Metabolite sorting of a germplasm collection reveals the Hydroxylase3 locus as a new target for maize provitamin A biofortification

Vitamin A deficiency, a global health burden, can be alleviated through provitamin A carotenoid biofortification of major crop staples such as maize and other grasses in the Poaceae. If regulation of carotenoid biosynthesis was better understood, enhancement could be controlled by limiting beta-carotene hydroxylation to compounds with lower or no nonprovitamin A activity. Natural maize genetic diversity enabled identification of hydroxylation genes associated with reduced endosperm provitamin A content. A novel approach was used to capture the genetic and biochemical diversity of a large germplasm collection, representing 80% of maize genetic diversity, without having to sample the entire collection. Metabolite data-sorting was applied to select a 10 line genetically diverse subset representing biochemical extremes for maize kernel carotenoids. Transcript profiling led to discovery of the Hydroxylase3 locus that coincidently mapped to a carotene QTL, thereby prompting investigation of allelic variation in a broader collection. Three natural alleles in 51 maize lines explained 78% of variation and ~11-fold difference in beta-carotene relative to beta-cryptoxanthin and 36% of the variation and 4-fold difference in absolute levels of beta-carotene. A simple PCR assay to track and identify HYD3 alleles will be valuable for predicting nutritional content in genetically diverse cultivars found world-wide.


INTRODUCTION
Vitamin A deficiency is a global health problem affecting 140-250 million children and accounts for increased childhood mortality and disease (World Health Organization (Geneva), 1995; Underwood, 2004;Black et al., 2008). Humans and animals are unable to synthesize their own vitamin A and rely on dietary provitamin A carotenoid pigments; plant-derived carotenes are metabolized to produce vitamin A, two molecules from β-carotene and only one from α-carotene 5 pathway continues with hydroxylation of the carotenes which depletes the provitamin A pool by converting provitamin A compounds to nonprovitamin A xanthophylls (Matthews and Wurtzel, 2007). Therefore, pathway branching and hydroxylation are key determinants in controlling provitamin A levels. Both of these aspects have been targets for metabolic engineering of endosperm carotenoids in other species (Diretto et al., 2006;Diretto et al., 2007). However, transgenic solutions to manipulate the carotenoid biosynthetic pathway, as achieved in Golden Rice (Ye et al., 2000) and other plants (Giuliano et al., 2008), are not always acceptable as observed by the overwhelming public resistance to GMO food crops. Moreover, transgene incorporation into one genetic variety used for laboratory transformation produces variable carotenoid phenotypes (Aluru et al., 2008); and the desired transgene-produced trait is not predictably transferred to other genetic backgrounds of cultivars worldwide. This lack of predictability could be overcome given a better understanding of the rate-controlling steps and tools to genotype and/or phenotype varieties for predicting the outcome of transgene introduction as we report here. Given the extensive natural diversity inherent in maize, it is conceivable to predictably breed high provitamin A maize in a wide range of genotypes given a thorough understanding of pathway bottlenecks and development of corresponding breeding alleles.
Our collaborative effort in the recent development of LCYE-based breeding markers for maize demonstrated feasibility of a nontransgenic, traditional breeding approach to control the pathway branching step and force pathway flux towards β-carotene and its nonprovitamin A derivatives (the β-branch) (Harjes et al., 2008). However, unless hydroxylation is also controlled, nonprovitamin A xanthophyll compounds will predominate. Given that lycopene cyclization can be forced towards the β-branch, the next challenge in breeding high β-carotene in maize 6 endosperm is to block β-carotene hydroxylation to increase levels of β-carotene relative to βcryptoxanthin and downstream xanthophylls. Therefore, we embarked on a study capitalizing on maize germplasm resources to characterize the maize β-carotene hydroxylase genes and to discover breeding markers for enhancing the relative levels of seed β-carotene.

