Keywords

1 Introduction

l-Ascorbic acid (ascorbate, AsA) is the most abundant water-soluble antioxidant found in plants. Ascorbate content varies widely among species but in general AsA concentration in plants is in the mM range (Noctor and Foyer 1998). Ascorbate has a wide variety of physiological roles. It functions as an enzyme cofactor, as a modulator of the cell cycle, and as a scavenger protecting tissues against damage caused by reactive oxygen species (ROS) produced from a multitude of environmental stresses including water deficit, soil salinity, cold, and heat, among others. Alteration of the redox balance resulting in accumulation of ROS (e.g., singlet oxygen, superoxide, and hydrogen peroxide) is a feature common to multiple environmental stresses. When the intensity of those stresses exceeds the antioxidant and repair capacity of cells, ROS can accumulate and cause oxidative damage to all macromolecules, promoting apoptosis and senescence. On the other hand, there is growing evidence of the important role of low levels or ROS as signaling molecules in pathways that stimulate adaptive hormonal and metabolic responses. Therefore whether ROS induce oxidative damage or trigger acclimation depends on the rate and site of production and is controlled by the antioxidant system, of which ascorbate and glutathione are central components (Foyer et al. 1994; Foyer and Noctor 2011).

A better understanding about how ascorbate is made in plants and the factors that regulate its content is of chief importance because this molecule provides the dietary source of vitamin C to humans, primates, and other animals. Mutations in the terminal enzyme of the animal vitamin C pathway (l-gulono-1,4-lactone oxidase) are what render its synthesis impossible for humans and make this an essential vitamin that we must acquire through diet (Linster and Van Schaftingen 2007). Deficiency of ascorbic acid is often associated with human health problems such as anemia, infections, bleeding gums, scurvy, poor wound healing, capillary hemorrhage, muscle degradation, atherosclerotic plaques, and neurotic disturbances (Chambial et al. 2013). Vitamin C deficiency is a common issue for malnourished people in developing countries, and also for various subpopulation groups in developed countries including those who are incarcerated, smokers, widows, and people who suffer various eating and nutrient absorption disorders (Beal et al. 2017).

Vitamin C is involved in essential physiological processes that impact the agronomical value of crops and its physiology. The growth and the world population present unique challenges to agriculture. We need to find ways to make crops more productive with less inputs. One of the potential solutions to this great challenge is the development of improved crops with enhanced nutritional value, higher tolerance to environmental insults, and superior yields.

There are five routes for the production of AsA, one in animals and four in plants (Fig. 1). The biosynthetic pathway for AsA in animals was elucidated in the early 1950s (Ishikawa et al. 2006). On the other hand, plants possess four possible ways to make this key molecule. These are the d-mannose/l-galactose (Wheeler et al. 1998), l-gulose (Wolucka and Van Montagu 2003), d-galacturonate (Agius et al. 2003), and myo-inositol (Lorence et al. 2004) routes (Fig. 1). The best characterized of these pathways is the d-mannose/l-galactose route. The degree of which the so called “alternative routes” is operational in each plant species and the tissue and temporal specificity of these alternative routes is currently unclear.

Fig. 1
figure 1

The ascorbate metabolic network. There is a single biosynthetic pathway for l-ascorbic acid (ascorbate, vitamin C) in some animals, while four distinct routes lead to its formation in plants: MI (myo-inositol) pathway, Gul (l-gulose) pathway, Man/Gal (d-mannose/l-galactose) pathway, and GaIU (d-galacturonic acid) pathway. The enzymes involved in the four different routes are: MI pathway: myo-inositol oxygenase (MIOX); glucuronate reductase (GlcUR); gluconolactonase (GNL); and l-gulono-1,4-lactone oxidase (GuILO). Man/Gal pathway: phosphomannose isomerase (PMI); phosphomannose mutase (PMM); GDP-mannose pyrophosphorylase (VTC1); GDP-3′5′-epimerase (GME); l-galactose guanyltransferase (VTC2); l-galactose-1-phosphate phosphatase (VTC4); l-galactose dehydrogenase (GalDH); l-galactono-1,4-lactone dehydrogenase (GLDH). GaIU pathway: d-galacturonate reductase (GaIUR). The enzymes in the recycling pathway are monodehydroascorbate reductase (MDHAR) and dehydroascorbate reductase (DHAR). Adapted from Suza et al. 2010

