The Silencing of Carotenoid β-Hydroxylases by RNA Interference in Different Maize Genetic Backgrounds Increases the β-Carotene Content of the Endosperm

Maize (Zea mays L.) is a staple food in many parts of Africa, but the endosperm generally contains low levels of the pro-vitamin A carotenoid β-carotene, leading to vitamin A deficiency disease in populations relying on cereal-based diets. However, maize endosperm does accumulate high levels of other carotenoids, including zeaxanthin, which is derived from β-carotene via two hydroxylation reactions. Blocking these reactions could therefore improve the endosperm β-carotene content. Accordingly, we used RNA interference (RNAi) to silence the endogenous ZmBCH1 and ZmBCH2 genes, which encode two non-heme di-iron carotenoid β-hydroxylases. The genes were silenced in a range of maize genetic backgrounds by introgressing the RNAi cassette, allowing us to determine the impact of ZmBCH1/ZmBCH2 silencing in diverse hybrids. The β-carotene content of the endosperm increased substantially in all hybrids in which ZmBCH2 was silenced, regardless of whether or not ZmBCH1 was silenced simultaneously. However, the β-carotene content did not change significantly in C17 hybrids (M7 × C17 and M13 × C17) compared to C17 alone, because ZmBCH2 is already expressed at negligible levels in the C17 parent. Our data indicate that ZmBCH2 is primarily responsible for the conversion of β-carotene to zeaxanthin in maize endosperm.