RESULTS AND DISCUSSION
Two structurally distinct classes of carotene hydroxylases are known: the P450 heme-thiolate CYP97A and CYP97C enzymes and the nonheme diiron monoxygenases [reviewed in (Matthews and Wurtzel, 2007;Giuliano et al., 2008)]. As found in rice (Quinlan et al., 2007), maize contained one gene each for CYP97A and CYP97C, respectively. A total of six unlinked maize paralogs encoding nonheme diiron β-carotene hydroxylases (HYD) were identified in contrast to three rice genes ( Fig. 2; Tables S1 and S2); phylogenetic analysis suggested that the gene duplications found in the grasses occurred after the monocot and dicot evolutionary split.
Maize HYD1 and HYD2 are pseudogenes while HYD3-6 encode enzymes with characteristic hydroxylase domains and plastid targeting signals (Sun et al., 1996) (Fig. S1). HYD3 and HYD4, are syntenous with rice HYD1 (Fig. 3), and are predicted to encode proteins with markedly different isoelectric points, but functional testing confirmed that both encode carotene β-ring hydroxylases (Fig. S2). The CYP97A and CYP97C genes have been functionally tested in rice (Quinlan et al., 2007); phylogenetic analysis of this ancient evolutionary clade would suggest that the maize encoded enzymes behave similarly.
In the maize inbred B73 which has been used extensively for investigating regulation of endosperm carotenogenesis (Li et al., 1996;Matthews et al., 2003;Li et al., 2008a;Li et al., 7 2008b;Li et al., 2009;Vallabhaneni and Wurtzel, 2009), mRNA abundance for the six functional carotene hydroxylases genes varied among tissues and during endosperm development, where comparison was also made with carotenoid accumulation (Fig. 4). To test which, if any, carotene hydroxylase gene showed a statistical correlation between mRNA levels and endosperm provitamin A β-carotene content, transcript abundance was quantified in a genetically diverse maize germplasm collection (Liu et al., 2003;Islam, 2004;Harjes et al., 2008). However, it was impractical to screen endosperm developmental samples in staged, hand-pollinated plants for the entire collection representing 80% of maize genetic diversity. Therefore, a uniquely selected subset of lines for testing was chosen from 148 lines with known carotenoid content and composition (Islam, 2004). From existing metabolite data, we sorted lines for highest total carotenoid content, highest ratio of β-carotene to β-cryptoxanthin, highest ratio of βcryptoxanthin to zeaxanthin, highest ratio of β-carotene to zeaxanthin, highest ratio of α-carotene to lutein, and highest ratio of lutein to zeaxanthin. The resulting ten-line subset (Table S3; Table   S4) exhibited the most extreme carotenoid biochemical phenotypes and likely contained the most favorable and informative alleles for breeding carotenoid content and composition. For example, a block in carotene hydroxylase activity would be predicted to manifest as a high ratio of βcarotene to β-cryptoxanthin, while factors that controlled flux would be predicted to influence levels of total carotenoids. The lines are genetically diverse, spanning four major genetic diversity groups and eight subgroups (Table S4) among the 260 Goodman Diversity lines (Liu et al., 2003). The 10 lines were then used for quantitative transcript profiling of endosperm samples at different developmental stages (days after pollination, DAP). To demonstrate adequacy of sample size and to validate use of this biochemically diverse subset to infer pathway regulation, we applied Pearson correlation analysis to show that PSY1 but not PSY2 or PSY3 transcript levels 8 showed statistically significant correlation with seed carotenoid content (Li et al., 2008b;Vallabhaneni and Wurtzel, 2009). The findings were consistent with molecular and genetic studies associating PSY1 and control of pathway flux in endosperm (Randolph and Hand, 1940;Palaisa et al., 2003;Wong et al., 2004;Pozniak et al., 2007;Li et al., 2008b). The subset has also been recently tested to examine additional genes that encode enzymes for steps that represent carotenoid biosynthetic pathway bottlenecks and those steps that do not . For example, the finding of carotenoid bottlenecks in the upstream isoprenoid biosynthetic pathway is consistent with other studies in the literature. The validated germplasm subset was then used to assess the possible role of carotene hydroxylase gene expression in controlling seed β-carotene composition. Transcript abundance was quantitatively measured, over a range of endosperm developmental stages, for the five carotene β-ring hydroxylase genes, HYD3, HYD4, HYD5, HYD6, and CYP97A and the one ε-ring hydroxylase gene, CYP97C. Time points chosen were based on earlier studies that investigated carotenoid accumulation during endosperm development (Li et al., 2008b). HYD3 was the only gene for which transcripts were found to correlate with carotenoids (Table S3). Interestingly, some of the lines showed a steep reduction in HYD3 transcripts between 15 and 25 DAP ranging from 0.7 to >5000-fold. A Pearson correlation analysis comparing β-carotene content and transcript levels for all carotene hydroxylase genes was conducted and HYD3 was the only gene for which there was a statistically significant (p value ≤ 0.05) correlation between transcript levels and provitamin A βcarotene content (Fig. 5). In the correlation plots shown, where the colored lines denote a 95% confidence interval, the "fold reduction" of HYD3 transcripts over endosperm development (15 to 25 DAP) positively correlated with seed β-carotene (µg/g) levels (r=0.79, p=0.005) (Fig. 5A).
As would be predicted for a critical endosperm hydroxylase gene, HYD3 transcripts at 25 DAP also positively correlated with zeaxanthin (µg/g) levels (r=0.78, p=0.007) (Fig. 5B). That is, the greater the reduction in transcripts over development, the higher the β-carotene content, as would be predicted from a correspondingly lower level of hydroxylase activity; conversely, the higher the transcripts at 25 DAP, the greater the level of zeaxanthin, as would be predicted from increased hydroxylase activity.