Full size image

In this chapter, we present an overview of the successful strategies to metabolically engineer elevated AsA content in model plants and crops and summarize the studies in which the stress tolerance of those high ascorbate plants has been analyzed. We also discuss how the use of high-throughput phenotyping or phenomics is revolutionizing the way plant phenotypes are characterized and illustrate the power of digital phenotyping in the development of model plants with elevated AsA and enhanced ability to withstand stresses.

2 Successful Metabolic Engineering Strategies to Enhance Ascorbate Content in Plants

During the last 20 years, we have witnessed an unprecedented progress in the understanding of the multiple pathways leading to AsA formation in plants. Table 1 summarizes the successful strategies that have been reported in the engineering of elevated AsA content in plants.

Table 1 Metabolic engineering strategies to increase ascorbate in plants
Full size table

All genes in the d-mannose/l-galactose pathway have been cloned and characterized, and several of those have been used to increase AsA content in more than one species. Overexpression of those genes has led to a 1.2-fold to 6-fold AsA increase.

Phosphomannose mutase (PMM) is one of the key enzymes in the regulation of the initial steps of the AsA synthesis via the d-mannose/l-galactose pathway (Fig. 1). Badejo et al. (2009) developed transgenic tobacco (Nicotiana tabacum) overexpressing PMM from acerola (Malpighia glabra), leading to a 2-fold AsA increase compared to controls.

Overexpression of kiwifruit GDP-l-galactose phosphorylase increased AsA content 6-fold in tomato fruits, while overexpression of this gene in potato increased AsA content in the tubers (Bulley et al. 2012). This gene has also been expressed in Arabidopsis (Zhou et al. 2012) and rice (Zhang et al. 2015).

The importance of GDP-galactose guanylyl transferase (GGT) as a rate-limiting step in the production of AsA production was confirmed with the overexpression of the GGT gene in Arabidopsis resulting in a 2.9-fold increase in AsA content (Zhou et al. 2012).

The overexpression of tomato GME leads to an increase of 1.2-fold in the AsA content of leaves and a boost of 1.6-fold in fruits (Zhang et al. 2011). This gene has also been successfully expressed in Arabidopsis leaves (Zhou et al. 2012; Huang et al. 2014) and alfalfa shoots (Ma et al. 2014).

GDP-d-Mannose pyrophosphorylase has been widely studied in a number of plants including tobacco (Badejo et al. 2008), tomato (Cronje et al. 2012), and Arabidopsis (Zhou et al. 2012; Sawake et al. 2015). The AsA content of transgenic tobacco plants expressing acerola GMP was 2-fold higher compared to wild type (Badejo et al. 2008). Overexpression of l-galactose dehydrogenase (GalDH) in Arabidopsis resulted in a modest increase (1.2-fold) in AsA content (Zhou et al. 2012).

l-Galactono-1,4-lactone dehydrogenase (GLDH) catalyzes the last step in the d-mannose/l-galactose pathway to AsA (Fig. 1). Liu et al. (2011) showed that suppression of GLDH expression resulted in a loss of chlorophyll, lower ribulose-1,5-bisphosphate carboxylase (RUBISCO) protein content, and a lower rate of CO2 assimilation. As a consequence, a slower rate of plant growth and lower seed set were observed. However, overexpression of this enzyme maintained high levels of chlorophyll, RUBISCO protein, and a higher rate of net photosynthesis, resulting in higher seed set. Also increased GLDH expression correlated with the reduced lipid peroxidation, indicating that this enzyme functions protecting the plant against ROS (Liu et al. 2011, 2013). GLDH has been overexpressed in multiple plants including rice (Liu et al. 2011; Zhang et al. 2015), Arabidopsis (Zhou et al. 2012), tobacco (Liu et al. 2013), and lettuce (Landi et al. 2015) leading to 1.3-fold to 2.1-fold AsA content compared to controls.

