Roots are essential for plant productivity and serve a variety of functions, such as water and nutrient uptake, forming symbioses with other microorganisms in the rhizosphere, anchoring the plant to the soil, and acting as storage organs. The different interactions of a root with its environment depend on its organization and structure, from the cellular to whole-plant level. The root contains a stele, comprised of the xylem, the phloem, and the pericycle (Smith and De Smet, 2012). The stele is encircled by concentric layers of epidermal, cortical, and endodermal tissues. The root apical meristem forms the basic stem cell pool from which other cell types develop. This root apical meristem also holds the quiescent center (QC), with rarely dividing cells that signals the surrounding cells to organize and maintain the initial stem cells (Dolan et al., 1993). There are generally two types of roots: (i) those that are formed in the embryo, such as the primary and seminal roots in maize (Hochholdinger, 2009), tap or primary root in common bean (Lynch and Brown, 2012); (ii) those formed post-embryonically from consecutive nodes on shoots, normally referred to as adventitious roots (ARs). These include basal roots in beans, nodal roots in maize, ARs of sweetpotato, potato, cassava, as well as yam, and lateral roots (LRs; Lynch and Brown, 2012). LRs are formed post-embryonically from the pericycle of all root classes through auxin-dependent cell cycle activation. This cell cycle forms the LR founder cells that undergo several rounds of cell division to initiate LRs (Overvoorde et al., 2010). The elongation, growth angles from the main axis, lateral branching and longevity of all root classes forms the root system which is determined by genetic, physiological, and environmental factors (Lynch and Brown, 2012).

Root system architecture (RSA) therefore refers to the spatial configuration of the root system or the explicit deployment of root axes (Lynch, 1995). Under poorly understood genetic control, RSA exhibits plasticity and responds to external environmental conditions such as soil moisture, nutrients, temperature, pH, and microbial communities (Bao et al., 2014). The study of RSA is important for agricultural productivity because most soils have uneven distribution of resources and/or localized depletions that make spatial distribution of the root system an important determinant of a plant’s ability to exploit these resources (Lynch, 1995). Progress in the study of RSA in agricultural crops has consequently been realized, especially for cereals, and evidence for the genetic control of RSA and its relationship to increased productivity under stress is currently well-documented. Despite these achievements, information on RSA in root and tuber crops (RTCs), which form the second largest group of crops for global food security after cereals, is still lacking. A recent review by Villordon et al. (2014b) on root architecture and RTC productivity clearly indicates this gap. In the current review, we discuss RSA with special reference to RTCs, the genetic control of RSA, the relationship between RSA and abiotic stresses, and how RSA can be manipulated to confer tolerance to abiotic stresses. We then draw conclusions on the way forward for RSA studies in RTCs.

Understanding RSA and the mechanisms of its development will allow manipulation and exploitation of different root traits to improve plants’ adaptation to changing climates and increase yields for the growing global human population (Smith and De Smet, 2012). Vegetatively propagated RTCs such as potato (Solanum tuberosum), sweetpotato (Ipomoea batatas), cassava (Manihot esculenta), and yam (Dioscorea spp.) provide food security for vast populations, especially in sub-Saharan Africa where many resource-poor small holder farmers provide the majority of food. Of the four major RTCs, cassava and sweetpotato are storage roots, while potato and yam are tubers. Little literature is specifically targeted to root growth and development in RTCs compared to cereals. The little literature available also mainly focuses on the growth and development of the storage root or tuber, primarily at harvest and post-harvest evaluations, rather than the entire root system. In this section we describe the root systems in the four major RTCs, based on available literature (Figure 1).

Cassava, potato, sweetpotato, and yam have ARs originating from the shoot or subterranean stem, in contrast with the primary root in seed-propagated crops which originates from the embryo. In sweetpotato and cassava, RSA is composed of ARs, LRs and storage roots (SRs), whereas in potato, the ARs can be divided into basal (ARs in Figure 1) and stolon roots (STR). In yam, the ARs root system is the most pronounced. The simple recognition of the main AR axis and its spatial and temporal relationship to LRs and their initiation in RTCs would enable systematic investigations to further understand the mechanisms that trigger LR emergence and morphogenesis.

