Total tocopherol levels in maize grain depend on chlorophyll biosynthesis within the embryo

Abstract

Background

Tocopherols are a class of lipid-soluble compounds that have multiple functional roles in plants and exhibit vitamin E activity, an essential nutrient for human and animal health. The tocopherol biosynthetic pathway is conserved across the plant kingdom, but source of the key tocopherol pathway precursor, phytol, is unclear. Two protochlorophyllide reductases (POR1 and POR2) were previously identified as loci controlling the natural variation of total tocopherols in maize grain, a non-photosynthetic tissue. POR1 and POR2 are key genes in chlorophyll biosynthesis yet the contribution of the chlorophyll biosynthetic pathway to tocopherol biosynthesis is still not understood.

Results

We took two approaches to alter the activity of these two POR genes within kernel tissue, physiological treatments and CRISPR/Cas9-mediated knockouts, to determine the role of chlorophyll biosynthesis for tocopherol content. Since light is required for POR enzymatic activity, we imposed a dark treatment on developing kernels, which reduced chlorophyll a and tocopherols levels in embryo tissue by 92–99% and 87–90%, respectively, compared to the light treatment. In CRISPR/Cas9-mediated knockouts, the levels of chlorophyll a and tocopherols in embryos of the por1 por2 double homozygous mutant were reduced by 98–100% and 76–83%, respectively, compared to WT.

Conclusion

These findings demonstrate that tocopherol synthesis in maize grain depends almost entirely on phytol derived from chlorophyll biosynthesis within the embryo. POR1 and POR2 activity play crucial roles in chlorophyll biosynthesis, underscoring the importance of POR alleles and their activity in the biofortification of vitamin E levels in non-photosynthetic grain of maize.

Background

Suboptimal dietary intake of vitamin E, an essential micronutrient, can be prevalent in countries where maize grain significantly contributes to the caloric requirements of human populations [8, 15]. Increased dietary vitamin E intake has been associated with reduced mortality from cardiovascular disease and cancer [22, 44]. Improved crop nutritional quality through agronomic, conventional breeding, or bioengineering approaches, collectively known as biofortification, represents a potential avenue to enhance dietary vitamin E by increasing tocochromanols, a group of biosynthetically related compounds with varying levels of vitamin E activity [37]. Biofortification efforts can be maximized through a comprehensive understanding of the genetic and molecular basis underlying tocochromanol synthesis and accumulation in maize grain.

Tocochromanols, lipid-soluble compounds synthesized in plants, algae, and some cyanobacteria, protect lipids by quenching both reactive oxygen species and lipid peroxyl radicals [4, 30, 42]. They are structurally separated into two classes: tocotrienols and tocopherols. Each class has four types: α, β, δ, and γ (Fig. 1). Tocotrienols, predominantly produced in the endosperm in monocots and a few dicots [16, 21, 57], are associated with numerous health benefits due to their potent antioxidant properties [45]. Tocopherols, primarily produced in the embryo [16], generally have higher vitamin E activity, with α-tocopherol providing the highest vitamin E activity on a molar basis [26].

Fig. 1

Tocochromanol biosynthesis pathway. The two precursor pathways and the carotenoid pathway are represented as black boxes. The six quantified tocochromanols and chlorophyll a are indicated in black nonitalicized text. The names of key a priori genes are bolded and italicized at the pathway step(s) catalyzed by their encoded enzymes, with the POR genes in red text. The key tocopherol precursor phytol is bolded in red. Compound abbreviations: DMGGBQ, 2,3-dimethyl-5-geranylgeranyl-1,4-benzoquinol; DMPBQ, 2,3-dimethyl-6-phytyl-1,4-benzoquinol; GGDP, geranylgeranyl diphosphate; GG-Chlorophyll, geranylgeranyl-chlorophyll a; HGA, homogentisic acid; HPP, p-hydroxyphenylpyruvate; MGGBQ, 2-methyl-6-geranylgeranyl-1,4-benzoquinol; MPBQ, 2-methyl-6-phytyl-1,4-benzoquinol; Phytyl-DP, phytyl diphosphate; Phytyl-P, phytyl monophosphate. Gene abbreviations: chlorophyll synthase (CHS); chlorophyll dephytylase1 (CLD1); farnesol kinase (FOLK); geranylgeranyl reductase (GGR); homogentisate geranylgeranyl transferase (HGGT); light harvesting protein-like 3 (LIL3); protochlorophyllide reductase (POR1, POR2, and POR3); tocopherol cyclase (VTE1); homogentisate phytyltransferase (VTE2); MPBQ/MGGBQ methyltransferase (VTE3); γ-tocopherol methyltransferase (VTE4); phytol kinase (VTE5); phytol phosphate kinase (VTE6); α-/β-hydrolase (VTE7). Figure updated from [55]

The core tocopherol biosynthesis pathway has been well-characterized in Arabidopsis thaliana [34] and is conserved across the plant kingdom (Fig. 1). The committed step of tocopherol synthesis is catalyzed by VTE2 (homogentisate phytyltransferase), which condenses homogentisic acid (HGA) and phytyl diphosphate (PDP) to produce 2-methyl-6-phytyl-1,4-benzoquinol (MPBQ), the immediate precursor of the four tocopherol types (Fig. 1). This condensation reaction and the availability of its two substrates, HGA and PDP, primarily determine the level of tocopherol produced within a tissue [4, 37]. HGA is derived from the aromatic amino acid pathway, which is tightly controlled by feedback inhibition [50]. In Arabidopsis, the PDP required for tocopherol synthesis is derived from a two-step phosphorylation of free phytol. Phytyl-P synthesis depends on two kinases, VTE5 and FOLK, which when knocked out results in tocopherol deficiency in 4-week-old Arabidopsis seed and leaf tissues [40]. Phytyl-P is subsequently phosphorylated to PDP by VTE6. vte6 null alleles result in non-detectable levels of tocopherol [51]. While the synthesis of PDP from phytol is well understood, the mechanism of phytol production for tocopherol synthesis remains unclear.

