Content uploaded by Otmar Huber
Author content
All content in this area was uploaded by Otmar Huber
Content may be subject to copyright.
Available via license: CC BY 4.0
Content may be subject to copyright.
A Nonredundant Human Protein Chip for
Antibody Screening and Serum Profiling*
Angelika Lueking‡, Alexandra Possling‡, Otmar Huber§, Allan Beveridge‡,
Martin Horn¶, Holger Eickhoff¶, Johannes Schuchardt储, Hans Lehrach‡,
and Dolores J. Cahill‡**‡‡
There is burgeoning interest in protein microarrays, but a
source of thousands of nonredundant, purified proteins
was not previously available. Here we show a glass chip
containing 2413 nonredundant purified human fusion pro-
teins on a polymer surface, where densities up to 1600
proteins/cm
2
on a microscope slide can be realized. In
addition, the polymer coating of the glass slide enables
screening of protein interactions under nondenaturing
conditions. Such screenings require only 200-
l sample
volumes, illustrating their potential for high-throughput
applications. Here we demonstrate two applications: the
characterization of antibody binding, specificity, and
cross-reactivity; and profiling the antibody repertoire in
body fluids, such as serum from patients with autoim-
mune diseases. For the first application, we have incu-
bated these protein chips with anti-RGSHis
6
, anti-GAPDH,
and anti-HSP90

antibodies. In an initial proof of principle
study for the second application, we have screened serum
from alopecia and arthritis patients. With analysis of large
sample numbers, identification of disease-associated pro-
teins to generate novel diagnostic markers may be possi-
ble. Molecular & Cellular Proteomics 2:1342–1349, 2003.
Array-based technologies are increasing in significance in
genomic and proteomic research. DNA and oligonucleotide
microarrays are widely used tools (reviewed in Refs. 1–3).
Currently, one of the main applications of DNA chips is for
gene expression profiling, a key technology in genomics to
unravel differential gene expression (4 –7). Although DNA ar-
ray technology has been used very successfully for a variety
of applications, and protein chips would complement DNA
chips as a logical development, the current use of protein
chips is limited to only a small number of laboratories (re-
viewed in Refs. 8 and 9). To a large extent this is due to the
more complex nature of proteins, requiring laborious work in
cloning, recombinant expression, and purification. Here, we
have addressed this problem by developing an automated,
robust protein expression and purification protocol to provide
thousands of highly pure human proteins and used them to
generate high-density, high-content protein arrays on poly-
mer-coated glass slides, with up to 1600 proteins per cm
2
.
Current methods to determine antibody binding and specific-
ity are still labor-intensive and complicated. They involve
screening the antibodies on nonordered,
gt11 phage library
plaque-lift filters or on tissue extracts previously separated by
two-dimensional gel electrophoresis (10 –12).
Applying protein microarray technology will greatly simplify
this process, especially the protein arrays described here
contain large numbers of proteins derived from an ordered
recombinant source. This enables the direct connection from
the protein product of the individual expression clone to its
corresponding cDNA sequence information (13). Other advan-
tages include the short time and small screening volumes
required. Moreover, many experiments can be performed in
parallel, and the entire procedure is suitable for automation.
Bu¨ ssow et al. have previously shown that cDNA libraries,
cloned into an Escherichia coli expression vector, can be
screened for protein expression on high-density filter mem-
branes (14). Using robot technology (15), a human fetal brain
cDNA expression library (hEx1) was picked into microtitre
plates, and high-density protein arrays were produced on
filter membranes followed by in situ protein expression. This
approach was extended to automated spotting of protein
microarrays from crude lysates of the expression cultures
(16). In an initial analysis of this library, more than 66% of the
clones contain inserts in the correct reading frame. Sixty-four
percent of these clones comprise full-length proteins. Further-
more, these expressed proteins are suitable for either matrix-
assisted laser desorption/ionisation-time-of-flight mass spec-
trometry and functional screening assays (17, 18). This library
has been normalized to obtain a nonredundant, unique cDNA
set (Uniclone Set) using oligonucleotide fingerprinting (19).
Automated, parallel protein expression and purification were
performed on 2413 protein-expressing clones, as described
in “Experimental Procedures.” Purified proteins were immo-
bilized on modified glass surfaces. The protein microarrays
From the ‡Max-Planck-Institute for Molecular Genetics, Ihne-
strasse 73, 14195 Berlin, Germany; §Institute of Clinical Chemistry
and Pathobiochemistry, Universita¨ tsklinikum Benjamin Franklin, Hin-
denburgdamm 30, 12200 Berlin, Germany; ¶Scienion AG, Volmer-
strasse 7a, D-12489 Berlin, Germany; 储MicroDiscovery GmbH,
Immanuelkirchstra

