Kuijken et al., 2015

13 Pages·2015·0.59 MB·English

Journal of Experimental Botany, Vol. 66, No. 18 pp. 5389–5401, 2015
doi:10.1093/jxb/erv239 Advance Access publication 12 June 2015
REVIEW PAPER
Root phenotyping: from component trait in the lab to
breeding
René C.P. Kuijken1,2, Fred. A. van Eeuwijk3, Leo F.M. Marcelis4 and Harro J. Bouwmeester2,*
1 Wageningen UR, Greenhouse Horticulture, Wageningen, 6708 PB, The Netherlands
2 Wageningen UR, Laboratory of Plant Physiology, Wageningen, 6708 PB, The Netherlands D
3 Wageningen UR, Biometris, Wageningen, 6708 PB, The Netherlands ow
n
4 Wageningen UR, Horticulture and Product Physiology, Wageningen, 6708 PB, The Netherlands lo
a
d
e
d
* To whom correspondence should be addressed. Email: [email protected] fro
m
h
Received 20 January 2015; Revised 3 April 2015; Accepted 10 April 2015 ttp
://jx
b
Editor: Roland Pieruschka .ox
fo
rd
jo
u
Abstract rna
ls
.o
In the last decade cheaper and faster sequencing methods have resulted in an enormous increase in genomic data. rg
a/
High throughput genotyping, genotyping by sequencing and genomic breeding are becoming a standard in plant t U
n
breeding. As a result, the collection of phenotypic data is increasingly becoming a limiting factor in plant breeding. iv
e
Genetic studies on root traits are being hampered by the complexity of these traits and the inaccessibility of the rhizo- rs
ity
sphere. With an increasing interest in phenotyping, breeders and scientists try to overcome these limitations, resulting o
in impressive developments in automated phenotyping platforms. Recently, many such platforms have been thor- f W
oughly described, yet their efficiency to increase genetic gain often remains undiscussed. This efficiency depends on isc
o
n
the heritability of the phenotyped traits as well as the correlation of these traits with agronomically relevant breeding s
in
targets. This review provides an overview of the latest developments in root phenotyping and describes the environ- -M
a
mental and genetic factors influencing root phenotype and heritability. It also intends to give direction to future phe- dis
o
notyping and breeding strategies for optimizing root system functioning. A quantitative framework to determine the n
L
efficiency of phenotyping platforms for genetic gain is described. By increasing heritability, managing effects caused ib
ra
by interactions between genotype and environment and by quantifying the genetic relation between traits phenotyped rie
s
in platforms and ultimate breeding targets, phenotyping platforms can be utilized to their maximum potential. o
n
N
o
Key words: Genomic selection, heritability, phenotyping, rhizosphere, root exudation, root system architecture (RSA). ve
m
b
e
r 1
1
, 2
0
1
Introduction 5
World food production needs to double by 2050 to feed the measures will not be sufficient to reach the production aims
predicted population by that time (Tilman et al., 2011). At (Ray et al., 2013), creating the need for more rapid genetic
the same time, worldwide salinization and erosion of agricul- improvement of crop varieties that are also better adapted to
tural land and decline in phosphate fertilizer reserves pose future climate and agricultural management conditions. To
a threat to global yield increases. At this moment world- achieve an increased rate of genetic gain, new breeding tools
wide yields per hectare are far below their potential and and approaches are required. The introduction of new high-
intensification of agriculture by technology and knowledge throughput sequencing technologies provides such a tool. In
transfer are often preferred over conversion of ecologically the last decade, the costs of sequencing have dropped dra-
valuable land. Current yield improvements by technological matically from $800.00 per Mb of DNA sequence in 2008 to
© The Author 2015. Published by Oxford University Press on behalf of the Society for Experimental Biology. All rights reserved.
For permissions, please email: [email protected]
5390 | Kuijken et al.
$0.08 in 2014 (Wetterstrand, 2014). This decline in price and to grow a panel of individual plants, to acquire data and to
increase in throughput have resulted in tremendous amounts further process these data into quantitative morphological
of genomic data, creating many new opportunities, but at the or physiological traits. The second goal is to emphasize the
same time making phenotyping the major bottleneck in plant strong and complex influence of environmental and genetic
breeding and fundamental plant science. factors on root phenotype and heritability. We suggest how
Currently large public, private and academic efforts are these factors should be taken into consideration when devel-
aimed at increasing phenotyping capacity to keep up with the oping phenotyping strategies. The third goal is to offer a
developments in genotyping and sequencing. Rapid develop- quantitative framework to assess the efficiency of a pheno-
ments in robotics, information technology and automation typing platform for genetic improvement.
are reinforcing these efforts. In the past the majority of phe-
notyping efforts was focused on shoot traits such as yield,
shoot vigour, product quality and disease resistance, whereas Phenotyping platforms for root system
root phenotyping received less attention. The main reason for
architecture
this negligence is the technical difficulty in accessing the soil
when phenotyping root traits, especially by non-destructive Optimizing root system architecture (RSA) can facilitate Do
w
methods. higher crop yields (Wasson et al., 2012), especially in regions n
lo
The development of root phenotyping accelerated only of low-input agriculture or drought. The term RSA com- ad
e
rtieocnenintgly., Tahltihs ocurugchi atlh reo lreo oist ilslyussttermat eids bcryu tchiae l2 f0o–r5 0p%la notf ftuontca-l pasr isreoso at lbernogatdh ,r aronoget odfe nrsoiotyt, mrooorpt hborlaongcihcainl gp aarnadm etotetrasl sruocoht d from
fixed carbon that plants may translocate to their root system surface. The optimal phenotyping platform to quantify such http
(Lynch and Whipps, 1990; Kuzyakov and Domanski, 2000). In parameters is often a compromise between a range of desir- ://jx
addition, resistance to soil pathogens and pests, yield of root able platform properties. The priorities given to these com- b.o
crops, and nutrient and water uptake are root traits of great promised properties are best discussed in relation to three xfo
agronomic importance. These root traits are still relatively processes that constitute phenotyping: plant cultivation, data rdjo
u
easy to incorporate into breeding programmes. Phenotyping acquisition and data processing. For these three processes rn
a
for resistance against soil-borne pathogens is often done general desirable properties are (i) low developing and oper- ls
.o
through in vivo studies as is the case with Verticillium dahliae ating costs and (ii) the possibility to measure large numbers rg
(Bolek et al., 2005) Aphanomyces cochlioides, beet necrotic of individual replicates, genotypes or treatments with little at U/
yellow vein virus, Rhizoctonia solani and Pythium ultimum effort. Process-specific properties that may need to be com- niv
e
(Luterbacher et al., 2005). In vitro tests exist for resistance promised are discussed below. rs
against Colletotrichum graminicola (Planchamp et al., 2013), Specific desirable properties for a suitable cultivation sys- ity o
Pseudomonas syringae (Ishiga et al., 2011) and nematodes tem are agronomic relevance, the ability to grow plants in f W
(Williams et al., 2002; Haase et al., 2007b). Nutrient uptake is several developmental stages, the extent to which experi- isc
o
a trait that clearly affects the whole plant and can be pheno- mental noise can be reduced and the possibility and effort ns
in
typed by measuring shoot traits such as nutrient use efficiency to acquire good quality data. When agronomic relevance of -M
a
(NUE) (Moll et al., 1982; Good et al., 2004), relative growth the trait of interest is the main priority, soil is the preferred d
is
rate, and chlorophyll content (SPAD). Drought tolerance cultivation medium. Root systems grown in hydroponics, gel- on
L
can be phenotyped by analysing yield under arid conditions lan gum or soil can differ considerably (Hargreaves et al., ib
(Chapuis et al., 2012), turgor and a range of shoot imaging 2009; Wojciechowski et al., 2009; Clark et al., 2011) proving rarie
techniques (Berger et al., 2010) as proxy. Yield of root crops that artificial growth mediums are not always representative s o
n
can be quantified destructively. In conclusion, for these traits of field conditions. Strong nutrient, water, temperature and N
o
development of automated root phenotyping platforms is not soil strength gradients under field conditions complicate the ve
m
the main challenge as they can be measured relatively easily. mimicking of natural soil conditions or abiotic stresses under b
e
For root system architecture (RSA) and root exudation traits, controlled conditions, even in pots and rhizotrons (Whitmore r 1
1
however, large-scale phenotyping has only just started to and Whalley, 2009; Cairns et al., 2011). Nevertheless Hund , 2
0
1
emerge. The study of RSA traits, such as lateral root number et al. (2011) found consistent Quantitative Trait Locus (QTL) 5
and length, is complicated by the inaccessibility of the soil clusters for RSA across controlled and field environments.
