Opinion
Challenges in breeding for yield
increase for drought
Thomas R. Sinclair
Crop Science Department, North Carolina State University, Raleigh, NC, USA
Crop genetic improvement for environmental stress at
the molecular and physiological level is very complex
and challenging. Unlike the example of the current major
commercial transgenic crops for which biotic stress
tolerance is based on chemicals alien to plants, the
complex, redundant and homeostatic molecular and
physiological systems existing in plants must be altered
for drought tolerance improvement. Sophisticated tools
must be developed to monitor phenotype expression at
the crop level to characterize variation among genotypes
across a range of environments. Once stress-tolerant
cultivars are developed, regional probability distribu-
tions describing yield response across years will be
necessary. This information can then aid in identifying
environmental conditions for positive and negative
responses to genetic modification to guide farmer selec-
tion of stress-tolerant cultivars.
Genetic challenge of drought stress
There have been tremendous advances in understanding
the physiology, biochemistry and molecular genetics of
plant response to drought in the past half-century. Increas-
ing numbers of papers are being published reporting new
insights about the factors contributing to increased plant
development and growth rates [1]. Yet, this information
has had little or no impact to date in improving the
intrinsic behavior of plants resulting in increased crop
yield [2–5]. For example, the two highly visible, successful
transgenic crops – glyphosate tolerance and Bacillus thur-
ingiensis (Bt) insect tolerance – ‘defend’ against pest-me-
diated yield loss, but do not address fundamental, intrinsic
plant performance that would lead to improved crop re-
sponse to yield-limiting abiotic factors. Indeed, these two
commercial transgenes do not rely on acceleration or sup-
pression of fundamental biochemical or physiological path-
ways altering plant development or growth – they are
based on specific chemicals that are alien to plants and
reside outside the contributory mechanisms leading to crop
yield. These two cases do not offer a model for improving
crop performance when subjected to environmental stres-
ses, including drought.
Alteration of an entire synthetic process in plants to get
adequate changes at whole plant and field levels is ex-
tremely difficult, as proved in nearly a half century of
physiological studies [5,6]. Many individual processes,
each influenced by a myriad of reactions and molecular
controls, contribute and integrate to affect whole plant
Corresponding author: Sinclair, T.R. (trsincla@ncsu.edu).
1360-1385/$ – see front matter ß 2011 Elsevier Ltd. All rights reserved. doi:10.1016/j.tplants.2011.02.008 Trends in Plant Science, June 2011, Vol. 16, No. 6
performance. This is certainly true for drought-adaptive
traits [7]. The ‘bottom–up’ approach, starting at the molec-
ular level, is often promoted as a simple and direct means
for achieving major yield gains. However, plant modifica-
tions via transformation of an existing molecular pathway
will likely flounder, given the vast number of interacting
biochemical networks and the damping of any simple
molecular change when scaling-up the complex physiologi-
cal hierarchy leading to yield. The route to yield-change
expressed by crops is different for each geographical loca-
tion and new growing season, reflecting new combinations
and temporal dynamics in the biotic and abiotic environ-
ments [1]. Certainly the DNA code may carry information
that influences alternate plant responses to each new
environment, but how that information is integrated
throughout the growing season at the physiological level
and across the community of plants are the critical deter-
minants of crop yield.
An alternative to the bottom–up perspective is a ‘top–
down’ approach that starts with a whole-crop perspective
to consider how the plant community may perform in the
field across a range of environments. What traits can be
altered in individual plants to result in an increased yield
by a community of such altered plants that is a crop? The
advantage of the top–down approach is that the starting
point is the performance of intact crop plants that can be
studied for expression of a desired behavior, instead of an
abstract extrapolation of what a specific molecular-level
transformation might contribute at the whole crop level.
