Antioxidants are protective molecules that neutralize reactive oxygen species and prevent oxidative damage to cellular components such as membranes, proteins and nucleic acids, therefore reducing the rate of cell death and hence the effects of ageing and ageing-related diseases. The fortification of food with antioxidants represents an overlap between two diverse environments, namely fortification of staple foods with essential nutrients that happen to have antioxidant properties (e.g. vitamins C and E) and the fortification of luxury foods with health-promoting but non-essential antioxidants such as flavonoids as part of the nutraceuticals/functional foods industry. Although processed foods can be artificially fortified with vitamins, minerals and nutraceuticals, a more sustainable approach is to introduce the traits for such health-promoting compounds at source, an approach known as biofortification. Regardless of the target compound, the same challenges arise when considering the biofortification of plants with antioxidants, that is the need to modulate endogenous metabolic pathways to increase the production of specific antioxidants without affecting plant growth and development and without collateral effects on other metabolic pathways. These challenges become even more intricate as we move from the engineering of individual pathways to several pathways simultaneously. In this review, we consider the state of the art in antioxidant biofortification and discuss the challenges that remain to be overcome in the development of nutritionally complete and health-promoting functional foods.
Fortification is the process of adding essential micronutrients and other health-promoting compounds to foods. The fortification of processed foods such as flour, bread, packaged cereals, dairy products and salt is a public health policy in the industrialized world, aiming to reduce the number of people suffering from malnutrition and to increase general health and wellbeing in the population. In developing countries, fortification programmes are often unsustainable due to poor governance, inefficient food-distribution networks and the prevalence of subsistence agriculture in rural populations, which means that most agricultural products are not processed centrally before distribution and consumption. Approximately 50% of the global population is thought to be malnourished but the vast majority of malnourished people are the rural poor in developing countries, subsisting on a diet of milled cereal grains lacking many essential nutrients and other health-promoting compounds (Farré et al., 2011a,b; Yuan et al., 2011). In these settings, biofortification is a more sustainable strategy because this involves the fortification of crops at source either through the application of nutrient-rich fertilizers or the breeding or engineering of crops to synthesize and/or accumulate nutritionally important compounds, therefore avoiding the need to fortify processed food products (Gómez-Galera et al., 2010).
Biofortification programmes generally focus on essential micronutrients, which are either organic compounds (vitamins) or minerals required in amounts <1 mg/day. These compounds act as cofactors or metabolic precursors and are required for specific biological processes, such that insufficient intake results in characteristic deficiency diseases (Zhu et al., 2007; Gómez-Galera et al., 2010; Table S1). The major deficiency diseases in developing countries correspond to essential nutrients that tend to be present at low levels in milled cereal grains, for example vitamin A, iron, iodine, zinc, vitamin C and folic acid. As well as their requirement for particular metabolic processes, certain essential nutrients also act as antioxidants or promote the activity or availability of antioxidants, which help to prevent diseases that result from or that are exacerbated by the accumulation of oxidative damage to cells, including cancer, cardiovascular disease and neurodegenerative disorders. Many non-essential molecules consumed in the diet are also antioxidants with health-promoting effects, and hence there is an overlap between essential nutrients and non-essential compounds (sometimes described as nutraceuticals) that act as antioxidants.
A key example of such a ‘dual-purpose nutrient’ is vitamin A, which is obtained in the diet either as esters of retinol from meat and dairy products or as pro-vitamin A carotenoids such as β–carotene from plants. Vitamin A is converted into the visual pigment rhodopsin (retinal), in the retina of the eye, and acts as a co-regulator of gene expression (retinoic acid); β-carotene is also an antioxidant, as are many other (non-essential) carotenoids. Similarly vitamin C (ascorbate) is an essential cofactor for several enzymes and vitamin E (tocochromanol family) is a regulator of protein kinase activity and gene expression, but their potent antioxidant activities are arguably just as important as their essential and non-replaceable functions. Many non-essential compounds are known to have health-promoting antioxidant activities, including anthocyanins, resveratrol and flavonoids. Even metal ions, which are usually regarded as pro-oxidants, can be important to maintain antioxidant activity in humans, because they act as cofactors for certain antioxidant enzymes, for example iron as a cofactor for catalase. In the industrialized world, the fortification of processed foods with essential nutrients is taken for granted and additional fortification with non-essential but health promoting antioxidant compounds forms a large part of the luxury functional foods market (Espín et al., 2007).
Several recent reviews have focused on the fortification of staple foods with essential nutrients, predominantly as a strategy to alleviate micronutrient deficiency in developing country settings (Bai et al., 2011; Gómez-Galera et al., 2010; Yuan et al., 2011; Zhu et al., 2007). Here we discuss the development of strategies for the biofortification of plants with antioxidants or molecules that promote antioxidant activity, including essential nutrients and nutraceuticals.
Antioxidants confer health-promoting effects through a variety of mechanisms
Antioxidants inhibit the oxidation of other molecules and thereby prevent them from causing oxidative damage, which is a major contributory factor to diseases associated with ageing and with ageing itself. Antioxidant molecules are generally either lipophilic or hydrophilic, but both types employ common molecular mechanisms, including hydrogen atom transfer (HAT), single-electron transfer followed by proton transfer (SET-PT), sequential proton loss electron transfer (SPLET) and the formation of radical adducts (Ghanta and Chattopadhyay, 2011; Gülçin, 2012). These mechanisms are summarized in Figure 1. Some antioxidants have a single active mechanism, for example ascorbate/vitamin C uses HAT alone, whereas others employ multiple mechanisms, for example flavonoids can use HAT, SET-PT and SPLET depending on their molecular structure (Gülçin, 2012).
