Classification of protein profiles from antibody microarrays using heat and detergent treatment

Article (PDF Available)inNew Biotechnology 29(5):564-70 · October 2011with33 Reads
DOI: 10.1016/j.nbt.2011.10.005 · Source: PubMed
Abstract
Antibody microarrays offer new opportunities for exploring the proteome and to identify biomarker candidates in human serum and plasma. Here, we have investigated the effect of heat and detergents on an antibody-based suspension bead array (SBA) assay using polyclonal antibodies and biotinylated plasma samples. With protein profiles from more than 2300 antibodies generated in 384-plex antibody SBAs, three major classes of heat and detergent susceptibility could be described. The results show that washing of the beads with SDS (rather than Tween) after target binding lowered intensity levels of basically all profiles and that about 50% of the profiles appeared to be lowered to a similar extent by heating of the sample. About 33% of the profiles appeared to be insensitive to heat treatment while another 17% showed a positive influence of heat to yield elevated profiles. The results suggest that the classification of antibodies is driven by the molecular properties of the antibody-antigen interaction and can generally not be predicted based on protein class or Western blot data. The experimental scheme presented here can be used to systematically categorize antibodies and thereby combine antibodies with similar properties into targeted arrays for analysis of plasma and serum.
RESEARCH
PAPER
New
Biotechnology
Volume
29,
Number
5
June
2012
Classification
of
protein
profiles
from
antibody
microarrays
using
heat
and
detergent
treatment
Anna
Ha¨ggmark
1
,
Maja
Neiman
1
,
Kimi
Drobin,
Martin
Zwahlen,
Mathias
Uhle
´
n,
Peter
Nilsson
and
Jochen
M.
Schwenk
Science
for
Life
Laboratory
Stockholm,
School
of
Biotechnology,
KTH
-
Royal
Institute
of
Technology,
Box
1031,
SE-171
21
Solna,
Sweden
Antibody
microarrays
offer
new
opportunities
for
exploring
the
proteome
and
to
identify
biomarker
candidates
in
human
serum
and
plasma.
Here,
we
have
investigated
the
effect
of
heat
and
detergents
on
an
antibody-based
suspension
bead
array
(SBA)
assay
using
polyclonal
antibodies
and
biotinylated
plasma
samples.
With
protein
profiles
from
more
than
2300
antibodies
generated
in
384-plex
antibody
SBAs,
three
major
classes
of
heat
and
detergent
susceptibility
could
be
described.
The
results
show
that
washing
of
the
beads
with
SDS
(rather
than
Tween)
after
target
binding
lowered
intensity
levels
of
basically
all
profiles
and
that
about
50%
of
the
profiles
appeared
to
be
lowered
to
a
similar
extent
by
heating
of
the
sample.
About
33%
of
the
profiles
appeared
to
be
insensitive
to
heat
treatment
while
another
17%
showed
a
positive
influence
of
heat
to
yield
elevated
profiles.
The
results
suggest
that
the
classification
of
antibodies
is
driven
by
the
molecular
properties
of
the
antibody–antigen
interaction
and
can
generally
not
be
predicted
based
on
protein
class
or
Western
blot
data.
The
experimental
scheme
presented
here
can
be
used
to
systematically
categorize
antibodies
and
thereby
combine
antibodies
with
similar
properties
into
targeted
arrays
for
analysis
of
plasma
and
serum.
Introduction
The
exploration
of
the
human
proteome
using
affinity
reagents
is
an
attractive
means
of
defining
the
protein
content
of
a
given
sample
[1].
The
objective
for
such
studies
could
be
to
use
validated
antibodies
and/or
other
affinity
binders
for
the
discovery
of
new
protein
biomarkers
to
facilitate
improved
disease
detection,
diag-
nostics
and
monitoring.
These
efforts
have
been
previously
ham-
pered
by
the
lack
of
reagents
for
a
proteome-wide
coverage,
but
today
both
commercial
providers
and
academic
groups
are
work-
ing
on
a
systematic
generation
of
renewable
reagents
toward
all
human
proteins
[2].
These
are
several
international
initiatives
such
as
the
SH2-consortium
[3],
ProteomeBinders
consortium
[4],
efforts
focusing
on
the
antibodies
toward
cancer-related
targets
[5]
and
the
Human
Protein
Atlas
project
[6].
The
sample
type,
which
is
to
be
analyzed
by
these
reagents,
generally
defines
the
analytical
design
and
setup.
For
the
analysis
of
tissues
and
cells,
immunohistochemistry
[7],
confocal
micro-
scopy
[8]
or
reverse
phase
arrays
[9]
are
being
used,
whereas
fluids
such
as
serum
and
plasma
generally
demand
that
the
affinity
reagent
is
immobilized
on
a
solid
support
to
capture
its
target
from
the
solution.
Here,
antibody
microarrays
have
become
an
important
method
[10].
Their
exploratory
format
involves
direct
labeling
of
the
sample
[11]
as
compared
to
sandwich
assays
[12],
where
antibody
pairs
are
utilized
for
antigen
detection.
Both
bead-
based
and
planar
microarrays
are
being
used
today
for
studying
plasma
protein
profiles
with
antibody
arrays
and,
as
mentioned
above,
increasing
numbers
of
reagents
are
now
becoming
available
for
these
to
expand
the
application
spectra
deeper
into
the
pro-
teome.
In
a
previous
study,
we
described
the
use
of
heat
treatment
for
polyclonal
antibodies
and
their
binding
to
protein
targets
in
plasma
and
serum
[13,14],
where
we
also
showed
that
antibodies
Research Paper
Corresponding
author:
Schwenk,
J.M.
(jochen.schwenk@scilifelab.se)
1
These
authors
contributed
equally
to
this
work.
564
www.elsevier.com/locate/nbt
1871-6784/$
-
see
front
matter
ß
2011
Elsevier
B.V.
All
rights
reserved.
doi:10.1016/j.nbt.2011.10.005
of
monoclonal
origin
were
affected
in
similar
fashion.