In photosynthetic tissues such as Arabidopsis leaf and seed, the phytol supply for tocopherol synthesis was thought to depend on chlorophyll degradation, during which chlorophyll a is converted to pheophytin a and then hydrolyzed by pheophytin pheophorbide hydrolase (PPH) to produce pheophorbide and phytol. However, in an Arabidopsis pph null mutant, tocopherol levels did not differ from the wild type (WT) in leaf and seed tissue [60]. An alternative pathway for phytol could be the chlorophyll-salvage cycle, which releases both chlorophyllide a and phytol through the dephytylation of chlorophyll a. Yet when a key gene in the chlorophyll-salvage cycle, CHLOROPHYLL DEPHYTYLASE1 (CLD1), was silenced there was no significant difference in tocopherol levels compared to the WT [27].

A recent study of natural variation in total seed tocopherols in Arabidopsis identified a strong dependence of tocopherol content on chlorophyll biosynthesis, with phytol being hydrolyzed from chlorophyll biosynthetic intermediates rather than from the degradation of the bulk chlorophyll pool [1]. Specifically, Albert et al. [1] found that null alleles of a seed-specific, plastid-localized alpha/beta hydrolase (VTE7) decreased seed tocopherol content by 55% with concomitant increases in specific chlorophyll biosynthetic intermediates with partially reduced tails. There was no impact on tocopherol content of Arabidopsis leaves, but in maize a leaky vte7 Mu-allele decreased kernel and leaf tocopherol content by 38% and 50%, respectively, indicating differences in phytol provision for tocopherol synthesis between monocot and dicot leaves [1].

Despite maize grain being a non-green, non-photosynthetic tissue, tocopherol synthesis in grain is also strongly dependent on chlorophyll synthesis. Two protochlorophyllide reductases (POR1 and POR2), key chlorophyll biosynthetic genes, underlie the largest effect QTL for total tocopherols in maize grain [5]. Their involvement is supported by near-isogenic and single knockout lines in maize, further highlighting the role of chlorophyll in tocopherol synthesis in non-photosynthetic tissue [29, 56]. In addition to its expression being regulated by light, POR also requires light as a co-substrate to catalyze the reduction of protochlorophyllide to chlorophyllide [14, 47]. POR is part of a complex that includes geranylgeranyl reductase (GGR), chlorophyll synthase (CHS), and a non-enzymatic protein, light harvesting protein-like 3 (LIL3), which catalyzes the final steps of chlorophyll a synthesis [19, 49]. The dependence of tocopherol synthesis on chlorophyll synthesis is further supported by knockouts of LIL3, GGR, and CHS, which result in severely reduced levels of chlorophylls and tocopherols in leaf tissue [18, 19, 48, 49, 58]. Given the large molar ratio of tocopherol to chlorophyll in developing maize embryos, Diepenbrock et al. [5] hypothesized that instead of chlorophyll degradation, a chlorophyll-based biosynthetic cycle provides phytol for tocopherol synthesis through the repeated removal of phytol during chlorophyll a synthesis. The identification of VTE7 and its impact on tocopherol abundance and chlorophyll biosynthetic intermediates in both leaf and grain tissue in maize support this model [1].

In this study, we took two complementary experimental approaches to assess the role(s) of POR1 and POR2 and therefore, the contribution of chlorophyll biosynthesis, in tocopherol biosynthesis within maize kernels. In physiological experiments, we imposed light and dark treatments on developing kernels to alter the in vivo light-dependent activities of all PORs [17] and measured the impacts on tocochromanols, chlorophyll a, and the transcriptome within embryo tissue. In CRISPR/Cas9 knockout experiments, we generated null alleles of POR1 and POR2 and evaluated the metabolomic consequences in embryos of the single and double mutants. From these experiments, we demonstrated the central role of chlorophyll biosynthesis in tocopherol synthesis within the maize embryo.

Materials and methods

Experimental design for evaluation of physiological treatments

We conducted a set of physiological experiments by exposing developing kernels to light and dark treatments to alter the in vivo activities of POR1 and POR2. The field experiment conducted in 2018 included founder lines of the U.S. maize nested association mapping (NAM) panel with contrasting POR1 and/or POR2 allelic effects [5]. Relative to the common parent B73, these included B97 and M37W with large POR1 effects, NC358 and Ki11 with large POR2 effects, and MS71 and OH7B with large effects for both POR1 and POR2. We also included B73 as a reference line, given that it is the common parent for this panel. The seven genotypes were planted in a split-plot design with two replications at Cornell University’s Musgrave Research Farm in Aurora, NY. Each main plot of a single genotype comprised three one-row subplots, each with a length of 5.33 m. The light treatment, dark treatment, and control condition were randomly assigned to each subplot. Each subplot had an average of 20 plants, with multiple plants in each subplot having their primary ears self-pollinated and covered with medium-sized pollination bags (MIDCO Global, St. Louis, MO). The light and dark treatments were applied to self-pollinated ears at 12 DAP. For each light-treated ear, the pollination bag was removed, and the outer husks were peeled, leaving only three to four layers, allowing overexposure to light while still retaining moisture for kernel development. A foil bag was used to completely cover the pollination bag of each dark-treated ear, preventing the penetration of light to the developing kernels. The control ears remained in their pollination bags to represent typical self-pollinated ear conditions (Supporting Information Fig. S1).

We harvested self-pollinated ears from each subplot at two kernel developmental time points for RNA-seq (at 24 DAP) and metabolite profiling (at 24 DAP and at maturity). At 24 DAP, three ears per subplot were collected to provide developing kernels for dissection. The harvested ears were immediately placed on wet ice in a cooler with a sealed lid for transportation. For each subplot, 20 kernels from the middle section of each of the three dehusked ears were manually dissected with a scalpel to separate embryos from other seed tissues. This procedure was conducted in a dark room, with each dissector wearing a green headlamp to minimize the initiation of light-regulated reactions. The embryos from all three ears were bulked to form a representative sample, frozen in liquid nitrogen, and stored at − 80 °C. The remaining selfed ears in each subplot (3–13 ears) were harvested at physiological maturity, dried, individually shelled, and then stored separately as mature kernel samples.

In 2019, to generate an additional year of metabolite data from mature kernels, we repeated the physiological experiment identical to that of 2018. However, this time, we only included four of the seven NAM founders —B73, B97, Ki11, and OH7B— planted in a split-plot design at Musgrave Research Farm. Due to a weather-delayed field planting, all six subplots of Ki11 and one subplot under light treatment of OH7B matured too late and were consequently not harvested. Self-pollinated ears from each subplot (2–7 ears) were harvested at physiological maturity and processed individually to generate mature kernel samples, following the same procedure as in 2018.

