Phenomics—Technologies to relieve the phenotyping bottleneck

Abstract

Global agriculture is facing major challenges to ensure global food security, such as the need to breed high-yielding crops adapted to future climates and the identification of dedicated feedstock crops for biofuel production (biofuel feedstocks). Plant phenomics offers a suite of new technologies to accelerate progress in understanding gene function and environmental responses. This will enable breeders to develop new agricultural germplasm to support future agricultural production. In this review we present plant physiology in an 'omics' perspective, review some of the new high-throughput and high-resolution phenotyping tools and discuss their application to plant biology, functional genomics and crop breeding.

Full text

Feature
Review
Phenomics
–
technologies
to
relieve
the
phenotyping
bottleneck
Robert
T.
Furbank
1
and
Mark
Tester
2
1
High
Resolution
Plant
Phenomics
Centre,
Australian
Plant
Phenomics
Facility,
CSIRO
Plant
Industry,
GPO
Box
1600,
Canberra,
ACT
2601,
Australia
2
Australian
Centre
for
Plant
Functional
Genomics
and
The
Plant
Accelerator,
Australian
Plant
Phenomics
Facility,
University
of
Adelaide,
Hartley
Grove,
Urrbrae,
SA
5064,
Australia
Global
agriculture
is
facing
major
challenges
to
ensure
global
food
security,
such
as
the
need
to
breed
high-
yielding
crops
adapted
to
future
climates
and
the
iden-
tification
of
dedicated
feedstock
crops
for
biofuel
pro-
duction
(biofuel
feedstocks).
Plant
phenomics
offers
a
suite
of
new
technologies
to
accelerate
progress
in
understanding
gene
function
and
environmental
responses.
This
will
enable
breeders
to
develop
new
agricultural
germplasm
to
support
future
agricultural
production.
In
this
review
we
present
plant
physiology
in
an
‘omics’
perspective,
review
some
of
the
new
high-
throughput
and
high-resolution
phenotyping
tools
and
discuss
their
application
to
plant
biology,
functional
genomics
and
crop
breeding.
Plant
biology
faces
new
challenges:
a
role
for
plant
phenomics
Global
agriculture
and
the
plant
biology
underpinning
it
are
facing
major
challenges
which
require
new
approaches
to
functional
genomics
and
plant
breeding.
Global
food
security,
the
identification
of
appropriate
and
efficient
plant-based
biofuel
feedstocks
and
coping
with
climate
change
are
foremost
in
the
minds
of
scientists,
politicians
and
the
general
public.
To
address
these
issues,
we
need
new
high-yielding
genotypes
of
agricultural
crops
adapted
to
our
future
climate.
Agricultural
crops
fulfilling
future
food
and
fuel
needs
must
display
both
high
intrinsic
yield
and
yield
stability
under
abiotic
stresses.
Annual
increases
in
yield
achieved
from
traditional
breeding
programs
worldwide
are
no
longer
sufficient
to
meet
projected
de-
mand
for
all
three
major
cereal
crops:
rice
(Oryza
sativa),
maize
(Zea
mays)
and
wheat
(Triticum
aestivum)
[1–3].
With
the
burgeoning
world
population,
cereal
grain
yields
alone
must
increase
by
at
least
70%
before
2050.
In
fact
rice
demand
has
already
exceeded
supply
for
the
past
2
years
[1].
Demand
for
biofuel
feedstocks
will
also
undoubtedly
increase
over
the
next
decade
[4],
resulting
in
potential
competition
for
arable
land
between
food
and
fuel
crops.
At
the
same
time
the
impacts
of
climate
change
on
global
temperatures
and
rainfall
patterns
are
likely
to
lead
to
reduction
in
yields
due
to
abiotic
stress
[3].
Although
climate
change
can
also
have
a
positive
impact
on
yields
through
the
CO
2
fertilizer
effect
on
photosynthesis
[5,6],
the
benefit
of
high
CO
2
on
yield
shows
large
inter-
and
intra-specific
genetic
variation
and
modern
cultivars
show
particularly
poor
responses,
which
has
stimulated
interest
in
developing
‘climate
change-adapted’
agricultural
germ-
plasm
[5,6].
There
has
been
a
degree
of
confidence
over
the
past
decade
that
the
genomics
revolution
and
gene
technology
can
provide
solutions
to
the
current
challenges
in
plant
breeding.
Homozygous
genome
wide
knockout
lines
are
available
in
Arabidopsis
(Arabidopsis
thaliana)
[7],
1001
Arabidopsis
ecotypes
are
currently
being
sequenced
to
provide
comparative
genomic
databases
[8],
the
Medicago
(Medicago
truncatula)
genome
has
been
sequenced
and
genetic
resources
are
publicly
available
[9]
and
the
genome
sequence
of
Brachypodium
(Brachypodium
dystachion),
a
model
cereal,
has
also
recently
been
reported
[10].
The
genome
sequences
of
rice,
maize,
sorghum
(Sorghum
vul-
gare),
barley
(Hordeum
vulgare)
and
many
dicot
and
mono-
cot
crops
are
either
sequenced
or
soon
will
be,
as
sequencing
costs
spiral
downwards.
‘Re-sequencing’
of
genomes
of
agricultural
crops
to
assess
allelic
variation
will
become
even
more
commonplace
in
the
next
few
years.
To
harness
this
wealth
of
genomic
information
for
agri-
cultural
application,
it
has
to
be
carefully
and
comprehen-
sively
linked
to
phenotype
in
a
‘real
world’
environment.
For
model
systems,
linkage
of
genotype
to
phenotype
has
been
often
illusive.
Phenotypic
descriptions
of
genome-
wide
knockout
collections
are
limited
(see
[11]
for
one
of
the
most
comprehensive
examples).
The
frequency
with
which
the
annotation
‘no
visible
phenotype’
occurs
is
symp-
tomatic
of,
among
other
things,
a
lack
of
capacity
for
the
plant
community
to
analyze
the
subtle
and
sometimes
complex
phenotypic
effects
of
genetic
modification.
A
com-
bination
of
systematic
‘industrial
scale’
phenotyping
at
high-throughput
[12]
to
mine
candidate
germplasm
for
genes
underlying
traits
of
agricultural
importance
and
high-resolution
examination
of
subtle
phenotypes
of
smal-
ler
subsets
is
clearly
required.
