Identification of GB1, a gene whose constitutive overexpression increases glycinebetaine content in maize and soybean

Abstract Efforts to increase glycinebetaine (GB) levels in plants have been pursued as an approach to improving plant performance under stress conditions. To date, the impact of engineered levels of GB has been limited by metabolic constraints that restrict the achieved increases. We report the identification of a novel gene, GB1, that is differentially expressed in high and low GB accumulating maize genotypes. The predicted GB1 protein shows 60% identity to a putative C‐4 sterol methyl oxidase from rice. Overexpression of GB1 in maize and soybean led to dramatically higher leaf GB content in most of the transgenic lines compared to wild‐type. These results suggest that the GB1 protein is an important component of the biochemical pathways controlling GB accumulation in plants.

The capability to accumulate GB varies among plants, with some genera classified as accumulators and some as nonaccumulators (Rhodes & Hanson, 1993). Several of the genes involved in GB biosynthesis have been isolated (Lamark et al., 1991;Rathinasabapathi et al., 1997;Weretilnyk & Hanson, 1990) and have been engineered into plants with the object to improve stress tolerance by increasing GB levels. Successful GB metabolic engineering has been reported for several plant species, mainly through overexpression of the bacterial or plant genes responsible for choline oxidation (Hayashi, Alia Mustardy, Deshnium, Ida, & Murata, 1997;Nuccio et al., 1998;Park et al., 2004;Sakamoto & Murata, 1998). Although the increase in GB in the reported studies was significant, accumulation was considerably lower than the GB level in high accumulator species. One possible explanation for this is that choline availability may limit GB accumulation in some plants. In fact, transgenic tobacco plants overexpressing CMO were able to accumulate large amounts of GB only when choline was supplemented (Nuccio et al., 1998).
Zea mays is considered a GB accumulator species, but certain genotypes are relatively deficient in GB. Brunk, Rich, and Rhodes (1989) demonstrated that maize genotypes could be generally classified into two distinct groups: low GB (L-GB) and high GB (H-GB) genotypes, with GB levels of less than 0.09 or more than 1.0 lmol/g fresh weight (FW), respectively. Rhodes and Rich (1988) reported that the L-GB trait behaved as a single recessive gene, and Lerma et al. (1991) conducted complementation analysis to demonstrate that the L-GB phenotype across several genotypes was attributable to a single locus, the recessive allele of what Yang et al. (1995) designated the Bet1 gene. Although investigations of GB-relevant metabolites and enzymes in near-isogenic maize lines that differ at the Bet1 locus have been conducted, the precise biochemical defect in L-GB lines has not been determined (Peel, Mickelbart, & Rhodes, 2010;Yang et al., 1995).
We investigated further the genetic basis of the GB phenotype difference in maize to identify novel genes for GB engineering.
Under the assumption that the GB difference in maize might be regulated at the gene transcription level, a gene expression analysis experiment was conducted to identify genes associated with the H-GB phenotype. This work describes the discovery of a novel gene (GB1) that, when overexpressed, significantly increases GB accumulation in two tested species, maize and soybean.

| Gene expression analysis
For gene expression analysis, 98 inbred lines characterized for GB levels were used; of these, 59 were categorized as H-GB and 36 as L-GB. Ten replicates per genotype were planted and grown to the V7 stage of development. At V7, irrigation was withheld from half of the replicates to impose a drought condition from V8 to V10. For the irrigated section, irrigation was maintained during the entire course of the experiment. Leaf tissue was sampled from one plant per replicate when drought plots showed clear symptoms of dehydration stress (leaf rolling and leaf grayish cast). Total RNA was isolated from leaf tissue; cDNA for hybridization was synthesized from 1 lg of mRNA. The hybridization array used for analysis consisted of 5,749 elements, of which 4,433 represented unique genes.