The three major lipophilic antioxidant classes in mammals are carotenoids, tocochromanols and coenzyme Q10, all of which are derived from terpenoids. Carotenoids and tocochromanols are obtained from the diet whereas coenzyme Q10 is synthesized de novo in a multistep pathway starting with acetyl-CoA. All three classes act primarily by scavenging lipid peroxyl radicals (ROO˙) and by disrupting free radical chain reactions in membranes. Carotenoids are tetraterpenoids (C40 isoprenoids) produced mainly by photosynthetic organisms and some fungi, bacteria, algae and archaea, and they can be classed either as carotenes or xanthophylls, the latter distinguished by the presence of oxygen atom(s) (Fraser and Bramley, 2004). The best known carotenoids are those with pro-vitamin A activity (particularly β-carotene) because these represent a major dietary source of vitamin A, but others with important antioxidant effects include lycopene, lutein and zeaxanthin. Lycopene is the red pigment present at high levels in tomatoes and has been shown to reduce the risk of cardiovascular disease and some types of cancer (Fraser and Bramley, 2004). Lutein is very abundant in green tissue, but zeaxanthin is restricted to a few sources like pepper and a number of corn varieties. The latter are not particularly abundant in western diets. They are particularly important in the macula and lens of the human eye, where they filter high-energy blue light and counteract oxidative damage to protect against age-related macular degeneration (Fraser and Bramley, 2004). The tocochromanols are a group of eight structurally related compounds collectively known as vitamin E obtained primarily from green leaves and oilseeds. The most potent form in humans is α-tocopherol, not because it is intrinsically any more active than the others but because it is absorbed most efficiently (Brigelius-Flohe and Traber, 1999). Coenzyme Q10 is present in the mitochondria where it functions in the electron transport chain and helps to prevent oxidative damage to membranes. Higher levels of coenzyme Q10 may help to prevent cardiovascular and neurological disorders as well as certain cancers and is used as an orphan drug for the treatment of congestive heart failure in children (Dhanasekaran and Ren, 2005).
There are many water-soluble antioxidants in mammalian diets, ranging from simple compounds such as ascorbate to complex molecules such as the flavonoids. The tripeptide glutathione has an antioxidant activity which is conferred by the sulfhydryl group of cysteine (Rennenberg, 1980). Lipoic acid uses a thiol functional group to neutralize free radicals (Goraca et al., 2011). As well as scavenging free radicals, hydrophilic antioxidants may also chelate free metal ions that promote oxidation reactions. In plants, flavonoids, phenolic acids and glutathione are responsible for this activity, whereas uric acid and lipoic acid perform the same function in animals (Goraca et al., 2011; Gülçin, 2012). However, the consumption of antioxidant phytochemicals in the diet increases the ability of mammals to counteract oxidative stress, reducing the frequency of cell death promoted by oxidative damage and therefore providing benefits such as a healthy immune system, a lower risk of disease and increased longevity (Ghanta and Chattopadhyay, 2011).
Biofortification increases the availability of essential nutrients and health-promoting compounds at source
Three staple crops—rice, maize and wheat—provide 60% of the calories consumed by humans (FAO, 2010). Milled cereals are deficient in many nutrients, including essential amino acids (particularly lysine), essential fatty acids, vitamins and minerals. Human populations that subsist entirely or predominantly on milled cereals are therefore the most at risk of deficiency diseases, and these tend to be the rural poor in developing countries, who are also the least likely to benefit from organized programmes to distribute fortified processed foods or vitamin and mineral supplements (Gómez-Galera et al., 2010; Underwood and Smitasiri, 1999).
The limited impact of conventional interventions in developing country settings has promoted the use of biofortification as a sustainable approach that is equally beneficial to subsistence farmers and consumers of processed foods (Zhu et al., 2007). Biofortification means that nutrients and other health-promoting compounds are incorporated while the plant is still growing, and are therefore present in the harvested material and at all subsequent stages en route to the consumer. Mineral biofortification can be achieved in limited cases by using mineral-rich fertilizers although the impact of this approach depends on the mobility of mineral ions in the soil and the ability of plants to concentrate them in edible tissues such as cereal grains. There are two general strategies that are suitable for biofortification with any organic or inorganic nutrient, that is conventional breeding and genetic engineering. These are similar in aim, albeit different in scope. Both attempt to create plant lines carrying genes that favour the most efficient biosynthesis and/or accumulation of essential micronutrients and other health-promoting compounds. Conventional breeding achieves this by crossing the best performing plants and selecting those with favourable traits over many generations, sometimes in combination with biotechnology tools such as mutagenesis or marker assisted selection, whereas genetic engineering introduces the traits as recombinant DNA and allows the best-performing plants to be selected in a single generation (Harjes et al., 2008; White and Broadley, 2005). Conventional breeding is limited to genes that can be sourced from sexually compatible plants and requires lengthy breeding programmes to introduce traits into locally adapted elite varieties, whereas genetic engineering has no such limitations and novel genes can be introduced directly into local cultivars. First-generation engineered crops with modified input traits have already shown their potential to enhance agricultural productivity and reduce poverty in developing countries (Christou and Twyman, 2004; Farré et al., 2010a, 2011a), and now second-generation crops that address nutritional requirements through the direct modification of output traits are under development (Pérez-Massot et al., 2012; Yuan et al., 2011). Genetic engineering also allows nutritional traits to be targeted to specific organs (e.g. cereal seeds) and multiple traits can be combined in the same plants without complex breeding programmes (Naqvi et al., 2009, 2010).