Engineering increased AsA content using a combination of genes has been more successful than using a single enzyme. Zhou et al. (2012) overexpressed the combination of GGT + GPP and GGT + GLDH separately in Arabidopsis. Their results indicate that the GGT + GPP transgenic lines performed best leading to a 4.1-fold AsA content.

Ripening strawberry has been reported to synthesize AsA from d-galacturonic acid and its methyl ester (Mapson and Isherwood 1956; Loewus and Kelly 1961). In 2003, Agius et al., discovery of a d-galacturonic acid reductase (GaIUR) gene from strawberry, open the possibility to engineer AsA content using this new gene. GalUR is the only enzyme identified in this pathway, which has been shown to be cytosolic (Agius et al. 2003). GalUR overexpression has led to a 2- to 3-fold AsA increase in Arabidopsis (Agius et al. 2003). In plant storage organs, possible sources of AsA include in situ synthesis and long-distance transport of AsA synthesized in other tissues. Hemavathi et al. (2009) developed transgenic potatoes via overexpression of GalUR. These resulting potato tubers had two times more AsA than controls. Transgenic tomato hairy roots overexpressing GalUR had 2.5-fold AsA (Wevar Oller et al. 2009) while overexpression of GalUR in fruits led to 1.4- to 2.5-fold AsA increases (Amaya et al. 2014; Cai et al. 2014; Lim et al. 2016).

Biochemical and molecular data indicate that myo-inositol can also be a precursor for the biosynthesis of AsA in Arabidopsis (Lorence et al. 2004; Zhang et al. 2008). This pathway involves four enzymes, starting from the oxidation of myo-inositol to d-glucuronic acid and further reduction to l-gulonic acid and to l-gulono-1,4-lactone, and further conversion to AsA. These reactions are catalyzed by myo-inositol oxygenase (MIOX) , glucuronate reductase (GlcUR) , gluconolactonase (GNL) , and l-gulono-1,4-lactone oxidase (GuILO , a.k.a. GLOase), respectively. These four enzymes have been already characterized by the Lorence Laboratory (Table 1). Arabidopsis lines overexpressing MIOX contained up to 3-fold ascorbate in leaves and presented increased biomass accumulation, growth rate compared to controls (Lorence et al. 2004). These results are similar with the findings by Kulkarni (2012) that showed a 3-fold increase in AsA content after transient AtMIOX4 expression in tomato fruits. Glucuronate reductase (GlcUR) is the second enzyme in the inositol pathway to AsA, and an isoform from Arabidopsis has been already characterized (Lorence and Nessler 2007; Yactayo-Chang 2011). The overexpression of GlcUR in Arabidopsis led to a 2-fold ascorbate content in leaves compared to the wild-type controls (Yactayo-Chang 2011). The first two enzymes in the inositol pathway to AsA seem to be cytosolic. The Lorence Laboratory has evidence indicating that some isoforms of the last two enzymes in this pathway reside in the chloroplast (Yactayo-Chang 2016) and the endoplasmic reticulum (ER) (Aboobucker 2014; Aboobucker et al. 2017), respectively. Our group recently characterized a GNL isoform that is localized in the plastids. When overexpressed in Arabidopsis, transgenics GNL over-expressers have a 5-fold AsA content in leaves (Yactayo-Chang 2016; Yactayo-Chang and Lorence 2016). The last enzyme in the inositol pathway to AsA is l-gulono-1,4-lactone oxidase (GuILO) . GuILO belongs to the aldonolactone oxidoreductases protein family. GuILO participates in both the l-gulose and inositol pathways to ascorbate. Transgenic tobacco and lettuce plants expressing rat GuILO accumulated up to 7-fold more ascorbate than untransformed plants (Jain and Nessler 2000), while Arabidopsis lines overexpressing the rat isoform were reported to contain a more modest AsA increase (Lisko et al. 2013).

