PUBLISHED VERSION
Nadia Al-Tamimi, Chris Brien, Helena Oakey, Bettina Berger, Stephanie Saade, Yung Shwen
Ho, Sandra M. Schmöckel, Mark Tester and Sónia Negrão
Salinity tolerance loci revealed in rice using high-throughput non-invasive phenotyping
Nature Communications, 2016; 7:13342-1-13342-11
© The Author(s) 2016. This work is licensed under a Creative Commons Attribution 4.0
International License. The images or other third party material in this article are included in the
article’s Creative Commons license, unless indicated otherwise in the credit line; if the material
is not included under the Creative Commons license, users will need to obtain permission from
the license holder to reproduce the material. To view a copy of this license, visit
http://creativecommons.org/licenses/by/4.0/
Published version
http://dx.doi.org/10.1038/ncomms13342
http://hdl.handle.net/2440/102678
PERMISSIONS
http://creativecommons.org/licenses/by/4.0/
22 November 2016
ARTICLE
Received 11 Feb 2016
| Accepted 25 Sep 2016 | Published 17 Nov 2016
Salinity tolerance loci revealed in rice using
high-throughput non-invasive phenotyping
Nadia Al-Tamimi
1
, Chris Brien
2,3
, Helena Oakey
1
, Bettina Berger
3
, Stephanie Saade
1
, Yung Shwen Ho
1
,
Sandra M. Schmo
¨
ckel
1
, Mark Tester
1
&So
´
nia Negra
˜
o
1
High-throughput phenotyping produces multiple measurements over time, which require new
methods of analyses that are flexible in their quantification of plant growth and transpiration,
yet are computationally economic. Here we develop such analyses and apply this to a rice
population genotyped with a 700k SNP high-density array. Two rice diversity panels, indica
and aus, containing a total of 553 genotypes, are phenotyped in waterlogged conditions.
Using cubic smoothing splines to estimate plant growth and transpiration, we identify four
time intervals that characterize the early responses of rice to salinity. Relative growth rate,
transpiration rate and transpiration use efficiency (TUE) are analysed using a new association
model that takes into account the interaction between treatment (control and salt) and
genetic marker. This model allows the identification of previously undetected loci affecting
TUE on chromosome 11, providing insights into the early responses of rice to salinity, in
particular into the effects of salinity on plant growth and transpiration.
DOI: 10.1038/ncomms13342
OPEN
1
King Abdullah University of Science and Technology (KAUST), Division of Biological and Environmental Sciences and Engineering (BESE), Thuwal
23955-6900, Saudi Arabia.
2
University of South Australia, Phenomics and Bioinformatics Research Centre, Adelaide, South Australia 5001, Australia.
3
University of Adelaide, Australian Plant Phenomics Facility, The Plant Accelerator, Urrbrae, South Australia 5064, Australia. Correspondence and requests
for materials should be addressed to M.T. (email: mark.tester@kaust.edu.sa).
NATURE COMMUNICATIONS | 7:13342 | DOI: 10.1038/ncomms13342 | www.nature.com/naturecommunications 1
F
or more than half of the world’s population, rice
(Oryza sativa L.), the most salt-sensitive cereal
1–3
,isa
dietary staple. It is estimated that B20% of irrigated lands
are affected by salt (http://www.fao.org/water/en/). For example,
the Indo-Gangetic Basin in India and the Indus Basin in Pakistan
suffer losses in rice yield as high as 45% and 36–69%, respectively,
from soil salinity
1,4
. Moreover, climate change is foreseen to
increase saltwater ingress in coastal regions of Southeast Asia,
where rice is the primary cultivated crop
5
. With the global
population rising, a 26% increase in rice yield is predicted to be
required to meet global demands in the next 25 years
6
. Hence,
there is a vital requirement to significantly increase rice
productivity on salinized lands.
Exposure of plants to soil salinity rapidly reduces their growth
and transpiration rates (TRs) due to the ‘osmotic component’ of
salt stress (sensu Munns and Tester)
2
, which is hypothesized to be
related to sensing and signalling mechanisms
7
. Over time,
toxic concentrations of Na
þ
and Cl
accumulate in the cells
of the shoot, known as the ‘ionic component’ of salt stress,
which causes premature leaf senescence
2,8,9
. Both osmotic
and ionic components of salinity stress are likely to impact
yield. Despite significant advances in our understanding of the
ionic components of salinity tolerance, little is known about
the early responses of plants to salinity stress
7
. Therefore, the
discovery of new quantitative trait loci (QTL) contributing to
salinity tolerance, with a focus on the ‘osmotic component’, has
the potential to substantially improve crop productivity.