The top–down view may sound a bit old fashioned, but it
must be remembered that it is the use of field experiments
and careful observations that have enabled successful
selection and genetic gain for increased crop yield. Indeed,
numerous studies have demonstrated yield to be geneti-
cally complex and under control of many genes of small
impact. The challenge is to integrate the modern tools of
molecular genetics and physiology to bear on a practical
approach for yield advances in the field [8]. How can these
new insights be used to improve plant phenotype, thereby
increasing crop yield? Can this new information be used in
genotype selection to alter the expression of specific traits
and achieve yield increase? The answers to these critical
questions are not resolved. There are only a few examples
where the top–down approach has been successfully ap-
plied to intrinsically alter biochemical or physiological
activity to increase crop yield [2,3]. The few examples
resulting in cultivar releases with increased yield based
on this approach include heat tolerance in cowpea (Vigna
unguiculata) [9], improved water use in wheat (Triticum
289
Opinion
aestivum) [10], and drought tolerance of nitrogen fixation
in soybean (Glycine max) [11] which is the example consid-
ered in more detail later in this paper.
Here I discuss four topics concerning the application of
molecular and physiological knowledge that may help to
resolve these questions for improving adaptation to
drought: (i) assessing potential yield benefit and variation
within a germplasm population for a proposed drought
trait, (ii) developing suitable means of phenotyping the
specific trait of interest, (iii) developing techniques to track
expression of a desired trait through the breeding process,
and (iv) creating new criteria for cultivar selection and
marketing. While the focus is specially on developing crop
drought tolerance, it is anticipated that many of the con-
clusions of this discussion apply to other abiotic stresses.
Assessment of benefit
It is necessary to move beyond intuitive anticipation of a
significant benefit from a putative trait before launching a
major research program to develop an improved cultivar.
Intuition is not a reliable guide in the face of the complexity
of response to soil drying, and it gives no insight about the
amount of yield change that might be expected. An exam-
ple illustrating the failure of intuition in developing
drought tolerance is the substantial investment in trans-
genic lines that display tolerance under severe water defi-
cit. Experiments are reported that impose drought stresses
resulting in the death or near death of wild-type and
survival of transgenic sister lines. The problem is that
simply enhancing survival of grain crops is irrelevant in
commercial production. Any situation where water avail-
ability is so low to threaten survival means that even if the
plants survive, crop yield will necessarily be very low and
the farmer will be economically devastated. Rather than
survival, realistic yield improvements likely need to ad-
dress one of the following traits: (i) increase plant access to
water, most likely through deeper rooting, to support
greater crop growth; (ii) conserve soil water for use during
late-season water deficits; or (iii) overcome special sensi-
tivities in the plant that limit yield formation under water-
deficit.
The initiation point to genetically enhance crop yield
when subjected to environmental stress should be the
collection of evidence giving a solid basis for anticipating
a relevant benefit under production conditions and for
quantifying the expected yield gain. Unfortunately, many
current studies started with expensive and sophisticated
molecular studies without any direct evidence that modifi-
cation would have practical benefit. Important sources of
such evidence are experimental results in which trait
response is induced experimentally in high-yield commer-
cial cultivars to mimic anticipated genetic modification.
How large is the yield-increase observed in such experi-
ments? In studies on increasing drought tolerance of nitro-
gen fixation, such results were obtained by applying high
amounts of nitrogen fertilizer to soybean plots subjected to
drought stress [12,13]. In these experiments, removing
plant dependence on symbiotically-fixed nitrogen during
water deficits resulted in yield gains of 15–20%.
A very useful approach to fully understand possible
yield changes as a result of trait modification is to use
290
Trends in Plant Science June 2011, Vol. 16, No. 6
mechanistic crop models to simulate yield response over a
range of growing seasons and geographical locations [8,14].
We have used a relatively simple model of soybean devel-
opment, growth and yield to simulate soybean yield
changes that could be expected across the USA with modi-
fied drought traits ([15], Box 1).
Trait phenotyping
The expression of any particular physiological trait or
transformation event is highly dependent on the environ-
ment in which the crop is grown. The existence of desired
genetic code in a genotype offers no assurance about the
expression of plant phenotype in a given environment. The
challenge to make progress in developing higher-yielding
crops is that physiological phenotype other than yield must
be documented at some point in the breeding process
[14,16].