We noted the presence of a QTL for endosperm β-carotene content (Chander et al., 2007) that maps together with HYD3, between markers umc1506 and bnlg1028 on maize chromosome 10.05 (Fig. 3). Therefore, we sequenced all HYD3 alleles in the 10 line subset. Variation consistently seen in the high β-carotene lines was found in a ~ 40 bp region adjacent to the transcript start site (Fig. 6, blue box). A conserved transcript start site and first ATG were mapped by aligning available paralog-specific ESTs and genomic DNA for maize B73 HYD3 and sister paralog HYD4, and rice synteny partner HYD1 (Fig. 6). B73 and the other six low βcarotene lines had identical sequence in this region (variant "A"); A619 had sequence variant "B" while CI.7 and DE3 shared variation "C". "B" and "C" in the high β-carotene lines appear to contain a duplicated sequence from a downstream region that replaced the progenitor sequence seen in "A", although allele C shows a more complete match (14/17 nt) as compared to allele B (10/17 nt). Diversity analysis (Liu et al., 2003) of the 260 Goodman lines placed A619 and CI.7, which carried different HYD3 variations, in two genetically distinct subgroups of the maize non-Stiff stalk lines; such diversity grouping is consistent with finding that they do not share the same polymorphism. The identical "C" polymorphism seen in DE3 and CI.7 may be a result of a possible overlapping pedigree since DE3 was placed in a "mixed" group as an indication of genetic structure shared with more than one maize diversity group. However, further sequencing of the entire HYD3 gene of DE3 and CI.7 revealed some minor sequence differences between HYD3 alleles in these two inbreds (Supplementary Figure S3).
To test whether HYD3 allelic variation could explain β-carotene variation, a PCR assay was developed to rapidly genotype a broader maize collection. The assay distinguished between A and C or between A and B alleles and generated a common HYD3 paralog-specific product which could be sequenced to reveal new alleles; the assay correctly genotyped all 10 lines for their HYD3 allele (Fig. 7). We then selected an additional 41 lines spanning maize genetic diversity and range of carotenoid content and composition (Table S4). The combined group of 51 lines included 9 new lines identified by "metabolite-sorting" to exhibit high ratios of β-carotene to β-cryptoxanthin, as possible indicator of a block in β-carotene hydroxylation. The common HYD3 paralog-specific product was amplified and sequenced to characterize HYD3 alleles in this broader maize collection (Fig .7). In the combined 51 lines, we identified fourteen "high carotene" alleles: 7 C alleles, including one new C-like allele, and 7 B alleles. The fourteen B and C alleles were widely distributed among genetic diversity groups, as were the 37 A alleles (Table S4). One-way ANOVA showed that the mean ratios of β -carotene/ β -cryptoxanthin exhibited a statistically significant difference among the three HYD3 alleles (F 2,48 = 89.3, p< 0.0001) and that the HYD3 allele explained 78% of the variation (R 2 ) in the β -carotene/βcryptoxanthin ratio (Fig. 8). The absolute amount of β -carotene (ug/gm) also showed a statistically significant difference among the three alleles (A, B, C) in one-way ANOVA (F 2,48 = 13.7, p< 0.0001) and HYD3 allele explained 36% of total variation (R 2 ) in level of β -carotene (ug/gm). It is reasonable that HYD3 allele will explain less of the variation in absolute levels of β -carotene compared to ratio of be influenced not only by LCYE and HYD3 activities, but also by multiple factors controlling pathway flux ). In comparison, β -carotene/β-cryptoxanthin ratio will be influenced most directly by the HYD3 enzyme activity.
We next compared the means of the three alleles using the Tukey-Kramer HSD test at a 0.05 level. We observed a significant difference (p<.0001) between alleles C and B, and C and A for the ratio of The most favorable HYD3 C allele was associated with ~11-fold higher levels of β -carotene relative to β -cryptoxanthin and 4-fold higher levels of β-carotene (ug/gm). The most likely explanation is that during endosperm development a reduction in HYD3 transcripts leads to reduced conversion of β-carotene to downstream xanthophylls causing β−carotene to accumulate. To test this possibility, we selected CI.7 and DE3, lines carrying the C allele, and measured β -carotene at each of three developmental stages and compared to HYD3 transcript level for triplicate replicates of endosperms collected at 15, 20 and 25 DAP. As shown in Fig. 9, levels of β -carotene increased during endosperm development (for equal amounts of extracted carotenoids), as HYD3 mRNA levels decreased in the same endosperm developmental samples.
Therefore, there is a direct link between HYD3 transcript level and accumulated β-carotene. degradation. Previous development of LCYE markers (Harjes et al., 2008) provided the first step in tracking alleles to control pathway branching through traditional nontransgenic breeding selection. However, the LCYE markers were insufficient. Even if pathway branching could be controlled through selection of optimal LCYE alleles, there remained the problem that the provitamin A carotenes produced in the pathway would automatically be hydroxylated to nonprovitamin A compounds, as typically seen for maize endosperm composition of zeaxanthin or lutein (Kurilich and Juvik, 1999). Other upstream steps may influence carotenoid content ), but enhancement of provitamin A composition will specifically require the combined use of tools that target both LCYE and HYD3.
The C allele can be monitored by the simple PCR assay and its presence in genetically diverse lines will facilitate use in different geographical regions. This PCR assay may now be used to test predictability of the HYD3 allele in controlling β-carotene content in diverse cultivars. With 51 lines analyzed for HYD3 allele, LCYE allele data was available for 48 of those lines (Harjes et al., 2008). However, none carried optimal alleles for both HYD3 and LCYE. Therefore, the HYD3 assay may be used in combination with the previously described LCYE assays (Harjes et al., 2008) to select parental lines containing optimal HYD3 or LCYE alleles and to screen progeny at the seedling stage and identify those that are homozygous for optimal alleles of both genes. Thus it is predicted that the combination of HYD3 and LCYE alleles will lead to higher beta-carotene levels in maize endosperm than having optimal alleles of either gene alone. Many ongoing studies have investigated pathway regulation through transgene expression in one cultivar (Giuliano et al., 2008); the resulting phenotype is dependent on the genotype used and the resulting data are not predictably transferred to other genotypes, perhaps because of limited understanding of pathway regulation. Here we took advantage of natural variation inherent in a large germplasm collection; by judicious sampling of the collection, we were able to pinpoint a specific gene family member for which favorable alleles were discovered for predicting enhanced β-carotene composition. Metabolite-based sorting of a germplasm collection provided accessibility to a valuable resource that yielded critical knowledge and tools that will be needed for breeding high provitamin A in maize, an important food staple in vitamin A-deficient sub-Saharan Africa and Latin America, where vitamin A deficiency is a serious health problem. This approach can be easily adapted to other metabolite targets in other species to rapidly discover pathway regulation and to develop tools for predictive breeding.