Metabolite analysis of physiologically treated embryo and kernel samples

The concentrations of tocochromanols, carotenoids, and chlorophyll a were measured in the 24 DAP embryo samples following previously described methods [28, 36]. Briefly, metabolites were extracted from ~ 15 mg of ground 24 DAP embryos, and their quantification was performed using HPLC and fluorometry. Similarly, to assess only the tocochromanol profile of the mature kernel samples, tocochromanols were extracted from ~ 30 mg of ground kernels and quantified using the same method as for the embryo samples. The tocochromanol compounds measured were α-tocopherol (α-T), δ-tocopherol (δ-T), γ-tocopherol (γ-T), α-tocotrienol (α-T3), δ-tocotrienol (δ-T3), and γ-tocotrienol (γ-T3). Additionally, the sum traits total tocopherols (ΣT, ΣT = α-T + δ-T + γ-T), total tocotrienols (ΣT3, ΣT3 = α-T3 + δ-T3 + γ-T3), and total tocochromanols (ΣTT3, ΣTT3 = ΣT + ΣT3) were calculated. The carotenoid compounds measured were neoxanthin, violaxanthin, lutein, zeaxanthin, β-carotene, and other carotenes. In addition, the total carotenoid content was calculated by summing all individual carotenoids.

The mature kernel samples were obtained from individual ears rather than multiple ears. This approach helped in identifying potential outliers at the level of an individual ear, which could result from factors such as premature drying of immature kernels or light leakage due to improper application of the light and dark treatments. Consequently, the raw data underwent visual inspection for extreme outliers using box plots. No samples, when considered collectively on a sub-plot basis, indicated human errors.

To test the degree to which light and dark treatments affected the metabolite profiles of developing and mature kernels, we conducted statistical analyses on three datasets (Dataset S1): metabolite profiles of mature kernel samples averaged by subplot from 2018 and 2019, and embryo samples from 2018. The two datasets of mature kernel metabolites were treated as separate experiments due to the varying number of genotypes evaluated in 2018 and 2019. Levene’s test was conducted to test for equality of variances with the SAS PROC GLM procedure in SAS studio release 3.8 (https://www.sas.com/en_us/software/on-demand-for-academics.html). The genotype and treatment terms were tested separately for each phenotype, with the most significant term (α = 0.01) selected to correct for unequal variances. With the metabolite levels for each subplot from the 2018 and 2019 mature grain (averaged by subplot) and 2018 embryo samples, we screened the three datasets separately for outliers. Studentized deleted residuals [35] were estimated using the SAS PROC GLIMMIX procedure, with correction for unequal variances of each phenotype in a mixed linear model (Eq. 1), as follows [9, 46].

$${\text{Y}}_{ijk} =\upmu + {\text{genotype}}_{i} + {\text{rep}}_{j } +\text{ genotype }\times {\text{rep}}_{ij} + {\text{treatment}}_{k} +\text{ genotype }\times {\text{ treatment}}_{ik} + {\upvarepsilon }_{ijk}$$

(1)

in which Y_ijk is an individual phenotypic observation; μ is the grand mean; genotype_i is the fixed effect of the ith genotype; rep_j is the random effect of the jth replicate; genotype × rep_ij is the random effect of the interaction between the ith genotype and jth replicate; treatment_k is the fixed effect of the kth treatment, genotype × treatment_ik is the fixed effect of the interaction between the ith genotype and kth treatment; and ε_ijk is the residual error effect assumed to be independently and identically distributed according to a normal distribution with mean zero and variance σ_ε², that is ~ iid N(0, σ_ε²). No significant outlier was identified after a Bonferroni correction of α = 0.01. An ANOVA test for the significance of model main effects (Table S1) and Tukey’s Honestly Significant Difference (HSD) test for pairwise differences among treatments within each genotype (Table S2) were subsequently performed using Eq. 1 in the SAS PROC GLIMMIX procedure with a correction for unequal variances. For each metabolite phenotype, the Eq. 1 model was used to separately generate best linear unbiased estimator (BLUE) values for the 2018 (mature grain and 24 DAP embryo) and 2019 (mature grain) datasets (Table S3). We calculated a Pearson’s correlation coefficient (r) between BLUE values of each pair of metabolites measured in the 24 DAP embryo samples using the ‘cor’ function and visualized the results with the ‘corrplot’ package version 0.92 [53] in R version 4.4.0 [38]. All other visualizations were created with the package “ggplot2” [54].

RNA-seq analysis of physiologically treated embryo samples

The embryo samples collected from the physiological experiment conducted in 2018 were utilized for RNA-seq analysis. We ground 30 embryos in liquid nitrogen and then subsampled 100–200 mg of the ground tissue for RNA extraction using a modified hot borate method [52]. RNA was treated with DNase and assessed for purity and quality using a NanoDrop spectrophotometer (Thermo Fisher Scientific, Wilmington, DE). High-quality RNA was then used to construct Illumina TruSeq Stranded mRNA libraries (Illumina, San Diego, CA), which were subsequently sequenced in single-end 50 nt mode on an Illumina HiSeq4000 (Illumina, San Diego, CA) at the Resource Technology Support Facility at Michigan State University.

RNA-seq reads were trimmed using Cutadapt version 1.18 [33] with a quality cutoff score of 20 and a minimum read length of 30 nt. Trimmed RNA-seq reads were aligned to the B73 RefGen_v4 reference genome [23] using HISAT2 version 2.1.0 [24] with the following parameters: –min-intronlen 20’ and –max-intronlen 60,000 in stranded mode. Read counts per gene were calculated using the count function of HTSeq version 0.6.1p1 [3] with the following parameters: –format = bam, –order = pos, –minaqual = 10, –idattr = ID, –type = gene, and –mode = union in stranded mode.

Differentially expressed genes (DEGs) were identified using the R package DESeq2 version 1.40.2 [32] for each pair of comparisons (control vs. dark, light vs. dark, and control vs. light) within each of the seven genotypes, resulting in a total of 21 comparisons. A false discovery rate corrected P-value threshold of 0.05 was applied. Principal component analysis (PCA) was performed on the standardized rlog values obtained from the DESeq2 package [32] with the ‘prcomp’ function from the R package ‘stats’ version 4.3.0 and visualized with ggbiplot version 0.55.