matrix. Phenotyping of root exudation is hampered by bind- The most common practice of growing plants for root phe-
ing of exudates to the same soil matrix (Jones and Darrah, notyping in soil is the use of soil-filled tubes (Nagel et al.,
1994; Jones and Edwards, 1998; Kirk et al., 1999; Radersma 2009; Ytting et al., 2014), flat cartridges (Nagel et al., 2012;
and Grierson, 2004) and by degradation of exudates by the Dresboll et al., 2013) or cultivation in regular soil (Burton
microbial community around the root system (Kuijken et al., et al., 2012; Bucksch et al., 2014). A disadvantage of soil-
2015). based cultivation methods is soil heterogeneity, which aug-
The first goal of this review is to give an overview of the ments environmental noise. Indoor cultivation systems in
rapidly expanding number of root phenotyping platforms artificial medium can reduce environmental noise by allowing
and to pinpoint the challenges and common denominators a higher throughput and a more standardized micro-envi-
in the approaches used. The definition of phenotyping plat- ronment. The most commonly used method to reduce envi-
form employed here is the combination of hard and software ronmental noise and to easily observe the roots is to grow
Root phenotyping: from component trait to breeding | 5391
seedlings on agar plates (Nagel et al., 2009; Yazdanbakhsh (Mooney et al., 2012), creating a possible additional noise
and Fisahn, 2009; Slovak et al., 2014). Alternative 2D rhi- factor. An alternative imaging strategy is magnetic resonance
zotron systems are growth pouches (Hund et al., 2009; Adu imaging (MRI) (Rascher et al., 2011). Combining MRI with
et al., 2014), growth between paper (Le Marie et al., 2014) or positron emission tomography (PET) allows monitoring of
between fabric cloths in bins (Chen et al., 2011). The forced the distribution of photosynthetically fixed carbon in the
2D growth conformation and unnatural chemical and physi- root system (Jahnke et al., 2009; Nagel et al., 2009). However
cal properties of the medium come at the cost of agronomic these methods face the same shortcomings as X-ray CT.
relevance, making this the major shortcoming of these meth- These shortcomings combined make the current tomography
ods. 3D cultivation in agar or gellan gum (Fang et al., 2009; approaches less suitable for breeding purposes. A promis-
Iyer-Pascuzzi et al., 2010; Clark et al., 2011; Topp et al., 2013; ing method under development is biospeckle imaging, which
Ribeiro et al., 2014) or aqueous solution (Clark et al., 2013; reveals changes in heterogeneity of root tissue (Ribeiro et al.,
Pace et al., 2014) offers a hybrid solution between soil and 2D 2014). Ground-penetrating radar (GPR) and electrical resis-
strategies. The use of transparent soil is a promising and ele- tivity tomography (ERT) (Zenone et al., 2008) are at this
gant method that combines good visibility with a mechanical moment suitable techniques for imaging of the rhizosphere
D
resistance resembling sand (Downie et al., 2012). So far, most but need further development for a broad application in plant o
w
of the techniques developed for RSA phenotyping involve physiological and genetic studies. n
lo
the of use seedlings. Although there are examples in which After acquisition of one or several images of the mor- ad
e
the early stage root phenotype has some predictive value for phological state of a root system, the challenge is to process d fro
later developmental stages (Tuberosa et al., 2002a), the seed- these images into quantitative data. One of the things that m
ling root phenotype may not always be representative of the complicates extraction of a priori uncomplicated traits such http
mature plant (Watt et al., 2013). A cultivation system flex- as lateral root length and number is distinguishing overlap- ://jx
ible enough to allow growing plants in several developmental ping roots in 2D images in an automated way. This prob- b.o
x
stages will increase agronomic relevance. lem manifests itself typically with adult plants grown in 2D fo
Specific properties that may require a compromise when or 3D growth systems. In contrast, traits that are complex rdjo
u
developing the optimal data acquisition method are: reso- to quantify by the human eye such as total root area and rn
a
lution, processability of data, match with the desired culti- length can be imaged and processed fairly easily without ls
.o
vation system and possibility of measuring all desired root manual intervention since they do not require a distinction rg
a/
parameters. Many of the newly developed high-throughput between root zones and types. Data processing and extrac- t U
phenotyping platforms use 2D imaging with cameras (Nagel tion is preferably complete, fast, without errors and without niv
e
et al., 2009; Clark et al., 2013; Le Marie et al., 2014) or flatbed the need for human intervention. In the last decade a broad rs
scanners (Hund et al., 2009; Chen et al., 2011; Shi et al., 2013; range of software applications has been developed that meets ity o
Adu et al., 2014; Slovak et al., 2014). The main advantage of these demands to a greater or lesser extent. The major focus f W
2D growth and image acquisition is the very high throughput. points of these packages are hierarchical ordering of roots isc
o
An alternative but low-throughput 2D imaging strategy is 2D based on morphology, physiological function, topology, root ns
in
neutron radiography and tomography (Oswald et al., 2008; growth rate and anatomy. Image J (Schneider et al., 2012) -M
a
Moradi et al., 2009; Leitner et al., 2014). One of the major was one of the first open source image analysis packages that d
is
difficulties in root data acquisition is imaging overlapping allowed measurement of many of these traits. It offers full on
L
roots in complex and branched root systems in such a way control over data extraction but requires manual interven- ib
ra
that they can be distinguished easily. Untangling roots and tion. In the last five years, this program has been followed by rie
s
organizing them into 2D conformation is tedious and often an impressive number of (semi-) automated software pack- o
n
causes root damage. To solve this problem, root systems can ages (Table 1) (Lobet et al., 2013). Although many of these N
o
be grown in 3D in gel followed by 2D optical imaging with software packages are able to process images in a fully auto- ve
m
cameras and reconstruction of 2D images into 3D models mated way, manual correction of the image analysis and a b
e
(Clark et al., 2011; Topp et al., 2013). Through the use of check on processing quality is often necessary. Many of the r 1
1
laser scanners, 3D-grown roots in gel can be imaged in 3D listed software packages feature also ready to go plug-ins or , 2
0
1
(Fang et al., 2009). Several 3D-imaging methods allow in situ the ability to adapt or extend the software as the source code 5
visualization of root systems grown in soil without compro- has been made available.
mising temperature gradients, porosity, biochemical proper- The current wide range of customized RSA phenotyping
ties and mechanical resistance of natural soil. Most reported systems offers great choice and flexibility, yet also reveals
methods for data acquisition of 3D-grown roots in soil are plenty of overlap in research efforts. A combination of high-
based on X-ray computed tomography (CT), which has been throughput in vitro screens and verification of these screens
extensively reviewed by Mooney et al. (2012). Disadvantages using in situ 3D-MRI or tomography technology could com-
of this method are the potential loss of information due to a bine the best of both worlds in the medium term, i.e. in a
somewhat lower resolution and the complex 3D reconstruc- few years. In the long run, the ultimate phenotyping systems
tion of root systems (Mairhofer et al., 2013), the low through- should be flexible, fully automated and should integrate cul-
put, and the immobile and expensive equipment. Moreover tivation, image acquisition and data processing. Although
the automated 3D reconstruction of the root system from developments on integrated shoot phenotyping platforms
tomography frames is mostly based on modelling approaches are ahead of root phenotyping (Araus and Cairns, 2014),
5392 | Kuijken et al.