Organic antioxidants, such as carotenoids, tocochromanols, ascorbic acid and flavonoids, are synthesized de novo by plants, so engineering strategies aiming to enhance the availability of such compounds involve the modification of endogenous plant metabolism (Capell and Christou, 2004). This can be achieved using a variety of strategies as outlined in Figure 2. The most popular strategy has been to overexpress a known rate-limiting enzyme thus alleviating a metabolic bottleneck, preferably using an enzyme devoid of feedback inhibition (Shewmaker et al., 1999; Ye et al., 2000; Zhu et al., 2008). The first committed step in a metabolic pathway is often a suitable intervention point because this ensures flux is delivered to all downstream steps (Enfissi et al., 2005; Morris et al., 2006). However, it is also possible to overexpress multiple enzymes to ensure there is adequate flux throughout the entire pathway (Ravanello et al., 2003; Zhu et al., 2008), or express regulatory proteins to coordinately induce an entire endogenous pathway without the introduction of heterologous enzymes (Butelli et al., 2008; de Vos et al., 2000). Alternative strategies to achieve the accumulation of a specific desired metabolite include the suppression of a competitive pathway or branch point to ensure flux is directed in the appropriate direction (Diretto et al., 2006; Yu et al., 2008) or the creation/enlargement of a metabolic sink, which reduces feedback inhibition and allows the desired product to accumulate in a stable manner (Lopez et al., 2008; Lu et al., 2006).
Different approaches must be used to achieve the accumulation of minerals because these cannot be synthesized de novo and must instead be taken up from the environment (Gómez-Galera et al., 2010; Zhu et al., 2007). Strategies to enhance mineral accumulation include the introduction of genes that improve (i) the efficiency of mineral uptake from the soil, (ii) the efficiency of transport form the roots to storage organs, (iii) the storage capacity and (iv) the bioavailability of the stored mineral (Gómez-Galera et al., 2010). These strategies are summarized in Figure 3.
It is important to emphasize the difference between bioaccumulation (the amount of a particular nutrient that can be stored in plant tissues) and bioavailability (the amount that can be absorbed when the plant tissue is consumed as food). Whereas most studies have focused on bioaccumulation, the bioavailability of nutrients in engineered crops is a more important indicator of its nutritional quality (Hirschi, 2008). The food matrix plays an important role in the bioavailability of organic and inorganic compounds. For example, 12 mg of β–carotene in a food matrix must be ingested to gain the same benefit as 1 mg of pure β-carotene dissolved in oil. Similarly, vitamin E absorption requires the presence of bile salts, pancreatic enzymes and oils or fats to promote solubility (Jeanes et al., 2004), and the bioavailability of ascorbate is enhanced by copresentation with proteins in the food matrix (Vinson and Bose, 1988). In the case of minerals, the presence of antinutritional compounds such as phytate and oxalate in vegetables can inhibit mineral absorption because they act as chelating agents (Gómez-Galera et al., 2010), whereas nutritional enhancers such as inulin can promote mineral absorption by slowing down the movement of food through the gut (Gómez-Galera et al., 2010). Reducing the quantities of antinutritional compounds and/or increasing the quantities of nutritional enhancers can therefore increase the bioavailability of nutrients. Bioavailability may also depend on the chemical form in which a nutrient is presented, for example selenium is absorbed more efficiently when presented in an organic form such as selenomethionine rather than as inorganic metal ions (Combs, 2001), and iron presented as a complex with ferritin is less susceptible to the effects of antinutritional compounds than non-heme iron (Lönnerdal, 2009).
Transgenic flavonoid-enriched tomato intake reduced C-reactive protein in human C-reactive protein transgenic mice expressing markers of cardiovascular risk more than in wild-type tomato (Rein et al., 2006). Moreover, the life span of cancer-susceptible mice fed on a diet supplemented with high-anthocyanin tomatoes was increased substantially (Butelli et al., 2008). Recently, it has been demonstrated that β-carotene in biofortified rice (Golden Rice) and maize has good bioavailability as a plant source of vitamin A in humans (Li et al., 2010; Muzhingi et al., 2011; Tang et al., 2009). Toxicity assessment in mice fed with multivitamin maize showed no adverse health effects and did not induce any clinical sign of toxicity (Arjó et al., 2012).
Genetic engineering for biofortification with lipophilic antioxidants
The carotenoid biosynthesis pathway in plants (and equivalent bacterial enzymes) is shown in Figure 4a. The key aspects of the pathway in the context of metabolic engineering are that the first committed step is catalysed by the enzyme phytoene synthase (PSY/CRTB) and that the next four desaturation and isomerization steps which produce all-trans lycopene are carried out by four enzymes in plants but by a single bacterial enzyme known as CRTI (phytoene desaturase), which is therefore preferred for genetic engineering to reduce the number of transgenes required. Lycopene represents a branch point in the pathway, where the competing activities of beta and epsilon cyclases (LYCB/CRTY and LYCE) result in the production of β-carotene (pro-vitamin A) and α-carotene. These products are then converted into lutein and zeaxanthin, respectively, by the activity of carotene hydroxylases.