Plant materials
Maize inbred lines (A619, B73, B37, CI.7, C131A, DE3, KUI2007, NC300, SD44 and TZI18) (Harjes et al., 2008) were field grown in Bronx, NY and sibling pollinated. Unfertilized ear, embryo (20 days after pollination, DAP) and endosperm (10 -30 DAP) were collected from fieldgrown plants, whereas leaf and root samples were collected from 6-leaf-stage seedlings grown in a green house (16 h day at 25°C). Samples were stored at -80°C prior to use. Cloning and DNA sequence analysis: A maize B73 genomic BAC library (Gallagher et al., 2004) was probed with the maize HYD4 cDNA, GenBank AY844956. Eight BAC clones representing three groups were identified and a representative of each group was sequenced by primer walking (DNA Sequencing Facility, University of Chicago Research Center) and the data deposited in GenBank (HYD1, EU638325; HYD2, EU638326; HYD3 AY844957). The HYD3 cDNA sequence (Genbank# AY844958) was used as a query to identify additional Zea mays paralogs and orthologs from Oryza sativa (www.gramene.org) and Sorghum bicolor (www.phytozome.org). Maize methyl filtered contig sequences, and corresponding BAC clones harboring carotene hydroxylase genes (diiron type) were identified (Table S1). Gene models were drawn using Genscan (www.genscan.com), Softberry (www.softberry.com) and Vector NTI Suite 9.0 (Invitrogen, Carlsbad, CA). Transcription start sites were estimated by EST and promoter analysis (www.softberry.com). Maize ESTs corresponding to CYP97A (DV169913) and CYP97C (BE552887) were found by searching plantGDB maize database using orthologous gene sequences of CYP97A (LOC_Os02g57290) and CYP97C (LOC_Os10g39930) from Oryza sativa (Quinlan et al., 2007). As found for rice and sorghum, only one maize ortholog was found for each gene encoding these two CYP97 enzymes (Supplementary Table S1).