Table 1. Overview of available data processing software. Table adapted from http://www.plant-image-analysis.org (Lobet et al., 2013)
Software package Main root traits Automated Availability Publication
2D root images
Aria Convex-hull, count length, shape, Automated Freeware (Pace et al., 2014)
ARTT Localisation, orientation, velocity-profile Automated On-demand (Russino et al., 2013)
BRAT Diameter, length, orientation Automated Open-source, freeware (Slovak et al., 2014)
DART Insertion, length, topology, Manual Freeware (Le Bot et al., 2010)
DIRT Branching angle, density, diameter, Automated Open source, freeware (Bucksch et al., 2014)
distribution, length
ElonSim Length Automated Freeware (Benoit et al., 2014)
EZ-Rhizo Insertion, insertion-angle, length, nr. of Semi-automated Freeware (Armengaud et al., 2009)
branches, topology
GiA Roots Convex-hull, depth, length, nr. of branches, Automated Freeware (Galkovskyi et al., 2012)
perimeter, surface, volume D
o
GROW Map-Root Velocity profile Automated On-demand (Walter et al., 2002) w
n
GrowScreen-Root Insertion-angle, length, nr. of branches Semi-automated On demand (Nagel et al., 2009) lo
a
Growth Explorer Velocity-profile Automated On-demand (Basu and Pal, 2012) de
d
IJ-Rhizo Diameter, length Automated Open Source (Pierret et al., 2013) fro
KineRoot Gravitropism, growth Automated Freeware (Basu et al., 2007) m
h
PlantVis Velocity Automated On-demand (Roberts et al., 2010) ttp
RNQS Count, length, nodules, Automated Freeware (Remmler et al., 2014) ://jx
Root System Analyzer Count, diameter, insertion-angle, insertion, length Automated Freeware (Leitner et al., 2014) b.o
Root Tip Detection Count, diameter Automated Freeware (Kumar et al., 2014) xfo
RootFlowRT Growth, velocity-profile Automated Freeware (van der Weele et al., 2003) rd
jo
RootFly Colour, diameter, length, Manual Open source, freeware (Zeng et al., 2008) u
rn
RootNav Convex-hull, count, insertion, insertion-angle, Semi-automated Open source (Pound et al., 2013) als
length .o
rg
RootReader2D Depth, length, number of branches, topology, Semi-automated Freeware (Clark et al., 2013) a/
RootScape Shape Semi-automated Freeware (Ristova et al., 2013) t U
n
RootTip Trace Length Semi-automated Freeware (Geng et al., 2013) iv
e
RootTrace Curvature, length, number of branches Automated Open source, freeware (French et al., 2009) rs
ity
SmartRoot Diameter, insertion, insertion-angle, length, num- Semi-automated Freeware (Lobet et al., 2011) o
ber of branches, orientation, topology f W
Anatomy of cross-sections isc
o
Cell-O-Tape Cell number, cell-size Manual, Open source, freeware (French et al., 2012) ns
semi-automated in-M
CellSeT Cell geometry, cell number Automated Freeware (Pound et al., 2012) a
d
RootScan Cell number per tissue, tissue area, vessel area Automated Freeware (Burton et al., 2012) iso
n
Software dedicated to 3D root systems L
iRoCS Toolbox 3D reconstruction, diameter, localisation, volume Automated, Open source, freeware (Schmidt et al., 2014) ibra
semi-automated rie
s
RootReader3D Convex-hull, depth, insertion-angle, length, Automated On-demand (Clark et al., 2011) o
n
number of branches, orientation surface, N
o
v
volume, width e
m
RooTrak 3D reconstruction Semi-automated Freeware (Mairhofer et al., 2012) b
e
r 1
1
, 2
0
1
efforts to create an RSA phenotyping platform that inte- (Kuzyakov and Domanski, 2000), although lower (Jones 5
grates cultivation, data acquisition and processing do exist et al., 2004) and higher (Lynch and Whipps, 1990) shares
(Yazdanbakhsh and Fisahn, 2009). As we will describe later, have been reported. In any case, exudation represents a sig-
the ideal phenotyping platform causes little environmental nificant factor in the total carbon balance of the plant.
noise between genotypes, treatments or individual replicates Under highly controlled conditions, root exudation could be
and results in high heritability for the measured traits, which considered a loss whereas under field conditions exudation
are preferably agronomically relevant. could be considered an opportunity. Quantity and quality of
root exudates play a significant role in nutrient acquisition
in marginal soils (Radersma and Grierson, 2004; Hoffland
Phenotyping platforms for root exudation
et al., 2006; Richardson et al., 2009; Oburger et al., 2011),
As discussed above, plants translocate 20–50% of the pro- phytoremediation (Wenzel, 2009), avoidance of aluminium
duced assimilates to their root system. 10–18% of this large toxicity (Kochian et al., 2005; Ryan et al., 2009), generation
carbon investment is released into the soil by root exudation of electricity (Strik et al., 2011), production of proteins or
Root phenotyping: from component trait to breeding | 5393
pharmaceuticals (Borisjuk et al., 1999; Gontier et al., 2002) Alternative practices are the use of ion exchange resin (Shane
and the interaction with parasites (Bouwmeester et al., 2003), et al., 2008) or filter paper (Haase et al., 2007a). Aeroponics is
pathogens (Hartmann et al., 2009; Baetz and Martinoia, a useful technique to collect root exudates under high oxygen
2014), predators of these pathogens (van Tol et al., 2001) and availability (Kohlen et al., 2012).
growth promoting micro-organisms (Hartmann et al., 2009). The choice of data acquisition platform depends on
Many of these roles are specific to a single exudate or a chem- whether as many metabolites as possible in the root exudates
ical subgroup of exudates, which indicates the need for exu- need to be analysed or just a certain compound or chemi-
date identification and targeted quantification to exploit their cal class. In both cases taking an aqueous sample of the
potential for breeding. Elucidating the genetic background of rhizosphere is most convenient and low cost, making this
exudate composition and efflux rate allows the creation of approach by far the most popular (Neumann et al., 2009).