CRISPR/Cas9 vector construction and transformation for POR knockout generation

We employed CRISPR/Cas9 mutagenesis to create functional knockouts of both POR1 and POR2. In maize, POR1 (Zm00001d032576) and POR2 (Zm00001d013937) have 95.4% CDS sequence identity and 95.2% amino acid sequence identity, assessing sequences obtained from http://maize.uga.edu/ with Blast 2 sequences [2]. The high similarity enabled us to design a single gRNA construct (CRISPR-A) and a double gRNA (CRISPR-B) construct targeting conserved regions within the third exon of both genes (Fig. S2, Methods S1). Agrobacterium-mediated transformation of B104 embryonic callus was performed with the assembled vector (Methods S1) at Iowa State University’s Plant Transformation Center as described [12, 13]. The 24 T₀ seedlings from nine events (CRISPR-A generated six events and CRISPR-B generated three events) were shipped to Cornell University where they were self-pollinated or backcrossed with B104 to create BC₁F₁ individuals in a greenhouse at Cornell University’s Guterman Bioclimatic Lab in Ithaca, NY, in 2018.

Identification of homozygous POR knockout mutants

We used PCR to identify six transgene-free BC₁F₁ individuals with the desired CRISPR-A-induced knockout (Fig. S3, Methods S1), which were then used to generate BC₁F₂ plants homozygous for por1 and/or por2 knockout alleles at Cornell University’s Musgrave Research Farm in Aurora, NY, in 2019. Twelve seeds were planted in each plot, and plots were 3.05-m long. Sanger sequencing was employed to analyze purified PCR fragments obtained using POR1-specific and POR2-specific primer pairs (Tables S4, S5), revealing not only the presence or absence but also the type of induced mutations in each BC₁F₂ plant. The primary ear of each BC₁F₂ plant was self-pollinated following rigorously controlled pollination procedures (Methods S1). Selfed ears were harvested at physiological maturity, shelled, and the kernels of each ear were individually maintained for metabolite analysis. Among the six BC₁F₂ families, only CR9.3–5 and CR9.3–41 provided a sufficient number of samples for the different genotype classes, so they were prioritized for HPLC analysis. However, these two families had double mutant (por1/por1;por2/por2) plants that produced harvestable ears with only minimal seed.

Experimental design for evaluation of POR double knockout mutants

Due to the severity of the double knockout mutant phenotype, we planted seeds from self-pollinated por1/ + ;por2/por2 plants of the CR9.3–5 and CR9.3–41 families in a greenhouse environment. The por1/ + ;por2/por2 mutant was selected for the experiment due to its visibly stronger leaf color mutant phenotype compared to the por1/por1;por2/ + mutant. In 2020, seeds from self-pollinated BC₁F₂ WT (POR1/POR1;POR2/POR2) and por1/ + ;por2/por2 plants of the CR9.3–5 and CR9.3–41 families were sown in pots in the Guterman greenhouse to generate BC₁F₃ plants. For each of the two families, leaf tissue from seedlings derived from seeds of self-pollinated ears of por1/ + ;por2/por2 BC₁F₂ plants was collected for genomic DNA isolation, followed by PCR and Sanger sequencing as previously described, to classify BC₁F₃ plants as por2/por2, por1/ + ;por2/por2, or por1/por1;por2/por2. All individual BC₁F₃ plants of the four genotype classes, including WT, from both families, were self-pollinated. For each of the two families, a self-pollinated primary ear was harvested from each of four WT and four por1/ + ;por2/por2 BC₁F₃ plants at 24 DAP, followed by immediate freezing in liquid nitrogen after removing the husk leaves. For each frozen ear, the middle section was hand-shelled on dry ice, and the collected kernels were stored at − 80 °C. The remaining BC₁F₃ plants were retained for mature kernel harvest. However, only the CR9.3–5 family had por1/por1;por2/por2 plants with harvestable ears. Consequently, only the selfed primary ears of plants from this family (4 to 11 plants per genotypic class) were harvested at physiological maturity and processed individually to generate mature kernel samples for analyzing metabolite levels.

After genotyping the endosperm tissue (Methods S1), frozen kernels were dissected on a metal plate over dry ice to collect embryo and endosperm tissues. Kernels from each plant contributed 15 embryos and five endosperms yielding four sets of samples for every genotype class and tissue type. To ensure ample tissue for metabolite extraction, the first two sets and the last two sets were combined, resulting in the creation of two biological replicates for each genotype class, each comprising either 30 embryos or 10 endosperms. For RNA extraction, 10 embryos were collected from dissected frozen kernels from each plant in the CR9.3–5 family.

Metabolite analysis of POR knockout mutant embryo, endosperm, and kernel samples

We performed a metabolite analysis of embryo, endosperm, and mature kernel samples collected from field and greenhouse evaluations of WT and POR knockout mutants (Dataset S1). The procedures for tocochromanol, carotenoid, and chlorophyll a quantification are described in an earlier section. The only modification is that the total carotenoids phenotype was quantified as total area/mg fresh tissue.

Statistical analysis was performed to assess differences between CRISPR/Cas9 mutants and/or WT. All tests were performed in R (version 4.4.0). The developing embryos and endosperms from the 2020 greenhouse experiment and the mature kernels from the 2019 field experiment were analyzed with the following model:

$${\text{Y}}_{ij} =\upmu +{\text{ family}}_{i} + {\text{genotype}}_{j} + {\text{family}}_{i} \times {\text{genotype}}_{j} + {\upvarepsilon }_{ij}$$

(2)

in which Y_ij is an individual phenotypic observation; μ is the grand mean; family is the fixed effect of the ith family; genotype_i is the fixed effect of the jth genotype; genotype × family is the fixed effect of the interaction between the ith family and jth genotype; and ε_ij is the residual error effect assumed to be independently and identically distributed according to a normal distribution with mean zero and variance σ_ε², that is ~ iid N(0, σ_ε²). In the analysis of mature kernels from the 2020 greenhouse experiment, the variables of family and the interaction between family and genotype were removed, as only the family CR9.3–5 was analyzed. This is because the family CR9.3–41 did not produce enough seed for mature kernel analysis after the harvest of developing kernels. Studentized residuals were calculated using a Bonferroni correction with α = 0.01 in the package ‘car’. Two outliers were identified and removed: two chlorophyll a data points from endosperm tissue of the CR9.3–4.1 family, causing the removal of the family and the interaction term in subsequent analysis for this metabolite-tissue combination. Levene’s test and ANOVA (type III) (Table S6) were performed with the package ‘car’ version 3.1–2 [11]. Tukey’s HSD test (Table S7) was performed with the package “agricolae” [10].