As discussed above, there are multiple strategies available to enhance carotenoid production and one popular approach has been the overexpression of PSY/CRTB to alleviate the metabolic bottleneck at the first committed step in the pathway. One successful example of this approach in canola resulted in a 50-fold increase in carotenoid levels in the seeds (to 1617 μg/g fresh weight, FW), predominantly represented by α-carotene and β-carotene (394 and 949 μg/g FW, respectively), and was achieved by expressing the CRTB gene under the control of a seed-specific promoter (Shewmaker et al., 1999). The engineered canola was the pioneering work for the successful enhancement of carotenoid production in crop plants. The success of this approach relied on the fact that enzymes acting downstream of PSY were not limiting in canola seeds and the phytoene was successfully converted into downstream carotenoids.
However, the expression of daffodil PSY in rice endosperm only resulted in the accumulation of phytoene because the subsequent desaturation steps were also limiting (Burkhardt et al., 1997; Schaub et al., 2005). This is in contrast to canola which contains a native carotenogenic pathway. Therefore, to boost carotenoid levels in rice grains, it was necessary to express multiple enzymes from the pathway. Initially, daffodil PSY was combined with the multifunctional bacterial enzyme CRTI resulting in the accumulation of β-carotene and xanthophylls in ‘Golden Rice’ (Ye et al., 2000). These results indicated that the endogenous levels of lycopene cyclases and carotene hydroxylases were sufficient to convert the lycopene generated by CRTI into downstream products (Schaub et al., 2005). Later, the amount of β-carotene was increased to 31 μg/g dry weight (DW), a 17-fold improvement, by replacing the daffodil PSY with its more efficient corn ortholog, resulting in the higher performance Golden Rice II (Paine et al., 2005). Tissue-specific transgene expression is critical for successful carotenoid modulation. For example, the first report of tomato engineered to expressed tomato PSY1 constitutively resulted in a dwarft phenotype due to the depletion of the endogenous precursor pool of geranylgeranyl diphosphate (GGPP) leading to a shortage of gibberellins (Fray et al., 1995). This was overcome subsequently by fruit-specific expression of bacterial CRTB (Fraser et al., 2002). Flux through the carotenoid pathway can also be increased by making more precursors available by engineering upstream pathways [e.g. by overexpressing 1-deoxy-d-xylulose 5-phosphate (DXP) synthase (DXS) to provide more DXP in the methylerythritol phosphate (MEP) pathway; Enfissi et al., 2005] but this also increases flux through unrelated pathways that utilize the same precursors, so is less efficient and targeted.
Whereas the results described above appear to reflect the additive effects of transgenic and endogenous metabolic capabilities, more complex interactions have been observed when feedback effects occur between the superimposed pathways. The expression of CRTI in tomato was envisaged to enhance the lycopene content but actually resulted in a threefold increase in β-carotene levels while reducing the overall level of carotenoids including lycopene (Römer et al., 2000). The elevated β-carotene content was unexpected because of the low LYCB levels normally found in tomato fruits, but further investigation revealed that the endogenous LYCB gene had been induced in the transgenic plants (Römer et al., 2000). The outcome of the above strategy therefore depends on the relative activities of endogenous LYCB and LYCE which vary in different plants and even in different cultivars, so more predictable results can be achieved by deliberately expressing one or other of the enzymes to increase flux in the appropriate direction beyond the branch point. As an example, the overexpression of LYCB in tomato fruits increased the accumulation of β-carotene and zeaxanthin at the expense of α-carotene and lutein (D'Ambrosio et al., 2004; Rosati et al., 2000). Similarly, transgenic canola seeds expressing CRTB, CRTI and CRTY accumulated more carotenoids than wild-type seeds and the β:α-carotene ratio increased from 2:1 to 3:1, showing that the additional lycopene β-cyclase (CRTY) activity skewed the competition for the common precursor lycopene and increased flux specifically towards β-carotene (Ravanello et al., 2003). In corn, the outcome of such experiments depends on competition and complementarity between the transgenic and endogenous pathways. For example, in the white inbred M37W which lacks significant amounts of carotenoids in the endosperm, the expression of PSY, CRTI and LYCB increased the β:α–carotene ratio from 1.21 to 3.51 (Zhu et al., 2008) and zeaxanthin levels were 29.64 μg/g DW, but also enhanced flux through the competing branch, resulting in nearly 25-fold the normal levels of lutein (up to 10.76 μg/g DW) showing that LYCE activity is not a limiting step in corn endosperm. In contrast, the introgression of the same pathway into two yellow-endosperm varieties with opposing β:α–carotene ratios (0.61 for EP42 and 1.90 for A632) generated one hybrid accumulating lutein (23.41 μg/g DW) with a β:α–carotene ratio of 2.05 and the other accumulating high levels of zeaxanthin (56.9 μg/g DW) with a β:α–carotene ratio of 6.8 (Naqvi et al., 2011a).