This
pro-
cedure
is
equivalent
to
the
‘epitope
retrieval’
methods
commonly
used
for
pretreatment
in
the
field
of
immunohistochemistry
[15]
and
used
for
analysis
of
cancer
patients’
sera
in
hospital
pathology
departments.
Presumably,
epitopes
normally
hidden
by
the
native
fold
of
the
protein
target
are,
in
some
cases,
made
accessible
for
binding
to
the
antibody
through
partial
denaturation
of
the
protein
target
by
heat
and/or
detergent
treatment.
Similarly,
we
showed
that
pretreatment
of
plasma
and
serum
samples
with
heat
makes
the
recognition
of
some
protein
targets,
but
not
all,
more
efficient
[16].
Here,
we
have
extended
this
observation
by
a
more
in-depth
study
for
a
large
set
of
polyclonal
antibodies
generated
as
part
of
the
Human
Protein
Atlas
project
[17],
which
have
been
raised
against
denatured
protein
fragments.
Using
labeled
plasma
in
assays
with
both
heat
and/or
SDS
treatment,
we
have
been
able
to
show
that
different
treatments
allow
classification
of
the
anti-
bodies
according
to
their
susceptibilities
and
that
these
effects
did
not
appear
to
be
directly
related
to
the
class
of
target
proteins.
Materials
and
methods
Antibodies
Protocols
for
antigen
selection,
cloning,
expression,
purification
and
immunization
of
rabbits,
followed
by
affinity
purification
to
yield
monospecific
polyclonal
antibodies,
and
their
characteriza-
tion
with
Western
blots
(WBs)
and
antigen
microarrays
were
applied
as
described
previously
[18,19].
All
protein
fragments
used
for
immunization
were
produced
as
fusions
with
an
albumin
binding-tag
and
a
target
protein
part
of
approximately
80–120
amino
acids.
In
total
2304
antigen-purified
polyclonal
antibodies
and
an
albumin
binding
protein
(HisABP)
were
utilized.
Bead
coupling
Antibodies
were
coupled
in
microtiter
plates
to
magnetic,
carboxy-
lated
beads
(MagPlex,
Luminex-Corp.)
as
previously
described
[13]
and
antibody
concentrations
were
normalized
using
a
liquid
handling
system
(EVO150,
Tecan).
The
coupling
solution
of
one
bead
identity
did
not
contain
any
antibody
or
protein,
later
referred
to
as
the
‘bare
bead’,
and
for
another
identity
it
contained
IgG
from
a
nonimmunized
rabbit.
A
384-plexed
bead
mixture
was
created
by
combing
equal
volumes
from
all
coupled
beads
and
the
coupling
efficiency
for
each
antibody
was
determined
via
R-Phy-
coerythrin
labeled
anti-rabbit
IgG
antibody
(Jackson
ImmunoR-
esearch).
Sample
labeling
EDTA-treated
plasma
samples
were
provided
by
Atlas
Antibodies
AB,
Sweden.
The
sample
used
was
a
mixture
of
plasma
collected
from
three
males
and
three
females
all
without
known
disease
diagnosis.
This
sample
was
labeled
and
analyzed
in
accordance
with
previous
studies
[13].
At
first,
plasma
was
thawed
at
48C
and
3
ml
were
added
to
22
ml
1
PBS
using
a
liquid
handler
(SELMA,
CyBio).
N-Hydroxysuccinimidyl
ester
of
biotinoyl-tetraoxapenta-
decanoic
acid
(NHS-PEO4-Biotin,
Pierce)
was
then
added
at
10-fold
molar
excess
over
sample
protein
concentration
followed
by
an
incubation
over
two
hours
at
48C.
The
reaction
was
stopped
by
the
addition
of
a
250-fold
molar
excess
over
biotin
of
1
M
Tris–HCl,
pH
8.0
and
incubated
for
another
20
min
at
48C
before
a
final
storage
at
808C.
As
a
second
sample,
a
pool
of
labeled
plasma
from
96
individuals,
which
was
obtained
through
the
EU
project
BIO-
NMD,
was
used.
Assay
protocol
The
samples
were
used
without
removing
unincorporated
biotin
and
diluted
1/50
(SELMA,
CyBio)
in
an
assay
buffer
composed
of
0.5%
(w/v)
polyvinylalcohol
and
0.8%
(w/v)
polyvinylpyrrolidone
in
0.1%
casein
(all
from
Sigma)
in
PBS
(PVXC)
supplemented
with
0.5
mg/ml
rabbit
IgG
(Bethyl).
The
samples
were
heat
treated
in
a
thermocycler
for
30
min
at
238C
or
568C.
Then
45
ml
was
added
to
5
ml
of
bead
mixtures
in
a
microtiter
plate
(Greiner)
and
incuba-
tion
took
place
overnight
on
a
shaker
(Grant)
at
an
ambient
temperature.
Beads
were
either
washed
in
wells
with
3
100
ml
PBS
pH
7.4,
0.05%
Tween20
(PBS-T,
also
denoted
Tween)
or
3
100
ml
0.1%
SDS
in
PBS
(denoted
SDS)
and
all
beads
washed
a
final
time
with
100
ml
PBS-T
on
a
plate
washer
(EL406,
Biotek).
Beads
were
treated
with
0.4%
paraformaldehyde
in
PBS
for
10
min,
washed
again
in
3
100
ml
PBS-T
and
resuspended
in
50
ml
PBS-
T
containing
0.5
mg/ml
R-Phycoerythrin
labeled
streptavidin
(Invitrogen)
and
incubated
for
20
min.
Finally,
wells
were
washed
with
3
100
ml
and
fluorescence
measured
in
100
ml
PBS-T.
In
the
following
we
refer
to
‘assay
conditions’
as
the
combination
of
pretreating
the
samples
at
a
defined
temperature
(238C
or
568C)
and
washing
the
beads
with
detergent
(Tween
or
SDS).
Read-out
and
statistical
analysis
The
measurements
were
performed
using
a
FlexMap3D
instrument
(Luminex
Corp.).