RT-qPCR and RNASeq analysis of POR expression in double knockout mutant embryos

The four biological replicates of collected por1/por1;por2/por2 and WT embryos from the CR9.3–5 family were used for isolating total RNA. Briefly, each replicate comprising 10 embryos underwent grinding in a 1.5 mL conical bottom microcentrifuge tube using a pellet pestle (Fisher scientific, Pittsburgh, PA, USA). The grinding process ensured that the bottom of the tube remained submerged in liquid nitrogen, maintaining low temperatures throughout the procedure. We subsampled 100 mg of ground tissue for total RNA extraction using a modified hot borate method [52], followed by DNA removal employing an Ambion TURBO DNA-free Kit (Thermo Fisher Scientific, Waltham, MA, USA). The concentration and purity of RNA samples were assessed using an Epoch 2 Microplate Spectrophotometer (Agilent, Santa Clara, CA, USA), and the integrity of RNA samples was evaluated through gel electrophoresis. One microgram of total RNA was reverse-transcribed using the PrimeScript 1st strand cDNA Synthesis Kit (Takara, Kyoto, Japan) and oligo-dT according to manufacturer’s recommendations. Primer pairs were designed to amplify selected regions of POR1, POR2, and POR3 (Zm00001d001820) using Primer Premier 5 [25] (Table S4). RT-qPCR was performed in 96-well plates with a QuantStudio 7 Pro Real-Time PCR system (Thermo Fisher Scientific, Waltham, MA, USA), employing Power SYBR Green PCR Master Mix (Thermo Fisher Scientific, Waltham, MA, USA) (Table S5). The amplification conditions were as follows: 95 °C for 10 min, followed by 40 cycles of 95 °C for 15 s and 60 °C for 60 s. Each RT-qPCR analysis was performed in quadruplicate. Relative expression levels were calculated by normalizing to the reference gene ACTIN1 (Zm00001d010159) using the comparative CT method [43]. The 2^−ΔΔCT method [31] was used to analyze RT-qPCR in R (version 4.4.0), with the average expression of POR1 from WT embryos used as the control. ANOVA and Tukey HSD test were performed to compare the relative quantification of WT to double knockout mutant within and across each gene using the ‘stats’ package in R (version 4.4.0).

High-quality RNA from two of the four biological replicates of collected por1/por1;por2/por2 and WT embryos was used to construct Illumina TruSeq Stranded mRNA libraries (Illumina, San Diego, CA), which were subsequently sequenced in paired-end 150 nt mode on an Illumina Novaseq 6000 (Illumina, San Diego, CA) by Psomagen (Rockville, MD). RNA-seq reads were trimmed using Cutadapt version 4 [33] with a quality cutoff score of 30 and a minimum read length of 75 nt. Trimmed RNA-seq reads were aligned to the B73 RefGen_v4 reference genome [23] using HISAT2 version 2.1.1 [24] with the following parameters: –dta-cufflinks –max-intronlen 5000 and –rna-strandness RF. Read counts per gene were calculated using the count function of HTSeq version 0.13.5 [3] with the following parameters: –format = bam, –order = pos, –stranded = reverse, –minaqual = 10, –idattr = ID, –type = gene, and –mode = union. Differential expression analysis was performed as previously described.

Results

Tocopherol synthesis is light-dependent within developing embryo tissue

POR enzyme activity requires light as a co-substrate [17] and to understand the relationship between chlorophyll synthesis and tocopherol levels in non-photosynthetic kernel tissue, we conducted a physiological experiment on developing grain by exposing the ear to light and dark treatments (Fig. S1). At 12 days after pollination (DAP), we applied both treatments to a subset of maize NAM founder lines by covering the ears with aluminum foil for the dark treatment or removing 3–4 outer husks for the light treatment. In 2018, we tested seven NAM founders with contrasting allelic effects [5] of one or both POR loci, while a further subset of three NAM founders was evaluated in 2019. Metabolite analysis of mature kernels revealed that in both 2018 and 2019, the dark treatment reduced total tocopherols compared to light-treated samples by a mean of 82% and control samples by 78% across genotypes (Figs. S4, S5). Similar trends were observed for individual tocopherols (α-tocopherol, δ-tocopherol, γ-tocopherol) (Figs. S4-S6). We dissected embryos from 24 DAP developing kernels (from 2018) and found both chlorophyll a and tocopherols were reduced by a mean of 96% and 88%, respectively, in dark-treated samples compared to light-treated samples (Fig. 2a, b). There was a strong positive correlation (r = 0.6) between the concentrations of chlorophyll a and total tocopherol (Fig. S7). In 24 DAP embryos, total tocopherol levels were 40- to 200-fold higher than chlorophyll a levels across all genotypes in both the control and light treatments (Fig. 2). While the reduction in tocopherol abundance between dark-treated and control samples was consistent across genotypes, years, and tissues, the changes in total tocopherols between light-treated and control samples were inconsistent. For example, in 2018, six of the seven genotypes had a mean increase of 54% in embryo tocopherol levels in response to light treatment compared to control treatment, but in only two lines (B97 and MS71) did the significant increase in tocopherols persist in mature kernels (Figs. 2, S4).