Carotenoid profiles can also be modulated by inhibiting endogenous carotenogenic enzymes, for example by a tuber-specific antisense approach against LYCE in potato which increased β-carotene levels by up to 14-fold and total carotenoid levels by up to 2.5-fold. However, there was no corresponding reduction in lutein levels again suggesting that LYCE is not a rate-limiting step (Diretto et al., 2006). Similar results were observed in canola seeds by RNA interference against LYCE (Yu et al., 2008). The inhibition of β-carotene hydroxylase (BCH) prevents β-carotene from being converted into zeaxanthin, and this strategy has achieved 38-fold increase in β-carotene levels in potatoes, in concert with a 3.7-fold increase in lutein levels and a 0.5 fold reduction in zeaxanthin levels (Diretto et al., 2007). Targeting the next step in the pathway (zeaxanthin epoxidase, ZEP) prevents zeaxanthin from being converted into downstream products, and this strategy increased total carotenoid levels 5.7-fold, β-carotene levels 3.4-fold, lutein levels 1.9-fold and zeaxanthin levels 133-fold (Römer et al., 2002). Vitamin E (α–tocopherol) was increased two to threefold in the transgenic potatoes. Fine-tuning these alterations has the potential to significantly enhance the nutritional value of potatoes. These studies show that carotenoid levels in plants can be increased both by alleviating early bottlenecks and introducing bottlenecks at later steps or in competitive branches.
A final strategy that has been used to enhance β-carotene and total carotenoid levels in plants is to increase the number of storage compartments (Lopez et al., 2008; Lu et al., 2006). Carotenoids are stored in adapted plastids, so increasing the number of these organelles or encouraging their differentiation can act as a metabolic sink, shifting the metabolic equilibrium towards carotenoid synthesis. The cauliflower Or gene was identified through the discovery of a dominant allele that causes the curd to become orange, and the overexpression of this allele in potato under the control of the granule-bound starch synthase (GBSS) promoter produced potato tubers with bright orange flesh and carotenoid levels of up to 31 μg/g DW, including 3.75 μg/g DW β-carotene (Lopez et al., 2008).
In some of the examples discussed above, the introduction of heterologous enzymes has enhanced the production of carotenoids that are already produced in moderate to large amounts or has modified the carotenoid profile of edible plant tissues, whereas in other cases the endogenous tissues do not produce carotenoids and the entire pathway must be introduced de novo (e.g. Golden Rice). Similarly, most plants do not produce ketocarotenoids such as astaxanthin and canthaxanthin, which are the pink/red pigments most frequently found in seafood and have potent, health-promoting antioxidant activities (Zhu et al., 2009). Ketocarotenoids are currently produced in bacteria and are added artificially to fish feed pellets to increase the pigmentation of for example farmed salmon and trout. The expression of bacterial β-carotene ketolases in corn endosperm has resulted in the production of astaxanthin corn lines that could be valuable sources of ketocarotenoids in the aquaculture, nutraceutical and cosmetic industries (Zhu et al., 2008).
The tocochromanol biosynthesis pathway in plants is shown in Figure 4b. The key aspects of this pathway in the context of metabolic engineering are that it begins with the prenylation of homogentisic acid (HGA), derived from the shikimate pathway, with phytyldiphosphate (PDP), derived from the MEP pathway, and that prenylation may be carried out by two different enzymes—homogentisate phytyltransferase (HPT) or homogentisate geranylgeranyl transferase (HGGT)—to generate alternative intermediates that give rise to the tocopherol and tocotrienol branches of the pathway, respectively. These intermediates are substrates for the same three enzymes, yielding eight different products, so genetic engineering can be used not only to increase the total tocochromanol content, but also to alter the relative levels of the different forms. This provides a strategy to boost levels of the most potent form, α-tocopherol (Farré et al., 2012a).
As with carotenoid engineering, total tocochromanol levels in plants can be increased by expressing single or multiple rate-limiting enzymes. HGA is the key target because it is the last common precursor before the pathway splits into multiple branches, and HGA levels can be boosted by expressing enzymes in the shikimate pathway such as TyrA and/or HPPD which increase total tocochromanol levels without significantly affecting the balance between different forms (Karunananda et al., 2005). In contrast, the expression of downstream enzymes such as MPBQ-MT and γ-tocopherol methyltransferase (γ-TMT) has the effect of interconverting different forms of vitamin E without a significant impact on total tocochromanol levels. For example, the expression of MPBQ-MT (VTE3) in Arabidopsis seeds increased the total tocopherol content marginally but caused the preferential accumulation of γ-tocopherol (75%–85% of total tocopherols; Van Eenennaam et al., 2003). Similarly, the expression of γ-TMT had no impact on total tocochromanol levels but resulted in 85%–95% of the total tocopherol pool being converted into α–tocopherol, representing an 80-fold increase over wild-type seeds and a ninefold increase in total vitamin E activity indicating that flux was diverted into the α-branch of the pathway (Shintani and DellaPenna, 1998). Transgenic soybeans expressing MPBQ-MT and γ-TMT accumulated >95% α-tocopherol, resulting in a greater than eightfold increase of α-tocopherol and an up to fivefold increase in seed vitamin E activity (Van Eenennaam et al., 2003). The result was a fivefold increase in vitamin E activity.
Similar increases in vitamin E activity were achieved following the overexpression of γ-TMT in lettuce (Cho et al., 2005), but the most promising results were achieved by crossing transgenic lettuce lines expressing HPT (thus a higher flux through the entire pathway) with those expressing γ-TMT (skewing the pathway in favour of α-tocopherol). This resulted in an increase of both the total tocopherol content and the α/γ tocopherol ratio (Cho et al., 2005). More recently, lettuce plants simultaneously transformed with the Arabidopsis genes encoding HPT and γ-TMT also showed both quantitative and qualitative increases in vitamin E activity, a sixfold increase in the total tocopherol content and also a sixfold increase in the α/γ ratio (Li et al., 2011).