Ben Miflin [17] argued that ‘‘Undue or sole emphasis on
genomics will lead to an ever increasing gap between the
genetic information acquired and an understanding of the
phenotype, a ‘phenotype gap’’’. Indeed, the lack of ability to
phenotype plants for specific trait performance has been a
critical limitation in applying physiological information for
more than half a century [6]. The challenge is that it is very
difficult to accurately phenotype plants for anything but
the most obvious trait. Most of the physiological traits that
impact on response to environmental stress require de-
tailed, sophisticated and usually expensive techniques to
phenotype plants, and can be applied only to a very limited
number of genotypes. Not surprisingly, many physiological
studies searching for genetic variation have frequently
involved no more than 20 genotypes, and almost never
more than 50 genotypes. Conversely, there have been a few
attempts to search a large number of genotypes for genetic
variation, and these attempts have failed due to the lack of
sophistication in the screen.
A solution to the dilemma of phenotypic screens either
being too difficult and sophisticated, or too crude and with
poor resolution is a multi-tiered screening sequence involv-
ing both types of screens [7]. A simple but less accurate
screen that allows a large number of genotypes to be
examined is a first-tier screen followed by tiers of more
sophisticated screens of decreasing numbers of genotypes.
Such multi-tiered screens for crop improvement are not
new in classical breeding efforts. ‘Tandem’ selection breed-
ing is used regularly in some crops. We developed a three-
tiered sequence of physiological screens (Box 2) in our
efforts to identify candidate genotypes for use as parents
in breeding efforts to sustain nitrogen fixation activity
during soil water deficit [18].
Tracking of phenotypic improvement
Unfortunately, the approach of using multi-tiered screens
is likely only suited for identifying parental lines with
desired physiological traits. This approach is too laborious
for tracking a trait through each stage of a breeding effort.
How can the status of a physiological trait in a breeding
effort be resolved?
Of course, one very appealing approach is to identify a
genetic marker or maybe a specific nucleotide sequence
completely linked with the desired trait. The challenge, as
Opinion
Box 1. Model assessment of enhanced nitrogen fixation drought tolerance
A mechanistic soybean growth and yield model was used to simulate
over growing seasons the soil water budget and its impact on crop
development, growth and yield [15]. A Geographical Information
System data base of at least 50 years of weather data for individual
30 km  30 km grid locations across the U.S. were the input for the
model simulations. To simulate yield for all grid locations and all
years across the U.S., the model was run for >130,000 environments.
The model was used to first simulate a ‘standard’ soybean and then
re-run after modifying the model to simulate the behavior of a
specific drought trait hypothesized to influence crop yield in
response to drought. The probability of yield gain is color coded
[()TD$FIG]
45
40
35
30
P(x ≥ 0) g m -2
0.0, 0.05
0.1, 0.15
0.2, 0.25
0.3, 0.35
0.4, 0.45
0.5, 0.55
0.6, 0.65
0.7, 0.75
0.8, 0.85
0.9, 0.95
Figure I. Simulation results for incorporation of the trait for drought tolerance of nitrogen fixation in soybean. (a) Probability of yield gain and (b–d) absolute yield
change for yield percentiles of 75% (wet), median and 25% (dry). Reproduced with permission from [15].
discussed above, is that response to water-deficit is not
likely to be under the control of a single or even a few genes.
One possibility is that a collection of gene markers might
be identified that contribute to physiological phenotype.
Nevertheless, genetic markers only offer information about
the genetic potential related to expression of physiological
trait(s) for drought tolerance. While markers offer impor-
tant information, they do not provide insight about the
level of expression of a trait over a range of field environ-
ments. Trait expression is likely to be very much depen-
dent on the biotic and abiotic environment in which the
plant is grown. Within the plant the level of redundancy,
location of the gene within the genome, and regulation by
other genes may also impact trait expression. Therefore,
the genetic tag could be useful in ensuring the possibility of
phenotypic expression, but unfortunately, such tags do not
Trends in Plant Science June 2011, Vol. 16, No. 6
for all locations in Figure I a as a result of the trait of nitrogen fixation
tolerance to drought. Yield gain was achieved in >85% of the
growing seasons, and in many regions it was >95%. The absolute
amount of the yield change is color coded in Figure I b, c and d for the
75 percentile, median and 25 percentile years for each grid. The
greatest absolute yield gains were in the drier years (25 percentile).