High-throughput
phenotyp-
ing
has
become
more
common
in
laboratories
in
the
commercial
sector,
but
this
information
rarely
enters
the
public
domain,
and
then
often
after
delay
and
with
un-
known
completeness.
Marker-assisted
selection
(MAS)
in
plant
breeding
has
become
more
common
in
recent
years
and
is
now
used
routinely
for
traits
such
as
pathogen
resistance
conferred
Review
Corresponding
author:
Furbank,
R.T.
(Robert.Furbank@csiro.au).
1360-1385/$
–
see
front
matter.
Crown
Copyright
ß
2011
Published
by
Elsevier
Ltd.
All
rights
reserved.
doi:10.1016/j.tplants.2011.09.005
Trends
in
Plant
Science,
December
2011,
Vol.
16,
No.
12 635

by
single
genes.
Bi-parental
recombinant
inbred
lines,
doubled
haploid
sets
or
genetically
characterized
acces-
sions
for
population
genetics
can
be
used
to
identify
quan-
titative
trait
loci
(QTLs)
and
ultimately
clone
genes
of
interest
under
these
QTLs
[3].
Although
assembling
the
necessary
genetic
resources
is
in
itself
a
challenge,
pheno-
typing
the
populations
is
widely
recognized
as
the
most
laborious
and
technically
challenging
part
of
this
process.
For
a
population
to
be
screened
for
a
valuable
agricultural
trait
(such
as
grain
size,
abiotic
stress
tolerance,
product
quality
or
yield
potential),
replicated
trials
are
necessary
across
multiple
environments
over
a
number
of
seasons.
Phenotyping
tools
currently
in
common
use
require
de-
structive
harvests
at
fixed
times
or
at
particular
phenolog-
ical
stages
and
are
slow
and
costly.
Furthermore,
if
a
promising
candidate
gene
is
to
be
tested
for
allelic
varia-
tion
in
a
mapping
population,
this
phenotyping
work
needs
to
be
done
precisely.
The
labor-intensive
and
costly
nature
of
conventional
field
phenotyping
have
meant
that
many
crop
breeding
programs
make
a
single
measurement
of
final
yield
for
replicated
plots
in
contrasting
environments
over
multiple
seasons.
However,
yield
itself
is
one
of
most
poorly
inher-
ited
traits
in
crop
breeding
(see
[13]
and
references
there-
in).
The
bottleneck
in
field
phenotyping
has
driven
intense
interest
over
the
past
decade
in
applying
remote
sensing
technologies
to
field
crop
monitoring
and
in
this
regard
field
phenomics
is
more
advanced
in
many
respects
than
controlled-environment,
high-throughput
analysis.
The
‘phenotyping
bottleneck’
described
above
can
now
be
addressed
by
combining
novel
technologies
such
as
non-
invasive
imaging,
spectroscopy,
image
analysis,
robotics
and
high-performance
computing.
Phenomics
could
be
de-
scribed
as
simply
‘high-throughput
plant
physiology’.
As
a
result,
field
evaluation
of
plant
performance
is
much
faster,
and
facilitates
a
more
dynamic,
whole-of-lifecycle
measure-
ment
less
dependent
on
periodic
destructive
assays.
Fur-
thermore,
application
of
these
tools
in
dedicated
high-
throughput,
controlled-environment
facilities
has
the
po-
tential
to
improve
precision
and
reduce
the
need
for
repli-
cation
in
the
field.
This
brings
us
to
the
age
of
‘phenomics’
(Box
1).
High-throughput
phenomics
of
model
systems:
the
phenomics–genomics
pipeline
A
clear
goal
of
phenomics
is
to
bridge
the
gap
between
genomics,
plant
function
and
agricultural
traits.
Particu-
larly
in
the
context
of
model
systems,
where
availability
of
genomic
sequence
is
burgeoning,
there
is
a
pressing
need
for
a
searchable
phenotypic
database
linking
gene
se-
quence
to
plant
structure,
development,
composition
and
performance,
all
measured
in
a
clearly
defined
environ-
ment.
To
generate
a
meaningful
database
of
phenotypic
characters
in
a
model
species,
the
first
requirement
must
be
to
have
data
objectively
described,
preferably
in
a
mathematical,
easily
digitized
and
searchable
format
(using
ontologies,
or
controlled
vocabularies,
for
example,
[14]).
Second,
the
information
on
how
the
experiment
was
carried
out,
the
plant
material
used
and
the
growth
condi-
tions
used
must
be
recorded
in
a
standardized
format
(the
‘metadata’).
Third,
standardization
or
at
least
adequate
reporting
of
the
phenotyping
techniques
employed
is
desir-
able
to
allow
comparison
of
data
sets.
The
inadequacy
of
these
descriptions
and
standardization
of
experimental
treatments
in
the
literature
and
in
current
databases
has
severely
limited
the
utility
of
both
transcriptional
and
metabolomic
data
(see
[15]
and
references
therein).
High-throughput
phenotyping
of
model
plants
such
as
Arabidopsis
using
non-invasive
imaging
technologies
is
a
rapidly
advancing
field
(www.plantphenomics.org.au
and
[1,16–18]).
Image
analysis
and
mathematical
treatment
of
imaging
data
to
extract
growth
dynamics,
morphological
characters,
and
spatially
described
photosynthetic
param-
eters
are
challenging
and
require
sophisticated
storage
and
linkage
of
primary
images
and
calculated
data.
Utili-
zation
of
digital
imaging
and
image
analysis
to
estimate
growth
rates
from
projected
leaf
area
is
relatively
simple
Box
1.
What
is
phenomics?
Phenome
=
Gene

Environment
or
the
expression
of
the
genome
as
traits
in
a
given
environment.
Plant
phenomics
Plant
phenomics
is
the
study
of
plant
growth,
performance
and
composition
(Figure
I).
Forward
phenomics
uses
phenotyping
tools
to
‘sieve’
collections
of
germplasm
for
valuable
traits.
The
sieve
or
screen
could
be
high-throughput
and
fully
automated
and
low
resolution,
followed
by
higher-resolution,
lower-throughput
mea-
surements.