elite lines with favourable exudation profiles. However, phe- Exudate concentrations in such aqueous samples tend to be
notyping for root exudate composition has shown to be very low. Therefore, depending on the target trait and the analyti-
difficult so far. Despite these difficulties, claims of genetic cal platform used, samples have to be concentrated or puri-
variation in exudation have been made (Aulakh et al., 2001; fied. High-performance liquid chromatography (HPLC) and
D
Hoekenga et al., 2003; Yan et al., 2004; Hoffland et al., 2006; capillary electrophoresis (CE) are fast, low cost and low tech o
w
Gao et al., 2009; Jamil et al., 2011, 2012). Often used meas- platforms to quantify concentrations of organic acids and n
ures for genetic variation are genetic variance (σG2), coefficient carbohydrates. The limited number of compounds that can load
e
of genetic variation (CVg), coefficient of additive genetic vari- be analysed in a targeted way and the relatively low resolu- d fro
ation (CV ) or the phenotypic range of the measured trait. tion are disadvantages of these platforms. Platforms for high m
Of the meAntioned studies only Yan et al. (2004) use such resolution chemical analysis usually combine chromatogra- http
numeric measures to quantify the genetic variation in exuda- phy with mass spectrometry, as with liquid chromatography ://jx
tion. They report phenotypic ranges for total acid exudation coupled to mass spectrometry (LC-MS) (Strehmel et al., b.o
(TAE, µmol plant-1) and acid exudation rate (AER, μmol g-1 2014) and gas chromatography coupled to mass spectrometry xfo
h-1) of 3.60–20.60 and 1.65–64.00, respectively for Phaseolus (GC-MS). Nuclear magnetic resonance (NMR) is another rdjo
u
vulgaris. Although research on phenotyping of root exuda- option for high-throughput untargeted root exudate analysis rn
a
tion has been intensified, automated phenotyping platforms (Fan et al., 1997) although it has lower sensitivity than the ls
.o
for exudation have yet to be described. Parallel to phenotyp- MS-based methods. Mass spectrometry is also used for tar- rg
a/
ing for RSA, phenotyping for exudation is best discussed by geted analysis, especially of low abundant compounds such t U
three processes: plant cultivation, data acquisition and data as signalling molecules. For this purpose, triple-quad mass niv
e
processing, all having their specific challenges. spectrometry coupled to HPLC, or preferably UPLC, is supe- rs
Finding the optimal cultivation method for phenotyping rior. This is illustrated by the detection of ultralow concen- ity o
for exudation entails almost the same compromise between trations of strigolactones in root exudates (Sato et al., 2005; f W
desirable cultivation and data acquisition properties as for Lopez-Raez et al., 2008). An alternative analytical approach isc
o
RSA phenotyping. Two additional difficulties complicate to quantify total rhizodeposition is the use of carbon isotopes ns
in
plant cultivation for exudation phenotyping and have con- (Kuzyakov and Domanski, 2000). The ability to phenotype -M
a
flicting solutions. First of all, catabolism of exudates in non- exudation from intact root systems in a more natural rhizos- d
is
sterile rhizospheres (Kuijken et al., 2015) and the dynamic phere is the major advantage of this approach. However, car- on
L
adhesion of exudates to the soil (Radersma and Grierson, bon isotope studies are not high-throughput, are limited to ib
ra
2004) obscure net exudation quantity and quality. Secondly, untargeted exudate analysis and have low resolution. rie
s
the combined effects of the root environment on RSA (Clark Root exudate data acquired by targeted approaches such o
n
et al., 2011) and exudation (Veneklaas et al., 2003; Mimmo as HPLC or CE does not require complicated data analysis, N
o
et al., 2011) and the effect of RSA on exudation (Lambers in contrast to data from untargeted approaches. In untar- ve
m
et al., 2006) result in a complex three-way interaction mak- geted approaches metabolite identification is usually done b
e
ing exudation a complex and plastic process. Soil is therefore only after statistical analysis of the data (Hall, 2011). But r 1
1
a commonly used medium because it represents a growth even then identification is sometimes cumbersome, making , 2
0
1
medium that approximates natural growth environments metabolite identification the most important challenge in the 5
or field conditions and allows for the cultivation of large analysis of root exudates. Unfortunately, the LC-MS technol-
numbers of plants at a low price. Hydroponics is often used ogy, which is most suitable for the untargeted analysis of root
(Veneklaas et al., 2003; Hoffland et al., 2006; Pearse et al., exudates, suffers from low reproducibility between machines
2007) because it allows easy sample collection, reduces the and labs, and does not produce highly informative mass data
micro-environmental variation caused by micro-organisms in contrast to GC-MS. Besides, the large diversity of chemical
and the soil matrix, facilitates the concomitant study of structures and concentrations within a crude exudate sample
RSA and approximates hydroponic production methods in does not allow for capture of the entire exudate metabolome
glasshouse horticulture. Soil and hydroponics are occasion- with a single purification procedure and only one standard-
ally combined with sterilization agents (Imas et al., 1997; ized setting of the analytical platform. These issues discour-
Gherardi and Rengel, 2004; Gent et al., 2005), sterile cultiva- age the development of databases that can be used for analysis
tion (Paterson et al., 2005; Sandnes et al., 2005; Kuijken et al., on every machine. Nevertheless, public databases are being
2015) or microsuction cups (Dessureault-Rompré et al., 2006). developed such as the MoTo database of semi-polar tomato
5394 | Kuijken et al.
fruit metabolites (Moco et al., 2006), which offer a starting 2001), gravitropic set-point angle (Malamy, 2005) responsive-
point for exudate identification. Meanwhile, studies that phe- ness to the environment (Malamy, 2005) and a broad range
notype the root exudate metabolome are scarce but do exist of signal synthesis, signal transduction and signal sensitivity
(Strehmel et al., 2014). Identifying differentially exuded com- processes that involves many developmental genes (Giehl et al.,
pounds between genotypes and elucidating the biosynthetic 2014). These examples of endogenous factors result in a count-
pathways behind these compounds will help to bridge the gap less number of potential genotypic differences that obscure
between phenotypic exudation data and genetics. the effect of a single gene on a certain root phenotype. The
complexity of endogenous root phenotype regulation is even
further increased by pleiotropy and genetic interactions such
Environmental and genetic factors influencing root
as epistasis and crosstalk. An example of a possible pleiotropic
phenotype and heritability
effect is the Root-ABA1 QTL (Giuliani et al., 2005; Landi
Traits measured with a phenotyping platform may dis- et al., 2007), causing resistance to root lodging, increased leaf
play macro- and micro-environmental variation. Macro- ABA concentrations and lower yields. An example of cross-
environmental variation concerns all variation between runs talk between signal transduction pathways is the interac-
on a single platform for a particular genotype panel. Here tion between transcript levels of P responsive genes and low Do
w
a run is defined as a single measurement series on a single P-induced regulation of root development (Sanchez-Calderon n
lo
panel of individuals. Runs on a single platform under the et al., 2006). Examples of epistasis are the various reports of ad
e
sleavmele o fc ognefingoutyrpatiico mnse atnhsa rte vperaold iuncteer aloctwio nc obrertewlaeteionn gse naot tytphee i(nZthearancgt ientg  aQl.T, L20s 0f1o;r Zrohout terta aitls. ,i n2 0r0ic6e;, Bmoauizteei lalné de tA aral.b, i2d0o1p2si)s. d from
and environment (G×E). In such a case, high G×E implies In conclusion, numerous interactions between genes, between http
low reproducibility for the platform, which is an undesirable environmental factors, and between genes and environments ://jx
situation. In this case runs on the same platform, under the influence the very plastic root phenotype. These interactions b.o
same configurations, intended to be exact replicates, should obscure the contribution of individual genes to RSA and exu- xfo
be considered as separate environments. Different platform dation phenotypes, making breeding for these traits difficult. rdjo
configurations, plant nursing conditions or other deliberate urn
a
treatments between runs constitute a different type of macro- ls
.o
environmental variation. This type of variation does not cor- Assessing usability of a phenotyping rg
rupt reproducibility of a platform but can still cause G×E platform for genetic improvement at U/
effects between genotypes across runs. Micro-environmental n
iv
variation comprises differences between repeated samples of There are two main motivations for measuring root traits ers
individual genotypes within a run on a platform. High micro- on a phenotyping platform instead of in the target environ- ity
o
environmental variation will lead to low heritability, which is ment, i.e. the future growing conditions. The first motivation f W
elaborated in the next section. Both macro- and micro-envi- is a higher efficiency or increased heritability when the trait isc
o
ronmental factors can interact with a multitude of genetic of interest can be measured directly on the phenotyping plat- ns
in
factors, thereby undermining heritability of traits measured form. Higher efficiency can be achieved when a root trait is -M
on a platform and obscuring the relation between traits meas- hard to measure under future growing conditions. The phe- ad
is
ured on a phenotyping platform and the breeding targets in notyping platform may then be cheaper and less time- and on
L
the field. resource-consuming. A higher efficiency is also achieved when ib
Environmental factors known to influence RSA are tem- a phenotyping platform allows measurement of root traits at rarie
perature (Nagel et al., 2009), microbes (Lambers et al., 2009), an earlier developmental stage, which shortens the selection s o
n
water stress (Bengough et al., 2011), soil structure such as cycle and speeds up genetic improvement. A phenotyping N
o
soil strength (Cairns et al., 2011; Acuna and Wade, 2012) and platform may increase heritabilities by reducing noise levels. ve
m
porosity (Bengough et al., 2011) and availability of nutrients This is particularly beneficial when the future growing condi- b
e
such as nitrogen, phosphorus, iron and sulphate (Lopez- tions are such that genotypic differences are hard to detect. r 1
1
Bucio et al., 2003; Manschadi et al., 2014). Nutrient enriched The second motivation for using a phenotyping platform , 2
0
patches in the rhizosphere cause localized morphological is when the trait of interest is not the trait as measured at the 15
responses of the root system (Giehl et al., 2014; Hodge, 2004; platform itself, but a breeding target in the field, such as yield.