The paucity of work on ‘osmotic tolerance’ is likely to be due,
at least in part, to the need for the development of new methods
for the analysis of plant growth and transpiration. Despite
progress in analysing the image-based phenotyping data collected
non-invasively with high time and spatial resolution, few
statistical methods have accurately modelled plant growth and
transpiration. Here we report a new statistical method for
quantifying plant growth and transpiration using the data
generated by high-throughput non-invasive phenotyping. We
apply this method to precisely quantify the effects of salt stress on
the growth and transpiration of rice plants.
A genome-wide association study (GWAS) was undertaken
with the aim of identifying new loci that contribute to the early
responses of rice to soil salinity. Two rice diversity panels, indica
and aus, were phenotyped at The Plant Accelerator. Rice plants
were grown in waterlogged conditions, to represent this aspect of
irrigated rice fields that should be included in salinity tolerance
studies, as the effects of hypoxia on salinity tolerance has been
well documented
10
. In addition, we explored a model for
analysing GWAS that enabled interactions between treatment
groups (control and salt) and the genetic marker of interest. This
new model substantially improves the detection of significant
single-nucleotide polymorphisms (SNPs) that are specifically
related to the treatment. By combining the analysis of
high-throughput phenotyping data and GWAS, we investigated
the effects of salinity on relative growth rate (RGR), TR
and transpiration use efficiency (TUE), and identified several
new salinity tolerance loci associated with these previously
uncharacterized traits.
Results
Indica maintains growth better than aus in saline conditions.
To assess the early responses of two rice diversity panels to saline
conditions, we exposed a total of 553 rice accessions (297 indica
and 257 aus varieties) from 24-day-old plants to 150 mM NaCl.
Over 13 days, several physiological responses of plants exposed
to high salinity were monitored and compared with those of
plants maintained in low salt concentrations. This was done using
high-throughput, non-destructive imaging. From three RGB
images (one top and two side views), we made daily measure-
ments of the total number of pixels for each plant, as a proxy for
shoot biomass. There are numerous mathematical models to
describe growth curves
11–13
, but most of these models make
assumptions about the shape of the curve; for example,
exponential growth models are typically used for young
seedlings and short growth intervals. However, we observed
that in our experiment, plant growth was neither exponential nor
logistic throughout the imaging period, and in particular,
salt-treated plants did not show exponential growth.
To avoid these erroneous assumptions and to accurately
describe plant growth and to better estimate RGR, we fitted cubic
smoothing splines (hereafter referred to as splines) to smooth the
trend in the projected shoot area (PSA) for each plant.
Examination of the plots of the smoothed values for PSA,
absolute growth rate (AGR) and RGR (Supplementary Fig. 1)
indicates that neither exponential nor logistic curves would
accurately describe the growth of plants in this experiment,
particularly in the case of the salt-treated plants. This approach
has the advantage of making no a priori assumptions about the
shape of the curve; to our knowledge, this is the first time that
cubic smoothing splines have been used to provide an unbiased
analysis of high-throughput phenotypic data. Although several
decades ago, splines were fitted to data to characterize growth
from destructive harvesting
14
, they have seldom been used since.
Shipley and Hunt
15
advocated their use for characterizing growth
and Li and Sillanpa
¨
a
¨
12
suggested their application to describe
complex growth trends
12,15
.
We found that PSA strongly positively correlates with shoot
biomass when using the squared Pearson correlation coefficient
(for example, r
2
¼ 0.945 for indica and r
2
¼ 0.91 for aus in the
northeast (NE) Smarthouse, using Pearson’s correlation;
Supplementary Fig. 2), confirming our experimental set-up as
suitable to monitor plant growth. From the smoothed PSA, we
were able to calculate RGRs and AGRs between imaging days
(Supplementary Fig. 1). As expected, RGR decreases through time
to a greater extent under saline conditions than in control
conditions (Fig. 1a,b). More specifically, a rapid reduction in
biomass production was observed immediately after salt applica-
tion, suggesting that the rice plants responded to the ‘osmotic
component’ of salt stress, before a build-up of salt in the leaves
could impact plant growth, as occurs after several days, in the
later ‘ionic phase’
2
.