Screens
might
include
abiotic
or
biotic
stress
challenges
and
must
be
reproducible
and
of
physiological
relevance.
Reverse
phenomics
is
the
detailed
dissection
of
traits
shown
to
be
of
value
to
reveal
mechanistic
understanding
and
allow
exploitation
of
this
mechanism
in
new
approaches.
This
can
involve
reduction
of
a
physiological
trait
to
biochemical
or
biophysical
processes
and
ultimately
a
gene
or
genes.
Chromosome
Cells
TRENDS in Plant Science
Nucleus
Cell
Gene
DNA
Figure
I.
Plant
phenomics
is
the
study
of
plant
growth,
performance
and
composition.
Review Trends
in
Plant
Science
December
2011,
Vol.
16,
No.
12
636

and
accurate
for
rosette
plants
such
as
Arabidopsis
and
can
be
achieved
in
high-throughput
using
trays
of
20
or
more
individuals
[19,20].
Such
a
growth-imaging
approach
has
already
been
used
to
screen
for
drought
tolerance
in
Ara-
bidopsis
accessions
[19]
and
for
QTLs
linked
to
biomass
increases
induced
by
heterosis
in
more
than
400
recombi-
nant
inbred
lines
of
Arabidopsis
[21].
Recently,
this
tech-
nique
has
been
extended
to
include
pulse-modulated
chlorophyll
fluorescence
imaging
as
a
tool
for
examining
photosynthetic
responses
to
drought
stress
in
Arabidopsis,
in
addition
to
growth
rate
response
[17].
Morphological
descriptors
for
herbarium
identification
and
plant
devel-
opment
are
well
established
and
based
on
vectorization
of
2-D
digital
images
[22].
Three-dimensional
plant
models,
more
appropriate
to
cereals
and
larger
dicots,
have
been
developed
using
math-
ematical
approaches
known
as
‘L-systems’
[23,24],
which
simulate
plant
development
with
a
series
of
generative
rules
for
plant
organs.
Although
such
generative
models
have
been
used
successfully
to
describe
floral
development
[25]
and
to
generate
realistic
rendering
of
trees
in
3-D
[26],
utilization
of
L-systems
approaches
for
quantitative
high-
throughput
phenomics,
functional
genomics
and
plant
breeding
is
in
its
infancy.
Geometric
descriptors
of
plant
organs
were
primarily
developed
for
leaf
shape
analysis
and
comprise
terms
such
as
projected
area,
center
of
mass,
eccentricity
or
symmetry,
and
statistical
moments,
which
allow
accurate
mathemat-
ical
representation
of
plant
shape
without
the
need
to
store
large
images
for
comparison
[22].
This
approach
has
re-
cently
been
adapted
in
the
analysis
software
of
a
commer-
cial
imaging
device
(MAT
or
Morphological
Analysis
Tool;
LemnaTec
Scanalyser,
www.lemnatec.com)
and
similar
information
can
also
be
extracted
from
images
collected
using
simple
imaging
systems
and
public
domain
software
[27].
The
potential
utility
of
this
shape
characterization
for
phenomics
of
Arabidopsis
is
illustrated
in
Figure
1.
In
this
experiment,
the
heterotic
behavior
of
progeny
of
a
cross
between
the
Arabidopsis
ecotypes
C24
and
Ler
was
exam-
ined
by
digital
growth
analysis
using
projected
leaf
area
(derived
from
a
series
of
top
camera
images
over
time)
and,
using
the
mathematical
shape
analysis
tools
described
above,
by
extraction
of
rosette
morphology
at
fixed
time
points.
(In
this
case,
the
means
for
20
plants
at
14
days
after
planting
are
shown;
R.T.
Furbank
and
X.
Sirault,
unpublished.)
Typical
images
of
the
parent
plants
and
the
cross
are
also
shown
after
a
thresholding
algorithm
was
applied.
Even
this
simple
experiment
can
provide
insights
into
the
genetics
controlling
rosette
morphology
if
carried
out
across
large
collections
of
genetic
material.
In
Figure
1,
shape
analysis
is
plotted
in
a
‘radar’
graph
where
data
for
each
ecotype
are
normalized
against
the
largest
value
for
a
given
character
in
the
entire
data
set,
to
give
values
between
0
and
1.
The
parents
of
the
cross
are
clearly
different,
with
Ler
having
both
a
greater
projected
area
and
much
greater
rosette
eccentricity.
The
progeny
of
the
inter-ecotype
cross
obviously
has
a
much
larger
projected
leaf
area
at
this
developmental
stage
than
both
parents
and
considerably
reduced
eccentricity
compared
to
Ler.
If
such
an
analysis
were
to
be
carried
out
over
a
large
number
of
ecotypes
with
sequenced
genomes,
recombinant
inbred
lines
or
genome-wide
gene
inactivation
lines
grown
under
controlled
conditions,
considerable
progress
could
be
made
in
understanding
the
genetics
of
plant
growth
and
mor-
phology.
As
discussed
above,
a
database
of
phenomics
informa-
tion
requires
comprehensive
metadata
description
and
agreed
ontologies.
It
is
unlikely
that
large-scale
standardi-
zation
of
experimental
conditions
and
techniques
will
be
possible
(although
an
International
Plant
Phenomics
Ini-
tiative
has
been
established
to
address
these
issues;
www.plantphenomics.com).
Well-described
metadata
miti-
gates
the
issue
of
standardized
experimental
conditions
to
some
degree
and
recently
there
have
been
two
attempts
to
provide
ontology-based
solutions
to
combining
metadata
repositories
with
phenotypic
databases
and
search
tools.
Total
leaf area
Diameter
Eccentricity
Roundness
(a) (b)
Circumference
Surface
coverage
15000
10000
5000
0
3836.20 5432.35
9769.00
C24 Ler C24XLer
Total area (pixels)
1.000
0.800
0.600
0.400
0.200
0.000
TRENDS in Plant Science
Figure
1.
Shape
analysis
of
C24,
Ler
ecotypes
of
Arabidopsis
and
a
C24

Ler
cross.
An
example
of
the
extraction
of
morphological
features
from
images
of
Arabidopsis.