e 12, D-10405 Berlin, Germany; and ‡‡Centre for
Human Proteomics, Royal College of Surgeons in Ireland, Dublin 2,
Ireland and Protagen AG, Emil-Figge-Stra

e 76a, D-44227 Dortmund,
Germany
Received, March 10, 2003, and in revised form, August 25, 2003
Published, MCP Papers in Press, September 29, 2003, DOI
10.1074/mcp.T3000010-MCP200
Technology
© 2003 by The American Society for Biochemistry and Molecular Biology, Inc.1342 Molecular & Cellular Proteomics 2.12
This paper is available on line at http://www.mcponline.org
generated were tested for suitability to determine antibody
specificity, and their applicability for profiling of serum from
patients with autoimmune diseases was analyzed.
EXPERIMENTAL PROCEDURES
Expression Vector and Bacterial Strain—A cDNA library (hEx1) was
generated from human fetal brain poly(A)
⫹
RNA by oligo(dT) priming,
was size fractionated by gel filtration and directionally (SalI-NotI)
cloned in the expression vector pQE30NST (GenBank Accession No.
AF074376) and transformed into Escherichia coli strain SCS1 (Strat-
agene, La Jolla, CA) (14). The protein expression subset of this library
was normalized by oligo-fingerprinting and re-arrayed into a nonre-
dundant subset. A total of 2413 clones of this nonredundant set were
selected and used for high-throughput expression.
Sequence and Data Analysis of hEX1 cDNA Expression Library—A
total of 2303 clones of the expression library were sequenced, and
the sequences obtained were translated in the three possible reading
frames (⫹1, ⫹2, ⫹3) by the bioinformatics package GCG (Wisconsin
Package Version 10.3, Accelrys Inc., San Diego, CA). All translated
amino acid sequences were stringently searched for the epitope tag
sequence, RGSHHHHHH (RGSH
6
-tag), and the final reading frame of
the expression clone was determined.
Subsequently, the 2303 cDNA sequences were blasted against the
nonredundant protein database, NRPROT. The BLAST results for each
cDNA clone were searched, and the top five hits were tabulated. The
final reading frame determined by the RGSH
6
-tag was determined, and
the results were analyzed by searching for BLAST hits, which match the
correct reading frame. Four sets of results were generated:
1. sequences that had at least one top-five BLAST hit that cor-
responded to the correct reading frame;
2. sequences that produced BLAST hits that did not correspond
to the correct reading frame;
3. sequences that generated BLAST hits, but where the correct
reading frame is unknown; and
4. sequences that did not match anything in NRPROT.
Protein Expression and Purification in High-throughput—Each pro-
tein was expressed twice in 1-ml cultures in deep-well microtitre
plates. The protein extracts of each culture were generated as de-
scribed (16) and combined. Proteins were purified as described with
the modification, that the proteins were eluted in 30
l 35% (v/v)
acetonitrile, 0.1% (v/v) trifluoroacetic acid (17).
Generation of Protein Chips—A microscope slide treated with
Bind-Silane (Amersham Pharmacia Biotech, Piscataway, NJ) and a
covering slide treated with Repel-Silane (Amersham Pharmacia Bio-
tech) were arranged together, separated by a 30-
m thick metal
spacer. The space between the slides was filled with solution of 8%
(w/v) polyacrylamide-bisacrylamide (30:0.8). After polymerization, the
slides were washed with water and dried. The microscope slides were
placed in a Q-Array System (Genetix, New Milton, UK), equipped with
humidity control (50%) and 16 blunt-ended stainless steel print tips
with a tip diameter of 150
m.A13⫻ 13 spot pattern with a
center-to-center distance of 420
m was printed onto the slides as
duplicates. Each spot was loaded five times, resulting in a total
transfer volume of 10 nl. These ready-to-use slides can be used
without loss of signal giving reproducible results for at least 4 weeks
when stored at 4 °C.
Detection Procedure—After spotting, the protein chips were
blocked in 2% (w/v) bovine serum albumin/Tris-buffered saline, 0.1%
(v/v) Tween 20 (TBST)
1
at room temperature and incubate with pri-
mary antibody (mouse-anti-RGSH
6
(Qiagen, Valencia, CA) 1:2000
dilution; mouse-anti-GAPDH (Research Diagnostics, Inc., Flanders,
NJ) clone 6C5, 1:5,000 dilution; mouse-anti-HSP90