Based on characteristic growth patterns of smoothed PSA, RGR
and AGR, we separated the response of the rice plants to salinity
stress into four intervals for further analysis (Fig. 1a,b;
Supplementary Fig. 1a–f). In interval I, 2–9 days after treatment,
AGR increases in control conditions and then plateaus. During
interval II, 2–6 days after treatment, the RGR declines less rapidly
under control and saline conditions than during interval III
(6–9 days after treatment). During interval IV, 9–13 days after
treatment, AGR increases less in control compared with salt-
treated plants. The first day after salt treatment was excluded
from further analyses because of reduced confidence in its values,
inevitable from the spline fitting occurring across an interval that
had clearly distinct properties.
We found that RGR was lower in the salt-treated accessions
than in the control plants for both rice panels (Fig. 1). Comparing
the percentage decrease in salt-treated plants relative to control
plants at each interval, consistent with the results of Campbell
et al.
16
, we found that indica lines maintained better growth than
aus lines (Fig. 1c,d; Supplementary Table 1); for example, in the
interval 2–6 days after treatment, growth of indica due to salinity
decreased by 21%, while that of aus decreased by 29%.
The early growth response index (EGRI) provides an estimate
for the early responses of plants to salinity. For all earlier
ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13342
2 NATURE COMMUNICATIONS | 7:13342 | DOI: 10.1038/ncomms13342 | www.nature.com/naturecommunications
intervals, indica had a higher EGRI than aus (Fig. 1e). Although
rice plants suffer a significant and rapid decrease in growth under
salinity stress, indica accessions are better able to maintain their
growth compared with aus accessions. In addition, accessions
previously reported to be salt-tolerant, such as ‘Pokkali’ (an indica
rice), have a higher EGRI than accessions previously reported as
salt sensitive, such as ‘IR28’ (refs 17,18) (Supplementary Table 2).
This result suggests that these early responses are likely to be an
important component of overall plant salinity tolerance in
the field.
Indica maintains transpiration under saline conditions. Our
study also examines the previously unstudied traits of TR and
TUE in response to moderate salinity stress under waterlogged
growth conditions. To accurately describe TR, we fitted splines to
daily measures of transpired water for each plant. As expected, we
observed a clear acceleration in TR over time in control plants,
and only a small increase over time in salt-treated plants
(Fig. 2a,b). These results are consistent with a previous study of
wheat and barley under saline conditions
19
. Notably, the aus
panel had a greater average decrease for TR in all four intervals
when compared with indica (Supplementary Table 1). The
substantial decrease in TR in the aus panel 9–13 days after
treatment for both control and salt-treated plants can be
explained by the sudden decrease in hours of sunshine at this
point in the experiment. According to the Bureau of
Meteorology’s daily weather observations in March 2015, 9.1 h
of sunshine on average were recorded during the first 10 days of
imaging, but only 4.4 h of sunshine for the last 3 days (9–13 days
after treatment).
TUE in this work is defined by the ratio of aboveground
biomass produced per unit of water transpired and depends on
the characteristics of the plants and on the environment where
the plants grow. We calculated TUE as a third-order-derived trait
(TUE is estimated from transpiration and growth from PSA, and
these are, in turn, estimated from measures of water loss and pixel
counts, respectively.) On average, TUE decreases marginally over
time for control plants and more rapidly for salt-treated plants
(Fig. 2c,d; Supplementary Fig. 3). Salinity reduced TR propor-
tionally more than TUE, similar to wheat and barley
19
. The indica
panel had a lower average decrease in TUE compared with the
0.4
0.3
0.2
0.1
0.0
–1 0 1 2 3 4 5 6 7
8
9101112
13
Time after salting (days)
RGR (per day)
Indica
Control
Salt
Control
Salt
Control
Salt
aus
0.4
0.3
0.2
0.1
0.0
RGR (per day)
–1
0
1
2
34
567
8
910111213
Time after salting (days)
Control
Salt
aus
Indica
Control
Salt
Treatment
0.3
0.2
0.1
0.0
2 – 9 2 – 6 6 – 9 9 – 13
Time intervals for days after treatment
Relative growth rate (per day)
0.3
0.2
0.1
0.0
Relative growth rate (per day)
2 – 9 2 – 6 6 – 9 9 – 13
Time intervals for days after treatment
Treatment
Control
Salt
Trait/interval (days)
EGRI.2–9
EGRI.2–6
EGRI.6–9
EGRI.9–13
Mean
0.72
0.79
0.61
0.51
0.43
0.51
0.25
0.14 0.95
0.99
1.11
1.06
Max
s.d.