Fig. 2

Untransformed best linear unbiased estimators (BLUEs) of metabolites from developing embryo tissue. a Total tocopherol concentration. b Chlorophyll a concentration. c Total tocotrienol concentration. d Total carotenoid concentration. Different letters denote P-value < 0.05 in Tukey’s HSD for within-genotype treatment comparisons. Error bars represent standard error estimates

Tocopherols and tocotrienols rely on many of the same enzymes for their synthesis (Fig. 1), but in contrast to tocopherols, the four tocotrienol traits (α-tocotrienol, δ-tocotrienol, γ-tocotrienol, and total tocotrienol) showed no consistent significant differences among different treatments across tissue types (Figs. 2c, S3-S5). Embryo samples from the dark treatment exhibited an average increase of 156% in total tocotrienols compared to light-treated or control samples in all but one genotype (NC358). However, this trend was again not apparent in the mature kernels (Figs. S4, S5). To assess whether the physiological treatment affected an unrelated biochemical pathway, we also measured carotenoids (β-carotene, lutein, zeaxanthin, neoxanthin, violaxanthin, and total carotenoids) in embryo samples. None of the carotenoid traits showed consistent patterns of change between treatments (Figs. 2d, S8), consistent with the treatment effects being specific to tocopherol synthesis.

POR1 and POR2 RNA levels in the embryo are not light-dependent

Substrates for tocopherols, tocotrienols, and carotenoids are derived from the same pathways: HGA from the shikimate pathway for tocopherols and tocotrienols, and for all three compound classes, GGDP from the MEP pathway. Given the impact of the dark treatment on tocopherol synthesis in both embryo and mature kernels and the lack of impact on carotenoid or tocotrienol levels, our results are consistent with the shikimate and IPP pathways being unaffected. We hypothesized that the expression of a priori genes in chlorophyll biosynthesis pathways, including POR1 and POR2, and other related pathways (Table S8) could explain the changes in tocopherols resulting from light and dark treatments. To test this hypothesis, we performed RNA-seq on bulked 24 DAP embryos dissected from fresh kernels for each genotype and experimental condition combination for the seven NAM founders in the 2018 experiment. Unlike tocopherols and chlorophyll a, which showed light-dependent variation, principal component analysis showed that the variation in gene expression was genotype- rather than treatment-dependent (Fig. S9). We assessed gene-level differences between treatments within a genotype to determine if differences in light-dependent tocopherol synthesis could be explained by differentially expressed genes (DEGs). Surprisingly, genes encoding activities of the POR/LIL3/CHS/GGR complex were not found to be differentially expressed between treatments using DESeq2 [32] (Fig. 3). However, the expression of POR genes within treatments and across genotypes did show strong positive correlations (r-values between 0.76–0.94) with joint-linkage allelic effect estimates [5] (Fig. S10), indicating that the genotypic effect remained consistent across treatments. Of the 125 a priori genes assessed (Table S8), only two genes (a coproporphyrinogen oxidase (NEC4) from the chlorophyll a biosynthesis pathway; and a tyrosine aminotransferase (NAAT1) from the tyrosine degradation pathway) were differentially expressed in only one of three treatment pairs (dark vs. control) in only one of the seven genotypes (M37W). Overall, the RNA-seq data indicates the light-dependent chlorophyll and tocopherol changes are not due to changes in the expression of the 125 a priori genes.

Fig. 3

Expression differences of genes within a priori pathways (Table S8) between physiological treatments in 24 DAP embryo tissue. Box plots show the average log fold change in gene expression across genotypes on the y-axis between two treatments as denoted on the x-axis. Each point represents the average log-fold change in expression between treatments for a single gene. Differences in the expression of CHS, GGR1, GGR2, LIL3, POR1, POR2, and VTE2 between treatments are denoted by color

Generating por1/por2 knockout mutants by CRISPR/Cas9

We generated CRISPR/Cas9 knockouts of POR1 and POR2 to directly determine their contributions to chlorophyll synthesis within the embryo and any effects on tocopherol abundance. Two CRISPR lines were selected for detailed analysis, CR9.3–5 and CR9.3–41 (Figs. S11 and S12). POR1 contained a one-base deletion in both lines, causing a frameshift at amino acid 138 and a premature stop at amino acid 188. In line CR9.3–5, POR2 had a one-base insertion, resulting in a frameshift at amino acid 138 and premature stop at amino acid 253. In line CR9.3–41, POR2 had a nine-base deletion that removed amino acids 138–140, along with a one-base insertion resulting in a frameshift at amino acid 203 and a premature stop at amino acid 213. All of these mutations eliminated conserved amino acids essential for substrate binding and catalysis, resulting in null mutations [6, 41, 59]. After the selection of suitable transgene-free por1 and/or por2 knockout BC₁F₁ individuals (Fig. S3), 380 transgene-free BC₁F₂ seedlings successfully germinated in the field and could be visually separated into three groups at the seedling stage: 308 green-leaf individuals, 44 individuals that exhibited a green-yellow leaf color, and 28 yellow-leaf individuals (Fig. 4). At maturity the green-yellow plants became indistinguishable from WT, while the yellow-leaf plants remained yellow with a small and stunted phenotype. PCR genotyping results revealed that the yellow-leaf plants were homozygous double mutants (por1/por1;por2/por2), the green-yellow leaf plants carried a single functional allele of POR1 (por1/ + ;por2/por2) and the green-leaf plant group contained all other possible genotypes including plants with a single functional allele of POR2 (por1/por1;por2/ +). This difference in leaf color suggests that the por2 mutation had a stronger effect on maize leaf chlorophyll content during development in the B104 background, consistent with results from [29]. Since por1 (por1/por1) and por2 (por2/por2) homozygous single mutants were also indistinguishable from WT (Fig. 4), we speculated that the function of POR1 and POR2 were complementary in leaves and that they had dose effects on maize leaf chlorophyll content.

Fig. 4

Green-yellow plant color segregation was observed after emergence in BC₁F₂ progeny planted in 2019. Scale bars are 2 cm. Genotypes are grouped according to leaf color

Metabolite analysis of por1/por2 mutants

We harvested mature kernels from the field-grown, self-pollinated BC₁F₂ plants with WT, por1 single mutant (por1/por1), por2 single mutant (por2/por2), and por1 or por2 monoallelic double mutant genotypes (por1/ + ;por2/por2 and por1/por1;por2/ + , respectively) from CRISPR families CR9.3–41 and CR9.3–5 (Figs. S11, S12) and analyzed them for tocochromanols and carotenoids by HPLC. All tocochromanol traits and total carotenoids in por1 and por2 single mutants were similar to WT (Figs. 5a, S13). Tocopherol levels in por1/ + ;por2/por2 and por1/por1;por2/ + kernels were reduced by ~ 41% and ~ 36%, respectively, compared to WT. Total tocotrienol and carotenoid levels in the kernel were ~ 30% and ~ 36% higher, respectively, in por1/por1;por2/ + mutants and unchanged in por1/ + ;por2/por2 mutants compared to WT (Fig. 5a). Kernels of these monoallelic double mutants are still segregating for multiple genotypes at their por1 and por2 loci, respectively, and the wild type segregants at each locus could partially obscure their impact on tocochromanols and carotenoids. Unfortunately, the por1 por2 double homozygous mutant plant (por1/por1;por2/por2) was severely stunted in the field and produced minimal seed.