Forde and Walch-Liu, 2009). Environmental effects known to Such breeding targets may show low heritability under field
influence root exudation are soil type (Veneklaas et al., 2003; conditions because they are composed of many small genetic
Mimmo et al., 2011), nutrient availability (Paterson et al., factors that interact with many environmental factors. In
2006; Pearse et al., 2006; Yoneyama et al., 2007; Lopez-Raez the case of low heritability, measurements on a phenotyping
et al., 2008), aluminium toxicity (Ma, 2000; Kochian et al., platform of one or several component traits that contribute
2005) and microbes (Meharg and Killham, 1991, 1995). As to the breeding target might be a solution. Requirements for
mentioned before, effects of environmental factors on RSA this approach to be successful are a substantial heritability of
can lead to an indirect effect on exudation. the component trait and a high correlation with the breeding
Besides environmental cues, RSA and root exudation are target in the target environment. A set of exudates or a cer-
controlled by various endogenous factors such as hormone tain root morphology that contributes to yield under a stress
balance (Fukaki and Tasaka, 2009), plant age (Aulakh et al., condition can be an example of such an approach.
Root phenotyping: from component trait to breeding | 5395
The efficiency and thus relevance of a phenotyping plat- (replicates, plots) taken for a particular genotype within a run
form depends on two requirements. The first requirement or experiment.
is a sufficiently high heritability for the traits measured on
the phenotyping platform. The second requirement is a suf- σ2 σ2
ficiently high correlation between a component trait analysed H2 =σG2 =σ2 +(σ2 /e)G+(σ2 /e*n) (3)
on the platform and the target trait in the target environment. P G GE e
The direct and indirect selection response theory (Falconer
and Mackay, 1996) provides a convenient quantitative frame-
In the case of strong G×E effects, the heritability in a phe-
work to establish if these requirements are met. According to
notyping platform can be increased by testing genotypes in
this theory we need to evaluate the criterion in Eq. 1:
an increased number of runs which reduces the influence of
the G×E variance σ2 in the overall phenotypic variance (Eq.
GE
HC2rC2T >HT2 (1) 3). The same is true for the breeding target in the target envi-
ronment, for which the influence of the G×E interaction is
This equation describes the relation between the heritabil-
reduced by running more experiments.
ity for the trait measured at the platform H2 (between 0 and D
C With the simplified quantitative framework as described ow
1); the heritability of the target trait in the target environment n
above, we can see that it is of no use to measure a platform trait lo
H2 (between 0 and 1) and the squared genetic correlation a
T with high precision, i.e. low noise level, when the (squared) de
b1)e.t wMeeeent icnogm tphoisn cernitt etrriaoitn a mnde atnarsg tehta ttr ainitd rCi2rTe c(tb estewleecetnio −n 1f oarn da genetic correlation rC2T between platform trait and target trait d from
in the target environment is low (Eq. 1). The problem of meas-
component trait at the platform is more effective than direct uring a platform trait with high precision and low r2 will http
selection for the target trait in the target environment. be more severe in the situation where the platform trCaTit is a ://jx
In the case of a single run on a phenotyping platform or a b
component trait than when it is a target trait. Although many .o
single experiment in the target environment, the heritability studies on automated phenotyping platforms do not mention xfo
of component trait and breeding target is estimated by the or quantify the genetic correlation between platform traits rdjo
ratio of genetic to phenotypic variance as described in Eq. 2: and target traits, many such relations have been described. urna
ls
Drought tolerance for example is an important breeding tar- .o
H2 =σσG22 =σ2 +σ(σG22 /n) (2) gsheot wfonr tmo abney ac rcoopms pionn tehnet twroairtl da.n Dd etehpuesr prroeodtiicntgiv eh afos rb teheins at Urg/
P G e n
breeding target (Tuberosa, 2012). The DRO1 locus alters root iv
e
with σG2= genotypic variance; σ2P = phenotypic variance; angle and rooting depth and is predictive for yield perfor- rsity
σ2 = error variance on a plot basis and n the number of inde- mance during drought (Uga et al., 2013). Another important o
e f W
pendent measurements (replicates, plots) taken for a partic- breeding target is yield in nutrient-poor soils. Shallower and
is
ular genotype. A phenotyping platform could be attractive more branching roots result in improved nutrient acquisition co
n
because it allows for more precise measurements, thus lower- in the topsoil layer (Lynch and Brown, 2012). Also exuda- sin
ing the micro-environmental variance σ2 and simultaneously tion of organic acids contributes to nutrient acquisition in -M
e a
d
allowing for a larger number of replicates n to be measured. poor soils (Dakora and Phillips, 2002; Oburger et al., 2011). is
o
Together this results in increased heritability. Breeding for increased shoot height or biomass can lead to n L
In the case of multiple runs on a platform or multiple increased susceptibility to root lodging (van Delden et al., ibra
experiments for the breeding target in the target environment, 2010). Root systems with high exploitation indices contribute rie
s
we need to take into account G×E variation across different to the breeding target root lodging tolerance (Giuliani et al., on
N
runs and/or experiments. An example of G×E for a root trait 2005). Mace et al. (2012) report associations between nodal o
v
is rooting depth of wheat genotypes in soils with different soil root angle QTLs and the stay-green trait and grain yield. Low em
b
strengths (Acuna and Wade, 2012, 2013). Examples of indi- strigolactone exudation was shown to contribute to the breed- e
r 1
vidual genes that show interaction with the environment are ing target Striga resistance (Jamil et al., 2011). Several studies 1
, 2
LPR1 (Camacho-Cristobal et al., 2008) causing a differential report correlations between root QTLs and yield (Tuberosa 0
1
5
reaction to P stress, and GmEXPB2 responding to Fe, P and et al., 2002b; Hund et al., 2011; Cai et al., 2012). The magni-
water stress (Guo et al., 2011). The question then is how to tude of genetic correlation r2 depends on the commonality
CT
define the platform and target environment? A common solu- in QTLs or underlying genes that drive phenotypic variation
tion is to take a weighted average across runs or experiments. in component traits and breeding targets.
Heritabilities can also differ across experiments as shown for It needs to be remarked that even when the genetic cor-
depth penetration rates of roots in winter wheat (Ytting et al., relation between a component trait in the platform and the
2014). Heritability for a component trait across multiple breeding target in the target environment is low, identifying
runs on the same platform, or for the breeding target across component trait QTLs is not completely useless. As long as
multiple experiments is described in Eq. 3 with σ2 = geno- the direct contribution of component QTLs to the target trait
G
typic variance across runs or experiments; σ2 = phenotypic can be shown they can be used in marker-assisted selection.