At
least
50
events
per
bead
ID
were
counted
and
binding
events
were
displayed
as
median
fluorescence
intensity
(MFI).
Data
processing
and
analysis
were
performed
in
R,
and
if
not
stated
otherwise,
the
intensity
values
were
processed
by
log
2
transformation
and
scaling
using
the
R
function
‘scale’.
Classifica-
tion
was
achieved
using
Self-organizing
Tree
Algorithm
(SOTA)
for
unsupervised,
hierarchical
divisive
clustering
[20].
Results
Assay
design
Antibody
suspension
bead
arrays
(SBAs)
were
performed
using
a
single,
labeled
sample
to
determine
the
susceptibility
of
the
gen-
erated
protein
profiles
to
heat
and
a
denaturing
washing
detergent.
Temperatures
of
238C
or
568C
for
heat
treatment
before
sample
incubation
were
used
and
combined
with
Tween
or
SDS
based
washing
buffers
after
the
incubation.
To
create
a
profile
across
the
four
conditions,
the
data
are
presented
in
the
order
238C
+
Tween,
568C
+
Tween,
238C
+
SDS
and
568C
+
SDS.
The
focus
was
on
arrays
built
with
randomized
selections
of
antibodies,
which
were
obtained
from
a
purification
routine
within
the
Human
Protein
Atlas
project.
Proof-of-concept
experiment
In
a
first
study
presented
in
Fig.
1,
the
alterations
from
antibodies
targeting
mainly
known
plasma
proteins
are
shown.
Experiments
were
performed
with
a
single
sample
and
triplicate
measurements
per
condition,
and
antibodies
targeted
proteins
such
as
fibrinogen
beta
chain
(FGB),
apolipoprotein
H
(APOH),
complement
factor
B
(CFB)
and
alpha-2
macroglobulin
(A2M).
In
this
small
pilot
set,
50%
of
antibody
profiles
were
reduced
in
intensity
by
heat
and
SDS
New
Biotechnology
Volume
29,
Number
5
June
2012
RESEARCH
PAPER
www.elsevier.com/locate/nbt
565
Research Paper
treatment
but
showing
different
profile
characteristics.
The
bead
with
conjugated
rabbit
IgG
revealed
a
profile
similar
to
an
uncon-
jugated
bead
(bare
bead)
wherein
heat
and
SDS
did
further
decrease
intensity
levels.
Intensities
peaked
at
568C
with
Tween
for
binders
targeting
A2M,
FGB
and
CFB,
and
a
more
gradual
decrease
across
the
conditions
was
observed
for
five
binders.
It
was
also
observed
that
different
profiles
were
shown
for
binders
raised
against
the
same
target
protein.
This
suggests
that
generated
protein
profiles
differ
in
their
susceptibility
to
heat
and/or
SDS
washing
thus
allowing
the
classification
of
antibodies
into
reactivity
classes.
The
SOTA
clustering
with
three
groups
revealed
that
the
majority
of
antibodies,
denoted
cluster-1,
were
sensitive
to
heat
and
SDS
treatment,
while
a
smaller
number
appeared
to
be
affected
with
elevated
intensity
levels
by
heat
(cluster-2)
or
SDS
treatment
(cluster-3),
respectively.
Protein
profile
clusters
Expanding
this
analysis
to
antibodies
chosen
based
on
concentra-
tion
and
specificity
testing
using
antigen
arrays
rather
than
which
protein
these
antibodies
are
targeting,
bead
arrays
were
created
with
384
different
bead
IDs,
here
referred
to
as
SBAs.
These
SBAs
were
created
with
the
support
of
liquid
handling
systems
for
antibody
concentration
normalization,
bead
washing
using
mag-
netic
separation
and
a
plate
washer
and
subsequent
bead
pooling.
We
used
2300
antibodies
in
six
bead
arrays
and
performed
the
experiments
on
the
same,
single
and
labeled
sample
as
above.
Unsupervised
clustering
was
performed
to
identify
groups
of
pro-
files
as
above.
Intensity
levels
for
the
different
antibodies
ranged
over
three
orders
of
magnitude
and
to
identify
common
suscept-
ibility
profiles
across
all
intensities,
the
measured
intensity
values
were
log
2
transformed
and
scaled.
To
investigate
the
interassay
variation
the
same
sample
was
retested
in
a
second,
independent
experiment
using
the
same
antibody
arrays
and
in
addition,
a
second
sample
was
also
analyzed
to
estimate
the
concordance
of
variations
within
and
between
samples.
Summarized
in
Fig.
2a
is
the
result
from
the
unsupervised
clustering
with
the
protein
profiles
for
one
SBA
over
the
four
conditions
with
all
replicates
represented.
The
left
and
middle
panels
show
the
same
sample
analyzed
twice
and
the
right
panel
the
independent,
additional
sample.
Both
the
cluster
profiles
and
the
number
of
antibodies
in
each
cluster
appeared
to
be
similar
in
all
three
cases
and
distinct
cluster
profiles
are
shown
in
Fig.
2b.
In
concordance
with
the
proof-of-concept
study,
the
majority
of
antibodies
were
sensitive
to
both
heat
and
SDS
treatment
while
antibodies
affected
by
only
one
of
the
conditions
clustered
in
two
smaller
groups.
When
observing
the
result
from
several
bead
arrays
separately,
a
compar-
able
distribution
of
antibodies
per
cluster
was
found
in
the
three
profile
clusters
(Fig.
3).