But the median and 75 percentile years also showed positive gains in
yield, indicating that even in these growing seasons there was
sufficient soil water deficit to impair nitrogen fixation activity.
Almost no penalty in yield was simulated from the nitrogen fixation
drought tolerant trait.
(a) (b)
-800, -50
-30, -25
-20, -15
-10, -5 (c)
0, 5
10, 15
20, 25
30, 50
70, 90
120, 150 (d)
-105 -100 -95 -90 -85 -80 -75
TRENDS in Plant Science
offer definitive information about phenotypic expression
being sought in superior lines [19].
The ultimate solution is to have ‘tools’ available to
document phenotypic expression at each stage of breeding
and in each test environment. However, such comprehen-
sive tools seem improbable, if not impossible, and unreal-
istic for practical breeding programs. What can be the
alternative to tracking phenotypic expression at each stage
in the breeding process? One possibility may be to focus
efforts on phenotyping for physiological performance only
at a few critical stages in the breeding cycle. Certainly, one
critical stage would be the initial identification of parent
lines, either existing genotypes or transgenic lines that
readily express the desired trait under a range of condi-
tions. This can be accomplished by the multi-tiered ap-
proach described above.
291
Opinion
Box 2. Multi-tier selection scheme for nitrogen fixation
drought tolerance
A three-tier selection scheme (outlined in Figure I) was developed to
identify candidate parental soybean lines with high tolerance of
nitrogen fixation under drought conditions [18]. The broadest, and
least accurate, screen was based on the concentration of ureides in
the petioles of well-watered soybean genotypes. High leaf ureide
concentrations had been shown to be associated with sensitivity to
nitrogen fixation to drought. From a screen of about 3500 plant
introduction (PI) lines, about 250 lines or slightly less than 10% with
low petiole ureide concentrations were selected for a field test of
nitrogen accumulation under dry conditions. The lines in the second
tier screen were grown on a sandy soil, which had both a low
nitrogen content and low water holding capacity. The lines were
lightly irrigated approximately every other day to maintain them for
about three weeks at or slightly above a soil water content resulting
in slight leaf wilting. Again about 10% of these field-tested lines (24
lines) were selected for the third tier of screening for intensive
measurement of nitrogen fixation response to soil drying in the
greenhouse. Ultimately, 11 PI lines were identified that had
[()TD$FIG]
substantial superiority in nitrogen fixation tolerance to soil drying.
Identification of N2 tolerant germplasm
~ 3 500 plant introductions
Field screen Low petiole ureide
~ 250 PI selections
Field screen N accumulation under drought
24 PI selections
Glasshouse test ARA with drying soil
11 tolerant PIs
TRENDS in Plant Science
Figure I. Diagram of three-tier screen to identify parents with drought
tolerance of nitrogen fixation.
The second critical stage for physiological phenotyping
might be after the breeding population has been decreased
to a more limited, yet substantial number of candidate
lines by using standard agronomic selection criteria. That
is, initially remove lines from the population that do not
contain the marker for the desired trait, show disease or
insect susceptibility, vulnerability to lodging, or have in-
appropriate agronomic characteristics. At this stage, the
number of candidate lines can be narrowed further using
second or third tier screens to a manageable number of
lines for yield test.