P
variance across runs or experiments; σ2  = G×E variance; e Ideally, such direct contributions of the component QTLs to
GE
the number of runs or experiments σ2 = error variance on a target traits prove to be based on causal relations instead of
e
plot basis and n the number of independent measurements correlations. For example Hund et al. (2011) report about a
5396 | Kuijken et al.
consensus map of RSA QTLs from 17 QTL mapping studies worthwhile to consider a phenotyping platform environment
in maize and grouped 161 QTLs in 24 clusters. These meta- as a training situation for fitting a genomic prediction model,
data showed that several QTLs with an effect on axile and lat- while the target trait in the target environment constitutes the
eral roots co-located, pointing towards QTLs for root system selection environment. In other words, the marker effects as
size instead of a specific root architecture. Such a meta-anal- estimated on the platform may be used to predict the target
ysis demonstrates the possibility of a correlation based on the trait in the target environment, under the condition that scal-
effects of yield on roots instead of roots on yield. In such ing issues are taken into account. Such predictions based on
case, selection for component traits will result in selection for marker effects may also be applicable for standard multi-QTL
pleiotropic genes or undesired physiological traits. Testing the models.
correlations between root traits and plant size, as has been Despite the fast development of selection strategies based
done for nodal root angle and root dry weight (Singh et al., on genomics, phenotyping will continue to play an important
2011), could prevent some of such problems. The common role in breeding. Although selection by GS is not directly
practice of root phenotyping on seedlings outside the target based on phenotyping, GS still relies on a relationship
environment represents another example in which the impor- between phenotype and genetic markers in the training popu-
D
tance of a direct contribution of a component QTL needs lation. Furthermore, GEBVs might become more accurate in o
w
to be emphasized. The combined environmental and plant- the near future when identified genes with known effects are n
lo
developmental gap between component and target traits easier to include in GS prediction models. Although we can ad
e
represents a double risk of reduced relevance of a compo- breed great crop varieties without cloning genes or annotat- d fro
nent trait. These examples demonstrate some of the pitfalls ing the genome, our fundamental understanding of plants m
in component trait selection and show that insight in causal will come to a stop if we stick with GS or comparable tools http
relations between component and target traits is required for and neglect functional genomics through forward genetics. ://jx
successful improvement of the target trait. This further underlines the continued need for improved phe- b.o
x
notyping facilities in plant science. fo
rd
Prospects for root phenotyping jo
u
Conclusions rna
So far QTL-mapping and genome-wide association stud- ls
.o
ies (GWAS) have proved to be valuable tools to improve the The number and speed of developments in automated phe- rg
a/
root system by yielding a broad range of root QTLs. Some notyping are impressive, and as a consequence now a broad t U
of these QTLs have been successfully introgressed in crops; range of root phenotyping platforms is available. The costly niv
e
examples are the QTLs for deep root mass (Shen et al., 2001) and labour-intensive development of phenotyping plat- rs
and root length (Steele et al., 2006) in near isogenic lines of forms should not be a goal in itself. Instead, these platforms ity o
rice. However, for many breeding programmes it is often not should support plant physiological research and breeding for f W
worth the effort to fine-map QTLs as their effects tend to be agronomically relevant breeding targets. Numerous authors isc
o
n
small and prone to G×E interactions. If these QTLs are con- describe their root phenotyping platform aiming to make s
in
nected to a component trait for yield, then their effects on the it a universal standard. In practice, such a standard is hard -M
a
ultimate breeding target might be even smaller. When iden- to establish since every research or breeding goal requires a d
is
o
tifying small-effect QTLs is difficult, an alternative to QTL different phenotyping approach, leading to many new phe- n
L
mapping or GWAS could be genomic selection (GS). notyping platforms. The large number of phenotyping plat- ib
ra
GS is based on a genomic estimated breeding value forms creates another issue: a lack of standardization in data rie
s
(GEBV), modelled from genetic relations between individu- storage. This deficiency prevents large-scale data exchange o
n
als in training and breeding or selection populations (Crossa between root researchers and the comparison of phenotyping N
o
et al., 2011; Desta and Ortiz, 2014). Phenotype and geno- platforms. In an effort to overcome this issue the XML-based ve
m
type information of the training population is used to esti- root system markup language (RSML) was developed (Lobet b
e
mate marker effects. These marker effects are then used to et al., 2015). Another omission in the use of root phenotyping r 1
1
select on the basis of markers alone in the breeding popula- platforms is that a description of G×E effects and reproduc- , 2
0
1
tion, i.e. without phenotyping. GS allows simultaneous esti- ibility of results is often missing. G×E effects are very likely 5
mation of many large and small marker effects without any to occur, since the root system is very plastic and responds
prior knowledge about their effect or function. In relation to strongly to environmental influences. It is therefore important
phenotyping and breeding for root traits, the added values of to quantify G×E effects when developing phenotyping plat-
GS are (i) better exploitation of minor genes, (ii) the ability forms and breeding strategies for improved root traits.
to phenotype only a subset of a population and to predict The choice for a phenotyping platform depends on the
the rest and (iii) the ability of pre-selection of a population breeding target. This breeding target can be the same trait
prior to phenotyping in the target environment, which pre- as measured on the platform, or the platform may measure a
vents phenotyping lines that are a priori known to be inferior. component trait of a breeding target in the field. The relation
The strategies as mentioned in (ii) and (iii) reduce phenotyp- between platform traits and field traits, combined with their
ing costs and efforts, which is especially helpful when traits heritabilities, determines the efficiency of a phenotyping plat-
are difficult to quantify. With the advent of high density form for breeding. Therefore we could reason that before we
SNP panels and genotyping by sequencing, it may even be invest in the creation of yet another phenotyping platform,
Root phenotyping: from component trait to breeding | 5397
we need to determine its added value for the heritability and Bouteillé M, Rolland G, Balsera C, Loudet O, Muller B. 2012.
Disentangling the intertwined genetic bases of root and shoot growth in
the contribution of the platform traits to the breeding targets
Arabidopsis. Plos One 7 doi: 10.1371/journal.pone.0032319
in the field. But without the phenotyping platform itself, these
Bouwmeester HJ, Matusova R, Sun ZK, Beale MH. 2003. Secondary
aspects are hard to determine. So we should continue devel- metabolite signalling in host-parasitic plant interactions. Current Opinion in
oping phenotyping platforms while keeping a keen eye on its Plant Biology 6, 358–364.
efficiency during development. Bucksch A, Burridge J, York LM, Das A, Nord E, Weitz JS, Lynch JP.
2014. Image-based high-throughput field phenotyping of crop roots. Plant
Choosing a breeding strategy largely depends on the herita-
Physiology 166, 470–486.
bility of the trait of interest. The fact that roots are difficult to
Burton AL, Williams M, Lynch JP, Brown KM. 2012. RootScan:
phenotype means that cloned genes or markers closely linked Software for high-throughput analysis of root anatomical traits. Plant and
to root traits can help avoiding expensive or laborious pheno- Soil 357, 189–203.
typing methods. To breed for optimized roots, QTL mapping Cai H, Chen F, Mi G, Zhang F, Maurer HP, Liu W, Reif JC, Yuan L.
2012. Mapping QTLs for root system architecture of maize (Zea mays
and GWAS combined with MAS will often be preferred with
L.) in the field at different developmental stages. Theoretical and Applied
regard to discrete large gene effects. The use of genomic selec- Genetics 125, 1313–1324.
tion will often be preferred with regard to many small gene Cairns JE, Impa SM, O’Toole JC, Jagadish SVK, Price AH. 2011.