Introduction
Humans are unable to synthesize vitamin A de novo and must obtain this essential nutrient in their diet, either directly as retinol from animal sources or as pro-vitamin A (PVA) carotenoids from plants. The major PVA carotenoid is β-carotene, which can be converted into two molecules of retinol. The other PVA carotenoids are α-carotene, γ-carotene, and β-cryptoxanthin, and these yield one molecule of retinol each [1]. Vitamin A in its reduced form (retinal) is necessary for rhodopsin biosynthesis, making this nutrient essential for normal vision, and it also helps to maintain healthy epithelial and immune system cells. The acidic form (retinoic acid) is a key developmental morphogen. Therefore, vitamin A deficiency (VAD) causes visual disorders in the form of night blindness and (in more serious cases) xerophthalmia, as well as epithelial and immune system defects that leave individuals, especially children, highly susceptible to infections [1][2][3]. The polished grain (endosperm) of cereal crops such as rice (Oryza sativa), maize (Zea mays), and wheat (Triticum aestivum) are staples in (endosperm) of cereal crops such as rice (Oryza sativa), maize (Zea mays), and wheat (Triticum aestivum) are staples in many parts of the world, but they contain very low levels of PVA carotenoids, hence populations that subsist on cereal-rich diets are prone to VAD [4,5]. Increasing the levels of PVA carotenoids in cereals is an effective way to combat this challenge [1,2,4,6,7].
PVA carotenoids are synthesized via the methylerythritol 4-phosphate (MEP) pathway, which is also known as the non-mevalonic acid (MVA) pathway [6][7][8]. The first committed step in carotenoid biosynthesis is the condensation of two molecules of geranylgeranyl diphosphate (GGPP) by phytoene synthase (PSY) to produce phytoene, which is then converted into all-trans-lycopene via four desaturation and isomerization steps. Lycopene is the branching point of this pathway ( Figure 1). In one branch, lycopene is cyclized to form the PVA carotenoids γ-carotene and β-carotene. The latter is oxidized to form another PVA carotenoid (β-cryptoxanthin) and subsequently the non-PVA carotenoid zeaxanthin, antheraxanthin, and violaxanthin. In the other branch, lycopene is cyclized to produce the PVA carotenoid α-carotene, which is oxidized to form the non-PVA carotenoids α-cryptoxanthin and zeinoxanthin, and finally lutein. The oxidation reactions that deplete the pool of PVA carotenoids are catalyzed by enzymes known as carotenoid β-hydroxylases. Xanthophyll biosynthesis pathway in maize endosperm. Enzymes responsible for the major activity at each step are shown in bold (blue). Identified maize locus names corresponding to each enzyme are shown in brown. Abbreviations: LYCB, lycopene β-cyclase; LYCE, lycopene ε-cyclase; CYP97C, heme-containing cytochrome P450 carotenoid ε-ring hydroxylase; CYP97A/B, hemecontaining cytochrome P450 carotenoid β-ring hydroxylase; BCH1/2, Non-heme di-iron β-carotenoid hydroxylases; ZEP, zeaxanthin epoxidase.
Two maize BCH genes (ZmBCH1 and ZmBCH2) (ZmBCH1, also termed ZmcrtRB3, or ZmHYD4; ZmBCH2 also termed ZmcrtRB1, or ZmHYD3) have been cloned and functionally characterized [16][17][18][19]. Candidate-gene association analysis identified 18 polymorphic sites in ZmBCH1 significantly associated with one or more carotenoid-related traits in 126 diverse yellow maize inbred lines, indicating that ZmBCH1 plays a role in hydroxylating the β-rings of α-carotene and β-carotene, but predominantly α-carotene [19]. In contrast, quantitative trait locus (QTL) mapping, genome-wide association studies (GWAS), and functional analysis revealed that ZmBCH2 strongly influences the conversion of β-carotene to zeaxanthin, with hypomorphic ZmBCH2 alleles correlating with higher β-carotene levels in the endosperm [16,18,21,22]. ZmBCH2 is also the only carotenoid hydroxylase gene expressed at high enough levels in maize endosperm to be detected by mRNA blot, and it is preferentially transcribed in the endosperm [17].
In potato (Solanuum tuberosum), silencing the BCH1 gene by RNA interference (RNAi) increased β-carotene levels from trace amounts in wild-type tubers to 3.31 µg/g fresh weight in the transgenic plants, with a concomitant loss of zeaxanthin and an increase in lutein [23]. The simultaneous silencing of both potato BCH genes increased β-carotene levels by up to 38-fold (with a concomitant loss of zeaxanthin) and the total carotenoid level increased by up to 4.5-fold [24]. Similarly, silencing BCH expression in sweet orange plants (Citrus sinensis) produced oranges with a deep yellow color and up to 36-fold more β-carotene [25]. In wheat, blocking the expression of an endogenous BCH gene promoted the conversion of β-carotene into xanthophylls, but nevertheless increased the β-carotene levels in the endosperm by up to 10-fold [26].
The ability to increase β-carotene levels in other crops by silencing one or more BCH genes suggests that a similar strategy could be deployed in maize to address VAD in developing country populations. However, given the diverse activities of carotenoid β-hydroxylases in different crops, a functional analysis of the corresponding genes is necessary during the development of biotechnology programs to increase PVA levels. This is particularly important in maize due to the low ratio of PVA to non-PVA carotenoids in the endosperm and the diverse carotenoid profiles of different maize varieties [27]. In order to evaluate the impact of silencing both of the BCH-type enzymes in different genetic backgrounds, we therefore introgressed ZmBCH1/2 RNAi silencing cassettes into different lines and investigated the endosperm carotenoid content in the resulting hybrids.

RNAi-Mediated Silencing of ZmBCH1 and ZmBCH2 in M37W White Maize
The ZmBCH1 and ZmBCH2 DNA fragments used for RNAi construct show 96.4% identity, so the pHorP-RNAi-ZmBCH2 silencing construct was expected to downregulate not only the primary target ZmBCH2 but also the closely related paralog ZmBCH1 in M37W maize. This was confirmed by mRNA analysis. T3 endosperm from homozygous T2 plants was analyzed using selective probes for each transcript, revealing negligible levels of endogenous ZmBCH2 mRNA in two transgenic lines (M1 and M7) and low levels in two additional lines (M9 and M13) (Figure 2A). The ZmBCH1 transcript could not be detected in any of these lines by mRNA blot analysis so quantitative RT-PCR was used instead. This revealed that ZmBCH1 was downregulated in all transgenic lines by at least two-fold, compared to M37W wild-type plants ( Figure 2B).