Sequence analysis and chromosome mapping
Mapping and Synteny: Chromosomal positions of maize genes in the Zea mays B73 inbred were mapped either by utilizing tools available at WebAGCoL package (Pampanwar et al., 2005) or Maize GDB. Orthologous genes from Oryza sativa ssp. Japonica were identified through synteny comparisons with maize (www.tigr.org/tdb/synteny/). Phenetic and Protein Analysis: HYD amino acid sequences were aligned using ClustalW and a neighbor-joining tree was constructed with bacterial crtZ from Erwinia herbicola (Genbank # AAA64983) as an out group with 500 bootstrap replication support using MEGA3 software (Kumar et al., 2001). Chloroplast transit peptide signal and transmembrane analysis were predicted using ChloroP 1.1 (Emanuelsson et al., 1999) and TMHMM 2.0 (Krogh A et al., 2001) respectively.

Transcript and total carotenoid analysis
RNA extraction and quantitative RT-PCR were performed using gene specific primers (Table   S1) and normalized to actin, as described (Li et al., 2008a). Values are expressed as the mean of three RT-PCR replicates +/-standard deviation. Total carotenoids were extracted as previously described (Li et al., 2008b) and the concentration (shown in Fig. 4) was calculated using the Lambert-Beer equation (Schiedt and Liaaen-Jensen, 1995).

Plasmids and Functional Complementation
E. coli BL21 (DE3) cells (Novagen, Madison, WI) containing pAC-BETA-04 accumulate β carotene and were used to test hydroxylase function (Sun et al., 1996). BM382572 (HYD3) and BG320875 (HYD4) cDNAs were sub-cloned for functional analysis into the EcoRI / NotI and EcoRI / XhoI sites of the pET23c expression vector (Novagen, Madison, WI), and renamed as pTHYD3 and pTHYD4 respectively. Transformants carrying the test plasmids or empty vector were grown on selective medium and pigments extracted and analyzed by HPLC as described (Gallagher et al., 2004).

Total carotenoid content measurement
The carotenoid extraction procedure was based on Kurilich and Juvik (Kurilich and Juvik, 1999).
Five hundred mg maize endosperm was ground in ethanol and incubated for 6 min (6 ml ethanol, 0.1% butylated hydroxytoluene) at 85 0 C, followed by 10 min saponification with 120 μl (1g/ml) KOH. Samples were vortexed, placed on ice and 4 ml cold dH2O added. Three ml of 2:1 PE:DE (v/v) were added to each sample, vortexed and centrifuged for 10 min at 3500 rpm. The upper layer was retrieved and the separation was repeated twice with 3 ml of 2:1 PE:DE (v/v). The combined fractions were made up to 10 ml, and carotenoids were measured spectrophotometrically at OD 450nm using a Lambda UV/VIS spectrophotometer (Perkin Elmer Life Sciences, Boston), and the concentration of carotenoids were calculated using the Lambert-Beer equation (Schiedt and Liaaen-Jensen, 1995).

Measurement of β -carotene in developing endosperm of high betacarotene lines
The carotenoid extraction procedure was based on Kurilich and Juvik (Kurilich and Juvik, 1999).
Five hundred mg maize endosperm was ground in ethanol and incubated for 6 min (6 ml ethanol, 0.1% butylated hydroxytoluene) at 85 0 C, followed by 10 min saponification with 120 μl (1g/ml) KOH. Samples were vortexed, placed on ice and 4 ml cold dH2O added. Three ml of 2:1 PE:DE (v/v) were added to each sample, vortexed and centrifuged for 10 min at 3500 rpm. The upper layer was retrieved and transferred to a tube with a known weight. The separation was repeated twice with 3 ml of 2:1 PE:DE (v/v) and upper layers combined and dried under nitrogen gas. linear gradient from 100% acetonitrile to 80% acetonitrile/20% isopropanol over 6 min, followed by 100% acetonitrile at 6.5 min. The solvent flow rate was 0.3 mL/min. Data were collected for 7 min total and analyzed with Empower Pro software to determine β -carotene peak area for each sample. All samples were run in triplicates and standard deviation calculated.