Plants
were
grown
at
300
mmol
m
2
s
1
quanta,
10-h
photoperiod
and
21
8C
for
14
days
in
trays
of
20
individual
pots
and
imaged
from
above
daily
using
a
LemnaTec
Scanalyser
3-D
(LemnaTec,
Wuerselen,
Germany).
Images
were
analyzed
for
morphological
descriptors
[61]
using
the
algorithms
in
the
LemnaTec
Morphological
Analysis
Tool
(www.lemnatec.com;
LemnaLauncher
and
Miner
manual
24.07.2010.pdf).
(a)
Radar
plot
of
five
morphological
characters
of
ecotypes
C24,
Ler
and
an
F
1
cross
between
the
two
ecotypes,
derived
from
images
of
20
plants
at
14
days
after
planting.
Values
were
the
mean
of
20
values,
normalized
to
the
largest
value
of
the
data
set.
(b)
Mean
and
standard
errors
of
projected
rosette
area
of
plants
and
an
example
of
images
after
thresholding
to
remove
pixels
containing
soil
and
pot.
Review Trends
in
Plant
Science
December
2011,
Vol.
16,
No.
12
637

The
first
of
these,
Xemlab
[15],
was
conceived
primarily
to
deal
with
metabolomic
data
and
associated
metadata,
but
has
the
potential
to
form
a
basis
for
wider
application
and
is
still
under
development.
Also
under
development
at
the
Australian
Plant
Phenomics
Facility
(APPF)
is
PODD,
a
Phenomics
Ontology
Driven
Database
(http://www.plant-
phenomics.org.au/PODDProject).
PODD
is
intended
to
service
both
plant
and
animal
phenomics
and,
although
similar
in
concept
to
Xemlab
in
focusing
on
metadata
and
ontologies
linked
to
a
web-
based
graphical
user
interface,
it
utilizes
different
software
solutions
to
achieve
these
aims.
For
plant
applications,
it
is
intended
to
provide
a
mechanism
for
archiving
and
retriev-
al
of
phenotypic
data
produced
from
imaging,
spectral
analysis
and
a
vast
array
of
physiological
(phenomics)
data
produced
from
the
Australian
Plant
Phenomics
Facility.
There
is
substantial
interest
in
integrating
such
databases
with
the
genomic
information
currently
available
via
in-
ternational
databases
such
as
TAIR,
TIGR
and
NCBI,
and
with
other
‘omics’
information
such
as
metabolomic,
prote-
omic
and
transcriptomic
data.
Phenotyping
for
abiotic
stress
tolerance
in
crop
plants
As
discussed
above,
the
challenges
agriculture
currently
faces
require
significant
advances
in
both
yield
potential
and
yield
stability.
For
yield
stability,
both
abiotic
and
biotic
stress
tolerance
are
crucial
traits.
Many
of
these
traits
can
be
screened
for
at
the
seedling
stage
in
single
pots
in
controlled
environments
or
in
the
field,
and
in
many
cases
the
same
phenomics
tools
can
be
used
across
all
these
scales
of
phenotypic
screening.
Two
important
abiotic
stress
tolerance
traits
in
many
environments
are
drought
and
salinity
tolerance
[28,29].
In
many
respects,
these
stresses
produce
quite
similar
phenotypic
effects
and
the
phenomics
approaches
to
screening
show
a
high
degree
of
crossover.
One
of
the
first
effects
of
exposing
a
crop
to
salinity
(hours
to
days)
is
stomatal
closure,
induced,
at
least
in
part,
by
the
deleteri-
ous
osmotic
effect
of
solutes
on
the
ability
of
roots
to
take
up
water
from
the
soil
(reviewed
in
[28]).
This
osmotic
stress,
which
is
similar
in
nature
to
drought
stress
and
salinity
stress,
has
been
termed
‘chemical
drought’
[29].
The
effect
of
stomatal
closure
is
a
reduction
in
photosynthesis,
but
screening
based
on
photosynthetic
parameters
or
stomatal
conductance
measurements
are
generally
slow
and
often
have
low
reproducibility
[29].
As
is
the
case
with
many
plant
phenomics
tools,
a
surrogate
measurement
can
be
used
to
screen
for
stomatal
or
photosynthetic
responses
under
osmotic
stress.
One
of
the
best
examples
of
using
a
phenomics
approach
with
a
‘surrogate’
measurement
is
the
success
of
carbon
isotope
discrimination
(termed
CID
in
plant
breeding)
as
a
reproducible
indicator
of
transpiration
efficiency
in
crop
physiology
and
plant
breeding
[13].
This
technique
is
derived
from
observations
made
more
than
25
years
ago
that
plants
discriminate
against
the
heavy
isotope
of
car-
bon
(
13
C)
naturally
present
in
atmospheric
CO
2
,
both
in
the
process
of
CO
2
diffusion
into
the
leaf
and
in
the
metabolic
processes
of
photosynthesis
[30].
This
isotopic
discrimina-
tion
is
reflected
in
the
isotopic
signature
of
plant
dry
matter
and
in
C
3
crops,
CID
values
are
strongly
related
to
stomatal
conductance
and
transpiration
efficiency
for
a
given
photosynthetic
capacity
[13,30].
Not
only
has
this
proven
to
be
a
useful
research
tool,
it
has
also
been
successfully
used
to
find
genetic
variation
in
transpiration
efficiency
in
wheat
and
to
breed
commercial
varieties
with
greater
water-use
efficiency
and
yield
[30].
The
utility
of
this
approach
is
that
samples
can
be
collected
at
the
end
of
the
growing
season
and
the
isotopic
composition
reflects
the
integrated
effect
of
the
entire
growing
season,
avoiding
the
issues
of
measuring
leaves
and
plant
organs
at
key
phenological
stages.
The
disadvantages
of
CID
are
expense
(up
to
US$30
per
sample),
and
the
need
to
normalize
data
to
photosynthetic
capacity
or
yield
potential
to
obtain
varieties
which
are
both
good
performers
in
terms
of
growth
and
yield
and
conservative
consumers
of
water.
It
is
not
known
whether
CID
measurements
can
be
scaled
up
to
the
field
from
seedling
measurements
in
controlled
environments.
Recently,
infrared
thermography
has
been
successfully
used
at
the
young
seedling
stage
in
wheat
and
barley
to
select
genotypes
capable
of
maintaining
stomatal
conduc-
tance
under
osmotic
stress
(Box
2;
[31]).
