Vitamin A exists in several forms known as retinoids. Human can synthesize retinal from the abundant provitamin A carotenoids present in fruits and vegetables such as oranges (Citrus aurantium), broccoli (Brassica oleracea), spinach (Spinacia oleracea), carrot (Daucus carota), squash (Cucurbita maxima), sweet potato (Ipomoea batatas), and pumpkin (Cucurbita maxima) (Harrison, 2005). Plants produce four types of provitamin A carotenoids from isopentenyl diphosphate (IPP), which can form the intermediate product phytoene. Among carotenoids, α-carotene and β-carotene accumulate in greater amounts than γ-carotene and β-cryptoxanthin, which tend to be converted rapidly into downstream products in several forms known as retinoids (reviewed by Bai et al., 2011; Figure 1).

The RDI of vitamin A is 700 μg retinol equivalents per day²; however, provitamin A is non-detectable in rice (Oryza sativa), millet (Panicum miliaceum), and sorghum (Sorghum bicolor), and low in wheat (Triticum aestivum), barley (Hordeum vulgare), and potato (Solanum tuberosum)¹. In plants, β-carotene yields β-cryptoxanthin, which is further converted into zeaxanthin, whereas α-carotene yields lutein; and these chemicals are non-provitamin carotenoids (Figure 1). Both conventional and biotechnological approaches have been used to enhance the biosynthesis of provitamin A carotenoids in crops. The most commonly used gene for carotenoids biofortification is the gene encoding a phytoene synthase (PSY in plant; CRTB in bacteria; Figure 1). In Golden Rice, the introduction of a Pantoea ananatis phytoene desaturase (CrtI) and a Narcissus pseudonarcissus phytoene synthase (PSY) into rice results in an accumulation of 1.6 μg total carotenoids per gram of dry seeds (Ye et al., 2000). In Golden Rice 2, a maximum of 37 μg g^-1 total carotenoids with β-carotene preferentially accumulated in the endosperm is achieved when a maize (Zea mays) PSY1 gene is utilized (Paine et al., 2005). In these two events, the enhanced phytoene synthesis guarantees an adequate metabolic flux throughout the pathway. Similar strategies, i.e., overexpression of the phytoene synthase gene alone, phytoene desaturase gene alone, or combination of these two genes, are applied to other crops including canola (Brassica napus; both genes in Ravanello et al., 2003; Aluru et al., 2008; crtB in Shewmaker et al., 1999), flax (Linum usitatissimum; crtB in Fujisawa et al., 2008), potato (crtB in Ducreux et al., 2005), tomato (Solanum lycopersicum; crtB in Fraser et al., 2002; both genes in Aluru et al., 2008), maize (both genes in Naqvi et al., 2009), soybean (Glycine max; crtB in Schmidt et al., 2015), and wheat (both genes in Wang et al., 2014), among which the highest fold change of β-carotene is observed in soybean (Figure 1). A chimeric gene consisting of a chloroplast signal from pea (Pisum sativum) and a crtB from bacterium Pantoea is introduced into soybean via biolistics, resulting in an accumulation of 845 μg/g β-carotene, about 1500-fold increase as compared to wild type, in dry seeds (Schmidt et al., 2015).

Some other approaches are also used to enhance carotenoid biosynthesis in plants. Lycopene β-cyclase (LYCB), lycopene 𝜀-cyclase (LYCE), and HYDB are key enzymes in the β-branch of the carotenoid biosynthesis pathway. To achieve a high level of β-carotene accumulation, suppressing the activity of LYCE by silencing StLCY-e in potato or the activity of HYDB by silencing StCHY-𝜀 in potato is adopted to ensure the appropriate direction of the metabolic flux (Diretto et al., 2007; Van Eck et al., 2007). In wheat, simultaneous overexpression of CrtB and silencing of HYDB achieves an up to 31-fold increase of β-carotene up to 5.06 μg g^-1 (Zeng et al., 2015). Another approach is to make use of the Orange (Or) gene from cauliflower (Brassica oleracea var. botrytis, Lu et al., 2006). The OR protein is responsible for the accumulation of carotenoids in plants, and interacts with PSY to increase the stability and activity at a post-translational level for carotenoid biosynthesis (Zhou et al., 2015). Transgenic potato tubers expressing Or exhibit elevated carotenoids, and the β-carotene is increased continuously during long-term cold storage (Lopez et al., 2008; Li L. et al., 2012). Similar strategy is successful in tomato (Bai et al., 2014).