| Plant transformation
The GB1 gene was amplified and cloned into a plant expression binary vector under control of the rice actin promoter or the 35S promoter, for maize and soybean, respectively. Agrobacterium-mediated transformation of maize and soybean proprietary lines was carried out as previously described (Edgerton, Chomet, & Laccetti, 2003;Martinell et al., 2002).

| GB1 mapping
A B73xMo17 recombinant inbred line mapping population (Stuber, Lincoln, Wolff, Helentjaris, & Lander, 1992) was used to map the GB1 locus in maize. RFLP markers asg48 and phi053, defining the chromosome region 3.04 boundaries, were used to select recombinant inbred lines (RIL) homozygous for one of the parental alleles for both markers. Additional proprietary markers that map to the chromosome 3.04 region were used to confirm the absence of double recombination between the asg48 and phi053 markers. Two pools for each parental allele, each one composed of more than 15 individuals, were created (see supplemental information). To confirm that the individual pools were appropriate to map cDNA probes to the 3.04 chromosome region, the two RFLP probes umc10 and bnl15.20 were used as controls.

| GB analysis
Approximately 30 mg FW of leaf tissue was collected from maize or soybean plants, freeze-dried, and ground to fine powder. Powdered tissue was extracted with 80% ethanol, 0.1% formic acid, and 1 mM deuterated GB (d 9 gb) as a standard for 20 min and centrifuged to remove cell debris. Following filtration, GB content in the extract was determined by LC/MS-MS analysis using an API 2000 system (Applied Biosystems, Foster City, CA, USA) equipped with an AllTech Alltima C18 column. GB concentration of the tissue sample was calculated as the GB/d 9 gb peak area ratio. For the genotype survey analysis, deuterated d 9 Val was added to the extraction buffer and the relative amount of GB was calculated as ratio with d 9 Val standard.
GenBank Accession Number: KU232555. To demonstrate that the higher GB1 expression in H-GB genotypes was not simply the result of the higher GB levels, we tested if exogenously provided GB had any effect on GB1 mRNA accumulation. Plants from one H-GB genotype and one L-GB genotype were grown in a greenhouse and irrigated with a 50 mM GB solution starting 6 weeks after planting. GB analysis of plant tissue revealed that the compound was rapidly absorbed and transported to the leaves, but RNA blot analysis showed that the added GB did not increase GB1 transcript accumulation ( Figure S2).

Glycinebetaine accumulators Nonaccumulators
F I G U R E 1 Maize genotypes divide into two GB phenotypic classes. Relative GB concentration in leaves for 98 genotypes used in the gene expression experiment is shown in rank order. Each bar represents the mean of 10 replicates, and the relative GB concentration was estimated as the ratio of GB peak area per internal standard peak area. One unit of [GB] is taken as the threshold distinguishing GB accumulators and nonaccumulators CASTIGLIONI ET AL. | 3 3.2 | GB1 maps to the bet1 locus region Cosegregation of the GB1 with the previously mapped bet1 locus  was investigated to establish the relationship between the sequence reported here and the previously reported major determinant for GB accumulation in maize. A recombinant inbred line (RIL) mapping population developed by a cross of two H-GB lines was used for the analysis. As described in the Section 2, two pools of individuals were created to map cDNA probes to the maize 3.04 genome region. The Southern blot hybridization patterns of the selected RIL pools were compared to the parental and F1 generation. Two RFLP probes, umc10 and bnl15.20, mapped inside and outside, respectively, the 3.04 chromosome region boundaries and confirmed that the constituted pools were suitable to establish if cDNA probes were in tight linkage with the selected genome region. The probe umc10 shows a hybridization pattern characteristic for a linked probe. RFLP analysis using GB1 cDNA as probe revealed the hybridization pattern expected for probes located in the 3.04 chromosome region (see Figure S2), suggesting that bet1 and GB1 may be tightly linked.