(Transduction
Laboratories, Lexington, KY) clone 68, 1:2,000 dilution), followed by
two 10-min TBST washes and incubation with the secondary anti-
body (rabbit-anti-mouse-IgG-Cye3 (Dianova, Hamburg, Germany),
1:800 dilution) in 2% (v/v) bovine serum albumin/TBST. Subsequently,
the protein chips were washed three times for 10 min in TBST-T (0.5%
(v/v) Triton X-100) followed by a 5-min wash with Tris-buffered saline.
Detection of the signals obtained on these protein microarrays was
performed using a 428
TM
Arrayscanner System (Affimetrix, Palo Alto,
CA). Protein microarrays used for serum profiling were blocked in 5%
(w/v) fish gelatin/TBST at room temperature, and the serum was
added (diluted 1:20 in 5% (w/v) fish gelatin/TBST). After two 10-min
TBST washes and subsequent incubation with the secondary anti-
body (mouse-anti-human immunoglobulin G (IgG; Sigma, St. Louis,
MO) 1:5,000 dilution) in 5% (w/v) fish gelatin/TBST, the protein chips
were washed three times for 10 min in TBST. This was followed by
incubation with the tertiary antibody (rabbit-anti-mouse-IgG-Cye3
(Dianova) 1:800 dilution) in 5% (w/v) fish gelatin/TBST. Subsequently,
the arrays were washed three times, each in TBST-T for 20 min.
Detection was performed as previously described. All incubation
steps were performed for 1 h. All antibody dilutions were in blocking
buffer unless otherwise stated.
Western Blot—For Western blot analysis, proteins were purified
from 35-ml bacterial cultures. Protein expression was induced with 1
m
M isopropyl