Carotenoid Profiles in Hybrids Derived from BCH-Silenced Parents
Transgenic lines M7 and M13 were chosen to represent lines with near-complete ZmBCH2 silencing and partial ZmBCH2 silencing, respectively. Each transgenic line was crossed with a range of maize inbred lines to generate hybrids in different genetic backgrounds. The wild-type inbred parents were selected based on their diverse carotenoid profiles (Table S2). Inbred lines B73 and C17 accumulated lutein as the predominant carotenoid in the endosperm, whereas lines NC356, O1-3, O2-9, and PSY1 accumulated zeaxanthin instead. The purpose of these experiments was to investigate the impact of partial and near-complete ZmBCH gene silencing in diverse genetic backgrounds representing different expression levels of genes in the endogenous carotenoid biosynthesis pathway.
The endosperm carotenoid content and composition in transgenic lines M7 and M13 was similar to wild-type M37W plants. The transgenic lines accumulated ~4.5 µg/g dry weight (DW) total carotenoids compared to ~3 µg/g DW in the M37W parent. In both transgenic lines, zeaxanthin was much more abundant than lutein. Traces of antheraxanthin, a metabolic derivative of zeaxanthin ( Figure 1), were detected in line M13 but not M7 ( Figure 3, Table S1).

Carotenoid Profiles in Hybrids Derived from BCH-Silenced Parents
Transgenic lines M7 and M13 were chosen to represent lines with near-complete ZmBCH2 silencing and partial ZmBCH2 silencing, respectively. Each transgenic line was crossed with a range of maize inbred lines to generate hybrids in different genetic backgrounds. The wild-type inbred parents were selected based on their diverse carotenoid profiles (Table S2). Inbred lines B73 and C17 accumulated lutein as the predominant carotenoid in the endosperm, whereas lines NC356, O1-3, O2-9, and PSY1 accumulated zeaxanthin instead. The purpose of these experiments was to investigate the impact of partial and near-complete ZmBCH gene silencing in diverse genetic backgrounds representing different expression levels of genes in the endogenous carotenoid biosynthesis pathway.
The total carotenoid content of hybrids derived from lines B73 and C17 did not differ substantially from the levels detected in the inbred parents. In hybrid M7 × B73, the total carotenoid content increased by 1.3-fold to ~42 µg/g DW, whereas in hybrid M13 × B73 the total carotenoid content increased by 1.1-fold to ~34 µg/g DW. In hybrid M13 × C17, the total carotenoid content increased by 1.1-fold to ~45 µg/g DW, and the total carotenoid content of the M7 × C17 hybrid was ~38 µg/g DW, very slightly lower than the C17 parent. Although the total carotenoid levels showed little variation, we observed interesting changes in the carotenoid profiles of the hybrids compared to their parents. There were no significant changes in the level of β,ε-carotenoids in the B73 hybrids (M7 × B73 and M13 × B73), with lutein remaining the predominant carotenoid (18-19 µg/g DW) and lower levels of α-cryptoxanthin (2-3 µg/g DW). In contrast, the M7 × C17 and M13 × C17 hybrids accumulated traces of α-cryptoxanthin (~1 µg/g DW) that were not detected in the C17 parent, and lutein levels declined to 8-9 µg/g DW, representing a 1.3-1.4-fold decrease compared to C17.
The total carotenoid content of hybrids derived from lines B73 and C17 did not differ substantially from the levels detected in the inbred parents. In hybrid M7 × B73, the total carotenoid content increased by 1.3-fold to~42 µg/g DW, whereas in hybrid M13 × B73 the total carotenoid content increased by 1.1-fold to~34 µg/g DW. In hybrid M13 × C17, the total carotenoid content increased by 1.1-fold to~45 µg/g DW, and the total carotenoid content of the M7 × C17 hybrid was~38 µg/g DW, very slightly lower than the C17 parent. Although the total carotenoid levels showed little variation, we observed interesting changes in the carotenoid profiles of the hybrids compared to their parents. There were no significant changes in the level of β,ε-carotenoids in the B73 hybrids (M7 × B73 and M13 × B73), with lutein remaining the predominant carotenoid (18-19 µg/g DW) and lower levels of α-cryptoxanthin (2-3 µg/g DW). In contrast, the M7 × C17 and M13 × C17 hybrids accumulated traces of α-cryptoxanthin (~1 µg/g DW) that were not detected in the C17 parent, and lutein levels declined to 8-9 µg/g DW, representing a 1.3-1.4-fold decrease compared to C17.