D
effects. QTL-mapping, GWAS and functional genomics do Influence of the soil physical environment on rice (Oryza sativa L.) response o
to drought stress and its implications for drought research. Field Crops w
not have to compete with GS, but instead they can support n
Research 121, 303–310. lo
it by providing input for GS models. With increasing com- ad
Camacho-Cristobal JJ, Rexach J, Conejero G, Al-Ghazi Y, Nacry e
pisl eax citoyn otifn uGoSu sm noedeedl sf oarn dm oavrea ialanbdi liitnyc roefa sgienngolym sioc pdhaitsati cthateerde Pin,v Doloveudm ian sr oPo. t2 a0r0ch8i.t ePcRtuDr,a al cnh Aarnagbeidso inp sriess ApIoNnTsEeG toU MphEoNsTpAh-altikee s gtaernvea,t ioisn . d from
phenotyping platforms. Planta 228, 511–522. http
CRehsailpieunicse Rs ,t oD ewlalutecr Cd,e fiDceitb ien uaf pRh,e Tnaortdypieinug F p, lWatefolcrmke arn Cd .i n2 0th1e2 .fi eld: ://jx
b
How related are they in maize? European Journal of Agronomy 42, .o
x
References 59–67. fo
Acuna TLB, Wade LJ. 2012. Genotype × environment interactions for CDehveenlo YpLm, eDnut nofb aa bnionv eVlM se, mDii-ghgylder oApJo,n Sicid pdhieqnuoet yKpHinMg s, yRsetenmge fol rZ s. t2u0d1y1in.g rdjou
root depth of wheat. Field Crops Research 137, 117–125. root architecture. Functional Plant Biology 38, 355–363. rna
Acuna TLB, Wade LJ. 2013. Use of genotype × environment interactions Clark RT, Famoso AN, Zhao KY, Shaff JE, Craft EJ, Bustamante CD, ls.o
tsoo iulsn. dAenrnstaalsn do fr oBoottinagny d 1e1p2th, 3a5n9d– t3h6e8 a.bility of wheat to penetrate hard MtwcoC-doimucehn sSioRn,a Al rnoeost hsaysntselmey p DheJn, oKtoypcihngia pnl aLtVfo. r2m0 1fa3c. iHlitiagthe-st hgreonuegthicp ut at Urg/
Adu MO, Chatot A, Wiesel L, Bennett MJ, Broadley MR, White PJ, analysis of root growth and development. Plant, Cell & Environment 36, n
Dupuy LX. 2014. A scanner system for high-resolution quantification of 454–466. ive
vEaxrpiaetriiomne innt arol oBto gtaronwy t6h5 d, y2n0a3m9–ic2s0 o4f8 B.rassica rapa genotypes. Journal of CAnlaerskh RaTn,s Mleay cDCJu, rKdoy cRhBia, nJ uLVn.g 2 J0K11, .S Thharfefe J-Edi,m MecnsCioonuacl hro SoRt , rsity o
Araus JL, Cairns JE. 2014. Field high-throughput phenotyping: the new phenotyping with a novel imaging and software platform. Plant Physiology f W
crop breeding frontier. Trends in Plant Science 19, 52–61. 156, 455–465. is
c
o
Armengaud P, Zambaux K, Hills A, Sulpice R, Pattison RJ, Blatt Crossa J, Pérez P, de los Campos G, Mahuku G, Dreisigacker n
s
MR, Amtmann A. 2009. EZ-Rhizo: integrated software for the fast and S, Magorokosho C. 2011. Genomic selection and prediction in plant in
accurate measurement of root system architecture. Plant Journal 57, breeding. Journal of Crop Improvement 25, 239–261. -M
a
945–956. Dakora FD, Phillips DA. 2002. Root exudates as mediators of mineral dis
Aulakh MS, Wassmann R, Bueno C, Kreuzwieser J, Rennenberg H. acquisition in low-nutrient environments. Plant and Soil 245, 35–47. on
2001. Characterization of root exudates at different growth stages of ten Dessureault-Rompré J, Nowack B, Schulin R, Luster J. 2006. Lib
rice (Oryza sativa L.) cultivars. Plant Biology 3, 139–148. Modified micro suction cup/rhizobox approach for the in-situ detection ra
Baetz U, Martinoia E. 2014. Root exudates: the hidden part of plant of organic acids in rhizosphere soil solution. Plant and Soil 286, ries
defense. Trends in Plant Science 19, 90–98. 99–107. on
Basu P, Pal A. 2012. A new tool for analysis of root growth in the spatio- Desta ZA, Ortiz R. 2014. Genomic selection: genome-wide prediction in No
temporal continuum. New Phytologist 195, 264–274. plant improvement. Trends in Plant Science 19, 592–601. ve
m
Basu P, Pal A, Lynch JP, Brown KM. 2007. A novel image-analysis Downie H, Holden N, Otten W, Spiers AJ, Valentine TA, Dupuy b
e
technique for kinematic study of growth and curvature. Plant Physiology LX. 2012. Transparent soil for imaging the rhizosphere. Plos One 7 doi: r 1
145, 305–316. 10.1371/journal.pone.0044276 1, 2
Bengough AG, McKenzie BM, Hallett PD, Valentine TA. 2011. Root Dresboll DB, Thorup-Kristensen K, McKenzie BM, Dupuy LX, 01
5
elongation, water stress, and mechanical impedance: a review of limiting Bengough AG. 2013. Timelapse scanning reveals spatial variation in
stresses and beneficial root tip traits. Journal of Experimental Botany 62, tomato (Solanum lycopersicum L.) root elongation rates during partial
59–68. waterlogging. Plant and Soil 369, 467–477.
Benoit L, Rousseau D, Belin E, Demilly D, Chapeau-Blondeau F. Falconer DS, Mackay TFC. 1996. Introduction to Quantitative Genetics .
2014. Simulation of image acquisition in machine vision dedicated to Harlow, UK: Longman.
seedling elongation to validate image processing root segmentation Fan TWM, Lane AN, Pedler J, Crowley D, Higashi RM. 1997.
algorithms. Computers and Electronics in Agriculture 104, 84–92. Comprehensive analysis of organic ligands in whole root exudates using
Berger B, Parent B, Tester M. 2010. High-throughput shoot imaging to nuclear magnetic resonance and gas chromatography mass spectrometry.
study drought responses. Journal of Experimental Botany 61, 3519–3528. Analytical Biochemistry 251, 57–68.
Bolek Y, El-Zik KM, Pepper AE, Bell AA, Magill CW, Thaxton PM, Fang SQ, Yan XL, Liao H. 2009. 3D reconstruction and dynamic
Reddy OUK. 2005. Mapping of verticillium wilt resistance genes in cotton. modeling of root architecture in situ and its application to crop phosphorus
Plant Science 168, 1581–1590. research. The Plant Journal 60, 1096–1108.
Borisjuk NV, Borisjuk LG, Logendra S, Petersen F, Gleba Y, Raskin Forde BG, Walch-Liu P. 2009. Nitrate and glutamate as environmental
I. 1999. Production of recombinant proteins in plant root exudates. Nature cues for behavioural responses in plant roots. Plant Cell and Environment
Biotechnology 17, 466–469. 32, 682–693.
5398 | Kuijken et al.
French A, Ubeda-Tomas S, Holman TJ, Bennett MJ, Pridmore T. Hund A, Trachsel S, Stamp P. 2009. Growth of axile and lateral roots
2009. High-throughput quantification of root growth using a novel image- of maize: I. Development of a phenotying platform. Plant and Soil 325,
analysis tool. Plant Physiology 150, 1784–1795. 335–349.
French AP, Wilson MH, Kenobi K, Dietrich D, Voss U, Ubeda-Tomas Imas P, BarYosef B, Kafkafi U, GanmoreNeumann R. 1997.
S, Pridmore TP, Wells DM. 2012. Identifying biological landmarks using Phosphate induced carboxylate and proton release by tomato roots. Plant
a novel cell measuring image analysis tool: Cell-o-Tape. Plant Methods 8:7 and Soil 191, 35–39.
doi: 10.1186/1746-4811-8-7
Ishiga Y, Ishiga T, Uppalapati SR, Mysore KS. 2011. Arabidopsis
Fukaki H, Tasaka M. 2009. Hormone interactions during lateral root seedling flood-inoculation technique: a rapid and reliable assay
formation. Plant Molecular Biology 69, 437–449. for studying plant-bacterial interactions. Plant Methods 7:32 doi:
Galkovskyi T, Mileyko Y, Bucksch A, et al. 2012. GiA Roots: software 10.1186/1746-4811-7-32
for the high throughput analysis of plant root system architecture. BMC Iyer-Pascuzzi AS, Symonova O, Mileyko Y, Hao YL, Belcher H,
Plant Biology 12:116 doi: 10.1186/1471-2229-12-116 Harer J, Weitz JS, Benfey PN. 2010. Imaging and analysis platform
Gao XP, Zhang FS, Hoffland E. 2009. Malate exudation by six aerobic for automatic phenotyping and trait ranking of plant root systems. Plant
rice genotypes varying in zinc uptake efficiency. Journal of Environmental Physiology 152, 1148–1157.