0.11
0.1
0.13
0.15
indica
95% CI
0.012
0.011
0.015
0.017
Mean
0.64
0.71
0.53
0.52
Min
0.43
0.48
0.26
0.17 0.88
0.91
0.98
0.92
Max
aus
s.d.
0.09
0.09
0.1
0.12 0.015
0.012
0.011
0.011
95% CI
Min
ab
c
d
e
Figure 1 | Relative growth rate (RGR) of salinity-induced responses comparing indica and aus. (a) Smoothed RGR values were obtained from projected
shoot area (PSA) values to which splines had been fitted, as shown in Supplementary Fig. 2. This was applied to the data from individual indica and (b) aus
plants. The solid line represents the grand average of control conditions (blue) and saline conditions (red). In each panel, the RGB image of a rice plant on
the left is representative of a plant 1 day before salt application. The RGB image on the top right side represents the same plant after 13 days of salt
treatment, while the RGB image on the bottom right represents the same plant genotype at 13 days under control conditions. (c) Values of RGR at different
time intervals for indica (n ¼ 528; partially replicated; median ¼ 0.13, 0.15, 0.11 and 0.10 for intervals: 2–9, 2–6, 6–9 and 9–13 days after salting, respectively)
and (d) Values of RGR at different time intervals for aus (n ¼ 226; fully replicated; median ¼ 0.15, 0.17, 0.13 and 0.09 for intervals: 2–9, 2–6, 6–9 and 9–13
days after salting, respectively). (e) Table comparing the mean early growth response index (EGRI) at different time intervals for indica and aus. Min and
max refer to the minimum and maximum means, respectively. s.d. refers to standard deviation. CI, confidence interval.
NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13342 ARTICLE
NATURE COMMUNICATIONS | 7:13342 | DOI: 10.1038/ncomms13342 | www.nature.com/naturecommunications 3
aus panel (16.7% versus 24.4% for interval 2–6 days after
treatment; Supplementary Table 1), and TUE was positively
correlated with biomass production over time (RGR;
Supplementary Fig. 4). To quantify the relative performance of
plants with respect to TUE, we used a simple salt/control index
20
of the ratio of TUE in salt-treated plants relative to TUE in
control plants over the same time period. A box plot of this index
(Fig. 2e) shows that indica tends to maintain a higher
salt/control index for TUE than aus throughout the first three
intervals.
Association analysis of salinity-induced responses. We used
GWAS to identify genetic loci associated with the early responses
of rice to salinity stress. We compared two sets of genotypic
information for association analyses in the indica panel—‘GBS
300
200
100
0
–10123456789101112
13
Time after salting (days)
TR (mL water transpired per plant per day)
Indica
Control
Salt
300
200
100
0
TR (mL water transpired per plant per day)
aus
Control
Salt
Control
Salt
Control
Salt
aus
1.00
0.75
0.50
0.25
0.00
TUE (increase in kpixes per mL water transpired)
Indica
–1012345678910111213
Time after salting (days)
–1012345678910111213
Time after salting (days)
–10123
4
567
8
9
10 11 12
13
Time after salting (days)
1.00
0.75
0.50
0.25
0.00
TUE (increase in kpixes per mL water transpired)
1.6
1.2
0.8
0.4
2 – 9 2 – 6 6 – 9 9 – 13
Time intervals for days after salt treatment
TUE (salt/control)
Indica
aus
Panel
a
b
c
d
e
Figure 2 | Transpiration of salinity-induced responses comparing indica and aus. Spline curve fits of transpiration rate (TR) through time for individual
(a) indica and (b) aus plants and transpiration use efficiency (TUE) through time for individual (c) indica and (d) aus plants. The solid blue lines represent
the grand average spline in control conditions and the solid red lines represent the same in saline conditions. (e) Box plots of the TUE salinity tolerance
index (salt/control), comparing indica (n ¼ 528; partially replicated; median ¼ 0.78, 0.84, 0.71 and 0.69 for intervals: 2–9, 2–6, 6–9 and 9–13, respectively)
and aus (n ¼ 226; fully replicated; median ¼ 0.71, 0.75, 0.64 and 0.70 for intervals: 2–9, 2–6, 6–9 and 9–13 days after salting, respectively).
ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms13342
4 NATURE COMMUNICATIONS | 7:13342 | DOI: 10.1038/ncomms13342 | www.nature.com/naturecommunications