Silencing the Endogenous ZmBCH1 and ZmBCH2 Genes Causes β-Carotene to Accumulate in the Endosperm of Diverse Maize Hybrids
Transgenic maize plants in which both ZmBCH1 and ZmBCH2 were silenced simultaneously were generated because the RNAi cassette targeted a sequence conserved in both genes with 96.4% identity (Figure 2A,B). The potency of ZmBCH2 silencing varied among the transgenic lines, and we selected one line (M7) with near-complete ZmBCH2 silencing and another (M13) with partial ZmBCH2 silencing. ZmBCH1 expression was already much lower than ZmBCH2 in the wild-type M37W endosperm and it declined by 2-fold in both M7 and M13. The M37W inbred line used for transformation only accumulates traces of carotenoids because the ZmPSY1 gene is expressed at low levels in the endosperm [28,29]. Consequently, primary transformants in an M37W genetic background cannot be used to evaluate the impact of BCH gene silencing in the endosperm. We therefore introgressed the RNAi cassette present in lines M7 and M13 into different maize backgrounds selected on the basis of their endosperm carotenoid content and composition, to determine how the near-complete or partial silencing of ZmBCH2 affected carotenoid accumulation.
The carotenoid profiles of hybrids generated using transgenic lines M7 and M13 revealed diverse carotenoid profiles ( Figure 3, Table S1). Silencing ZmBCH1 and ZmBCH2 had a more profound impact on the β,β-carotenoid branch, given that the endosperm of parental lines B73, O1-3, and O2-9 contains no β-carotene but the corresponding hybrids accumulated up to~5 µg/g DW of this molecule. Furthermore, the NC356 and PSY1 hybrids accumulated up to 3.4-fold more β-carotene than their corresponding parents. However, no significant changes in β-carotene levels were detected in the C17 hybrids because ZmBCH2 is already expressed at minimal levels in the C17 parent ( Figure 3). The levels of β-cryptoxanthin increased in all C17 hybrids but decreased in all O1-3 and O2-9 hybrids. Interestingly, the levels of β-cryptoxanthin increased when B73, NC356 and PSY1 were hybridized with the strongly-silenced line M7 but decreased when the same parents were hybridized with the partially silenced line M13, suggesting that the degree of BCH activity is a key determinant of β-cryptoxanthin levels. The M7 parent has a lower hydroxylation capacity than M13, so the monohydroxylated carotenoid β-cryptoxanthin accumulates at the expense of the dihydroxylated carotenoid zeaxanthin. Accordingly, zeaxanthin levels decreased when NC356, O1-3, and O2-9 were hybridized with either of the RNAi lines, but increased in all hybrids of B73 and PSY1, showing a correlation between the depleted zeaxanthin pool and the increase in β-carotene levels due to ZmBCH2 downregulation. This agrees with previous reports that BCH2 is the key determinant of the conversion of β-carotene into zeaxanthin in maize endosperm based on QTL mapping, GWAS and functional analysis [16][17][18]21,22].
In the α-branch of the pathway, the β-hydroxylation of α-cryptoxanthin produces lutein. We found that α-cryptoxanthin accumulated in hybrids of C17, O1-3, O2-9, and PSY1 even though the parental lines do not produce this metabolite. In the M7 × B73 hybrid, the level of α-cryptoxanthin was similar to that detected in the B73 parent, but the M13 × B73 hybrid produced more α-cryptoxanthin than B73. Furthermore, both NC356 hybrids accumulated more α-cryptoxanthin than the NC356 parent. The B73 and NC356 lines accumulated the highest levels of ZmBCH1 mRNA, but the presence of α-cryptoxanthin suggests that additional hydroxylase activity is required for complete conversion to lutein. The quantity of lutein was reduced in the hybrids of lines C17, O1-3, and O2-9 compared to the parent lines, but increased in the NC356 hybrids and did not change in the PSY1 hybrids, again suggesting that additional β-hydroxylase activity (most likely provided by ZmCYP97A and/or ZmCYP97B) is necessary for the complete conversion of α-cryptoxanthin into lutein. ZmCYP97A and/or ZmCYP97B orthologs in A. thaliana [9,10], tomato [13], and rice [14,15] are known to be involved in lutein biosynthesis.

