTT-3′; and GSP3, 5′-TTGCCACCCAACGCCTGACA-3′. Nine fusion arbitrary degenerate primers (FP1-9) and FP-specific primers (FSP1 and FSP2) were designed according to Wang et al. (2011). LA Taq (Takara Bio USA Inc., Mountain View, CA) was used in the first round of PCR. The 20 μl reactions consisted of 10 μM FP primer and 2 μM GSP1 with all other reagent concentrations following the recommended LA Taq protocol. The primary round of PCR in the FPNI-PCR procedure includes high stringency PCR cycles (94°C for 10 s, 62°C for 30 s, 72°C for 2 m, repeated two times), followed by one low stringency PCR cycle (94°C for 15 s, 25°C for 1 min, 28°C ramping up 0.2C/sec for 3 min, and 72°C 2:30 min). This high and low stringency cycle pattern was repeated six times. The primary round of PCR was finished with another two high stringency cycles followed by one cycle of 94°C for 10 s, 44°C for 30 s, 72°C for 2 min, and a final extension with 72°C for 5 min. The PCR product was diluted to one half and used as template for the next round of PCR. Phusion Hot Start Flex Polymerase (NEB, Ipswich, MA) was used for both secondary and tertiary rounds of PCR in 20 μl reactions containing 10 μM of FSP primer, 2 μM of GSP primer and all other reagents with the recommended concentrations by the manufacturer's protocol. A 1/40 dilution of the secondary round PCR product was used as template for the tertiary round of PCR. Samples were visualized on 1% agarose gel stained with ethidium bromide under UV illumination, and amplified bands were extracted and purified using a QIAquick Gel Extraction Kit (Qiagen, Hilden, Germany) for sequencing.

Transmission electron microscopy analysis
Leaf samples collected from Bowman and BW397 plants at the booting stage were fixed with 2.5% glutaraldehyde in 0.1 M sodium phosphate buffer, pH 7.35 (Tousimis Research Corporation, Rockville, MD) at 4°C. Sample pretreatment and section preparation followed the protocol described in [54]. Stained with lead citrate for two minutes, dried sections on copper grids were visualized on a JEOL JEM-100CX II electron microscope (JEOL Inc., Peabody, MA).