Most genetic studies on RSA have been carried out using ‘model’ plant Arabidopsis and a few ‘model’ cereals including maize and rice. In this section we review the hormonal/genetic control of RSA as reported in Arabidopsis and/or cereals followed by what is known in RTCs for storage root and tuber development. Formation of LRs as an important aspect of RSA which is a result of cell division of a specific subset of pericyle cells (Casimiro et al., 2001). The cell division process is under the control of both root- and shoot-derived auxin, where cellular levels of auxin contribute to the regulation of gene expression, which then impacts root branching (Overvoorde et al., 2010). Auxin activates Cyclin-dependent-kinases (CDKs) (Himanen et al., 2002) and D-type cyclin (CYCD) (Nieuwland et al., 2009). These two are cell-cycle genes involved in pericycle division during the LR initiation process and whose inhibition leads to a reduced number of LRs. Also related to the cell cycle initiation is the E2Fa/Dpa transcription factor, which promotes the G1-S transition by controlling the expression of genes required for DNA replication (Boudolf et al., 2004). However, initiation of the cell cycle alone is not enough to initiate LR formation but, as Vanneste et al. (2005) showed, LR initiation requires fine tuning by both negative and positive mechanisms regulating auxin homeostasis and signal transduction in the pericycle. These processes are under the control of auxin-responsive genes dependent on Auxin/indole-3-acetic acid-auxin response factors (AUX/IAA-ARFs) auxin signaling pathways. Genes containing auxin-responsive elements (AREs) in the promoter region are directly regulated by ARFs. In the absence of auxin, the ARFs combine with AUX/IAA proteins (AUX/IAA-ARFs) and are therefore not active. In the presence of auxin, however, the AUX/IAA proteins are degraded by auxin-receptor proteins TIR1 and AFBs through the SCF ^TIR1/AFBs complexes and 26S proteasomes (Goh et al., 2012). This degradation leaves the ARFs active to either positively or negatively regulate auxin responsive transcription. There are several of these AUX/IAA-ARF modules which are proposed to successively coordinate different developmental processes by regulating distinct targets (De Smet et al., 2010). The exact number of such modules involved in LR development is however still unknown. De Smet et al. (2010) showed a bimodal auxin response where they found that in addition to the Solitary root/indole-3-acetic acid14 – auxin response factors7 and 19 (SLR/IAA14-ARF7-ARF19), the Bodenlos/indole-3-acetic acid12/monopteros-auxin response factor5 (BDL/IAA12/MP-ARF5), acting downstream of SLR/IAA14, was required to guarantee organized LR patterning. Goh et al. (2012) listed several modules responsible for different stages of LR initiation, including the IAA28-ARFs module, which regulates the specification of LR founder cells; the SLR/IAA14-ARF7-ARF19, which regulates nuclear migration and asymmetric cell division of the LR founder cells for LR initiation and the BDL/IAA12/MP-ARF5, which regulates LR initiation and organogenesis; the Short hypocotyl2/IAA3-ARF (SHY2/IAA3-ARF), which regulates primordia development and emergence after SLR/IAA14-ARF dependent LR initiation, and which also inhibits LR initiation. Each of these modules have target genes. Okushima et al. (2007), for example, showed that the SLR/IAA14-ARF7-ARF19 module regulates LR formation by directly activating lateral organ boundaries domain/ asymmetric leaves2-like (LBD/ASL) genes. Many other hormones interact with the auxin signaling pathways during LR initiation Cytokinin (CK) and exogenous abscisic acid (ABA) negatively affect LR development whereas Brassinosteroid (BR) positively affects LR formation. The pericycle founder cell cycling is blocked in the G2 to M transition phase by CK thereby inhibiting LR formation. In the presence of exogenous ABA, emergence of LR primordia from the parent root is inhibited before the LR meristem is activated. Despite this negative regulation of LR development by exogenous ABA, ABA signaling also has cross talks with auxin action via the ABA insensitive 3 (ABI3) and the enhanced response to ABA 1 (ERA1) genes which enhance auxin-regulated LR formation. Cross-talk is also indicated between BR and auxin-dependent LR formation, where it is thought to promote acropetal auxin transport (reviewed by Fukaki and Tasaka, 2009). Although most of these studies were carried out in Arabidopsis, Orman-Ligeza et al. (2013) compared some of these molecular control pathways in cereals and Arabidopsis and found that the AUX/IAA-ARF and the LBD/ASL regulatory pathways were conserved. Several reviews on genetic and hormonal control of RSA are available. Smith and De Smet (2012) have reviewed the genetic control of root branching, giving insights from Arabidopsis and cereals. Jung and McCouch (2013) also provide a comprehensive review on the genetic and hormonal control of RSA.