ZmBCH1 and ZmBCH2 Silencing Affects the Expression of CYP-Type ε-Hydroxylase Gene
In the M37W endosperm, ZmBCH2 is expressed more strongly than ZmBCH1, ZmCYP97A, ZmCYP97B, and ZmCYP97C ( Figure 4). ZmBCH1 is more important for the β-ring hydroxylation of α-carotene than β-carotene, because polymorphisms in this gene correlated more with variations in α-carotene levels rather than β-carotene levels [19]. ZmBCH1 mRNA levels varied widely in the different parental lines, with some lines accumulating up to 15-fold higher levels of this transcript than M37W. ZmBCH1 mRNA levels were up to 5-fold lower in the B73 and NC356 hybrids than their parents, suggesting that ZmBCH1 was strongly suppressed by the introgressed RNAi construct. However, there was little impact on ZmBCH1 mRNA levels in hybrids of O1-3, O2-9, and PSY1, and ZmBCH1 mRNA levels in C17 hybrids were higher than in the C17 and M37W parents, suggesting that endogenous ZmBCH1 alleles might influence the effectiveness of the RNAi construct introgressed into these hybrids. The strong suppression of ZmBCH1 expression in hybrids of B73 and NC356 compared to the corresponding parental lines did not correlate with higher α-cryptoxanthin and lower lutein levels, which suggests that additional carotenoid β-hydroxylase activity (probably ZmCYP97A and/or ZmCYP97B, as stated above) may be required for the complete conversion of α-cryptoxanthin into lutein. ZmBCH2 mRNA levels varied by only 2.5-fold among all the parental lines and hybrids. However, the ZmBCH2 locus has a strong impact on β-carotene accumulation in the endosperm [16,18] because it is the only hydroxylase gene strongly expressed in this tissue [17]. In most hybrids, ZmBCH2 mRNA levels were lower than in the corresponding parental lines, but the C17 hybrids were exceptional: the level of ZmBCH2 mRNA was lower than in the RNAi lines but higher than the C17 parent.
The different degrees of gene silencing in M7 and M13 also manifested in the hybrids derived from these lines. The levels of ZmCYP97A and ZmCYP97B mRNA in the parent lines and hybrids were too low to detect obvious differences, and any differences could reflect natural variations rather than the consequences of BCH gene silencing. This low level of variability also suggests that ZmCYP97A and ZmCYP97B are tightly regulated. The levels of ZmCYP97A and ZmCYP97B were similar in hybrids and their parents, whereas the carotenoid ε-hydroxylase gene ZmCYP97C varied by up to 19-fold, with lines B73 and C17 representing the extremes. ZmCYP97C was downregulated in the hybrids of B73, C17, O1-3, and O2-9 compared to the parental lines, but not in hybrids of NC356 or PSY1. Interestingly, the NC356 or PSY1 hybrids did not show a reduction in lutein levels, suggesting that the downregulation of BCH gene expression may have a feedback effect on the genes encoding carotenoid ε-hydroxylases.