The final critical stage is to document that those high-
yielding lines identified in field tests also readily express
the desired physiological trait. This information provides
assurance that the yield increase might be a result of the
putative physiological trait. (Of course, if the high-yielding
lines are shown not to have the desired traits then the
breeder is left with a high yielding line, and additional
studies will be needed to resolve the reason for the yield
improvement). We [18] employed such multi-tiered pheno-
292
Trends in Plant Science June 2011, Vol. 16, No. 6
Box 3. Cultivar selection for enhanced nitrogen fixation
drought tolerance
In the effort to develop cultivars expressing nitrogen fixation
tolerance to water-deficit conditions and high yield, plants were
phenotyped at critical stages in the breeding program. Initial effort
was given to identifying and characterizing the expression of
nitrogen fixation drought tolerance in the soybean cultivar Jackson
[21,22]. Jackson was then crossed with the high-yielding cultivar
KS4895 and the derived population was advanced by single seed
descent to the F3 stage. Dr. Pengyin Chen (University of Arkansas)
eliminated genotypes from the population, based on agronomic
traits for robustness and uniformity in plant maturity in several
environments. Then, 100 selected lines from this population were
tested by Drs. C.A. King and L.C. Purcell (University of Arkansas) in
the greenhouse for nitrogen accumulation by potted plants
subjected to a constant, mild drought stress for two weeks. This
relatively simple physiological screen resulted in the identification
of 17 candidate lines for field yield test in a number of locations and
years. From the field yield tests, two lines were identified as having
higher yield than the commercial checks under moderate drought
conditions. The final step was to test these two lines with controlled
soil drying in the greenhouse to determine directly that both lines
expressed nitrogen fixation drought tolerance [18].
typing at critical stages with success in efforts to develop
genotypes with nitrogen fixation tolerance to soil drying
(Box 3).
Cultivar selection and marketing
Due to complex interactions between physiological process-
es and the abiotic environment in the field, it is unlikely
that small changes in a physiological trait will result in a
uniform increase in yield across all environments. Physio-
logical trait improvement for drought conditions, unlike
disease or insect tolerance for which the presence of the
desired trait itself has virtually no negative consequences
on yield, can have quite variable impact on yield across
environments [15]. Instead of evaluating candidate culti-
vars based on grand mean yields over environments, which
is a common approach, the more useful information is a
probability response for yield change in the targeted envir-
onments. What are the environmental circumstances and
probability for yield gain and by how much, and conversely
what are the environmental circumstances and probability
for yield loss and by how much?
Results from field tests with soybean lines identified as
having drought tolerance of nitrogen fixation provided
information on yield response across environments [20].
Yield tests were made at a number of locations in the
southeast USA and the yield in each case was compared
with the average yield of commercial cultivars. The results
of these tests showed a yield advantage for one genotype
(R01-518F) in those environments where average yield of
commercial cultivars was 250–360 g m–2 (Figure 1). Out-
side this commercially important range of yield, there was
no consistent yield advantage or disadvantage for R01-
518F. Therefore, if there is a reasonable fraction of growing
seasons in the 250–360 g m–2 range, the nitrogen fixation
drought tolerance trait would be advantageous for farmers.
Farmers seem willing to digest probability data on yield
response if such information will improve their economic
wellbeing. Farmers need to know the fraction of growing
seasons a yield gain might be anticipated, the fraction of
()TD$FIG][ Opinion Trends in Plant Science June 2011, Vol. 16, No. 6

2 Campos, H. et al. (2004) Improving drought tolerance in maize: A view

Consistent
from industry. Field Crops Res. 90, 19–34
yield gain
3 Sinclair, T.R. and Purcell, L.C. (2005) Is a physiological perspective
90 relevant in a ‘genocentric’ age? J. Exp. Bot. 56, 2777–2782
4 Tuberosa, R. and Salvi, S. (2006) Genomics-based approaches to
60 improve drought tolerance of crops. Trends Plant Sci. 11, 405–412
Yield differ. (g m-2)
[R01-581F - check]

5 Passioura, J.B. (2010) Scaling up: the essence of effective agricultural

30
0 increase yield potential. Trends Plant Sci. 9, 70–75
-30
-60
100 200 300 400
Check yield (g m-2)
TRENDS in Plant Science
Figure 1. Graph of differences in yield for individual test environments between
R01-518F, a genotype with drought tolerance of nitrogen fixation, and the average
of commercial cultivars [20]. The developed cultivar R01-581F had consistently
greater yield than commercial cultivars in the yield range of 250–360 g m–2 which is
delineated in the figure with vertical dashed lines. This range of the consistent
yield gain is particularly important in commercial soybean production.