-D-thiogalactoside at an OD
578
of 0.6. After 4 h cells
were pelleted and resuspended in 3 ml 8
M urea, 0.1 M NaH
2
PO
4
,10
m
M Tris/HCl pH 8.0. After centrifugation, 1 ml of the cell lysate was
purified with spin columns according to the manufactures instruc-
tions. Equal amounts of protein were separated by SDS-PAGE and
transferred to polyvinylidene difluoride membrane by semi-dry blot-
ting. The membrane was blocked in TBST for 1 h, incubated with
patient serum diluted 1:500 in TBST for 1 h, and washed three times
in TBST. Goat-anti-human IgG-peroxidase (Dianova) diluted 1:10,000
in TBST was added for 30 min, and after washing chemoluminescent
detection was performed.
Image Analysis—Image and data analysis was performed using
proprietary software developed by MicroDiscovery (Berlin, Germany).
The clear and distinct signals and homogeneous spot shape obtained
(as seen in Fig. 1A) facilitated image analysis. Data normalization was
achieved by robust normalization techniques described in (20) and
(21).
RESULTS
For the generation of protein chips, we analyzed the protein
expressing resource, a nonredundant subset of a human fetal
brain cDNA expression library. Sequence analysis of 2303 of
the 2413 clones was performed. The reading frame of the
cDNA insert was stringently determined with respect to the
amino-terminal M-RGSH
6
-tag. In parallel, the 2303 cDNA se
-
quences were compared against the nonredundant protein
database NRPROT, and the top five BLAST hits for each
cDNA clone were found. The final reading frame was deter-
mined by the RGSH
6
-tag, and the results were analyzed by
searching for BLAST hits, which match the correct reading
frame. The sequence matching criteria used were highly strin-
gent in that there had to be a perfect match of all 27 nucleo-
tides of the RGSH
6
-tag. If there was a sequence error in this
region, the sequence would be called as not having the his-
tidine tag. Using these stringent criteria, no reading frame was
1
The abbreviations used are: TBST, Tris-buffered saline, Tween-
20; IgG, immunoglobulin G.
A Nonredundant Human Protein Chip
Molecular & Cellular Proteomics 2.12 1343
determined for 701 clones. The remaining 1602 sequences
gave the following results: 880 (55%) clones have their cDNA
insert in the RGSH
6
-tag corresponding reading frame, including
382 (43%) clones, comprising full-length proteins, and 724
(45%) clones that show a different reading frame to the frame
determined by the RGSH
6
-tag. Further analysis of the 701
clones, where the reading frame was not determined, showed
that 616 of these clones had no intact RGSH
6
-tag; 453 of these
clones give a BLAST result, whereas 248 clones show no match
at all.
To generate protein chips, proteins were expressed and
purified in parallel under denaturing conditions from 2413
bacterial clones obtained from a human fetal brain cDNA
expression library (14) that had been normalized by oligonu-
cleotide fingerprinting (19). We have chosen E. coli as the
expression system and optimized the culture volume and
purification conditions leading to expression of suitable
amounts for over 96.5% of the selected recombinant proteins
(Fig. 1C). Following high-throughput protein expression and
purification by nickel-nitrilotriacetic acid-immobilized metal
affinity chromatography (22), the expressed proteins were
spotted onto a modified glass surface, which has been coated
with a thin polyacrylamide layer. The 2413 purified, human
recombinant proteins were spotted in duplicate onto a glass
microscope slide (25 ⫻ 75 mm) using a transfer stamp con-
sisting of 16 pins with a 150-
m tip size. The duplicates were
ordered into two areas, each area consisting of a 13 ⫻ 13
pattern (quadrants), corresponding to a total number of 5408
spots, including 32 guide dots, on one slide. Rabbit-anti-rat-
IgG labeled with Cye5 served as a useful guide dot. Transfer-
ring 10 nl of each of the protein samples results in spots of
about 220
m in diameter, which enables theoretical spot
densities of up to 1600 spots/cm
2
. Due to the nonviscous and
noncrystalline eluting reagent (35% (v/v) acetonitrile, 0.1% (v/v)
trifluoroacetic acid), the protein spots, when loaded five times,
gave a sharp, round, and well-localized signal. In contrast to
DNA microarrays, ghost spots (i.e. spots with signal intensities
below background) or doughnut effects with a non-Gaussian
spot intensity profile were not observed on this surface.
To monitor the protein expression, purification, and arraying
procedures, the protein chips were screened with an antibody
specific for the RGSH
6
-tag epitope of these His
6
-tagged fu
-
sion proteins (Fig. 1A). When relative intensities of duplicates
were plotted against each other, the resulting diagonal indi-
cates a good reproducibility of the spotting and detection of
the immobilized proteins with a correlation coefficient of 0.94
(Fig. 1B). Relative intensities of detected signals were rather
homogeneously distributed in the range of 0–300 units with a
peak value at 225 units (Fig. 1C). The average detected signal
intensity was 132.5 units. We set a threshold of signals below
20 units as background; using this conservative threshold
only 3.5% of the proteins were not detected.
When a high-speed picoliter-spotting robot (ink-jet device)
was used for the deposition of proteins onto the polyacryl-
amide glass surface, the spot density could be increased to
5000 spots/cm
2
. Spotting different concentrations of purified
GAPDH assessed the sensitivity of detection of specific pro-
teins (Fig. 2A). The sensitivity of detection was calculated to
be 10 pmol/
l, corresponding 7.2 fmol or 288 pg of GAPDH in
a volume of 720 pl spotted in four drops. When the proteins
were spotted in a single step, a volume of 180 pl, with a
minimal concentration of 40 pmol/
l, was required. GAPDH
FIG.1. Protein microarray of 2413
recombinant, purified human pro-
teins. A, protein microarray probed with
anti-RGSH
6
antibody. All proteins posi
-
tively detected by the anti-RGSH
6
anti
-
body are represented in red and the
guide dots in green. B, correlation of
relative spot intensities of duplicates
(spot A versus its duplicate spot B). C,
distribution of relative spot intensities.
A Nonredundant Human Protein Chip
1344 Molecular & Cellular Proteomics 2.12
protein and purified proteins from four other uncharacterized
clones were piezoelectric sprayed onto the microchip and
detected either by anti-RGSH
6
and anti-GAPDH antibody (Fig.
2, B and C). These results demonstrate that our high-through-
put protein expression and purification procedure leads to
sufficient protein concentrations to enable the reproducible
generation of high-density protein microarrays.
As mentioned above, here we demonstrate the first of the
two protein chip applications to be described, the character-
ization of antibody binding, specificity, and cross-reactivity.
Antibodies are commonly used in diagnostic assays in many
different formats (23), requiring a high assay sensitivity and
specificity. We have analyzed and compared the specificity of
two mouse monoclonal antibodies, anti-GAPDH and anti-
HSP90