Heick et al. (1972) reported that AsA has been found in a number of yeasts including Saccharomyces cerevisiae. The arabino-1,4-lactone oxidase (ALO) enzyme catalyzes the last step in the formation of d-erythroascorbate in yeast. This enzyme uses l-galactono-1,4-lactone as efficiently as d-arabino-1,4-lactone to produce AsA. Overexpression of ALO resulted in elevated ascorbate (1.3-fold) in tomato fruits and stylo leaves (3.1-fold) (Cronje et al. 2012; Bao et al. 2016).

Once oxidized, AsA can also be recycled back to its reduced form. The reduced AsA is ready for its role again. Overexpression of dehydroascorbate reductase (DHAR) and monodehydro ascorbate reductase (MDHAR) has been implemented successfully to engineer plants with elevated AsA. Overexpression of DHAR has led to 1.3-fold to 6-fold AsA increase in various plant species (Table 1). Constitutive expression of MDHAR isoforms has also been successfully implemented to increase AsA content in plants (Table 1).

The existence of four different pathways leading to AsA formation in plants indicates the need of a complex regulatory mechanism for this network. Regulation within and between pathways in the network is largely uncharacterized. Six master regulators for AsA have been described in the literature. The ascorbic acid mannose pathway regulator 1 (AMR1) is one of these regulators. AMR1 negatively regulates the d-Man/l-Gal pathway (Zhang et al. 2009).

Another regulator of this network is the three KNOTTED-like homeobox gene (HBK3) . Homeobox genes encode a typical DNA-binding domain of 60 amino acids, known as homeodomain that characterizes a large family of transcription factors. These genes encode transcription factors which regulate important events in plant growth and development. The HBK3 gene from Norway spruce regulates somatic embryo yield through alterations in glutathione and AsA metabolism, which have been previously implicated in controlling development and maturation both in vivo and in vitro (Belmonte and Stasolla 2009).

The size of the AsA pool in plants is controlled by a combination of its biosynthesis, recycling, and transport and translocation. Some factors that regulate the AsA pool size by modulating biosynthesis are light, development, and environmental stress conditions. The ethylene response factor (ERF98) is considered a positive regulator of AsA. This regulator enhances tolerance to salt stress through direct activation of AsA synthesis in Arabidopsis (Zhang et al. 2012). Arabidopsis ERF98 transcriptionally activates gene expression of VTC1 to improve AsA content and tolerance to salt stress.

3 Effect of Ascorbate in the Ability of Plants to Withstand Abiotic Stresses

3.1 Low AsA Plants Are Sensitive to Abiotic Stresses

Ascorbate plays vital role in maintaining the plant health status. Mutation of genes involved in the d-mannose/l-galactose pathway led to the characterization of vitamin C defective (vtc) mutants. The vtc 1-1, vtc 2-1, and vtc 3 mutants have 50–75% lower AsA content than wild type. Plants with low AsA are sensitive to heat and light stress compared to control plants (Pavet et al. 2005; Conklin et al. 2013). Interestingly, the vtc 1-1 and vtc 2-1 lines are resistant to bacterial pathogen, Pseudomonas syringe (Tóth et al. 2011). The vtc mutant plants are sensitive to osmotic stress and oxidative stress (Cho et al. 2016) as shown in Table 2. This table illustrates that lowering AsA content makes plants more vulnerable to abiotic stresses.

Table 2 Effect of low ascorbate plants in response to stresses
Full size table

3.2 High AsA Lines Are Tolerant to Abiotic Stresses

Table 3 presents a summary of the studies in which diverse research groups have evaluated the tolerance to stresses of plants with enhanced AsA content. High AsA plants overexpressing genes involved in the d-mannose/l-galactose pathway have been found to be tolerant to salt stress (up to 100 mM NaCl), herbicide stress (up to 75 μM methyl viologen), cold stress (4 °C), and oxidative stress (wounding). Plant biomass, plant height, shoot length, root length, leaf color, survival rate, and seed germination rate are the most common readouts used to assess the tolerance to abiotic stresses in plants (Zhang et al. 2015, Zhang et al. 2011; Landi et al. 2015; Li et al. 2012).