In
this
case,
salt
was
used
to
induce
osmotic
stress,
but
this
technique
is
also
applicable
to
high-throughput
seedling
screening
for
drought
tolerance
in
the
vegetative
stages
of
crop
develop-
ment.
Such
screens
early
in
plant
development
allow
many
thousands
of
lines
to
be
assessed
rapidly
and
at
low
cost
relative
to
techniques
requiring
measurements
across
the
whole
lifecycle.
Although
traits
phenotyped
on
seedlings
in
isolated
pots
might
not
always
hold
up
when
scaled
to
the
field,
in
this
case
identical
ranking
of
genotypes
for
salinity
tolerance
was
observed
both
in
the
seedling
and
the
adult
plant
stage
[31].
The
respective
pros
and
cons
of
pot
versus
field
experiments
are
likely
to
depend
on
the
traits
being
measured
–
some
will
be
robust
to
the
reduced
reality
of
the
pot
in
a
controlled
environment
and
others
will
not.
It
is
likely
that
traits
such
as
yield
and
maintenance
of
yield
on
reduced
water
supply
will
be
more
greatly
affected
by
growth
in
pots
than
less
complex
traits
such
as
the
osmotic
component
of
salinity
tolerance.
Infrared
thermography
Infrared
thermography
or
even
simple
automated
spot
can-
opy
temperature
measurements
also
have
great
potential
for
low-cost,
high-throughput
field
phenotyping.
Carrying
out
porometry
in
the
field
to
assess
stomatal
response
to
low
soil
water
potential
is
laborious,
even
with
modern
tools
[29].
Canopy
temperature
has
been
widely
used
to
infer
crop
water
use
and
photosynthesis
and
in
some
cases
to
predict
yield
for
close
to
30
years
[32,33].
Handheld
thermopile-
based
infrared
thermometers
or
canopy
temperature
‘guns’
can
now
be
purchased
cheaply
but
are
still
not
routinely
used
in
crop
breeding
programs.
One
reason
for
this
is
the
long
time
taken
to
walk
hundreds
of
plots
logging
canopy
temperature,
and
the
associated
changes
in
environment
and
physiological
state
of
the
crop
during
the
period
of
measurement.
Furthermore,
the
inability
to
differentiate
between
signals
originating
from
the
plant
and
those
from
the
surrounding
soil
restricts
use
to
phenological
stages
after
canopy
closure.
Mounting
of
multiple
sensors
on
a
tractor
boom
and
passing
over
the
crop
can
improve
the
Review Trends
in
Plant
Science
December
2011,
Vol.
16,
No.
12
638

speed
of
deployment,
but
an
imaging
sensor
allows
a
much
larger
number
of
plots
to
be
assessed
simultaneously
[29].
Microbolometer-based
thermal
imaging
sensors
are
improv-
ing
rapidly
in
spatial
resolution
(cameras
capable
of
imaging
640
by
480
elements
are
now
affordable)
and
these
can
be
mounted
on
platforms
above
the
crop
on
model
aircraft,
helium
balloons
or
manned
aircraft;
however,
the
speed
of
acquisition
gained
by
raising
the
height
of
the
imaging
sensor
above
the
crop
obviously
reduces
spatial
resolution
[29].
The
complexities
of
diurnal
variation
of
the
radiation
load
on
the
crop,
angle
of
view
and
solar
angle
must
still
be
accounted
for
to
obtain
biologically
meaningful
and
repro-
ducible
results
[34].
Recently,
the
use
of
distributed
sensor
networks
which
continuously
monitor
canopy
temperature
with
wireless
communication
has
also
improved
the
utility
of
such
measurements
by
increasing
the
temporal
resolution
throughout
the
growing
season
[35].
Chlorophyll
fluorescence
analysis
Chlorophyll
fluorescence
has
also
been
used
as
a
surrogate
measurement
for
maintenance
of
photosynthetic
function
under
stresses
such
as
drought.
The
most
easily
measured,
and
hence
the
most
commonly
used,
fluorescence
parame-
ter
in
stress
studies
is
dark-adapted
Fv/Fm
(a
measure
of
the
intrinsic
photochemical
efficiency
of
light
harvesting
in
photosystem
II;
[36]).
This
measurement
is
now
possible
using
affordable
commercial
instrumentation
designed
for
imaging
of
whole
leaves
or
small
plants
using
pulse
am-
plitude-modulated
(PAM),
fluorometry
[36].
It
is
feasible
in
high-throughput
to
obtain
whole-plant
average
measure-
ments
or
to
target
leaves
at
the
same
developmental
stage
if
the
commercial
systems
and
software
are
adapted
to
this
purpose.
This
is
particularly
applicable
to
high-throughput
studies
of
stress
response
in
species
which
grow
predomi-
nantly
in
the
horizontal
plane
in
the
seedling
stage
[model
plants
such
as
Arabidopsis
and
tobacco
(Nicotiana
taba-
cum)
or
seedlings
of
dicots
such
as
canola
(Brassica
napus)
or
cotton
(Gossypium
ssp.)].
For
high-throughput
and
to
minimize
the
costs
of
plant
culture,
plants
can
be
grown
in
large
trays
or,
for
small
seedlings,
in
microtitre
plates
or
similar
vessels.
Fluorescence
imaging
also
allows
the
de-
termination
of
projected
leaf
area
and
hence
the
growth
rate
if
measurements
are
made
regularly
over
time
[37].
Fv/Fm
has
recently
been
measured
in
two
drought
studies
with
Arabidopsis
using
systems
scalable
to
high-through-
put
screening
[17,38].
One
study
reported
the
use
of
a
system
for
analyzing
drought
tolerance
which
measures
pulse-modulated
chlorophyll
fluorescence
and
digital
growth
analysis
from
projected
leaf
area
in
compartmented
trays
[17].
Comparisons
of
the
time
course
of
relative
growth
rate
(RGR)
and
Fv/Fm
in
the
Arabidopsis
ecotype
C24
and
PARP
mutants
after
withholding
of
water
clearly
show
that
Fv/Fm
is
relatively
insensitive
to
drought
stress,
whereas
RGR
falls
off
rapidly
after
watering
is
ceased.
The
Box
2.