Quality 38, 2315–2321. Jahnke S, Menzel MI, van Dusschoten D, et al. 2009. Combined MRI-
Geng Y, Wu R, Wee CW, Xie F, Wei X, Chan PMY, Tham C, Duan PET dissects dynamic changes in plant structures and functions. The Plant
L, Dinneny JR. 2013. A spatio-temporal understanding of growth Journal 59, 634–644.
D
regulation during the salt stress response in Arabidopsis. Plant Cell 25, Jamil M, Charnikhova T, Houshyani B, van Ast A, Bouwmeester H. o
2132–2154. 2012. Genetic variation in strigolactone production and tillering in rice and wn
Gent MPN, Parrish ZD, White JC. 2005. Nutrient uptake among its effect on Striga hermonthica infection. Planta 235, 473–484. loa
d
subspecies of Cucurbita pepo L. is related to exudation of citric acid. Jamil M, Rodenburg J, Charnikhova T, Bouwmeester HJ. 2011. Pre- ed
Journal of the American Society for Horticultural Science 130, 782–788. attachment Striga hermonthica resistance of New Rice for Africa (NERICA) fro
Gherardi MJ, Rengel Z. 2004. The effect of manganese supply on cultivars based on low strigolactone production. New Phytologist 192, m
exudation of carboxylates by roots of lucerne (Medicago sativa). Plant and 964–975. http
SGoieil h2l6 R0F, H27, 1G–r2u8b2e.r BD, von Wiren N. 2014. Its time to make changes: Jthoen meso bDilLiz,a Dtioanr roafh n PutRrie. n1t9s9 f4ro. mRo tlhee o rfh riozoost pdheerirvee.d P olarngta annicd- aScoidil s1 i6n6 , ://jxb
modulation of root system architecture by nutrient signals. Journal of 247–257. .ox
Experimental Botany 65, 769–778. Jones DL, Edwards AC. 1998. Influence of sorption on the biological fo
rd
Giuliani S, Sanguineti MC, Tuberosa R, Bellotti M, Salvi S, Landi utilization of two simple carbon substrates. Soil Biology & Biochemistry 30, jo
P. 2005. Root-ABA1, a major constitutive QTL, affects maize root 1895–1902. urn
architecture and leaf ABA concentration at different water regimes. Journal Jones DL, Hodge A, Kuzyakov Y. 2004. Plant and mycorrhizal als
of Experimental Botany 56, 3061–3070. regulation of rhizodeposition. New Phytologist 163, 459–480. .org
Gontier E, Clement A, Tran TLM, Gravot A, Lievre K, Guckert A, Kirk GJD, Santos EE, Santos MB. 1999. Phosphate solubilization by a/
Bourgaud F. 2002. Hydroponic combined with natural or forced root organic anion excretion from rice growing in aerobic soil: rates of excretion t U
permeabilization: a promising technique for plant secondary metabolite and decomposition, effects on rhizosphere pH and effects on phosphate niv
production. Plant Science 163, 723–732. solubility and uptake. New Phytologist 142, 185–200. ers
Good AG, Shrawat AK, Muench DG. 2004. Can less yield more? Is Kochian LV, Pineros MA, Hoekenga OA. 2005. The physiology, ity
rcerdopuc pinrogd nuucttriioenn?t iTnrpeuntd isn tion tPhlea netn Svicroiennmcee n9t, c5o9m7–p6a0ti5b.le with maintaining gPelannett iacnsd a Sndo ilm 2o7l4e,c u1l7a5r –b1io9l5o.gy of plant aluminum resistance and toxicity. of W
is
Guo W, Zhao J, Li X, Qin L, Yan X, Liao H. 2011. A soybean beta- Kohlen W, Charnikhova T, Lammers M, et al. 2012. The tomato co
erexsppaonnssine sg eton ea bGiomtiEcX sPtrBe2ss ienstr.i nTshicea Pllyla inntv Joolvuerdn ainl 6ro6o, t5 s4y1s–t5e5m2 .architecture CrhAizRoOspThEeNreO sIDig nCaLliEnAgV, ApGlanEt DaIrOchXiYteGcEtuNreA SanEd8 a(SffleCcCtsD r8e)p rreogduulactteivse nsin-M
Haase S, Neumann G, Kania A, Kuzyakov Y, Romheld V, Kandeler development through strigolactone biosynthesis. New Phytologist 196, ad
E. 2007a. Elevation of atmospheric CO2 and N-nutritional status modify 535–547. is
o
nodulation, nodule-carbon supply, and root exudation of Phaseolus Kuijken RCP, Snel JFH, Heddes MM, Bouwmeester HJ, Marcelis n L
vulgaris L. Soil Biology & Biochemistry 39, 2208–2221. LFM. 2015. The importance of a sterile rhizosphere when phenotyping for ib
Haase S, Ruess L, Neumann G, Marhan S, Kandeler E. 2007b. Low- root exudation. Plant and Soil 387, 131–142. rarie
level herbivory by root-knot nematodes (Meloidogyne incognita) modifies Kumar P, Huang C, Cai J, Miklavcic SJ. 2014. Root phenotyping by s o
rPolaont th aanird m Soorpil h3o0l1o,g 1y 5a1n–d1 r6h4iz.odeposition in host plants (Hordeum vulgare). rSoooilt 3ti8p0 d, e1t9e3c–tio2n0 9a.nd classification through statistical learning. Plant and n No
v
Hall RD. 2011. The Biology of Plant Metabolomics. Annual Plant Reviews Kuzyakov Y, Domanski G. 2000. Carbon input by plants into the em
Vol. 43 . Wiley-Blackwell. soil. Review. Journal of Plant Nutrition and Soil Science-Zeitschrift Fur be
Hargreaves CE, Gregory PJ, Bengough AG. 2009. Measuring root Pflanzenernahrung Und Bodenkunde 163, 421–431. r 1
1
traits in barley (Hordeum vulgare ssp. vulgare and ssp. spontaneum) Lambers H, Mougel C, Jaillard B, Hinsinger P. 2009. Plant-microbe- , 2
seedlings using gel chambers, soil sacs and X-ray microtomography. Plant soil interactions in the rhizosphere: an evolutionary perspective. Plant and 01
5
and Soil 316, 285–297. Soil 321, 83–115.
Hartmann A, Schmid M, van Tuinen D, Berg G. 2009. Plant-driven Lambers H, Shane MW, Cramer MD, Pearse SJ, Veneklaas EJ. 2006.
selection of microbes. Plant and Soil 321, 235–257. Root structure and functioning for efficient acquisition of phosphorus:
Hodge A. 2004. The plastic plant: root responses to heterogeneous matching morphological and physiological traits. Annals of Botany 98,
supplies of nutrients. New Phytologist 162, 9–24. 693–713.
Hoekenga OA, Vision TJ, Shaff JE, Monforte AJ, Lee GP, Howell Landi P, Sanguineti MC, Liu C, Li Y, Wang TY, Giuliani S, Bellotti
SH, Kochian LV. 2003. Identification and characterization of aluminum M, Salvi S, Tuberosa R. 2007. Root-ABA1 QTL affects root lodging,
tolerance loci in Arabidopsis (Landsberg erecta × Columbia) by quantitative grain yield, and other agronomic traits in maize grown under well-
trait locus mapping. A physiologically simple but genetically complex trait. watered and water-stressed conditions. Journal of Experimental Botany
Plant Physiology 132, 936–948. 58, 319–326.
Hoffland E, Wei CZ, Wissuwa M. 2006. Organic anion exudation by Le Bot J, Serra V, Fabre J, Draye X, Adamowicz S, Pages L. 2010.
lowland rice (Oryza sativa L.) at zinc and phosphorus deficiency. Plant and DART: a software to analyse root system architecture and development
Soil 283, 155–162. from captured images. Plant and Soil 326, 261–273.
Hund A, Reimer R, Messmer R. 2011. A consensus map of QTLs Le Marie C, Kirchgessner N, Marschall D, Walter A, Hund A. 2014.
controlling the root length of maize. Plant and Soil 344, 143–158. Rhizoslides: paper-based growth system for non-destructive, high