Table 3  Summary of studies that have evaluated the tolerance to stress of plants wutg enhanced ascorbate content
Full size table

Similarly plants overexpressing d-galacturonate reductase were found to be tolerant to salt stress (up to 600 mM NaCl), herbicide stress (up to 75 μM methyl viologen), cold stress (4 °C), and osmotic stress (up to 300 mM mannitol). Chlorophyll content, root length/biomass and shoot length/biomass, germination response, and malondialdehyde content have been the main readouts used to assess the tolerance to abiotic stresses in these plants (Hemavathi et al. 2009; Cai et al. 2014; Lim et al. 2016).

Plants overexpressing l-gulono-1,4-lactone oxidase (GulLO) , the enzyme involved in the l-gulose and the inositol pathway of AsA, have elevated AsA. When subjected to abiotic stresses, those plants were tolerant to herbicide stress (up to 20 μM methyl viologen), osmotic stress (up to 300 mM mannitol), and salt stress (up to 600 mM NaCl). Phenotypic changes, chlorophyll content, and malondialdehyde content were the variables used to assess the tolerance to stresses in these plants (Hemavathi et al. 2009; Lim et al. 2016).

Our group and others have characterized the response to stresses of high AsA Arabidopsis lines overexpressing MIOX4, one of the enzymes involved in the myo-inositol pathway to AsA. Increased level of AsA in those plants correlates with tolerance to salt stress (up to 200 mM NaCl), cold stress (16 °C for 3 weeks), heat stress (29 °C for 10 days/40 °C for 15 min), environmental pollutant (up to 200 ppb pyrene), and light stress (up to 300 μmol m−2 s−1). Photosynthetic efficiency, root growth, and shoot growth are the main readouts used to assess abiotic stress tolerance (Tóth et al. 2011; Lisko et al. 2013).

Arabidopsis plants overexpressing a recently characterized GNL containing 5-fold AsA content were found to display enhanced growth rate, tolerance to light stress, and improved photosynthetic efficiency and seed yield (Yactayo-Chang 2016; Yactayo-Chang and Lorence 2016).

Monodehydroascorbate reductase (MDHAR) and dehydroascorbate reductase (DHAR) are the recycling enzymes for AsA in plants. MDHAR and DHAR help to maintain redox status of AsA (Haroldsen et al. 2011). Overexpression of MDHAR from Lycopersicum spp. and Malpighia spp. resulted in 1.2–1.8-fold elevated AsA level in leaves compared to wild-type plants. Elevated AsA level correlated with tolerance to herbicide stress (up to 100 μM methyl viologen), heat stress (40 °C for 24 h), cold stress (4 °C for 24 h), and salt stress (up to 300 mM NaCl) (Li et al. 2010; Eltelib et al. 2012). Similarly, overexpression of DHAR resulted in 1.2–4.0-fold increase in AsA level in leaves, fruits, and roots compared to controls. High AsA plants were found to be tolerant to ozone stress (up to 0.2 ppm), drought stress (5% polyethylene glycol), herbicide stress (up to 3 μM methyl viologen), aluminum stress (up to 400 μM AlCl3), salt stress (up to 150 mM NaCl), light stress (up to 1000 μmol m−2 s−1), and cold stress (4 °C for 1 week). Photosynthetic efficiency, plant height, number of leaves, plant growth, ion leakage, H2O2 accumulation, transpiration, CO2 accumulation, and stomatal conductance were the parameters used to assess the abiotic stress tolerance on these plants (Kwon et al. 2003, Chen and Gallie 2005, Eltayeb et al. 2011, Li et al. 2012, Qin et al. 2015; Liu et al. 2011).

Overexpression of homologues of the plant GulLO enzymes from Saccharomyces cerevisiae, o-arabiono-1,4 lactone oxidase and Stylosanthes guianensis 9-cis-epoxycarotenoid deoxygenase, resulted in a boost of the AsA content in plants. These transgenics showed tolerance to herbicide stress (up to 0.1 mM methyl viologen), light stress (up to 1200 μmol m−2 s−1), aluminum stress (100 mM AlCl3), cold stress (3 and 6 °C), and water limitation stress (Chen et al. 2005; Bao et al. 2016).