Thermography
screening
for
osmotic
tolerance
in
cereals
Thermal
imaging
or
‘thermography’
can
be
used
to
extract
leaf
or
canopy
temperature
at
the
single
plant
level
(Figure
I)
or
the
plot
level
(Figure
II).
Figure
I
shows
thermal
images
of
durum
wheat
seedlings
treated
with
salt
for
3
days
before
measurement
compared
to
a
control
(adapted
from
[31]).
The
approximate
differential
in
tempera-
ture
for
the
control
and
treated
seedlings
is
1
8C.
With
careful
control
of
environmental
conditions,
the
magnitude
of
this
differential
can
be
used
to
screen
for
tolerance
to
osmotic
stress
and
has
been
validated
for
a
range
of
durum
wheat
genotypes
using
traditional
screening
based
on
biomass
accumulation
[31].
Thermography
can
also
be
used
in
the
field
as
a
remote
sensing
tool
to
capture
canopy
temperature
data
for
a
large
number
of
plots.
Figure
II
shows
a
thermal
image
of
a
wheat
trial
imaged
from
a
cherry
picker,
comprising
5
m

2
m
plots
per
genotype,
arrayed
in
a
grid
(Deery,
Sirault
and
Furbank,
unpublished).
Differences
in
water
extraction
or
stress
tolerance
can
be
detected
by
comparing
average
temperatures
of
each
plot
following
image
processing
to
remove
soil
signals
and
correction
for
solar
fluxes
and
heat
balance
[29].
Δ~1˚C
TRENDS in Plant Science
Figure
I.
Thermal
images
of
durum
wheat
seedlings
treated
with
salt.
˚C 33.3
19.9
Dist = 45 Trefl = 22.5
ε = 0.95
TRENDS in Plant Science
Figure
II.
Thermal
image
of
a
wheat
trial
imaged
from
a
cherry
picker.
Review Trends
in
Plant
Science
December
2011,
Vol.
16,
No.
12
639

drought
studies
show
that
dark-adapted
Fv/Fm
is
useful
mainly
as
an
osmotic
stress
‘survival’
screen,
which
is
of
limited
utility
for
many
annual
crops
but
might
be
of
more
use
in
perennials
or
where
sporadic
rain
events
are
expe-
rienced
through
the
growing
season.
A
similar
result
was
found
for
salinity
stress,
with
Fv/Fm
showing
no
more
stress
sensitivity
than
measurement
of
leaf
chlorophyll
content
[39].
In
the
case
of
tissue
tolerance
to
salt,
fluores-
cence
imaging
might
be
able
to
provide
valuable
leaf
level
information
on
the
pattern
of
accumulation
of
sodium
and
its
effects
on
the
photosynthetic
apparatus.
Alternative
quenching
parameters
which
can
be
derived
from
chlorophyll
fluorescence
analysis
are
electron
trans-
port
rate
(ETR)
or
non-photochemical
quenching
(NPQ).
Chlorophyll
fluorescence
has
not
been
widely
used
as
a
surrogate
for
ETR
under
abiotic
stress,
although
this
can
be
calculated
from
chlorophyll
fluorescence
quenching
if
the
incident
light
intensity
and
the
absorption
properties
are
known
[40].
Although
ETR
is
challenging
to
estimate
on
a
whole-plant
basis
owing
to
the
requirement
for
light
inter-
ception
to
be
quantified,
it
should
prove
much
more
sensitive
than
Fv/Fm
in
stress
tolerance
studies.
NPQ
has
frequently
been
used
as
an
indicator
of
stress
in
both
model
plants
and
crop
species
[36].
This
parameter
is
related
to
the
dissipation
of
energy
from
the
photosynthetic
apparatus
as
heat
and
is
a
sensitive
measure
of
photo
protection
through
the
xantho-
phyll
cycle
[36].
Under
stress
conditions,
NPQ
is
often
seen
to
rise
significantly,
either
because
of
photo
protection
of
photosystem
II
or
due
to
feedback
on
electron
transport
from
inhibition
of
carbon
metabolism
[36].
For
cold,
heat
and
UV
stress,
where
the
photosynthetic
light-harvesting
apparatus
is
often
the
first
point
of
dam-
age,
chlorophyll
fluorescence
can
be
a
useful
early
screen-
ing
system
for
stress
tolerance
[17].
In
the
case
of
cold
and
heat
stress,
effects
on
photosynthesis
and
even
changes
in
membrane
lipid
properties
can
lead
to
immediate
effects
on
chlorophyll
fluorescence
[41],
whereas
UV
stress
can
result
in
oxidative
damage
to
the
photosystems,
once
again
per-
ceived
as
a
loss
of
efficiency
of
light
harvesting,
useful
as
a
screening
tool
for
tolerance
to
UV-B
exposure
[42].
High-throughput
chlorophyll
fluorescence
screening
of
crop
plants
after
the
seedling
stage
is
somewhat
problem-
atic.
Cereals
and
dicot
crops
with
complex
vegetative
structure
are
difficult
to
accurately
image
without
con-
struction
of
a
full
3-D
model,
which
requires
images
to
be
acquired
from
multiple
viewing
angles
[43].
Using
pulse-modulated
fluorescence
imaging
to
acquire
this
in-
formation
is
also
challenging,
because
a
saturating
pulse
of
light,
usually
produced
by
LED
panels
in
commercial
instruments,
must
be
applied
to
the
entire
plant.
Obtain-
ing
a
‘whole-plant’
Fv/Fm
for
a
mature
crop
plant
is
possi-
ble
without
full
3-D
reconstruction,
but
the
value
of
this
is
yet
to
be
demonstrated.
The
technical
challenges
of
field
phenotyping
using
chlorophyll
fluorescence
are
even
great-
er.
PAM
fluorometry
is
at
present
limited
to
single
leaf
measurements
in
the
field
using
either
handheld
sensors
or
‘monitoring’
fluorometers
which
can
be
left
attached
to
a
single
leaf
over
a
period
of
weeks
or
months.
An
interesting
development
for
crop
application
is
the
LIFT
(laser
induced
fluorescence
transient)
system,
which
does
not
require
a
saturating
pulse
of
light
to
be
applied
in
order
to
deconvolute
the
chlorophyll
fluorescence
signal
[44].