In sweetpotato, the only study found specifically referring to genetic control of RSA is by Villordon et al. (2014a) who showed evidence that orthologs of genes associated with RSA in model crops were present in sweetpotato. They found increased expression of a putative nitrogen transporter and deceased expression of a high affinity nitrogen transporter as well as decreased expression of a MAD-box gene under low nitrogen (N) conditions. A substantial amount of information is however available for storage root formation which is part of RSA in sweetpotato and cassava. Cytokinin is important in regulating storage root development in sweetpotato (Hashizume et al., 1982). Zeatin Riboside (ZR), Trans-Zeatin Riboside (t-ZR) and 9-glucosyl-n-6-2-isopentenyl adenosine (i6Ado) are the major CK involved in developing and activating the primary cambium. Besides hormones, several genes are involved in storage root formation and development in sweetpotato. Tanaka et al. (2005) found SRF1 through SRF10 developmentally regulated genes to be involved in storage root formation. SRF1, SRF2, SRF3, SRF5, SRF6, SRF7, and SRF9 were upregulated while SRF4, SRF8, and SRF10 were downregulated during storage root formation. Tanaka et al. (2008) showed that knotted1-like homeobox (KNOX1) genes, Ibkn1, Ibkn2 and Ibkn3, are associated with storage root development in sweetpotato. Ibkn1 and Ibkn2 were upregulated in developing and mature storage roots relative to fibrous roots. Ibkn1 is homologous to shoot meristemless (STM) gene of Arabidopsis whose overexpression leads to higher CK levels, while Ibkn2 and Ibkn3 are homologous to Brevipedicellus gene of Arabidopsis which negatively regulates lignin biosynthesis. A group of MAD-box genes, IbMADS genes such as IbMADS3, IbMADS4, and IbMADS79 are also found in fibrous roots before thickening, mainly in the vascular cambium region where rapid cell division occurs during storage root thickening (Kim et al., 2002). Noh et al. (2010) found that a MADS-box protein copy DNA, SRD1 enhances the proliferation of the metaxylem and cambium cells during the auxin-dependent initial thickening and growth of storage roots. Storage root development in sweetpotato is enhanced when an expansin gene (IbEXP1) is down-regulated (Noh et al., 2012), but lignin biosynthesis is inhibited as starch biosynthesis is enhanced during early storage root formation (Firon et al., 2013). Details on the molecular regulation of storage root formation in sweetpotato have been reviewed by Ravi et al. (2009, 2014). No literature was available on the genetic control of ARs and LRs in cassava. However, de Souza et al. (2004) showed overexpression of the Mec1 gene which codes for a Pt2L4 glutamic acid-rich protein and a RING Zinc Finger and LEA protein genes in the secondary xylem tissue of storage roots relative to fibrous roots. Based on a correlation network, the relationship between KNOX1 genes, phytohormone biosynthesis and phytohormone-signaling genes was established, and it was hypothesized that phytohormones are involved in the initiation of storage root development in cassava (Sojikul et al., 2015). Both potato and cassava storage organs have been substantially studied, but not the genetic and hormonal control of RSA for either crop.

In addition to hormones, signaling components, and transcription factors, micro-interfering RNAs (miRNAs) and small-interfering RNAs (siRNAs) have been shown to affect RSA in plants, as reviewed by Meng et al. (2010) and Khan et al. (2011). The miRNAs and siRNAs are thought to be involved in auxin signaling, nutrition metabolism and stress response by mediating signal interactions. They have been identified in embryonic root development, radial patterning, formation of ARs and LRs. However, their role in RTCs has not yet been studied.