RESEARCH
PAPER
New
Biotechnology
Volume
29,
Number
5
June
2012
MFI [AU]
23°C+Tween
56°C+Tween
23°C+SDS
56°C+SDS
200
300
400
500
600
700
HPA002265 - A2M
MFI [AU]
23°C+Tween
56°C+Tween
23°C+SDS
56°C+SDS
150
200
250
300
350
400
450
500
HPA001654 - APOH
MFI [AU]
23°C+Tween
56°C+Tween
23°C+SDS
56°C+SDS
50
100
150
200
250
300
350
Bare bead
MFI [AU]
23°C+Tween
56°C+Tween
23°C+SDS
56°C+SDS
150
200
250
300
350
400
HPA000951 - CFB
MFI [AU]
23°C+Tween
56°C+Tween
23°C+SDS
56°C+SDS
1200
1400
1600
1800
2000
2200
2400
HPA001796 - CFB
MFI [AU]
23°C+Tween
56°C+Tween
23°C+SDS
56°C+SDS
200
300
400
500
600
700
HPA001900 - FGB
MFI [AU]
23°C+Tween
56°C+Tween
23°C+SDS
56°C+SDS
100
150
200
250
300
Rabbit IgG
MFI [AU]
23°C+Tween
56°C+Tween
23°C+SDS
56°C+SDS
5000
10000
15000
20000
25000
30000
HisABP - ALBU
3 2 10 1 2
relative Intensity [AU]
16
pAb
Cluster 1
3 2 10 1 2
relative Intensity [AU]
5
pAb
Cluster 3
3 2 10 1 2
relative Intensity [AU]
2
pAb
Cluster 2
23°C + Tween
56°C + Tween
23°C + SDS
56°C + SDS
Assay Conditions
A
B
FIGURE
1
Intensity
profiles
across
assay
conditions.
Using
a
single
sample,
a
set
of
antibodies
targeting
abundant
plasma
proteins
was
used
and
challenged
with
the
four
assay
conditions
(238C
+
Tween,
568C
+
Tween,
238C
+
SDS
and
568C
+
SDS).
(a)
The
unprocessed
intensity
profiles
are
summarized
by
average
and
standard
deviation
for
each
antibody
and
changes
in
intensity
levels
reveal
that
antibodies
appear
with
different
susceptibility
profiles.
(b)
Using
a
Self-organizing
Tree
Algorithm
(SOTA)
to
cluster
these
profiles,
three
clusters
have
been
chosen
to
summarize
the
different
profiles.
Data
from
triplicated
measurements
per
condition
are
shown
along
the
x-axis
as
238C
+
Tween,
568C
+
Tween,
238C
+
SDS
and
568C
+
SDS,
separated
by
dashed
vertical
lines.
Most
of
the
tested
antibodies
appeared
sensitive
to
heat
to
an
equal
extent
as
to
SDS
(cluster-1),
while
two
antibody
profiles
were
enhanced
by
heat
(cluster-2)
another
few
antibodies
were
not
sensitive
to
heat
(cluster-3)
but
to
SDS.
The
gray
lines
represent
antibody
profiles
while
the
red
line
is
the
profile
of
the
respective
node.
566
www.elsevier.com/locate/nbt
Research Paper
Variation
For
the
three
profile
groups,
the
coefficient
of
variation
(CV)
was
calculated
to
determine
the
intra-assay
variation
from
replicated
samples
in
the
same
experiment.
Using
unprocessed
data,
CV
values
in
the
respective
clusters
showed
congruent
pattern
over
the
four
conditions
(Fig.
4),
leading
to
the
conclusion
that
the
variation
within
the
assay
was
not
a
main
factor
in
defining
the
profile
groups.
Differences
in
CV
values
were,
however,
observed
between
the
assay
conditions.
The
median
of
the
variation
was
13%
without
heat
and
9%
with
568C
treatment;
when
SDS
was
included
in
the
washing
buffer
CV
values
decreased
further
to
5%
and
4%,
respectively.
There
was
a
decrease
in
intra-assay
variation
when
SDS
was
used
as
washing
detergent
with
overall
intensity
levels
decreasing
likewise
(Supplementary
Fig.
1).
Next,
the
inter-
assay
correlation
between
two
independent
assays
was
determined
for
4
SBAs.
Here,
washing
with
Tween
was
found
to
reveal
a
correlation
of
0.96
(238C:
0.96–0.97
and
568C:
0.96–0.99),
while
including
SDS
reduced
the
correlation
(238C:
0.42–0.67
and
568C:
0.63–0.75)
as
shown
in
Supplementary
Fig.
2.
It
was
shown
that
washing
with
SDS
did
greatly
reduce
signal
intensity
levels
and
variance
between
replicates
in
the
same
experiment
but
added
a
noticeable
variance
to
the
data
when
independent
experiments
were
compared.
By
contrast,
heat
appeared
to
improve
the
overall
correlation.
Cluster
affiliation
The
concordance
in
cluster
affiliation
was
studied
using
one
SBA
for
both
re-analysis
of
the
same
sample
and
an
independent
sample
(Supplementary
Tables
1
and
2).
It
was
shown
that
the
concordance
between
the
measurements
was
cluster-1
>
cluster-
2
>
cluster-3.
For
the
analysis
of
both
samples,
almost
an
equal
percentage
of
over
90%
of
antibodies
were
found
for
cluster-1
and
about
50%
of
antibodies
in
the
remaining
clusters
were
classified
nonconcordantly.
This
reclassification
suggests
that
antibodies
revealing
profiles
of
cluster-1,
where
a
similar
effect
is
achieved
from
heat
or
SDS
treatment,
were
most
concordant
between
the
New
Biotechnology
Volume
29,
Number
5
June
2012
RESEARCH
PAPER
FIGURE
2
Susceptibility
profiles
and
cluster
concordance.
(a)
Exemplified
are
clusters
of
susceptibility
profiles
obtained
from
SOTA
clustering
of
assays
of
one
SBA.
Here,
reproducibility
was
tested
twice
using
the
same
sample
(left
and
middle
panel)
and
using
an
independent
sample
(right
panel).
The
overall
distribution
of
these
numbers
of
antibodies
per
cluster
appeared
to
be
similar.
In
an
independent
analysis
of
the
same
sample,
there
was
an
87%
overlap
obtained
between
experiments
regarding
the
antibodies
found
in
the
three
clusters
(98%,
90%
and
66%).
For
two
independent
samples
an
overlap
of
65%
(92%,
51%
and
31%)
was
determined.