Developed
for
re-
mote
sensing
applications
and
not
an
imaging
system,
this
instrument
has,
however,
been
shown
to
provide
reproduc-
ible
results
similar
to
those
obtained
with
PAM
both
for
whole-plants
and
single
leaves
[44].
Another
established
optical
technique
related
to
chloro-
phyll
fluorescence
with
utility
in
stress-related
phenomics
is
leaf
spectroscopy
or
hyperspectral
reflectance
spectros-
copy
using
radiometric
or,
more
recently,
imaging
sensors
[34].
Leaves
of
crops
absorb
incoming
light
strongly
in
the
region
400–700
nm
but
reflect
light
strongly
in
the
near-
infrared
(700–1100
nm)
and
shortwave
infrared
(to
2500
nm).
Well-established
indices
such
as
NDVI
(normal-
ized
difference
vegetation
index)
have
been
developed
to
relate
leaf
chlorophyll
content
and
crop
biomass
to
spectral
reflectance
features
and
are
widely
used
in
remote
sensing,
crop
physiology
and
precision
agriculture
[34].
Similarly,
an
index
known
as
PRI
(photosynthetic
reflective
index)
has
been
proposed
as
a
crop
stress
indicator
based
on
the
absorption
features
of
the
photoprotective
pigments
known
as
xanthophylls
in
leaves
[45].
Leaf
reflectance
can
also
be
used
to
detect
steady-state
chlorophyll
fluorescence
in
the
field
[46]
using
spectral
features
between
699
and
710
nm.
Also,
since
reflectance
in
the
short-wave
infrared
region
is
strongly
influenced
by
tissue
water
content,
there
might
be
opportunities
to
use
this
region
of
the
reflectance
spectrum
as
a
surrogate
for
relative
leaf
water
content
[47].
There
is
also
scope
for
developing
indices
in
the
short-wave
infrared
for
other
plant
tissue
components
such
as
protein
nitrogen
and
carbohydrate.
Hyperspectral
reflectance
spectroscopy
has
not
com-
monly
been
used
in
plant
breeding,
although
reflectance
spectrometers
in
the
visible/near-infrared
region
are
now
very
affordable
and
commercial
LED-based
instruments
measuring
NDVI
are
in
common
agricultural
use.
A
major
limitation
to
the
utility
of
hyperspectral
data,
in
common
with
the
difficulties
in
interpreting
canopy
temperature
data,
is
variability
in
environmental
conditions
during
measurements
[34].
Most
spectrometers
are
passive
in
nature,
relying
on
solar
radiation
as
the
light
source,
leading
to
major
difficulties
in
quantitatively
comparing
results
between
plots
and
genotypes
owing
to
cloud
cover
and
changes
in
solar
angle
and
measurement
angle
during
the
photoperiod
[34].
The
availability
of
more
reasonably
priced
imaging
spectrometers
might
reduce
this
problem,
particularly
if
combined
with
low-level
aerial
observation
of
crops
in
the
field
[34].
Digital
growth
analysis
As
discussed
above,
one
of
the
least
complicated
but
useful
methods
for
quantitatively
determining
stress
tolerance
is
digital
growth
analysis.
Simple
analysis
of
projected
leaf
area
in
model
plants
has
proven
useful
and
the
avail-
ability
of
commercial
systems
for
carrying
out
quasi
‘3-D’
digital
growth
analysis
on
crop
species
have
meant
that,
such
approaches
are
becoming
more
popular
for
in
situ
crop
phenotyping
in
controlled
environment
facilities
(www.plantphenomics.com
and
www.plantphenomics.org.
au).
This
technique
uses
multiple
viewing
angles
(usually
two
side
views
and
a
top
view)
to
extract
a
mathematical
relationship
between
these
three
digital
images
and
Review Trends
in
Plant
Science
December
2011,
Vol.
16,
No.
12
640

biomass
or
leaf
area
[48].
The
correlation
between
digital
estimation
of
leaf
area
and
that
obtained
for
destructive
harvest
can
exhibit
an
r
2
value
of
greater
than
0.9
[48].
This
technique
will
have
greater
accuracy
early
in
plant
devel-
opment
and
will
reduce
in
utility
as
occlusions
become
problematic,
such
as
after
tillering
in
cereals.
Digital
imaging
of
growth
over
a
period
of
plant
devel-
opment
allows
assessment
of
the
sum
of
stress
response
mechanisms
and
offers
the
opportunity
to
tease
apart
many
of
these
responses.
For
example,
shortly
after
appli-
cation
of
salt,
just
as
stomata
shut
(discussed
above),
inhibition
of
plant
growth
also
occurs
rapidly,
which
is
independent
of
the
accumulation
of
the
salt.
This
provides
an
opportunity
to
separate
salt-dependent
and
salt-inde-
pendent
components
of
plant
responses
to
salinity.
After
longer
exposure
to
salinity,
leaf
senescence
can
be
quanti-
fied
by
separating
yellow
and
green
areas
of
the
leaf,
and
this
can
be
related
to
the
tolerance
of
tissues
to
accumu-
lated
salt
[28,48].
With
non-destructive
image
analysis,
these
components
of
salinity
tolerance
can
be
measured
on
a
single
plant.
Furthermore,
they
can
be
measured
rapidly
and
accurately,
so
these
components
can
be
measured
in
large
populations
–
such
as
mutant
populations,
or
map-
ping
populations
–
which
enables
a
genetic
approach
to
be
undertaken
to
identify
genes
underlying
variation
in
these
respective
components
of
tolerance.
Digital
imaging
in
visible
wavelength
regions
provides
information
not
only
on
plant
size,
but
also
on
the
color
of
the
plants,
thus
enabling
quantification
of
senescence
arising
from,
for
example,
nutrient
deficiencies
or
toxici-
ties,
or
pathogen
infections.
This
approach
has
recently
been
validated
in
an
experiment
where
digital
imaging
was
used
to
quantify
toxicity
of
germanium
(as
a
toxic
analogue
of
boron)
in
a
mapping
population
of
barley
[49]
to
identify
a
QTL
at
the
same
locus
as
previously
identified
for
boron
tolerance
using
a
visual
score
of
symptoms
[50].