The overexpression of HGGT should favour the tocotrienol branch of the pathway, but corn seeds overexpressing HGGT accumulated six times the normal levels of both tocotrienols and tocopherols (Cahoon et al., 2003). Recently, rice seeds expressing Arabidopsis HPPD resulted in a small increase in absolute tocotrienol synthesis (but no change in the relative abundance of the γ and α isoforms). In contrast, there was no change in the absolute tocopherol level, but a significant shift from the α to the γ isoform. These data confirm that HPPD is not rate limiting, and that increasing flux through the early pathway reveals downstream bottlenecks that act as metabolic tipping points (Farré et al., 2012b). The combined expression of HPPD and MPBQ-MT resulted in a threefold increase in γ-tocopherol levels without changing the total tocopherol content (Naqvi et al., 2011b). These experiments again showed that flux was directed into the α-branch but in this case it was blocked by low γ–TMT activity, forcing the accumulation of γ–tocopherol (Naqvi et al., 2011b).
The metabolic engineering of fat-soluble antioxidants has understandably focused on carotenoids and tocochromanols because of their status as essential nutrients. However, the beneficial properties of coenzyme Q10 have prompted investigations into good dietary sources of this compound and the development of strategies to increase its abundance by genetic engineering, potentially to provide additional sources for the nutraceutical industry. The most significant studies involved the expression of the Gluconobacter suboxydans decaprenyl diphosphate synthase gene in rice, with the corresponding enzyme targeted for import into the mitochondria which is the normal site of coenzyme Q10 activity in humans and the site of coenzyme Q9 synthesis in wild-type rice (Takahashi et al., 2006, 2009, 2010). This approach was chosen because it allowed the production of coenzyme Q10 through the extension of an existing metabolic pathway using a single enzyme, and because coenzyme Q10 is preferentially localized in the bran and germ suggesting that the ingestion of coenzyme Q10 could be increased by the consumption of unmilled brown rice as a health food (Takahashi et al., 2010).
Genetic engineering for fortification with hydrophilic antioxidants
Unlike carotenoids and tocochromanols, which are synthesized de novo in plants through a unique pathway, a key aspect of ascorbate biosynthesis in the context of metabolic engineering is that it may be synthesized de novo via several pathways or recycled from oxidation products, offering multiple intervention points for nutritional enhancement (Ishikawa et al., 2006). An overview of ascorbate synthesis and recycling is shown in Figure 4c. The major de novo synthesis pathway involves the intermediate l–galactose (Wheeler et al., 1998) but alternative routes via l-gulose (Wolucka and Van Montagu, 2003), d–galacturonic acid (Agius et al., 2003) and myo-inositol (Lorence et al., 2004) have been identified in particular species and tissues.
As discussed earlier, one of the key intervention methods is to overexpress a rate-limiting enzyme, and in the l-galactose pathway this has been achieved by targeting GDP-l-galactose phosphorylase (GGP), resulting in increases of between two and sixfold in crops such as tomato, potato and strawberry (Bulley et al., 2012). The first committed step in the l–galactose pathway is catalysed by GDP-mannose-3′,5′-epimerase, and transgenic tomato plants expressing two isoforms of this enzyme showed a modest 1.6-fold increase in ascorbate levels in ripe tomato fruits (Zhang et al., 2011). Success has also been achieved by targeting the alternative biosynthesis pathways, for example the rate-limiting l-gulonolactone oxidase in the myo-inositol pathway which resulted in a sevenfold increase in ascorbate in lettuce, although only a 1.4-fold increase in potato (Hemavathi et al., 2010; Jain and Nessler, 2000).
The ascorbate recycling pathway has been utilized for metabolic engineering in corn (Chen et al., 2003; Naqvi et al., 2009), potato (Goo et al., 2008) and tomato (Haroldsen et al., 2011) by overexpressing the key enzyme dehydroascorbate reductase (DHAR). Ascorbate levels in corn endosperm increased twofold following the constitutive overexpression of wheat DHAR (Chen et al., 2003) but sixfold following endosperm-specific expression of rice DHAR (Naqvi et al., 2009). A 1.6-fold increase in ascorbate was achieved in ripe tomato fruits constitutively expressing tomato DHAR (Haroldsen et al., 2011). Potato tubers constitutively expressing a sesame DHAR enzyme showed a 1.6-fold increase in ascorbate, while the use of a tuber-specific promoter resulted in a marginally lower 1.3-fold increase (Goo et al., 2008). It therefore appears that ascorbate metabolic engineering is exquisitely sensitive to the source of the heterologous enzyme and its spatiotemporal expression profile in the transgenic plants, which is likely to reflect the complex regulatory systems to control ascorbate accumulation.
Flavonoids, a large structurally diverse class of more than 9000 polyphenolic compounds, are synthesized via the phenylpropanoid pathway by plants and microbes. They include chalcones, flavones, flavonols, flavanones, anthocyanins and isoflavonoids (Wang et al., 2011). Although the phenylpropanoid pathway is complex and incompletely understood, key rate-limiting enzymes have been identified allowing the use of metabolic engineering to increase the overall flavonoid content, and the levels of particular subclasses (Wang et al., 2011).