Fig. 5

Effects of por1 and/or por2 knockout mutations on tocochromanols and carotenoids in mature kernels. a The metabolomic results of 53 samples from family CR9.3–41 (indicated by red dots) and CR9.3–5 (indicated by black dots) from the 2019 field experiment. b The metabolomic results of 31 samples from family CR9.3–5 from the 2020 greenhouse experiment. Different letters denote P-value < 0.05 in Tukey’s HSD for genotype comparisons. The boxplot color reflects the leaf color of the genotypes grown in the 2019 field experiment (Fig. 4)

To determine the consequences of the por1/por1;por2/por2 mutant genotype on tocopherol, tocotrienol, and carotenoid levels in mature grain, we grew and harvested mature kernels from WT, por2/por2, por1/ + ;por2/por2, and por1/por1;por2/por2 mutants in a greenhouse environment. Despite the reduced size of the por1/por1;por2/por2 mutant compared to WT (Fig. S14), it yielded substantial amounts of phenotypically similar mature kernels for HPLC analysis (Fig. S15). Tocopherol levels in greenhouse-grown plants generally agreed with those in field-grown genotypes (Figs. 5b, S16). Relative to the WT, the por2 single mutant showed the smallest reduction (18%), the por1/ + ;por2/por2 mutant had a ~ 55% reduction, and the por1/por1;por2/por2 mutant was the most severe with a ~ 93% reduction in tocopherol levels (Fig. 5b). In contrast, the carotenoid and tocotrienol levels of the three mutant genotypes were indistinguishable from WT (Fig. 5b).

To assess if the stunted phenotype of the por1/por1;por2/por2 mutant contributes to the exceedingly low tocopherol in mature kernels (Fig. 5b), we collected and genotyped developing kernels (24 DAP) from the self-pollinated ears of greenhouse-grown por1/ + ;por2/por2 mutants, allowing us to generate por1/por1;por2/por2 kernels from phenotypically WT plants (Fig. S17). We used single kernel genotyping to identify kernels carrying por1/por1;por2/por2, por2/por2, and por1/ + ;por2/por2 genotypes and then performed frozen kernel dissection to collect embryo and endosperm tissues for RT-qPCR analysis of the POR genes and metabolite analyses. Expression of POR1 and POR2 in por1/por1;por2/por2 mutant embryos was significantly decreased compared to WT, while average POR3 expression remained unchanged between the genotypes and was only 0.2% of WT POR1 expression. (Fig. S18). No other a priori genes besides POR1 and POR2 (Table S8), including VTE genes and GGR homologs, were differentially expressed between the por1/por1;por2/por2 mutant and the WT (Fig. S19), similar to the RNAseq results from the physiological experiment (Fig. 3). In embryo tissues, the por1/por1;por2/por2 mutant displayed a ~ 76–83% reduction in total tocopherols, along with a ~ 60–64% increase in total tocotrienols compared to the WT (Figs. 6a, S20). The por1/por1;por2/por2 mutant also had reduced tocopherol levels compared to the por2/por2 mutant and the por1/ + ;por2/por2 mutant (~ 76–77% and ~ 64–70% respectively). Additionally, the por1/por1;por2/por2 mutant embryos showed a > 98% reduction in chlorophyll a compared to the WT (Fig. 6a). Comparable levels of total carotenoids were found in double mutant embryos compared to WT. In contrast to embryo tissue, the metabolite levels in endosperm tissue were consistent across genotypes (Fig. 6b). Endosperm tissue accumulated chlorophyll a below our limit of detection, at least 100-fold lower than that observed in embryo tissue (Fig. 6b). By dissecting endosperm and embryo tissue from genotypically distinct kernels from the same ear, we show that tocopherol levels in maize grain are dependent on the genotype of the embryo. Overall, the functional knockout of POR1 and POR2 supports the reliance of tocopherol biosynthesis on chlorophyll biosynthesis within the embryo.

Fig. 6

Effects of por1 and/or por2 knockout mutations on tocochromanols, carotenoids, and chlorophyll a in developing kernel tissues. The metabolomic results from family CR9.3–41 (indicated by red dots) and CR9.3–5 (indicated by black dots) from 24 DAP embryo (a) and endosperm (b). Different letters denote P-value < 0.05 in Tukey’s HSD for genotype comparisons. The boxplot color reflects the leaf color of the genotypes grown in the 2019 field experiment (Fig. 4)

Discussion

The core tocopherol biosynthesis pathway is well known, but the mechanism(s) supplying the phytol group, a key precursor for tocopherol synthesis, has yet to be understood in both photosynthetic and non-photosynthetic tissues, such as maize grain. Despite maize embryos having extremely low chlorophyll levels (~ 500-fold lower than leaves) and tocopherol levels up to ~ 1000-fold higher than chlorophyll a, association studies in maize indicated that POR1 and POR2, which encode a key activity in chlorophyll biosynthesis, underlie major QTL for natural variation in total tocopherol levels in mature grain [5, 55]. To understand the contribution that chlorophyll biosynthesis makes to tocopherol biosynthesis within the maize embryo, we altered POR1 and POR2 activity by two complementary approaches: physiological inhibition of POR activity by darkness and single and double CRISPR/Cas9-mediated knockouts of the two genes. This resulted in up to a 93% reduction in tocopherol content in mature kernels, indicating tocopherol biosynthesis is almost entirely dependent on chlorophyll biosynthesis within the non-photosynthetic maize embryo.