No significant or relevant differences between nonaccumulators
and accumulators have been identified in gene sequences near the GB1 coding region. However, this analysis is incomplete; while highresolution sequence data have been gathered for some maize inbreds, the GB status of these has not been characterized. B73 and Mo17, two inbreds with extensive genomic information available, are both H-GB lines (see Figure S1).

| GB1 overexpression increased GB in maize and soybean
Transgenic maize plants constitutively overexpressing GB1 were generated by Agrobacterium tumefaciens transformation. Plants regenerated from in vitro culture were assayed at the V6-V8 developmental stage for leaf GB content. Average GB content for the GB1 transgenic lines was 7.1 mM, compared to 0.1 mM GB for the transformation genotype, a L-GB line. Although biological replication was not possible due to only one R0 plant per transgenic line, we could detect GB concentrations up to 20 mM for a few transgenic lines, corresponding to a more than 200-fold increase. The leaf GB content of the transgenic plants was notably higher than that observed in nontransgenic H-GB maize growing in the same environment.
To confirm that GB1 overexpression effectively increased GB in maize, 23 independently regenerated lines were selected for analysis. F I G U R E 2 The GB1 gene is differentially expressed in H-GB maize genotypes. Each panel reports the difference in gene expression between high and low GB genotypes for irrigated (a) and drought (b) conditions. Log-transformed array element hybridization signal intensity of each replication was normalized by subtracting the median of the noncontrol elements. Average signal values were calculated for high and low GB genotype classes and the difference between classes is presented in the scatter plot. The diagonal lines delimit the twofold difference in gene expression between the two GB classes. The differential signal for the GB1 element is circled H-GB D I D I D I D I L-GB D I D I F I G U R E 3 GB1 expression is induced by drought. Blot analysis of RNA isolated from H-GB and L-GB lines under drought (D) and irrigated (I) conditions, using GB1 cDNA as probe and L-GB testers throughout development, as revealed by the assessment of leaf tissue samples collected at the VT-stage (Figure 4).
To investigate whether GB1 expression could affect GB accumulation in other crop plant species, we also generated GB1 transgenic soybean lines. Nine soybean GB1 transgenic lines were assayed for GB leaf content 6 weeks after planting, and eight of the lines showed a substantial GB increase ( Figure 5) relative to the nontransgenic control.

| DISCUSSION
GB enhancement through metabolic engineering has been associated with an improvement in plant stress tolerance, but efficacy of this engineering in a field environment for important crops has not been demonstrated. One explanation for this could be the limited GB increases previously achieved through genetic engineering. We rationalized that factors other than the known GB metabolic enzymes could play a critical role for GB accumulation in maize. We thus decided to seek such additional factors by investigating the differences in gene expression for genotypes with opposite GB phenotypes to gain additional knowledge relative to the genetic control of GB accumulation in maize.
Our high-throughput GB assay enabled the survey of many genotypes. The unambiguous classification of the genotypes into the correct GB class was critical to design a powerful experiment to assess differential gene expression (Figure 1) GB concentration in leaves of soybean GB1 transgenic events compared to wild-type control (WT) 6 weeks after planting. Each bar represents the average of six replicates AE standard error found that PC is the major polar lipid of the outer leaflet of the chloroplast outer envelope membrane and therefore accessible from the cytosolic side of the membrane. Given that GB1 has putative transmembrane domains, but not organelle targeting signals, it is possible that GB1 is inserted from the cytosol to the outer organellar surface. Nuccio et al. (2000) built on evidence that cytosolic choline is precursor to GB synthesized in the chloroplast. Modeling established that choline availability is a factor that limits the synthesis of GB in tobacco expressing a chloroplastic GB pathway, suggesting that subcellular choline localization influences the overall accumulation of GB. It may be possible that choline released via a GB1-associated pathway is tied to choline transport into the chloroplast. This hypothesis suggests that the phosphatidylcholine pool or flux through the outer leaflet should be different between maize GB nonaccumulators and accumulators.
Our preliminary experiments showed that some of the GB1 transgenic lines developed negative visual phenotypes. It is probable that constitutive overexpression may negatively impact untargeted cellular functions.
We describe here a novel gene that plays a pivotal role control-