, respectively. Both antibodies are designed for their
use in Western immunoblotting and therefore they recognize
linear epitopes. Because the secondary mouse antibodies
that were used to detect the monoclonal antibodies (anti-
GAPDH and anti-HSP90

) might give some unspecific back-
ground signal, control incubations exclusively with the sec-
ondary antibody were performed. Clones, which were
detected in both the control as well as the anti-GAPDH and
anti-HSP90

incubations, were discarded. Moreover, the de-
tailed sequence analysis of these clones has shown that these
interactions were nonspecific.
The monoclonal anti-GAPDH antibody recognized its spe-
cific antigen GAPDH with the highest intensity of 305 units
(616F11; Fig. 3A). Additionally, clone 613F03 coding for ribo-
somal protein S20 (RPS20) was recognized with a relative
intensity of 201 units, respectively. The other weak signals
detected had an intensity of one-fifth to one-tenth of the
GAPDH signal. Ninety-nine percent of all clones show a signal
less than the set threshold of 20 units. In the anti-HSP90

antibody screen, its target antigen (clone 616P24; relative
intensity, 277 units; Fig. 3B) was preferentially detected, but in
addition this antibody showed medium cross-reactivity with at
least two other clones (with relative intensity of 151 and 149
units, respectively) and weak cross-reactivity to a further
seven clones. The sequence analysis of the medium-intensity
clones (homology to clone FLJ22122 fis, clone HEP19214 (no
known function) and Gallus gallus nuclear calmodulin-binding
protein) recognized by anti-HSP90

does not reveal any iden-
tical linear sequence motif. The proteins corresponding to the
weak-intensity clones were similarly analyzed, but showed no
homology to HSP90

.
Here, we demonstrate the second protein chip application,
profiling the antibody repertoire in serum. Screening protein
arrays with sera from large numbers of autoimmune patients,
for example with rheumatoid arthritis, diabetes, multiple scle-
rosis, or alopecia, would not only allow the identification of
potentially new autoantigens, but also the diagnosis and sub-
typing of the autoimmune disease based on the presence of
specific autoantibodies (24). In this proof of principle serum
FIG.2. High-speed picolitre spotting (ink-jetting) of recombi-
nant, purified proteins. A, determination of the detection limit. Puri-
fied GAPDH was spotted at 100 pmol/
l, 10 pmol/
l, 1 pmol/
l, 100
fmol/
l, and 10 fmol/
l and were subsequently detected by anti-
GAPDH antibody. B, detection of proteins purified in high-throughput
using the epitope specific anti-RGSH
6
antibody (top) and a specific
anti-GAPDH antibody (bottom). Lanes 1– 4, unknown proteins; lane 5,
GAPDH.
FIG.3. Determination of the specificity of antibodies against
GAPDH (A) and HSP90