Flavonoid metabolic engineering has highlighted two strategies for the production of antioxidant compounds. The first is the introduction of different structural genes from diverse plants to create a recombinant biosynthesis pathway. For example, genes encoding stilbene synthase, chalcone synthase, chalcone reductase, chalcone isomerase and flavone synthase from different sources were combined in transgenic tomatoes, resulting in the accumulation of stilbenes (resveratrol and piceid), deoxychalcones (butein and isoliquiritigenin), flavones (luteolin-7-glucoside and luteolin aglycon) and flavonols (quercetin glycosides and kaempferol glycosides) in the fruit peel (Schijlen et al., 2006). The total antioxidant capacity of the transgenic fruit peel was more than three times higher than wild-type fruits. These experiments show that genetic engineering can not only increase the levels of potentially health-promoting antioxidant compounds in fruits, but can also provide insight into the underlying metabolic pathways and allow the selection of particular flavonoid compounds.
The second approach involves the combinatorial use of different regulatory factors to increase flavonoid levels. For example, tomato fruits produce only low amounts of flavonols such as kaempferol and quercetin, which are concentrated in the fruit peel. However, the expression of corn regulatory genes Lc and C1 induced the production of kaempferol in the fruit flesh, increasing overall levels by 60% (de Vos et al., 2000). The additional expression of chalcone isomerase increased the flavonol concentration in the fruit peel by 78-fold, mainly due to the accumulation of rutin (Muir et al., 2001). Corn regulatory genes (this time C1 and R) have also been expressed in soybean, resulting in a twofold increase in isoflavonoids levels. The expression of these genes in conjunction with cosuppression of flavanone 3 hydroxylase (F3H) to block the anthocyanin and flavonol pathways resulted in a fourfold increase in the isoflavone content (Yu et al., 2003). Finally, tomato plants expressing two snapdragon-derived transcription factors accumulated anthocyanins at levels comparable to those found in blackberries and blueberries, which are among the best sources of dietary anthocyanins although they are more expensive and are consumed in smaller quantities than tomatoes (Butelli et al., 2008). The hydrophilic antioxidant capacity of the tomato fruits was threefold higher than wild type fruits, and the peel and flesh were both intense purple in coloration.
Melatonin is known as a terminal antioxidant because it forms stable oxidation products that cannot be recycled. It is synthesized de novo in humans and has a major role in the regulation of circadian rhythms, but it is also produced by plants where it fulfils an analogous role in the regulation of photoperiod responses as well as acting as an antioxidant. Melatonin is used as a drug to treat certain circadian disorders such as delayed sleep phase syndrome and seasonal affective disorder, but its antioxidant properties are also relevant as it has been indicated for protection against radiation and can also reduce the risk of certain forms of cancer. Melatonin is also sold as a dietary supplement in the USA.
The biosynthesis of melatonin in plants is incompletely understood, but several enzymes in the pathway have been expressed in transgenic plants including arylalkylamine N-acetyltransferase (AANAT), N-acetylserotonin O-methyltransferase (ASMT) and 2,3-dioxygenase (IDO) (Okazaki et al., 2009, 2010). For example, an algal AANAT enzyme has been expressed in Micro-Tom tomato (Okazaki et al., 2009) and human AANAT has been expressed in rice (Kang et al., 2010), in both cases leading to elevated melatonin levels in the transgenic plants. In contrast, melatonin levels decreased in transgenic rice plants expressing IDO, potentially reflecting the induction of a feedback mechanism (Okazaki et al., 2010). Because melatonin is a natural bioregulator in plants, its synthesis can also be induced by the application of exogenous chemicals such as benzothiadiazole (BTH) and chitosan (CHT) (Vitalini et al., 2011). Interestingly, it has been shown that plants with higher levels of melatonin have a normal level of β-carotenoids, but double the normal amount of vitamin E, and significantly higher quantities of vitamin C and glutathione, suggesting that part of the defense pathway induced by melatonin includes the induction of additional antioxidants (Wang et al., 2012).
Antioxidant activity can depend on the abundance of metal ions
A number of metals are required by humans in small amounts because they act as cofactors for enzymes, transcription factors and signalling proteins. This includes several enzymes whose function is to detoxify reactive oxygen species, which means metal ions are required as cofactors for antioxidant enzymes, including different forms of superoxide dismutase (requiring Cu, Mn and/or Zn; Johnson and Giulivi, 2005), catalase (requiring Fe) and glutathione peroxidase (requiring Se) (Mates et al., 1999). This presents an interesting quandary because metal ions are also responsible for oxidation reactions and the body has numerous mechanisms to sequester and/or compartmentalize metals to ensure they do not cause oxidative damage. A key example is iron, which generates hydroxyl radicals through the Fenton reaction, a process which is facilitated by metal ion reduction mediated by ascorbate. Iron is therefore both necessary and potentially toxic, which means it must be stored and transported in a bioavailable but inactive form using proteins such as transferrin and ferritin (Crichton and Charloteaux-Wauters, 1987; Theil, 1987).
The biofortification of crops with metals is nutritionally relevant because iron and zinc deficiencies are prevalent in developing countries, but selenium is also an important target for biofortification particularly in regions with selenium-depleted soils. The strategies used for mineral biofortification depend on the mobility of the metal in the soil and its metabolism by the plant. For example, iron tends to be immobilized by interactions with soil particles (particularly in alkaline soils) so biofortification strategies must increase its mobility as well as its transport and storage (Lee et al., 2012), whereas zinc is highly mobile in the soil and biofortification strategies focus on transport and storage alone (Palmgren et al., 2008; Stomph et al., 2009). Selenium is unusual because it can be incorporated into organic compounds in place of sulphur, such compounds being essential in humans but metabolic byproducts in plants.