The
gray
lines
represent
antibody
profiles
while
the
red
line
is
the
profile
of
the
respective
node.
Details
can
also
be
obtained
from
Supplementary
Tables
1
and
2.
(b)
Illustration
of
the
three
major
profile
clusters.
www.elsevier.com/locate/nbt
567
Research Paper
analyses.
The
variation
data
also
generally
suggest
that
Tween
and
heat
are
preferred
as
sample
treatment
and
assay
procedures
due
to
lower
variance,
but
SDS
may
well
serve
as
an
indicator
of
the
effect
that
heat
has
on
the
protein
profiles.
Comparison
to
Western
blot
As
part
of
the
Human
Protein
Atlas
project
workflow,
all
antigen-
purified
antibodies
have
been
validated
by
WB
analysis
using
a
pool
of
plasma
samples.
The
results
from
the
WB
analysis
are
scored
in
a
standardized
manner
as
supportive
(scores
1–3),
blank
(score
4)
and
nonsupportive
(scores
5–7)
in
accordance
to
the
theoretical
protein
masses
predicted
from
the
genome
sequence
[21].
In
Fig.
5,
the
results
from
the
WB
analysis
for
the
included
antibodies
are
compared
with
the
classification
obtained
from
the
bead
array
assay.
The
majority
of
the
antibodies
included
in
the
study
show
no
bands
in
the
WB
analysis
and
the
distribution
of
scores
over
the
three
clusters
are
similar,
even
though
the
heat
affected
group
(cluster-2)
contains
a
larger
portion
of
blank
WB
results
than
the
other
two.
As
an
additional
feature,
information
about
protein
classes
for
the
2000
protein
targets
in
this
assay
was
RESEARCH
PAPER
New
Biotechnology
Volume
29,
Number
5
June
2012
0%
20%
40%
60%
80%
100%
SBA-1
SBA-2
SBA-3
SBA-4
SBA-5
SBA-6
Proportion of antibodies in clusters
Cluster-3
Cluster-2
Cluster-1
FIGURE
3
Profile
distribution
across
different
bead
arrays.
A
set
of
6
suspension
bead
arrays
(SBAs)
were
classified
into
susceptibility
classes
using
SOTA
clustering.
These
2304
Human
Protein
Atlas
(HPA)
antibodies,
were
evaluated
with
one
sample
and
besides
SBA-5,
most
antibodies
appeared
sensitive
to
heat
and
SDS
treatment.
The
second
largest
proportion
were
antibodies
with
increased
intensity
levels
due
to
heat
treatment,
while
the
smallest
proportions
were
generally
those
antibodies
that
were
not
affected
by
heat
but
by
SDS
washing.
FIGURE
4
Within
profile
cluster
variation.
Three
clusters
were
formed
based
on
the
profiles
obtained
from
the
combination
of
assay
conditions
and
the
variation
determined
from
the
analysis
of
sample
replicates
was
calculated
from
unprocessed
data.
In
the
box
plots
the
distributions
of
the
CV
values
from
the
included
antibodies
are
summarized
and
it
was
found
that
the
variation
differed
between
the
assay
conditions
but
it
was
not
altered
between
in
the
clusters
and
for
the
respective
conditions.
FIGURE
5
Concordance
to
Western
blot
analysis.
For
classification
experiments
the
same
sample
as
employed
in
Western
blot
(WB)
analysis
and
scoring
[21]
was
used.
When
grouping
the
antibodies
in
accordance
to
the
their
clusters
and
WB
data
into
supportive
(scores
1–3),
blank
(score
4)
and
nonsupportive
groups
(scores
5–7),
it
shows
that
the
majority
of
antibodies
do
not
show
a
single
band
in
WB
analysis
irrespective
of
which
cluster
they
belong
to.
Nevertheless,
almost
70%
of
antibodies
found
in
cluster-2
were
blank
in
a
WB
analysis.
568
www.elsevier.com/locate/nbt
Research Paper
collected
and
summarized
for
the
clustering
groups
in
Supplemen-
tary
Table
3.
This
shows
no
particular
protein
class
enrichment
in
any
of
the
clusters
compared
to
the
overall
distribution.
This
and
the
previous
presented
coefficients
of
variation
for
the
different
clusters
over
the
conditions
indicate
that
no
apparent
technical
effects
may
have
driven
the
formation
of
the
three
clusters.
Thus,
the
determined
profiles
do
appear
to
rather
differ
due
to
the
nature
of
the
interaction
than
which
protein
or
class
of
proteins
they
may
target.
Discussion
The
presented
approach
shows
the
use
of
heat
and
SDS
washing
for
the
classification
of
protein
profiles
defined
by
antibodies
from
highly
multiplexed,
magnetic
bead
arrays.
The
study
makes
use
of
a
workflow
employing
384-plex
bead
arrays
for
the
analysis
of
biotinylated
and
nonfractionated
plasma
samples,
which
are
trea-
ted
at
238C
or
568C
before
the
incubation
and
washed
with
Tween
or
SDS
containing
buffers
afterwards.
The
results
show
that
the
intensity
values
obtained
from
the
protein
profiling
using
different
assay
conditions
allow
clustering
of
the
antibodies
into
three
major
susceptibility
groups.
Here
we
have
used
SDS
and
Tween
as
detergent
agents,
because
both
are
commonly
used
in
immunoassays,
with
Tween
being
generally
received
as
nondenaturing
and
previously
applied
in
the
bead
array
procedure.
By
contrast,
the
effect
of
SDS
on
proteins
is
complex
and
differs
for
different
proteins
and
protein
classes
[22].
In
the
presented
assay,
this
change
in
charge
and
conformation
will
occur
on
the
protein
of
interest
with
its
potentially
associated
binding
partners
as
well
as
the
capture
antibody.
It
was
observed
that
SDS
decreased
the
intensity
levels
of
nearly
all
antibodies
and
this
reagent
showed
low
intra-assay
variability,
but
an
extended
interassay
variability.
This
suggests
that
washing
with
SDS
at
the
previously
selected
concentration
of
0.1%
is
a
delicate
and
cur-
rently
not
a
robust
or
preferred
procedure.