On the other hand, overexpression of master regulators of the AsA metabolic network, ethylene response factor 8, SID 22, SIHZ24 and CSN5B from Arabidopsis thaliana and Solanum lycopersicum resulted in 1.2- to 2.0-fold increase in AsA content in leaves and fruits compared to controls. Plants with elevated AsA were found to be tolerant to salt stress (300 mM NaCl), light stress, herbicide stress (up to 50 mM methyl viologen), and oxidative stress. Shoot biomass, root biomass, aluminum accumulation, relative water content, ion leakage, chlorophyll content, and malondialdehyde content were the parameters used to assess the abiotic stress tolerance in those studies (Zhang et al. 2012; Wang et al. 2013; Cai et al. 2016; Hu et al. 2016).

In summary, increased AsA level in plants is associated with enhanced abiotic stresses tolerance, whereas decreasing AsA level has led to sensitivity to abiotic stresses. Surprisingly, plants with low AsA level (vtc mutants) have been shown to be resistant to some biotic stresses.

To date, the most common readouts of abiotic stress tolerance have been plant size, plant shape, and color and photosynthesis parameters. In all published studies to date research teams have relied on manual phenotyping to make these assessments. Manual phenotyping is time consuming and sensitive to bias and is limited to the resolution of the naked eye. Based on these drawbacks, it is important to implement the use of more advanced tools to analyze the phenotype of high AsA plants to fully realize the potential of these metabolic engineering strategies.

4 Plant Phenomics, a Modern Approach to Characterize Plant Phenotypes

4.1 Closing the Gap Between Genomics and Phenomics

Understanding crop adaptation to abiotic and biotic stress is very important, especially now in the face of climate change and global population growth. Powered by the advances in gene editing (Barrangou and Doudna 2016) and the study of natural genetic variability (Nunes-Nesi et al. 2016), plant genomics has been growing to select and improve crop adaptation and yield. This is why it is important to understand plant plasticity using phenomics. Plant phenomics is defined as the application and development of different methodologies to capture information related with performance, function, and structure of a large number of plants (Houle et al. 2010; Tardieu et al. 2017). The main purpose of plant phenomics is to understand plant behavior under a vast variety of scenarios and how exactly the genotypic traits are expressed through the plant phenotype.

Scientists that have had the opportunity to work with manual plant phenotyping understand how slow this process can be. Measuring plant area, plant height, seed number, panicle number, fruit color, leaf number, and other traits of interest can be time consuming and affected by human bias. Phenomics is a research area that is moving fast due to the progress in the development of new sensors and imaging techniques for several traits, plant organs, and responses (Furbank and Tester 2011). However, a major challenge that remains in this area is data handling and processing when analyzing the sensor information and translating that information into knowledge.

4.2 Plant High-Throughput Phenotyping

Being able to gather massive amounts of information based on high resolution images is very convenient. The capability of eliminating destructive measurements increases the capacity of adjusting the experimental design and allows gathering information throughout the life cycle of plants. Plant high-throughput phenotyping technology consists of taking images with different sensors using hundreds of plants per day (Fahlgren et al. 2015). Data acquisition and analysis have become more feasible due to advances in automation for plant phenotyping, specifically robotic and sensor based technology (Rahaman et al. 2015). Currently, this technology has expanded not only at the laboratory (controlled environment) and greenhouse levels but also to the field scale.

The adaptability of this new technology is overwhelming, and it has made it possible to have platforms such as the Field Scanalyzer System at Rothamsted Research, UK and the TERRA-REF (http://terraref.org/) project in Maricopa, AZ, examples of large field crop analytic robots in the world. These robots take images by moving on a surface of 100–200 m2 using different sensors. This technology can also be used on a smaller scale with options that are less expensive, such as taking images using a digital camera, a tripod, and a light box attached (Chitwood et al. 2014). The data gathered with this technology can then be analyzed using commercial software or open source algorithms. A compilation of software options for plant phenomics can be found in the online database plant-image-analysis.org/ (Lobet et al. 2013).