Reproduced with permission from [20].
years when there is no yield benefit, and the fraction of
years when the stress-tolerant cultivar will result in a yield
loss. Cultivars with stress-tolerant traits might be sold like
insurance in that the improved cultivar gives increased
yields in low-yield, low-income years but the insurance
premium is ‘paid’ in the very good years when a yield
decrease would result in a small income loss in years of
high revenue.
Acknowledgements
The author is grateful for the input offered by Greg Rebetzke (CSIRO,
Canberra, Australia), James Specht (Agronomy & Horticulture
Department, University of Nebraska, Lincoln, NE, USA) and Tommy
Carter (USDA-ARS, Raleigh, NC, USA) in their review of the draft of this
paper.
References
1 Herve, P. and Serraj, R. (2009) Gene technology and drought: A simple
solution for a complex traits? African J. Biotech. 8, 1740–1749
research. Func. Plant Biol. 37, 585–591
6 Sinclair, T.R. et al. (2004) Crop transformation and the challenge to
7 Salekdeh, G.H. et al. (2009) Conceptual framework for drought
phenotyping during molecular breeding. Trends Plant Sci. 14, 488–496
8 Messina, C.D. et al. (2011) Yield-trait performance landscapes: from
theory to application in breeding maize for drought tolerance. J. Exp.
Bot. 62, 855–868
9 Ehlers, J.D. et al. (2000) Registration of ‘California Blackeye 27’
cowpea. Crop Sci. 40, 854–855
10 Rebetzke, G.J. et al. (2002) Selection for reduced carbon isotope
discrimination increases aerial biomass and grain yield of rainfed
bread wheat. Crop Sci. 42, 739–745
11 Chen, P. et al. (2007) Registration of soybean germplasm lines R01-
416F and R01-581F for improved and yield and nitrogen fixation under
drought stress. J. Plant Registrations 1, 167–168
12 Purcell, L.C. and King, C.A. (1996) Drought and nitrogen source effects
on nitrogen nutrition, seed growth, and yield in soybean. J. Plant Nutr.
19, 969–993
13 Ray, J.D. et al. (2006) Influence of large amounts of nitrogen on
nonirrigated and irrigated soybean. Crop Sci. 46, 52–60
14 Tardieu, F. and Tuberosa, R. (2010) Dissection and modeling of abiotic
stress tolerance in plants. Curr. Opin. Plant Biol. 13, 206–212
15 Sinclair, T.R. et al. (2010) Assessment across the United States of the
benefits of altered soybean drought traits. Agron. J. 102, 475–482
16 Munns, R. et al. (2010) New phenotyping methods for screening wheat
and barley for beneficial responses to water deficit. J. Exp. Bot. 61,
3499–3507
17 Miflin, B. (2000) Crop improvement in the 21st century. J. Exp. Bot. 51,
1–8
18 Sinclair, T.R. et al. (2000) Identification of soybean genotypes with N2
fixation tolerance to water deficits. Crop Sci. 40, 1803–1809
19 Bernardo, R. (2008) Molecular markers and selection for complex traits
in plants: Learning from the last 20 years. Crop Sci. 48, 1649–1664
20 Sinclair, T.R. et al. (2007) Drought tolerance and yield increase of
soybean resulting from improved symbiotic N2 fixation. Field Crops
Res. 101, 68–71
21 Sall, K. and Sinclair, T.R. (1991) Soybean genotypic differences in
sensitivity of symbiotic nitrogen fixation to soil dehydration. Plant
Soil 133, 31–37
22 Serraj, R. and Sinclair, T.R. (1997) Variation among soybean cultivars
in dinitrogen fixation in response to drought. Agron. J. 89, 963–969

293