However, these strategies can also have disadvantages in some cases. For example, overexpression of PSY in tomato increases lycopene, β-carotene, and zeaxanthin, but decreases gibberellins, which results in plant dwarfism (Fray et al., 1995). This observation indicates that manipulation of the provitamin A metabolism is still limited by our incomplete understanding of the regulation of the endogenous pathways, including rate-controlling steps and timing of expression in carotenogenic tissues (Vallabhaneni and Wurtzel, 2009). Therefore, identification of the allelic variations that control carotenoid biosynthesis by means of candidate gene-based genome-wide association studies (GWASs) to better facilitate provitamin A biofortification through allele pyramiding has become an important alternative. Much work to identify the genes or QTLs responsible for the carotenoid accumulation has been done in maize, given its marked genetic diversity (Palaisa et al., 2003; Vallabhaneni and Wurtzel, 2009; Yan et al., 2010; Azmach et al., 2013; Owens et al., 2014; Figure 1). Besides, maize lines containing such natural variations can be used as donor parents to accelerate developing provitamin A-fortified tropical maize varieties, adapted to target growth conditions and consumer preferences. For example, the β-carotene-rich hybrid maize has developed through the marker-assisted introgression of the natural alleles in β-carotene hydroxylase gene by using a crtRB1-specific DNA marker for foreground selection. (Muthusamy et al., 2014).

To identify other metabolic bottlenecks in the carotenoid pathway, the genes of Arabidopsis 1-deoxy-D-xylulose-5-phosphate synthase (AtDXS) and Arabidopsis ORANGE (AtOR) are introduced into Golden Rice 2. Upon the transformation of AtDXS, accumulation of the carotenoids in the endosperm is significantly enhanced, confirming that the supply of isoprenoid precursors such as geranylgeranyl diphosphate (GGPP) is a rate-limiting step (Bai et al., 2016). In Sorghum, AtDXS was coexpressed with PSY1 and CRTI, and the transgenic events accumulated higher levels of total carotenoids, and the all-trans β-carotene levels ranged from 2.5 to 9.1 μg g^-1 DW in the mature-seed endosperm (Che et al., 2016). Upon the transformation of AtOR, levels of the carotenoids are also dramatically elevated, indicating that OR functions mainly through expanding the metabolic sink (Bai et al., 2016). Obviously, identification of the metabolic bottlenecks in the carotenoid pathway can help to refine the strategies for development of crops with specific carotenoids.

Folates, also called vitamin B₉, are essential water-soluble B-vitamins, including tetrahydrofolate (THF) and its derivatives. Folates play an important role as one-carbon donors and acceptors in all organisms. Folates are synthesized de novo in bacteria, fungi and plants. It is noteworthy that the folate biosynthesis pathway is split among cytosol, mitochondria, and chloroplasts in plants, whereas it is cytosolic in other organisms (reviewed by Storozhenko et al., 2005; Blancquaert et al., 2010, Figure 2). Usually, folates are abundant in dark-green leafy vegetables, but only tens of micrograms per 100 g weight in barley, corn, lettuce (Lactuca sativa), potato, rice, sweet potato, tomato, wheat, etc¹. Therefore, folate intake is prone to inadequacy, i.e., less than the RDI of 400 μg per day for adults², if the dietary pattern is not optimized.

Scientists have made lots of efforts to develop various folate-fortified crops using biotechnology in recent years, including tomato, rice, maize, lettuce, potato, and Mexican common bean (Phaseolus vulgaris) (Diaz de la Garza et al., 2004, 2007; Storozhenko et al., 2007; Naqvi et al., 2009; Nunes et al., 2009; Blancquaert et al., 2013; Ramirez Rivera et al., 2016; Figure 2). In general, two approaches for folate biofortification have been employed. One is to overexpress dihydrofolate synthetase (DHFS; FOLE in bacteria), a rate-limiting enzyme in folate synthesis, achieving an only 2-fold increase in folE-overexpressing corn endosperm up to 1.94 μg g^-1 (Naqvi et al., 2009). The other strategy, i.e., overexpression of the enzymes catalyzing the first committed step in the cytosol (CTPCHI), increases folates by 2 folds in tomato up to 2.99 nmol g^-1(Diaz de la Garza et al., 2004), 8.5 folds in lettuce up to 1.85 μg g^-1 (Nunes et al., 2009), respectively. Similar strategy is used in Mexican common bean and increased pteridine by 3 folds up to 3.25 μg g^-1 (Ramirez Rivera et al., 2016). A greater increase in tomato (25 folds, up to 8.40 μg g^-1) and rice (100 folds, up to 17.23 μg g^-1) is achieved by crossing the pteridine-overproducing traits with para-aminobenzoate (p-ABA)-overproducing traits (Diaz de la Garza et al., 2007; Storozhenko et al., 2007).