Statistical analyses
Pearson correlation analysis of transcript and carotenoid composition from 10-line subset of maize inbreds was performed using JMP v. 5.1.2 (SAS Institute, Cary, NC) to test the statistical significance (p ≤ 0.05) of the relationship. One-way ANOVA and Tukey-Kramer HSD test for comparisons between the alleles (A, B, C) and ratio of β -carotene/ β -cryptoxanthin or absolute levels of β-carotene in 51 lines was performed using JMP 5.1 (SAS institute Inc.).

Multiplex PCR assay for tracking HYD3 alleles
A multiplex PCR assay based on HYD3 promoter variation (A, B, or C) was developed to distinguish between HYD3 allele A (in the low carotene B73 inbred) and the HYD3 alleles B and C found in the high carotene inbreds A619 (allele B), CI.7 (allele C) and DE3 (allele C). The assay involved amplification of one HYD3 paralog-specific "control" PCR product produced by all alleles and a second product that distinguished allele A from C or allele A from B, in both homozygous and heterozygous samples. The PCR assay utilized four primers: (a) two external primers P1 (#1595, GACTTGTGAGCAAGGGGAAG) and P2 (#1592, GACGTGACTCCGAGGCTAGA) for amplification of a "control" product that was conserved in all alleles; and (b) two internal primers added for amplification of the HYD3 specific alleles. To track alleles C and A, the specific allele forward primer was C, (#2111, AACACTCCCGCTCCCGCGGCTCG, allele C), and the reverse primer was A (#2116, TTATATGGATAGTTCACATACCTC, allele A). Alternatively, to distinguish allele B from A, the forward primer used was primer B (#2109, AACACTCACGCTCCCGCG,     β OHase2, AT5G52570), tomato (crtR-b1, Y14809; crtR-b2, Y14810); Table S1 lists monocot accession numbers.

Figure 3. Synteny between maize and rice to map maize HYD orthologs
Orthologous genes encoding maize and rice nonheme diiron β-carotene hydroxylases were compared by synteny, where each chromosome of maize is syntenous to multiple rice chromosome regions numbered in green. Each maize chromosome is marked by bin (yellow box), contig (dark blue box), with its respective maize gene and linked genetic markers. Maize HYD3 (ZmHYD3) maps to chromosome 10 between DNA markers umc1506 and bnlg1028 which flank a β-carotene QTL (Chander et al., 2007) Table S3) in maize inbred lines (Harjes et al., 2008)   The sequence variation in HYD3 (Fig. 6, blue box) was used to develop a multiplex PCR assay to track HYD3 alleles (B and C) in the "high β-carotene" lines and to distinguish these from the A allele in the "low β-carotene" line. With this simple PCR assay, natural alleles can be tracked using traditional, nontransgenic breeding. The assay contained four primers to (a) amplify a control, HYD3 -specific product (*), which was conserved in all alleles and encompassed the allele-specific region ( Fig. 7A; Fig. 6, blue box), and (b) to amplify and distinguish specific alleles A from C or A from B. The multiplex PCR assay was next used to genotype all of the 10 inbred lines used in the correlation study (Fig. 7); as predicted, all lines having B or C alleles, also had a high β-carotene content (21-55% of total carotenoids) as compared to maize B73 (A allele) where kernels have only 3% β-carotene. A) Allele specific primer design. Top, HYD3 primers (P1 and P2) amplify all alleles, as product ( * ), the size of which is shown based on the A allele (B and C alleles generate a * product that differs by a few bp); allele-specific primers (A,

22
A619 or allele C from high β-carotene lines CI.7 and DE3, respectively. Bottom, The boxed region used to design the allele-specific primers is shown in more detail. Sequences and amplification direction are shown in corresponding colors for the three alleles. PCR product sizes are based on the A allele; products from the other alleles differ slightly. B) Screening of alleles in the maize diversity lines. Primers used were P1/P2/A plus either B (lane 1) or C (lanes 2-10).