Plants take up minerals from the soil in two different ways, one involving the direct absorption of mobile ions through specific transporters in the root plasma membrane, and the other involving the synthesis and secretion of chelating agents known as phytosiderophores which mobilize metals that have adsorbed onto soil particles, particularly iron. Genetic engineering can be used to enhance both mechanisms for example by overexpressing transporters to increase the uptake capacity of roots (Connolly et al., 2002; Grotz et al., 1998; Terry et al., 2000) and by expressing enzymes in the phytosiderophore biosynthesis pathway such as nitotianamine synthase (NAS; Johnson et al., 2011) or nicotianamine aminotransferase (NAAT; Takahashi et al., 2001) to increase the mobilization of metal ions in the soil. However, because such proteins are often promiscuous, these strategies often lead to the coaccumulation of iron and zinc. For example, the overexpression of rice NAS in transgenic rice plants under the control of a seed-specific promoter resulted in the accumulation of up to 19 μg/g DW of Fe and 52–76 μg/g DW of Zn in rice grains (Johnson et al., 2011). In cassava, the expression of the algal iron-transport protein FEA1 increased the level of Fe fourfold to 40 μg/g DW in transgenic tubers whereas the expression of Arabidopsis zinc transporters AtZAT1 and AtZIP1 increased the zinc content by up to fourfold and 10-fold, respectively (Sayre et al., 2011).
After absorption by the roots, metals are transported to sink tissues such as fruits or seeds through the phloem, so the overexpression of phloem plasma membrane transporters (YSL family transporters) can play an important role in biofortification (Ishimaru et al., 2010). Once they reach the sink tissues, metals must be stored in a bioavailable form if they are to be useful for nutritional purposes. This can be achieved by expressing metal storage proteins such as ferritin, which store both Fe and Zn (Drakakaki et al., 2005; Wirth et al., 2009) or enzymes such as selenocysteine methyltransferase (Ellis et al., 2004), which converts inorganic Se into an organic form. The overexpression of phytase, a fungal enzyme which degrades phytic acid, is a useful strategy to increase mineral bioavailability in rice (Wirth et al., 2009) and corn (Drakakaki et al., 2005) because this prevents phytic acid in the gut chelating metal ions and preventing absorption.
Conclusions and perspectives
Antioxidants are interesting targets for biofortification because they encompass essential nutrients for vulnerable populations in many developing countries as well as non-essential nutraceutical compounds which are presented as food additives or functional foods for the luxury markets in the West. Notwithstanding these differences, many of the same technical issues must be addressed in both environments, that is the need to modulate endogenous plant metabolic pathways to ensure that flux is directed to the appropriate compounds, the need to ensure such compounds accumulate in the most appropriate tissues and the focus on bioavailability rather than bioaccumulation. The examples discussed above provide a snapshot of the diverse approaches to antioxidant biofortification which are driven by the nature of the endogenous metabolic pathways. Where there is a single, well-characterized pathway (e.g. carotenoids and tocochromanols) it is often appropriate to target individual rate-limiting steps or key branch points, often through a combination of interventions to relieve early bottlenecks and reduce metabolic loss through downstream catabolism. In these cases, it is possible to affect the pathway either quantitatively (by targeting early, common steps) or qualitatively (by targeting branch points) to generate novel profiles of compounds. Where multiple pathways are present (e.g. ascorbate), there is also a tendency towards complex regulation that favours homeostasis. In these cases, it is often more productive to focus metabolic interventions in a localized manner (e.g. in seeds) to avoid feedback that can occur when transgenes are expressed constitutively. In contrast, the highly complex and multiple branched flavonoid pathway provides an example where targeting the regulatory genes is more beneficial.
In the future, it is likely that genetic engineering will be used to modify different pathways simultaneously, as recently demonstrated by the production of multivitamin corn with elevated levels of three vitamins (β-carotene, ascorbate and folic acid) but also other carotenoids, providing higher levels of several key antioxidants (Naqvi et al., 2009). As such ventures become more ambitious, it is likely that the complex interactions between pathways will become more apparent, meaning that interventions will need to become more refined and focused to avoid conflicts such as competition for substrates, precursors and intermediates (Zhu et al., 2007). It is therefore clear that as well as looking at the structural genes that represent each metabolic pathway, we also need to consider the regulatory factors, the interplay between metabolic flux and enzyme activity, the compartmentalization of different enzymes and the shuttling of intermediates, the interplay between endogenous and heterologous pathways in transgenic plants and the impact of metabolic interventions at the level of the metabolome.
Research at the Universitat de Lleida is supported by MICINN, Spain (BFU2007-61413; BIO2011-23324; BIO02011-22525; PIM2010PKB-00746); Acciones complementarias, BIO2007-30738-E, BIO2011-22525, MICINN, Spain; European Union Framework 7 Program-SmartCell Integrated Project 222716; European Union Framework 7 European Research Council IDEAS Advanced Grant (to PC) Program-BIOFORCE; COST Action FA0804: Molecular farming: plants as a production platform for high value proteins; Centre CONSOLIDER on Agrigenomics funded by MICINN, Spain.