Root system architecture has a central role in crop plants’ response to abiotic stresses. Since roots grow underground, they are the first to sense abiotic stresses and adjust their genetic program for post-embryonic development to survive the stress (Lynch, 1995). Plant roots obtain water and nutrients from the soil, which is a complex system with intrinsic properties, abiotic and biotic interactions. Modulation of RSA is therefore affected when changes in the plant nutritional status and external nutrient supply over time are perceived and integrated into the intrinsic root development program. The degree of root plasticity is based on variations in the number, extension, placement, and growth direction of individual components of the root system (Giehl et al., 2014). These changes in RSA consequently affect the growth and development of above-ground biomass (Paez-Garcia et al., 2015) by altering carbon allocation to shoots and/or triggering signaling pathways involving hormones, proteins, RNAs, among others (DoVale and Fritsche-Neto, 2015). In this case therefore, roots indirectly regulate leaf stomatal conductance and affect leaf blade posture and photosynthetic rate when exposed to abiotic stress.

Different abiotic stresses affect RSA in varied ways. Table 2 summarizes the root traits necessary for adaptation to different abiotic stresses. Deeper roots are associated with increased acquisition of water and mobile nutrients like N that may leach to lower soil layers (Lynch and Wojciechowski, 2015). LRs, the main determinants of ultimate RSA, are influenced strongly by moisture and nutrient distribution in the soil (Postma et al., 2014). Deak and Malamy (2005) showed that LR formation from LR primordia in Arabidopsis is repressed under drought stress when ABA and lateral root development (LRD2) gene, interact with auxin. Since ABA, LRD2 and auxin are also involved in RSA even without drought stress, it appears that such genes like LRD2 regulate the formation of LRs through promotive and repressive hormone signaling pathways depending on the environmental conditions. Repression of LR development under abiotic stress is of particular importance in root crops. In sweetpotato for example, the final storage root yield depends on the capacity of a genotype to develop LRs on the main ARs. Those with arrested or non-emerged LRs develop lignified steles, which inhibit localized swelling into storage roots (Villordon et al., 2012). Other Important contributors to RSA include single-cell projections from root epidermal cells referred to as root hairs (Tanaka et al., 2014). A high density of both root hairs and LRs is associated with increased nutrient uptake, especially in the top soil (Postma et al., 2014) but increased metabolic costs is a trade-off here (Zhan et al., 2015). There are other trade-offs associated with crop adaptation to individual abiotic stresses. Primary root length is inhibited under low-soil phosphorus (P) conditions, while LR development is promoted thereby leading to a shallower root system. This has negative effects under drought stress where deeper roots are necessary in order to have better access to water (Wasson et al., 2012). Reduced frequency of LR branching improves N uptake where genotypes with fewer but longer LRs have greater axial root elongation, deeper roots and better N uptake than those with a higher number of LRs (Zhan and Lynch, 2015). On the other hand, a larger number of LRs is required under P-limited conditions for topsoil foraging (Lynch and Brown, 2001). Since abiotic stresses normally occur in combination under field conditions, it is therefore evident that there is ‘no size fits all’ if adaptation to abiotic stress conditions is done considering each stress individually.

In RTCs, it is known that the root system is made up of ARs and LRs which presumably are involved in water and nutrient uptake and hence respond to abiotic stress. However, some RTCs have complex RSA because the harvestable part is also underground with several root classes, e.g., in potato, which may have different functions with regard to adaptation to abiotic stress. The potato root system is known to be shallow, with poor ability to penetrate soils thereby being drought susceptible (Porter et al., 1999). Despite having a shallow root system, potato is still not efficient in P and N uptake because the larger root system has a respiration carbon cost (Balemi and Schenk, 2009). Furthermore, most findings studied the root system as a whole without identifying possible roles for different root classes. An attempt at this was done by Wishart et al. (2013) who reported genetic variation for potato root traits without any specific abiotic stress. They suggested that basal roots were responsible for water uptake and anchorage while stolon roots were responsible for nutrient uptake and tuberization. Cassava and sweetpotato have less root classes compared to potato because the harvestable part is also a root. Pardales et al. (1999) studied the effects of high root zone temperature on root systems of cassava and sweetpotato. They showed a reduction in the total length of ARs, number and total length of first order LRs, under high root zone temperature in both crops. Pardales and Yamauchi (2003) also showed a suppression of AR and LR formation and development under drought stress in both sweetpotato and cassava. Recent studies in sweetpotato indicate direct influence on RSA of spatial and temporal availability of water and N availability similar to model systems (Villordon et al., 2014a). In yam, Hgaza et al. (2012) found no fertilizer response of tuber yield but a positive correlation between thinner, longer roots with tuber yield.

The aim of carrying out RSA studies in crop plants is to understand areas of interest within the root system and incorporate this information in crop improvement programs. Several approaches may be applied in manipulating RSA in order to adapt crops to changing climates.