Further
assay
develop-
ment
is
thus
required
and
should
include
other
detergents
of
similar
chemical
properties.
Nevertheless,
it
was
interesting
to
observe
that
heat
and
SDS
were
shown
to
decrease
profiles
to
a
similar
extent,
and
it
requires
further
investigations
if
similar
mechanisms
for
signal
intensity
reduction
can
be
identified
for
both
assay
conditions.
A
SOTA
was
used
to
cluster
the
data.
We
have
chosen
to
limit
the
classification
to
three
main
groups,
but
obviously
it
is
possible
to
extend
the
classification
to
larger
groups
if
appropriate.
In
Sup-
plementary
Fig.
3,
we
have
given
examples
of
classes
obtained
when
the
numbers
of
clusters
are
increased
from
3
to
4
and
6.
Here,
the
expected
decrease
in
diversity
within
a
cluster
becomes
appar-
ent
with
increasing
numbers
of
clusters
(antibody
profiles
become
more
concordant
with
the
node
profile).
By
contrast,
the
newly
introduced
node
profiles
present
variants
of
the
profiles
shown
in
Fig.
2b.
Differences
are
the
altered
distance
between
some
of
the
four
conditions
levels,
thus
representing
intermediates
of
cluster-1
and
cluster-3
or
cluster-2
and
cluster-3.
We
have
investigated
the
concordance
in
classifying
antibody
profiles
in
two
independent
experiments
involving
the
same
samples
as
well
as
sample
obtained
from
a
different
provider.
A
greater
concordance
was
hereby
determined
for
the
reclassifica-
tion
(87%)
compared
to
using
a
second
sample
(65%).
In
both
cases
a
concordance
greater
than
90%
was
seen
for
cluster-1,
which
shows
that
a
large
proportion
of
antibody
signals
decrease
with
heat
regardless
of
the
sample.
On
the
other
end,
the
greatest
differences
between
the
two
samples
were
found
for
antibodies
classified
as
‘heat
insensitive’.
As
also
discussed
previously
[16],
the
profile
enhancing
effects
of
heat
treatment
can
be
increased
epitope
accessibility
by
complete
or
partial
unfolding
events
revealing
the
epitope,
dissociation
of
hindering
complexes,
com-
pared
to
diminishing
effects
such
as
precipitation,
complex
aggre-
gation
and
complete
or
partial
folding
events,
which
would
alter
structural
epitopes.
The
present
investigation
shows
that
a
greater
extent
of
antibody
susceptibility,
which
is
shown
by
a
reduction
in
intensity
levels
after
heat
treatment,
appears
to
be
sample
independent.
In
comparison,
only
half
of
the
antibodies
with
increased
intensities
appeared
sample
independent
and
only
30%
of
the
‘heat
insensitive’
binders
found
common
between
different
samples.
Currently,
it
cannot
be
determined
whether
a
compo-
nent
common
for
the
two
samples
is
responsible
for
such
an
observation.
For
all
antibodies
employed
in
this
study,
WB
analysis
was
performed
on
a
single
sample
to
score
the
antibodies
based
on
the
recognized
bands
in
correspondence
to
the
predicted
mole-
cular
mass
of
the
target
[21].
Because
there
are
major
differences
in
sample
preparation,
analytical
concept
and
conditions,
applic-
ability
and
detection
levels
of
the
antibodies
may
differ
signifi-
cantly
between
the
orthogonal
methods.
The
WBs
make
use
of
a
sample
that
has
been
depleted
from
albumin
and
IgG,
which
changes
the
sample
composition
and
other
associated
proteins
are
eventually
being
codepleted.
By
contrast,
the
sample
complex-
ity
is
dissected
over
the
different
molecular
masses,
compared
to
the
arrays,
where
a
diluted
copy
of
neat
plasma
is
analyzed.
Labeling
samples
for
the
arrays
may
potentially
lead
to
epitopes
being
masked.
On
the
one
hand,
proteins
interacting
with
the
target
of
interest
may
additionally
hide
epitopes,
but
on
the
other
hand
these
partners
may
also
contribute
to
increased
intensity
levels
in
the
case
where
the
antibody
captures
its
target,
which
carries
along
these
associates.
When
investigating
the
protein
classes
of
the
antibody
targets
in
terms
of
their
cluster
affiliations,
no
apparent
relative
differ-
ence
between
the
clusters
could
be
observed
(Supplementary
Table
3).
While
our
current
data
so
far
support
the
idea
that
the
binding
characteristics
of
the
generated
antibodies
with
the
applied
assay
conditio ns
seem
to
be
responsible
for
the
observed
classifications,
it
will
be
necessary
to
investigate
further
if
any
other
leading
features
such
as
epitopes,
proteins,
antibody
type,
assay
design
and
analyzed
sample
can
be
identified
and
to
expand
possible
molecular
speculations.
Predicting
heat
and/
or
SDS
susceptibility
would
be
an
important
aid
in
systematically
selecting
antibodies
for
biomarker
studies
on
exploratory
affi-
nity
arrays,
but
using
the
currently
available
tools
and
informa-
tion,
such
a
prediction
appears
improbably
to
be
possible.
On
the
basis
of
our
findings
here,
it
is
the
molecular
property
of
an
antibody
that
defines
its
class
affiliation.
Thus,
further
investi-
gations
are
needed
to
identify
any
common
or
fundamental
antibody
or
antigen
attributes
that
could
allow
describing
anti-
body
susceptibility.
To
our
current
knowledge,
this
is
the
first
report
describing
the
use
of
384-plex
antibody
arrays
based
on
magnetic
beads,
while
others
have
developed
bead
arrays
of
higher
density
on
New
Biotechnology
Volume
29,
Number
5
June
2012
RESEARCH
PAPER
www.elsevier.com/locate/nbt
569
Research Paper
nonmagnetic
particles
[23].
The
use
of
magnetic
beads
is
favored
because
it
allows
a
straightforward
implementation
into
liquid
handling
systems.