In the first approach, we took advantage of the light-dependent enzyme activity of POR1/POR2 to inhibit chlorophyll synthesis in vivo [39]. By withholding light during kernel development, we could assess the dependence of tocopherol synthesis on chlorophyll synthesis. Using a diverse subset of the NAM founders with differing POR1 and/or POR2 allelic effects [5], we found that the abundance of chlorophyll a and total tocopherols in dark-treated embryo tissue decreased by 92–99% and 87–90%, respectively, compared to light-treated embryo tissue (Fig. 2a and b). Because light levels strongly impact POR gene expression in vegetative tissues [47], we assessed if light deprivation impacts POR1 and POR2 gene expression in embryo tissue. Surprisingly, compared to control, the dark treatment did not significantly reduce the mRNA levels of POR1, POR2, or other a priori genes for IPP synthesis, chlorophyll metabolism, and tocopherol biosynthesis (Table S8) within embryos across all genotypes (Fig. 3). This is consistent with the impact of the dark treatment on tocopherol and chlorophyll a abundance being caused by translational or posttranslational impacts on POR activity, or the stability/activity of other members of the POR/CHS/GGR/LIL3 complex [18, 48, 49, 58], as well as potential interactions with VTE biosynthetic enzymes. The low tocopherol levels remaining in dark-treated grain samples could arise from multiple sources including a trace amount of light leakage into the ear, the accumulation of tocopherols or their precursors prior to initiating dark treatment at 12 DAP, and/or from the availability of small amounts of GGDP-derived PDP that might not rely on chlorophyll-derived phytol.

In the second approach, we targeted the third exon of both POR1 and POR2 to create homozygous single and double knockout mutants with severely truncated, nonfunctional enzymes. Compared to the WT, the por1 por2 double homozygous mutant (por1/por1;por2/por2) had a metabolomic phenotype similar to that of dark-treated ears: a ~ 93% reduction of tocopherols in mature kernels (Fig. 5b) and a 76–83% and 98–100% reduction in tocopherols and chlorophyll a, respectively, in 24 DAP developing embryos (Fig. 6a). The low levels of tocopherol in mature grain of the por1/por1;por2/por2 double mutant could be due to the activity of POR3, the third POR homolog in maize that is mainly expressed during the first several days of grain development [7]. The similarity of reductions in grain tocopherol levels between the single por mutants (por1/por1 or por2/por2) and between the monoallelic double mutants (por1/ + ;por2/por2 or por1/por1;por2/ +) indicate that POR1 and POR2 play complementary roles in tocopherol synthesis within maize grain. In contrast, leaves of por1/por1;por2/ + mutant seedlings were visibly paler than the por1/ + ;por2/por2 mutant (Fig. 4), consistent with POR2 making a greater contribution to chlorophyll biosynthesis in leaf tissue than POR1 [56]. Both the genetic and physiological inhibition of POR provide conclusive evidence that nearly all tocopherol produced in maize kernels depends on chlorophyll biosynthesis within the embryo.

Tocopherols and tocotrienols are closely related chemically and biosynthetically with only a single gene differing in their pathways (Fig. 1); tocopherols require VTE2 for the condensation of HGA and PDP, while tocotrienols require HGGT for condensation of HGA and GGDP. Unlike the stark reduction in tocopherols, por mutations and dark treatments increased the tocotrienol content in developing embryos, suggesting that the availability of HGA was not decreased by the enzymatic activity of POR (Figs. 2d, and 6a). Similarly, GGDP levels also appear to be unaffected, as carotenoids—which depend on GGDP for their synthesis—remain unchanged in both experiments (Figs. 2c, and 6a). The increase in embryo tocotrienols is likely due to the ability of VTE2 to utilize its less preferred substrate, GGDP, when the levels of GGDP greatly exceed that of PDP [57]. The opposite response of tocotrienol and tocopherol synthesis to light deprivation is consistent with decreased availability of chlorophyll-derived PDP specifically for tocopherol synthesis being impacted rather than a reduction in other tocochromanol pathway substrates or VTE2 expression (Fig. 3). Given the large stoichiometric differences between chlorophyll a and tocopherol levels in developing maize kernels (Figs. 2a,b, and 6a), our data supports the hypothesis [5] that chlorophyll participates in a biosynthetic cycle within embryos. This biosynthetic cycle generates many molecules of phytol per molecule of chlorophyll—unlike chlorophyll degradation, an irreversible process that produces only a single molecule of phytol per molecule of chlorophyll [20].

A prior study [56] used tocopherol data from reciprocal hybrids with differing POR2 alleles, along with in vitro feeding of phytol to isolated, developing kernels, to hypothesize that the phytol for tocopherol synthesis in embryos is generated in leaf tissue and transported to the embryo by an unknown mechanism. Our data do not support this hypothesis, instead, they indicate the source of phytol for tocopherol synthesis in embryos is the embryo itself. The physiological experiment did not affect leaves, and if phytol was transported from leaves for synthesis in embryos, one would expect no impact on embryo tocopherol synthesis. Embryo and mature kernel tocopherol levels were instead severely decreased (Figs. 2, S4 and S5). In a second independent approach, we self-pollinated the phenotypically robust por1/ + ;por2/por2 mutant in the greenhouse to generate ears segregating for por1/por1;por2/por2, por1/ + ;por2/por2, and por2/por2 genotypes, which were identified by single kernel genotyping. In this experiment, kernel genotypes on each ear experience identical maternal effects and environmental conditions during development and, if phytol for tocopherol synthesis in embryos were transported from other tissue, one would again expect no impact on embryo tocopherol levels. Instead, a large reduction in tocopherol and chlorophyll levels was observed in por1/por1;por2/por2 embryos and mature kernels compared to por2/por2 and por1/ + ;por2/por2. These two experiments conclusively demonstrate that the phytol produced within the embryo, rather than the transport of phytol from maternal leaf tissue, determines embryo tocopherol levels (Fig. 6a).

In summary, we can now conclude that ~ 93% of total tocopherol in mature maize kernels is dependent on chlorophyll-derived phytol synthesized within embryo tissue. Defining the contribution of chlorophyll biosynthesis to tocopherol synthesis in the maize embryo indicates that POR1 and POR2 are primary breeding targets for elevating the total tocopherol content in maize, and likely other cereal grains. Selecting superior POR1 and POR2 alleles for total tocopherols, along with other loci that control the types of tocopherols accumulated (VTE3 and VTE4), should result in an optimized grain profile that has both increased tocopherol content and enhanced vitamin E activity to benefit agriculture and human health.