Since roots grow below ground, studying the entire root system naturally requires digging it out, a complex process in itself, and it is difficult to extract the entire system without breaking off the finer parts. For this reason, studies have mainly dwelled on above-ground traits related to abiotic stress tolerance. However, given the pressures on crop productivity caused by global climate change, with the associated abiotic stresses, and the notion that food production needs to double in the next few years to accommodate the growing global population, root manipulation seems to hold the key toward sustainable food production. Villordon et al. (2014a) suggested that a paradigm shift toward RSA studies would enable a truly inclusive green revolution and allow food-insecure, resource-poor farmers who depend on RTCs in developing countries to also benefit. With this mindset, plant biologists, geneticists, and breeders have now shifted some focus toward studying of root traits. Due to the aforementioned complexity of studying roots under the soil, plant scientists are now set on finding minimally intrusive, non-destructive, whole-root system evaluating platforms. Hydroponics and gels are the most widely used systems to phenotype root systems (Jung and McCouch, 2013). Although, they offer a simple way to study different root traits and have given insight into root growth and development, both are controlled and do not represent actual field environments, and correlation of the findings from such experiments with actual performance of a plant in its natural environment are limited. To address this bottleneck, plant scientists are continually seeking to develop methods that will allow study of RSA in a more natural environment. Several methods have been proposed and applied in various studies including ‘shovelomics’ (Trachsel et al., 2011), soil coring (Wasson et al., 2014), rhizolysimeters (Eberbach et al., 2013) and minirhizotrons (Maeght et al., 2013), which are all soil-based. However, these methods are also low throughput, slow, and not amenable to large numbers of genotypes like those required for genome-wide association mapping studies. Image-based systems have also been developed and proposed to study roots in their natural environments, including X-ray computed tomography (Tracy et al., 2010) where x-rays are used to acquire 3-D cross-sectional images of the roots within the soil, Laser (Braga et al., 2009) which allows collection of bio-speckle patterns of gel-grown roots, nuclear magnetic resonance (NMR: Menzel et al., 2007), ground penetrating radar (GPR; Hirano et al., 2012), infra-red (IR) imaging (Dokken and Davis, 2007), and near-infra-red (NIR) imaging (Tirlapur and König, 1999), among others. However, application of some of these methods is still limited by the costs involved and to a few genotypes. Another bottleneck associated with imaging methods is image analysis. Several root image analysis platforms have been developed to address this limitation². With these large numbers of imaging and image analysis platforms, the need for sharing and use of data requires establishment of trait ontology across them to allow development of root ideotypes for different environments. Efforts by Lobet et al. (2015) to develop a unified root architecture development language are therefore right on time. This, combined with scaling up of the image analysis methods mentioned above, will be able to provide further knowledge required to adapt crops to their highly variable environments.

The increasing global population requires increased food production on the same or even less agricultural land as used currently, if the effects of climate change render some of the available marginal lands unfit for agricultural production. Most of the present and past crop improvement efforts have focused on above-ground traits to adapt crop plants to different production constraints. Although great progress has been made, and food production significantly increased, by manipulating above-ground traits, an estimated 800 million people are still food insecure, whereas yields, especially in cereal, have reached their yield potential and are plateauing in certain regions of the globe. It is therefore time for crop scientists to tap into un-explored and less exploited diversity within RSA traits to ensure rapid genetic gains, and stable and enhanced productivity of agricultural systems for future environmental conditions and climate change scenarios. Due to the quantitative nature of RSA traits and complex interaction of several underlying pathways that control them, response of RSA to multiple individual stresses or combination of stresses is variable. Modeling of the responses of root traits to multiple stress scenarios in a combination of high-throughput root-trait phenotyping techniques, alongside a robust database and data analytical pipeline, could be a way to go. This proposed strategy is applicable to all crops, but is more urgent in RTCs, as the second largest source of food security after cereals, mainly growing in marginal areas where many cereals cannot survive. Also, for RTCs, the harvestable organs are part of the RSA. It is recommended to increase focus on RSA research by investing more resources. RTCs can learn from what has been found so far in cereals and adopt some of their methods, while developing high-throughput techniques to quantify RSA traits under optimal and stressful conditions.

Review was conceptualized and written by MK, DG, and AV.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.