A preview of this full-text is provided by Springer Nature.
Content available from Nature Methods
This content is subject to copyright. Terms and conditions apply.
Brief CommuniCation
https://doi.org/10.1038/s41592-018-0179-8
1Department of Immunology, University of Oslo and Oslo University Hospital, Oslo, Norway. 2K.G. Jebsen Center for Cancer Immunotherapy, University
of Oslo, Oslo, Norway. 3Department of Pathology, Oslo University Hospital, Oslo, Norway. 4K.G. Jebsen Inflammation Research Centre, University of
Oslo, Oslo, Norway. 5NMI Natural and Medical Sciences Institute, University of Tübingen, Reutlingen, Germany. 6CLIP—Childhood Leukemia Investigation
Prague, Department of Pediatric Hematology and Oncology, 2nd Faculty of Medicine, Charles University Prague and University Hospital Motol,
Prague, Czech Republic. 7SeekQuence, Mountain View, CA, USA. 8The Brain Institute, Universidade Federal do Rio Grande do Norte, Natal, Brazil.
9Norwegian Computing Center, Oslo, Norway. *e-mail: fridtjol@gmail.com
Western blotting (WB) is widely used to test antibody speci-
ficity, but the assay has low throughput and precision. Here
we used preparative gel electrophoresis to develop a capture
format for WB. Fractions with soluble, size-separated proteins
facilitated parallel readout with antibody arrays, shotgun mass
spectrometry (MS) and immunoprecipitation followed by MS
(IP-MS). This pipeline provided the means for large-scale
implementation of antibody validation concepts proposed by an
international working group on antibody validation (IWGAV).
Researchers can choose from among a plethora of antibodies
to the proteins they study, but antibody quality is unpredictable.
There is evidence that 50% of commercially available antibodies fail
in the applications they are specified for, and off-target binding is
common1,2. The consequence is that large amounts of resources are
wasted on unproductive experiments, and there are concerns that
cross-reactive antibodies contribute to false findings3–5.
Antibodies are often advertised with images of western blots to
document specificity6. However, a blot showing a single band at the
expected position is not definitive evidence, as many proteins have
similar masses. An IWGAV has proposed five conceptual pillars for
more rigorous quality assessment: (1) targeted disruption of DNA
or RNA to generate negative controls, (2) the use of MS data as a
reference, (3) the use of different antibodies to the same protein as
references for each other, (4) expression of tagged proteins to gener-
ate positive controls and (5) IP-MS7.
These IWGAV pillars represent an important step forward,
but their implementation is hampered by limitations intrinsic to
commonly used assays. For instance, WB is typically restricted
to testing of one antibody at a time, and the method involves
steps that are difficult to standardize, such as PAGE and pro-
tein transfer/blotting. Two blots stained with the same antibody
are therefore rarely identical, and measurement of the similarity
between images is inherently difficult.
To overcome the limitations of WB for antibody-validation
tests, we converted the assay into a capture format. We biotinyl-
ated sample proteins and subjected them to preparative PAGE to
obtain 12 fractions with soluble, size-separated proteins from each
of six cell lines8 (Fig. 1a). We incubated an aliquot of each fraction
with microsphere-based barcoded antibody arrays for detection by
flow cytometry (i.e., microsphere affinity proteomics (MAP))9,10
(Fig. 1b). In practice, PAGE-MAP is ‘capture WB’ in which peaks
in antibody reactivity across the fractions are analogous to bands
(Fig. 1b). We tabulated the results in a spreadsheet in which
thousands of rows corresponded to antibodies, and the 72 columns
contained values for signal intensity measured in 12 fractions from
each of the six cell lines (Supplementary Tables 1 and 2).
The sample preparation used for PAGE-MAP is also compatible
with shotgun MS and IP-MS. For unbiased detection, we immo-
bilized fractionated proteins on streptavidin beads and carried out
on-bead trypsin digestion to generate peptides for shotgun MS (Fig. 1c,
Supplementary Table 3). Finally, we used the results obtained by
PAGE-MAP and PAGE-MS to select antibodies and fractions for sin-
gle-plex IP for direct identification of antibody targets by MS (Fig. 1d).
IWGAV pillar 3 involves the correlation of results obtained when
two antibodies to the same protein are used to measure relative pro-
tein abundance in a series of samples7. Hereinafter we refer to our
data series with 72 data points per antibody as PAGE-MAP profiles.
We used a signal-to-noise ratio of 4 (maximum signal divided by
the median) as a cutoff to select profiles for further analysis. Results
obtained after hierarchical clustering showed that PAGE-MAP pro-
files of antibodies raised against the same protein often clustered
as nearest neighbors (Fig. 2a, Methods); a total of 1,046 antibodies
identified by this approach are referred to here as clustering anti-
bodies. The average correlation between neighboring reactivity
profiles was 0.97 (Fig. 2b, Supplementary Table 2). When we tested
3,544 antibodies in a second experiment, 74% of those that clustered
in experiment 1 also clustered in experiment 2 (Supplementary
Fig. 1). Those that clustered in only a single experiment had signal-
to-noise ratios in the lower range (Supplementary Fig. 1). The aver-
age correlation between technical replicates in PAGE-MAP was 0.92
(Supplementary Table 4), and correlations for signal-to-noise ratios
of antibodies were 0.8 and 0.97 for biological and technical repli-
cates, respectively (Supplementary Fig. 2, Supplementary Table 5).
Collectively, these results show that PAGE-MAP yields complex and
reproducible signatures for each specificity, analogous to finger-
prints. The method is therefore a powerful one for the identification
of antibodies that react with the same protein (pillar 3).
A reference antibody should be validated by an independent
method or react with an epitope different from that recognized by
the test antibody7. Most antibody manufacturers consider the exact
immunogen sequence as proprietary information, and among the
6,307 antibodies tested here, we identified only 210 that clustered
with an antibody raised against a nonoverlapping epitope in the
same protein (Fig. 2b, Supplementary Table 2). Validation based on
A high-throughput pipeline for validation of
antibodies
KrzysztofSikorski 1,2, AdiMehta1,3,4, MaritInngjerdingen1, FlourinaThakor3,4, SimonKling5,
TomasKalina 6, TuulaA.Nyman 1, MariaEkmanStensland1, WeiZhou7, GustavoA.deSouza8,
LarsHolden9, JanStuchly6, MarkusTemplin5 and FridtjofLund-Johansen 1,2*
NATURE METHODS | VOL 15 | NOVEMBER 2018 | 909–912 | www.nature.com/naturemethods 909
Content courtesy of Springer Nature, terms of use apply. Rights reserved