A similar strategy does not work in potato, regardless of the high expression of GTPCHI and Aminodeoxychorismate synthase (ADCS): folates in the immature potato tubers are mildly increased (∼ 2 folds, up to 1.23 μg g^-1), and expression of the endogenous GTPCHI and ADCS does not differ between the WT and transgenic lines (Blancquaert et al., 2013). These observations are markedly different from those reported in tomato and rice (Storozhenko et al., 2007; Waller et al., 2010). Therefore, probably there is a bottleneck in the folate biosynthesis pathway in potato tubers, and a further research is needed to investigate the mechanisms of regulating folate status in these plants that are different, lacking, or of minor importance in rice seeds and tomato fruit. Besides, the combination of folylpolyglutamate synthetase (FPGS), GTPCHI, and ADCS from Arabidopsis with folate binding proteins (FBP) from mammals has been successful in improving folate accumulation (up to 25.30 μg g^-1) and stability (folate levels were stable for 4 months at 28°C) in rice (Blancquaert et al., 2015). These results indicate that a better understanding of the folate pathway is required to formulate an engineering strategy useful for the majority of other staple crops (Blancquaert et al., 2013).

B-class vitamins, other than folates, include VB₁ (thiamine), VB₂ (riboflavin), VB₃ (niacin), VB₅ (pantothenic acid), VB₆ (pyridoxal, pyridoxine, pyridoxamine, and their phosphorylated derivatives), VB₇ (biotin), and VB₁₂ (cobalamin). To our knowledge, however, only VB₆ metabolic engineering has been mainly conducted in cassava (Manihot esculenta; Vanderschuren et al., 2013; Li et al., 2015).

VB₆ refers to a group of six water-soluble vitamers, among which pyridoxal-5′-phosphate (PLP) is of central importance because it is required as a cofactor for over 140 chemical reactions in the cell (Vanderschuren et al., 2013; Figure 3). The RDI for vitamin VB₆ is 1.3 mg per day for adults². Several groups reported overexpression of the genes either encoding pyridoxal phosphate synthase (PDX1) or encoding pyridoxal phosphate glutaminase (PDX2) in Arabidopsis, but the increase of VB₆ is not significant in most cases (Chen and Xiong, 2009; Leuendorf et al., 2010; Raschke et al., 2011). Later on, however, the simultaneous overexpression of Arabidopsis PDX1.1 and PDX2 genes in cassava achieved a 9.0-fold increase in the leaves to 54.74 μg g^-1 by CaMV35S promoter, and a 15.4-fold increase in the roots to 16.21 μg g^-1 by root-enhanced Patatin promoter, respectively (Li et al., 2015). Apparently, understanding of the regulatory mechanisms underlying the metabolic pathways can be improved in order for people to eventually benefit from the biofortification.

Vitamin C, also known as ascorbate, can be synthesized via four pathways. In plants, D-glucose 6-P, D-galacturonate, and myo-inositol are the substrates of the Smirnoff-Wheeler, pectin degradation, animal, and animal-like pathways, respectively (reviewed by Locato et al., 2013; Figure 4A). The RDI for vitamin C is 75 mg per day for adults². Potato and tomato contain one fourth to one fifth of the RDI for vitamin C per 100 g weight; while vitamin C is quite low and usually undetectable in the grains of cereal crops, including barley, corn, millet, rice, and wheat¹.

Vitamin C biofortification has been carried out in lettuce (Jain and Nessler, 2000), maize (Chen et al., 2003; Naqvi et al., 2009), potato (Hemavathi et al., 2009; Qin et al., 2011; Zhang et al., 2011; Bulley et al., 2012), tomato (Bulley et al., 2012; Cronje et al., 2012; Gest et al., 2013), and strawberry (Fragaria ananassa; Bulley et al., 2012) by either overexpressing the genes involved in biosynthesis or silencing the genes involved in ascorbate recycling (Figure 4). Most of the overexpression in Arabidopsis, tomato, strawberry and potato results in a 2- to 6-fold increase in ascorbate (Lorence et al., 2004; Hemavathi et al., 2009; Zhang et al., 2011; Bulley et al., 2012; Cronje et al., 2012). A 7-fold increase of ascorbate, the largest increase observed up to date, was achieved in transgenic lettuce when a rat L-Gulono-1,4-γ-lactone oxidase (GuLO) was overexpressed (Jain and Nessler, 2000). Unlike the overexpression, knockdown of the genes participating in recycling, such as monodehydroascorbate reductase (MDHAR) and dehydroascorbate reductase (DHAR), leads to less increase in tomato, maize, and potato (Chen et al., 2003; Naqvi et al., 2009; Qin et al., 2011; Gest et al., 2013; Figure 4B).