We
have
established
a
procedure
for
these
384-
plex
arrays
that
allows
analyzing
384
samples
in
a
single
experi-
ment
thus
generating
up
to
150,000
data
points
from
immunoas-
says
(Igel
U
and
Schwenk
JM,
data
unpublished).
These
profiling
efforts
will
consequently
enable
larger
scaled
and
more
extended
screenings,
but
demand
that
the
included
reagents
are
carefully
selected
and
evaluated.
Instead
of
additionally
analyzing
the
sets
of
384
or
more
antibodies
on
further
methods,
the
described
classification
procedure
presents
a
direct
and
integrative
approach
to
judge
and
compare
antibody
profiles.
The
presented
study
suggests
that
antibodies
on
affinity
arrays
can
be
classified
according
to
the
protein
profiles
they
generate
when
challenged
with
different
assay
conditions.
Affinity-based
approaches
in
combination
with
heat
treatment
have
the
poten-
tial
to
reveal
more
than
the
currently
described
plasma
biomarker
candidates
[14,16,24],
and
even
though
no
association
between
the
susceptibility
clusters
and
WB
performance,
protein
class
affiliations
could
yet
be
identified,
the
described
procedure
should
be
applied
to
all
affinity
reagents
to
eventually
reveal
preferred
protein
profiles
for
poly-
and
monoclonal
antibodies
retrospec-
tively.
With
such
information
at
hand,
the
route
to
optimized
experimental
conditions
and
increased
understanding
of
the
molecular
interactions
will
be
enhanced
significantly
and
aid
efforts
for
an
affinity-based
analysis
of
the
plasma
proteome.
Acknowledgements
We
would
like
to
thank
Per-A
˚
keNygren
for
a
fruitful
discussion
and
Fridtjof
Lund-Johansen
for
suggestions
regarding
SDS.
We
would
also
like
to
thank
the
entire
staff
of
the
Human
Protein
Atlas
for
their
great
work.
This
study
was
supported
by
AFFINOMICS
an
EC
FP7
Collaborative
Project,
by
the
ProNova
VINN
Excellence
Centre
for
Protein
Technology
(VINNOVA,
Swedish
Governmental
Agency
for
Innovation
Systems)
and
by
grants
from
the
Knut
and
Alice
Wallenberg
Foundation.
Appendix
A.
Supplementary
data
Supplementary
data
associated
with
this
article
can
be
found,
in
the
online
version,
at
doi:10.1016/j.nbt.2011.10.005.
References
1
Uhlen,
M.
(2008)
Affinity
as
a
tool
in
life
science.
Biotechniques
44,
649–654
2
Colwill,
K.
et
al.
(2011)
A
roadmap
to
generate
renewable
protein
binders
to
the
human
proteome.
Nat.
Methods
8,
551–558
3
Uhlen,
M.
et
al.
(2008)
A
pilot
project
to
generate
affinity
reagents
to
human
proteins.
Nat.
Methods
5,
854–855
4
Taussig,
M.J.
et
al.
(2007)
ProteomeBinders:
planning
a
European
resource
of
affinity
reagents
for
analysis
of
the
human
proteome.
Nat.
Methods
4,
13–17
5
Bjorling,
E.
et
al.
(2008)
A
web-based
tool
for
in
silico
biomarker
discovery
based
on
tissue-specific
protein
profiles
in
normal
and
cancer
tissues.
Mol.
Cell.
Proteomics
7,
825–844
6
Berglund,
L.
et
al.
(2008)
A
genecentric
Human
Protein
Atlas
for
expression
profiles
based
on
antibodies.
Mol.
Cell.
Proteomics
7,
2019–2027
7
Kampf,
C.
et
al.
(2004)
Antibody-based
tissue
profiling
as
a
tool
for
clinical
proteomics.
Clin.
Proteomics
1,
285–300
8
Barbe,
L.
et
al.
(2008)
Toward
a
confocal
subcellular
atlas
of
the
human
proteome.
Mol.
Cell.
Proteomics
7,
499–508
9
Grubb,
R.L.
et
al.
(2003)
Signal
pathway
profiling
of
prostate
cancer
using
reverse
phase
protein
arrays.
Proteomics
3,
2142–2146
10
Ayoglu,
B.
et
al.
(2011)
Systematic
antibody
and
antigen-based
proteomic
profiling
with
microarrays.
Expert
Rev.
Mol.
Diagn.
11,
219–234
11
Haab,
B.B.
et
al.
(2001)
Protein
microarrays
for
highly
parallel
detection
and
quantitation
of
specific
proteins
and
antibodies
in
complex
solutions.
Genome
Biol.
2
RESEARCH0004
12
Hsu,
H.Y.
et
al.
(2008)
Miniaturized
parallelized
sandwich
immunoassays.
Methods
Mol.
Biol.
428,
247–261
13
Schwenk,
J.M.
et
al.
(2008)
Antibody
suspension
bead
arrays
within
serum
proteomics.
J.
Proteome
Res.
7,
3168–3179
14
Schwenk,
J.M.
et
al.
(2010)
Comparative
protein
profiling
of
serum
and
plasma
using
an
antibody
suspension
bead
array
approach.
Proteomics
10,
532–540
15
D’Amico,
F.
et
al.
(2009)
State
of
the
art
in
antigen
retrieval
for
immunohistochemistry.
J.
Immunol.
Methods
341,
1–18
16
Schwenk,
J.M.
et
al.
(2010)
Toward
next
generation
plasma
profiling
via
heat-
induced
epitope
retrieval
and
array-based
assays.
Mol.
Cell.
Proteomics
9,
2497–2507
17
Uhlen,
M.
et
al.
(2010)
Towards
a
knowledge-based
Human
Protein
Atlas.
Nat.
Biotechnol.
28,
1248–1250
18
Agaton,
C.
et
al.
(2003)
Affinity
proteomics
for
systematic
protein
profiling
of
chromosome
21
gene
products
in
human
tissues.
Mol.
Cell.