4.3 Key Sensors in Plant Phenomics

Plant plasticity is very complex. When a genomic trait is adjusted, plants show very different architectures and phenotypic responses depending on the environmental conditions they face. Here is where different sensors come into play to help understand these behaviors. Sensor technology has improved since RGB images were used for the first time to determine plant growth, where top view RGB images were taken and correlated to obtain plant fresh weight (Leister et al. 1999).

The sensors available in the market work at different wavelength ranges of the light spectrum (Fig. 2). Color cameras are restricted to the 400–700 nm range and they have three colored sensors (red, green and blue) that are used to calculate the true color of each pixel (Fahlgren et al. 2015). These sensors use pixel-based maps to analyze several phenotypic characteristics such as biomass, area, diameter, flowering time, root architecture, plant height, and others. Although these sensors are the most commonly used, there is a disadvantage to them since they cannot distinguish between overlapping leaves or soil background unless some segmentation process is applied to the image (Li et al. 2014; Rahaman et al. 2015).

Fig. 2
figure 2

Representative images acquired with the visible (VIS), near infrared (NIR), and fluorescence (FLUO) sensors. (a–c) Raw images. (d–f) Analyzed images where the different parameters were quantified. (d) Shows geometrical parameters and color classification computed from VIS images. (e) In planta water content values computed from NIR images. (f) In planta chlorophyll fluorescence values computed from FLUO images. Values in E and F have been corrected to be relative to plant area

Full size image

Due to global warming, scientists have focused their research in finding solutions to reduce the negative effect of abiotic/biotic stresses in plant yield. When a plant is affected by high temperature or soil water is depleted, plants tend to close the stomata to avoid water loss. The physiology of plants is also affected due to the increase in leaf temperature, and responses such as leaf tissue and enzyme damage, lower photosynthetic rate, higher respiration rate, and others occur (Schauberger et al. 2017). For this reason, sensors that detect near-infrared (NIR) and infrared (IR) light are very important for plant high-throughput phenotyping. These sensors use pixel-based maps and are used to determine plant water content (900–1700 nm) and canopy/leaf temperature (700–1000 nm), respectively (Fahlgren et al. 2015).

Plants under stress conditions are affected internally, specifically in their photosynthetic machinery, which will affect crop yield. The use of fluorescence (FLUO) sensors allows for the quantification of chlorophyll molecules in photosystem II by artificial excitation, and this information is used to measure how efficient the plant is during photosynthesis (Jansen et al. 2009; Li et al. 2014). The sensor corresponds to a charge-couple device (CCD) camera with sensitive fluorescent signals. These signals occur when the sample is illuminated with visible or UV light (Rahaman et al. 2015). The FLUO sensors are pixel-based maps that are used to detect disease, infestation, photosynthetic status, carbon assimilation, quantum yield, non-photochemical quenching, and total chlorophyll content (Kuhlgert et al. 2016).

Agronomical traits such as tiller, leaf, and panicle number are key for yield estimation. This is why 3D sensors play an important role in plant phenomics. These type of sensors use depth maps and allow acquiring noncontact and nondestructive measurements. By combining a 3D camera with laser scanning, the amount of point clouds detected enables a valid and accurate description of the plant geometry (Paulus et al. 2013). Some of the phenotype parameters that can be detected with this sensor include shoot structure, leaf angle, canopy health status, leaf growth, coverage density, and panicle health (Li et al. 2014).

The latest sensor developed, and the most expensive one, corresponds to imaging spectroscopy, also called hyperspectral imaging. This technology combines features of RGB imaging based on pixel mapping and spectroscopy based on intensity from different spectral bands. The most important feature about this sensor is that it enables measuring plant chemical traits such as water, nutrients, lipids, sugars, and others (Pandey et al. 2017). This sensor is currently the slowest of all, and the image analysis requires additional computational knowledge due to the data dimensions and complexity.