Transpiration
of
plants
can
also
be
measured
over
time
by
automatically
monitoring
water
consumption
gravimetri-
cally.
When
combined
with
the
measurement
of
plant
growth
using
digital
imaging,
water-use
efficiency
(WUE)
can
be
monitored
through
the
life
of
the
plant,
and
effects
of
environment
and
genetic
make-up
on
WUE
can
be
tested.
The
first
attempts
at
this
have
been
recently
published
[51].
Phenomic
screening
for
biotic
stress
tolerance
Non-destructive
techniques
such
as
digital
imaging
in
the
visible
spectrum
and
imaging
of
chlorophyll
fluorescence
have
been
used
to
monitor
the
progress
of
disease
symp-
toms
in
leaves
for
some
years
[52].
Foliar
and
stem
fungal
pathogens
such
as
rusts
and
mildews
produce
large-scale
reprogramming
of
metabolism
soon
after
infection,
often
reflected
by
persistent
changes
in
ETR
and
NPQ
of
chloro-
phyll
fluorescence,
calculated
from
chlorophyll
fluores-
cence
images
of
the
affected
area
of
the
leaf
[52].
This
technique
allows
the
early
detection
of
symptoms
(before
symptoms
are
visible
to
the
eye),
quantification
of
the
area
of
infected
tissue
and
potentially
the
quantification
of
the
susceptible
and
resistant
response
to
pathogen
attack,
at
least
in
the
case
of
mildew
on
barley
leaves
[53,54].
Digital
imaging
in
just
the
visible
region
offers
no
advantage
in
sensitivity
over
the
detection
of
symptoms
by
eye,
but
it
provides
a
high-throughput
technique
to
quantify
lesions
or
chlorotic
areas
on
leaves.
Using
a
combination
of
careful
image
capture,
image
analysis
and
color
classification,
it
is
possible
to
follow
the
progres-
sion
of
lesions
over
time
quantitatively.
However,
this
approach
has
not
commonly
been
used
to
date
in
screens
for
pathogen
resistance
in
crop
plants.
One
of
the
reasons
that
application
of
plant
phenomics
to
pathology
has
only
recently
developed
might
be
that
the
genetics
of
resistance
to
major
pathogens
such
as
rusts
are
relatively
simple.
Rust
resistance
genes
of
the
NBS-LRR
type,
or
R-genes
[55],
act
in
a
gene-for-gene
specific
manner
with
respect
to
virulence
and
avirulence
genes
and
are
easily
scored
and
followed
in
crosses
of
crop
breeding
germplasm
using
MAS.
Recently,
however,
there
has
been
intense
interest
in
‘slow
rusting’
or
adult
plant
resistance
genes
for
rust
[56],
particularly
with
the
occurrence
of
‘super
strains’
of
rust
such
as
UG99
which
have
overcome
the
current
subset
of
R-genes
[57].
The
adult
plant
resis-
tance
genes
represent
greater
phenotypic
challenges
in
scoring
for
disease
symptoms
owing
to
the
need
to
examine
plants
at
multiple
time
points
during
the
progression
of
symptoms,
in
addition
to
genotypic
challenges,
because
these
traits
are
quantitative
and
can
be
non-race-specific
rather
than
‘gene-for-gene’.
Non-destructive
imaging
using
fluorescence
and
hyperspectral
reflectance
offers
great
promise
in
quantitative
scoring
of
such
adult
plant
resis-
tance
phenotypes.
Although
the
application
of
leaf-
and
shoot-level
phe-
nomics
to
pathogen
resistance
is
in
its
infancy,
at
least
this
screening
has
a
well-developed
basis
in
mechanistic
re-
search
on
foliar
symptoms.
High-throughput
phenomics
of
root
pathogens
has
received
little
attention
thus
far
and
the
methods
employed
to
achieve
uniform
infection
and
scoring
of
resistance
are
laborious
for
most
root
pathogens.
One
example
of
this
is
the
root
fungal
pathogen
Fusarium
oxysporum,
a
pathogen
of
many
crop
species
but
notably
of
cotton
[58].
This
pathogen
is
soil-borne,
infects
through
the
root
and
is
xylem-contained,
but
the
phenotypic
effect
is
blockage
of
the
xylem
tissue,
stunting
and
seedling
death
[58].
Currently,
disease
resistance
is
scored
by
percentage
survival
of
seedlings,
wilting
or
stunting
of
growth
at
a
fixed
time
after
germination
or
via
scoring
of
visible
symp-
toms
in
the
seedling
conductive
tissue
[58].
Disruption
of
xylem
tissue
causes
reduced
transpiration,
stomatal
clo-
sure
and
thus
potentially,
hotter
canopies,
offering
an
opportunity
for
a
rapid
non-invasive
screening
technique
for
tolerance
or
resistance
using
the
thermography
screen
for
transpiration
described
above
[31].
Pathogens
such
as
‘verticillium
wilt’
(Verticillium
dahliae)
and
‘black
root
rot’
(Thielaviopsis
basicola)
exhibit
similar
phenotypes
and
might
also
be
amenable
to
this
approach.
Direct
phenotyp-
ing
of
roots
for
pathogenic
effects
might
also
be
feasible
by
adaptation
of
imaging
techniques
to
measure
root
elonga-
tion
such
as
developed
for
screening
for
aluminum
(Al)
tolerance
[59].
Application
of
plant
phenomics
to
trait-based
physiological
breeding
The
challenge
for
comprehensive
and
quantitative
analysis
of
traits
for
physiological
breeding
has
been
the
application
Review Trends
in
Plant
Science
December
2011,
Vol.
16,
No.
12
641

of
appropriate
non-invasive
tools
to
directly
measure
these
traits
or
their
surrogates,
as
indicated
above.
Once
impor-
tant
traits
or
‘yield
components’
underpinning
a
superior
crop
variety
are
identified
(by
what
we
might
now
term
‘reverse
phenomics’),
either
a
genomic
region
needs
to
be
identified
to
select
for
this
trait
by
MAS
in
breeding
or,
in
the
case
of
multigenic
traits
commonly
encountered
in
quantitative
physiological
breeding,
a
robust
phenotypic
marker
is
pivotal.
Digital
imaging
can
be
used
for
obtaining
measures
of
biomass
and
growth,
and
for
probing
some
aspects