Vitamin E is important for human health, and dietary or supplemental vitamin E is absorbed and delivered to the liver (Traber, 2007). Plants are the primary source of dietary vitamin E, producing the tocopherol and tocotrienol derivatives that collectively constitute vitamin E (Chen et al., 2006). The RDI for vitamin E is 15 mg per day for adults^1,2. Tocotrienols are the major form of vitamin E in the seeds of most monocots and a limited number of dicots (Cahoon et al., 2003). Tocopherols have one saturated phytyl tail, while tocotrienols have 3-fold unsaturated side chains. Due to the presence of phenolic groups, vitamin E is easily oxidized and these derivatives are effective antioxidants. (Munne-Bosch and Falk, 2004; Figure 5). Among these derivatives, α-tocopherol has the highest biological activity for human health (Traber, 2007). Thus, the goal of vitamin E biofortification is to increase the contents of vitamin E and to convert all the other types into α-tocopherol.

Mostly, approaches to enhance vitamin E are to increase the activity of the enzymes in each step of the synthesis, including p-hydroxyphenylpyruvate dioxygenase (HPPD; Farre et al., 2012), homogentisate phytyltransferase (HPT1/VTE2; Seo et al., 2011), homogentisate geranylgeranyl transferase (HGGT; Cahoon et al., 2003), homogentisate solanesyltransferase (HST; Sadre et al., 2006), 2-methyl-6-phytyl-benzoquinol methyltransferase (MBPQ-MT/VTE3; Sattler et al., 2004; Tang et al., 2016), tocopherol cyclase (TC/VTE1; Kumar et al., 2005; Yabuta et al., 2013), and γ-tocopherol methyltransferase (γ-TMT/VTE4; Van Eenennaam et al., 2003; Cho et al., 2005; Tavva et al., 2007; Lee et al., 2008; Ghimire et al., 2011; Yabuta et al., 2013; Zhang et al., 2013; Figure 5).

Among the trials mentioned above, either a single gene or multiple genes in combination have been adopted. For example, when the barley HGGT gene is transformed into maize, the total tocotrienols and tocopherols increase around six folds (around 800 nmol g^-1) in the transgenic seeds (Cahoon et al., 2003). Barley HGGT gene also improves all-trans β-carotene stability in transgenic sorghum events containing AtDXS, PSY1, and CRTI (Che et al., 2016). The most commonly used single-gene approach is the overexpression of γ-TMT, which results in an increased proportion of α-tocopherol among the tocopherols, such as in canola (7.0-fold increase to over 70% of total; Van Eenennaam et al., 2003), lettuce (over 2-fold increase to 99% of total; Sattler et al., 2004; Cho et al., 2005), Perilla frutescens (1.8-fold increase to 99% of total with the CaMV 35S promoter, Ghimire et al., 2011; 26.6-fold increase to 75% of total with a seed-specific Vicillin promoter, Lee et al., 2008), and soybean(7.0-fold increase to 75% of total, Sattler et al., 2004; 10.4-fold increase to 88% of total,10.4 folds, Tavva et al., 2007) Compared with the single gene, the overexpression of multiple genes involved in vitamin E biosynthesis usually results in more massive accumulation of tocotrienols, and the total vitamin E activity is dramatically increased (Collakova and DellaPenna, 2003; Van Eenennaam et al., 2003; Sattler et al., 2004). For example, the combination of HPT and γ-TMT results in a 12-fold increase of vitamin E activity up to 47.4 mg α-tocopherol 100 g^-1 tissue in canola seeds (Collakova and DellaPenna, 2003); the combination of MT and γ-TMT results in a 5-fold increase in soybean (Van Eenennaam et al., 2003; Sattler et al., 2004); and the combination of tyrA (HPT in yeast), Arabidopsis HPPD and HPT results in an 11-fold increase in soybean (Karunanandaa et al., 2005). Besides, modulation of the activity of TC and MT also altered the α-/γ- tocopherol ratio in plants (Van Eenennaam et al., 2003; Kumar et al., 2005). Unfortunately, similar approaches do not bring about expected effects in maize and potato. For example, overexpression of Arabidopsis HPPD and MT results in an only 3-fold increase in γ-tocopherol in maize kernels, and other tocopherol isomers are undetectable (Naqvi et al., 2011). Similarly, constitutive overexpression of Arabidopsis HPPD or HPT does not affect the tocopherol composition in potato tubers (Crowell et al., 2008).

Genome-wide association study is also used to identify the natural allelic variations controlling vitamin E. Two insertion/deletions within VTE4, and a single nucleotide polymorphism (SNP) located 85 kb upstream of this gene are found to be significantly associated with α-tocopherol contents in maize kernels (Li Q. et al., 2012), and three genes, VTE1, HGGT1 and a prephenate dehydratase paralog, are also found to modestly contribute to tocotrienol variations in maize (Lipka et al., 2013). In tomato, a short interspersed nuclear elements (SINE) retrotransposon located in the promoter region of VTE3 is identified to be responsible for vitamin E accumulation in fruits (Quadrana et al., 2014). The above findings suggest the existence of as-yet-unknown candidate genes for use in vitamin E biofortification.