Proteomics
2,
405–414
19
Nilsson,
P.
et
al.
(2005)
Towards
a
human
proteome
atlas:
high-throughput
generation
of
mono-specific
antibodies
for
tissue
profiling.
Proteomics
5,
4327–
4337
20
Herrero,
J.
et
al.
(2001)
A
hierarchical
unsupervised
growing
neural
network
for
clustering
gene
expression
patterns.
Bioinformatics
17,
126–136
21
Bjorling,
E.
and
Uhlen,
M.
(2008)
Antibodypedia,
a
portal
for
sharing
antibody
and
antigen
validation
data.
Mol.
Cell.
Proteomics
7,
2028–2037
22
Otzen,
D.
(2011)
Protein–surfactant
interactions:
a
tale
of
many
states.
Biochim.
Biophys.
Acta
1814,
562–591
23
Wu,
W.
et
al.
(2009)
Antibody
array
analysis
with
label-based
detection
and
resolution
of
protein
size.
Mol.
Cell.
Proteomics
8,
245–257
24
Neiman,
M.
et
al.
(2011)
Plasma
profiling
reveals
human
fibulin-1
as
candidate
marker
for
renal
impairment.
J.
Proteome
Res.
10.1021/pr200286c
RESEARCH
PAPER
New
Biotechnology
Volume
29,
Number
5
June
2012
570
www.elsevier.com/locate/nbt
Research Paper
    • "After combination with biotinylated samples, bead identity and captured plasma proteins are detected using a flow cytometric analyzer. Previously, we have shown that limits of detection reach into lower ng/ml or higher pg/ml ranges while consuming less than 1 ml of plasma sample [16] for the profiling of 384 proteins [17]. The potential to screen hundreds of analytes in hundreds of patient samples simultaneously in one parallel assay allows for an effective exploration of potential candidates in a time-efficient manner. "
    [Show abstract] [Hide abstract] ABSTRACT: Author Summary Why do some malaria-infected children develop severe and lethal forms of the disease, while others only have mild forms? In order to try to find potential answers or clues to this question, we have here analyzed more than 1,000 different human proteins in the blood of more than 500 malaria-infected children from Ibadan in Nigeria, a holoendemic malaria region. We identified several proteins that were present at higher levels in the blood from the children that developed severe malaria in comparison to those that did not. Some of the most interesting identified proteins were muscle specific proteins, which indicate that damaged muscles could be a discriminatory pathologic event in cerebral malaria compared to other malaria cases. These findings will hopefully lead to an increased understanding of the disease and may contribute to the development of clinical algorithms that could predict which children are more at risks to severe malaria. This in turn will be of high value in the management of these children in already overloaded tertiary-care health facilities in urban large densely-populated sub-Saharan cities with holoendemic malaria such as in the case of Ibadan and Lagos.
    Full-text · Article · Apr 2014
    • "Only anti-C1QC capture revealed bands that indicate off-target binding. Even though it remains to be further understood whether denaturation (conformational changes) or changes in protein assemblies (dissolving complexes) altered epitope accessibility, this finding supports that the enhanced or reduced effects of heat treatment detected by single binder assays [6] for these antibodies relates to more or less capture of the intended target protein. MS-analysis of proteins captured by anti-C8B (Fig. 3D ), determined that among other proteins concordantly captured by normal goat IgG, the C8-complex including alpha, beta, and gamma chains was captured from a nonheated biotinylated plasma, as seen in Supporting InformationTable 4 and Supplementary Information. "
    [Show abstract] [Hide abstract] ABSTRACT: The increasing availability of antibodies towards human proteins enables broad explorations of the proteomic landscape in cells, tissues and body fluids. This includes assays with antibody suspension bead arrays that generate protein profiles of plasma samples by flow cytometer analysis. However, antibody selectivity is context dependent so it is necessary to corroborate on-target detection over off-target binding. To address this, we describe a concept to directly verify interactions from antibody-coupled beads by analysis of their eluates by Western blots and mass spectrometry. We demonstrate selective antibody binding in complex samples with antibodies towards a set of chosen proteins with different abundance in plasma and serum, and illustrate the need to adjust sample and bead concentrations accordingly. The presented approach will serve as an important tool for resolving differential protein profiles from antibody arrays within plasma biomarker discoveries. This article is protected by copyright. All rights reserved.
    Full-text · Article · Dec 2013
    • "Coupling of antibodies to beads was performed as previously described [25] and 30 ml of each bead identity was coupled to 1.6 mg of a different antibody. First, beads were washed twice on a magnet with activation buffer (AB) (0.1 M NaH2PO4 (Merck), pH 6.2) and subsequently beads were resuspended in 50 ml of AB. "
    [Show abstract] [Hide abstract] ABSTRACT: Patients with well-differentiated small intestine neuroendocrine tumors (WD-SI-NETs) are most often diagnosed at a metastatic stage of disease, which reduces possibilities for a curative treatment. Thus new approaches for earlier detection and improved monitoring of the disease are required. Suspension bead arrays targeting 124 unique proteins with antibodies from the Human Protein Atlas were used to profile biotinylated serum samples. Discoveries from a cohort of 77 individuals were followed up in a cohort of 132 individuals both including healthy controls as well as patients with untreated primary WD-SI-NETs, lymph node metastases and liver metastases. A set of 20 antibodies suggested promising proteins for further verification based on technically verified statistical significance. Proceeding, we assessed the classification performance in an independent cohort of patient serum, achieving, classification accuracy of up to 85% with different subsets of antibodies in respective pairwise group comparisons. The protein profiles of nine targets, namely IGFBP2, IGF1, SHKBP1, ETS1, IL1α, STX2, MAML3, EGR3 and XIAP were verified as significant contributors to tumor classification. We propose new potential protein biomarker candidates for classifying WD-SI-NETs at different stage of disease. Further evaluation of these proteins in larger sample sets and with alternative approaches is needed in order to further improve our understanding of their functional relation to WD-SI-NETs and their eventual use in diagnostics.
    Full-text · Article · Nov 2013
Show more