4.4 Phenomics to Understand Plant Abiotic/Biotic Stress Response

The power of phenomics relies on the fact that it is possible to extract a great deal of information from a single image. Figure 3 presents representative images of Arabidopsis plants that have been phenotyped using RGB, FLUO, and NIR sensors (panels A, B, and C). Once the images were analyzed through a commercial software (panels D, E, and F), the information was plotted and analyzed. Each of the sensors quantifies different phenotype parameters (Fig. 3). For example, from the RGB sensor, data such as plant size, architecture, biomass and growth can be obtained. From the NIR, quantification and localization of water can be obtained. More information can be gathered and analyzed, that will help understand in a more detailed manner plant adaptation to the new environment due to climate change. Subtle changes that are not detectable with the naked eye can be measured and quantified using sensors.

Fig. 3
figure 3

Comparison of different cameras used in high-throughput plant phenotyping and the parameters that can be measured with each of them. Each sensor captures signals from the visible and infrared spectrum of the light. The VIS (visible range) camera can quantify plant size, shape, and color. The IR (infrared range) camera can quantify plant stress. The PSII (photosystem II efficiency) sensor can quantify plant photosynthetic activity. The NIR/SWIR (near infrared and short-wave infrared) camera can detect plant relative water content. The TIR/LWIR (thermal infrared and long-wave infrared) sensor can detect plant canopy or leaf temperature. The hyperspectral camera can measure plant stress and chemical composition by detecting a wide variety of spectral bands

Full size image

Figure 4 illustrates the use of plant phenomics approaches in the characterization of high AsA Arabidopsis lines. As shown in this figure, AtMIOX4 over-expressers are salt tolerant as indicated by their higher growth rate and biomass accumulation compared to wild-type controls. Although not evident to the naked eye, but captured by the RGB camera AtMIOX4 displayed less chlorosis than wild-type plants exposed to NaCl stress.

Fig. 4
figure 4

Characterization of the phenotype of a high ascorbate line compared to controls. Arabidopsis seeds were sterilized, vernalized, and planted on Murashige and Skoog media. After germination vigorous seedlings were transferred to soil and grown in an environment controlled chamber (23 °C, 65% humidity, 16:8 h photoperiod, 200 μmol m−2 s−1 light intensity). After acclimation to soil, plants were subjected to salt stress (150 mM NaCl) 21 days after germination. (a, b) Raw RGB images after segmentation. (c) Projected leaf area of the Arabidopsis rosette; data are means ± SE, n = 15. (d, e) Color classification of the plants analyzed in (b). Values were corrected relative to plant area. This parameter distinguishes yellow color (chlorosis) from green color (healthy tissue). Chlorosis was not apparent by visual inspection, but detected by the RGB camera. MIOX Myo-inositol oxygenase over-expresser, WT wild-type control

Full size image

4.5 Data Analysis and Handling

Plant high-throughput phenotyping studies generate large-scale, multidimensional data sets that require proper analysis (Klukas et al. 2014). One of the main challenges in the field is standardization of the methodologies and data publication. Storage and data collection needs to be consistent and linked with specific phenotypes and genomic information (Krajewski et al. 2015).

Optimization of the experimental design is key to obtain good quality images and, therefore, reproducible data. First, it is necessary to determine the objective of the experiment before acquiring any information. Then, once the objective is determined, it is necessary to select the sensor/sensors to use. Moreover, having consistency in how the images are acquired is imperative. Improving the quality of the images facilitates the analysis. One simple example of this is adding a blue mesh as a soil cover, in order to improve background/object contrast and to reduce evaporation (Junker et al. 2014).

Plant development research has been positively affected by the advance in phenotyping technologies, which help to determine phenotypic traits that change depending on genotype and environment. Further analysis that integrates several sensors and overlaps the data collected is necessary to better understand plant plasticity and adaptation.

5 Conclusions

During the past two decades, we have significantly advanced our understanding about the different pathways plants use to make AsA, a key molecule involved in the modulation of plant growth and development, plant health, and stress tolerance. Full realization of the potential that this knowledge represents for the development of more nutritious and improved crops must involve the incorporation of novel approaches to engineer crops and to characterize their phenotype. We propose that gene editing (Barrangou and Doudna 2016) and plant phenomics are two approaches that are likely to revolutionize the way high AsA crops are made and characterized before some of them are able to reach widespread adoption.