Gene Cloning and Vector Construction
The ZmBCH2 cDNA fragment was cloned from maize (Zea mays L. cv. B73) endosperm total RNA by RT-PCR using forward primer 5 -GGA ATT CTC TAG ACT ATC GCT TCA GCT GGC AAA TGG AG-3 (EcoRI and XbaI sites are underlined) and reverse primer 5 -GAC TAG TGG ATC CAA CTT GTC CAT GTG GTG TAT CTT G-3 (BamHI and SpeI sites are underlined) based on sequence information in GenBank (accession number: AY844958). The ZmBCH2 sequence was transferred to the pHorP vector containing the barley D-hordein promoter, a 300-bp gusA gene fragment, and the ADP-glucose pyrophosphorylase terminator (ADGPP) using restriction enzymes XbaI and BamHI to yield the intermediate vector pHorP-ZmBCH2 sense. In a second step, the intermediate vector was digested with SpeI and EcoRI to introduce the antisense ZmBCH2 fragment between the gusA gene and ADGPP terminator, resulting in pHorP-RNAi-ZmBCH2 ( Figure S2).

RNA Extraction and Expression Analysis
Total RNA was extracted for mRNA blot analysis and cDNA synthesis as previously described [29]. Quantitative real-time PCR (qRT-PCR) was carried out in triplicate using the primers listed in Table S3. Real-time RT-PCR was performed on a BioRad CFX96TM system using 25-µL mixtures containing 10 ng of synthesized cDNA, 1× iQ SYBR Green Supermix (BioRad, Hercules, CA, USA), and 0.2 µM forward and reverse primers. Relative expression levels were calculated on the basis of serial dilutions of cDNA (125-0.2 ng), which were used to generate standard curves for each gene. Cycling conditions consisted of a single incubation step at 95 • C for 5 min followed by 30 cycles of 95 • C for 10 s, 58 • C for 35 s, and 72 • C for 15 s. Specificity was confirmed by product melt curve analysis over the temperature range 50-90 • C with fluorescence acquired after every 0.5 • C increase, and the fluorescence threshold value and gene expression data were calculated with BioRad CFX manager software version 3.1. Values represent the mean of three biological replicates ± SE. Amplification efficiencies were compared by plotting the ∆C t values of different primer combinations of serial dilutions against the log of starting template concentrations using the CFX96TM software. Differences among samples were observed by mean ± SE.

Carotenoid Extraction and UPLC (Ultra Performance Liquid Chromatography) Analysis
Carotenoids were extracted from 30 DAP maize endosperm for UPLC analysis in triplicate as described by Berman et al. [30]. Data were calculated from three replicates with error bars representing the standard error. Differences among samples were observed by mean ± SE.

Conclusions
The RNAi-ZmBCH2 cassette in transgenic lines M7 and M13 was introgressed into several maize genetic backgrounds with different carotenoid profiles. Knocking down endogenous ZmBCH2 expression by RNAi increased β-carotene levels in hybrids of the parental lines B73, N356, O1-3, O2-9, and PSY1, regardless of whether or not the endogenous ZmBCH1 gene was also affected. However, we detected no change in the level of β-carotene in hybrids of C17 (M7 × C17 and M13 × C17) as compared to the C17 parent because ZmBCH2 transcript levels were already very low. These results indicate that ZmBCH2 is the critical enzyme responsible for the conversion of β-carotene to zeaxanthin via β-cryptoxanthin in maize endosperm, and that its activity may be intimately linked with that of the carotenoid ε-hydroxylase encoded by ZmCYP97C.