(B). Yellow arrows indicate the correspond-
ing antigens (red spots); green arrows indicate clones showing cross-
reactivity with medium intensity. The guide dots appear as green
spots.
A Nonredundant Human Protein Chip
Molecular & Cellular Proteomics 2.12 1345
profiling experiment, we screened sera from two patients with
rheumatoid arthritis and one serum from a patient showing
characteristic symptoms of diffuse alopecia. The datasets
obtained from patient sera incubations were compared with
control sera from clinically nonremarkable persons and to
background incubations with human anti-IgG. From these
results, proteins detected by the autoimmune sera were iden-
tified (Fig. 4). 5⬘-tag sequences of these clones were deter-
mined, and their sequences with respect to an open reading
frame were analyzed. Three clones with false reading frames
had to be excluded. The remaining sequences were used for
BLAST searches against the public databases including Gen-
Bank and Unigene (25) (Table I). The relative intensities of the
clones were compared for each serum screened on the pro-
tein microarray. Higher relative intensities for sera from diffuse
alopecia and arthritis were obtained for the gene coding for
autoantigen p69. The two sera of patients with rheumatoid
arthritis recognized clones coding for a protein of the RAS
association domain family 1 and for diglycerol kinase
. The
alopecia serum recognized two additional proteins, the tumor
suppressor p33 ING homologue and a protein coded by
cDNA FLJ20427 with no known function. The microarray data
were confirmed by analyzing equal amounts of the purified
proteins (Fig. 5) on Western immunoblots with the diffuse
alopecia patient serum. The control protein ubiquitin B, in-
cluded in the Western immunoblot, was not detected by any
serum. Controls with a serum obtained from an alopecia
areata patient, another hair loss disease with clearly different
clinical features, and with the second antibody alone were
also screened (Fig. 5). The diffuse alopecia serum detected
the autoantigen p69 and the tumor suppressor p33 ING ho-
mologous protein, whereas the gene product of cDNA
FLJ20427 is not detected on the Western blot (Fig. 5A). This
might be due to changes in the conformation after SDS treat-
ment or to effects of the immobilization either on the chip or
during Western immunoblot procedure. Additionally, weaker
signals were obtained for the homologue of mouse synapto-
tagmin VI and the protein belonging to the RAS association
domain family I. Both of these proteins were also recognized
by the serum of the patient with characteristics specific for
alopecia areata, whereas autoantigen p69 and tumor sup-
pressor p33 ING homologous protein were not detected, sug-
gesting different types of hair loss (Fig. 5B).
DISCUSSION
Here, a new method for the generation of protein microar-
rays on coated glass slides with a high sensitivity for antibody
screening and serum profiling has been developed. Combin-
ing cDNA expression libraries and robot technology, 2413
different human proteins have been expressed, purified, and
spotted onto glass chips.
Similar to DNA arrays, glass slides for protein chips offer a
low fluorescent background, superior handling properties, re-
quire small screening volumes, and are suitable for automa-
tion. Recently, a microarray consisting of 10,800 spots of two
single/distinct proteins (protein G an FKBP12 binding domain)
has been described (26). Brown and coworkers have analyzed
115 antibody-antigen pairs arrayed on poly-
L-lysine-coated
glass slides and determined detection limits in the range of
100 pg/ml, confirming previously specified detection levels
(16, 27). Because microarrays allow highly parallel analysis,
there is increasing interest in arraying a large number of
different targets. We have chosen E. coli as a robust, fast, and
efficient protein expression system. Because the proteins
were expressed in E. coli, no post-translational modifications
such as phosphorylation or glycosylation were present. How-
ever, if protein expression systems introducing post-transla-
tional modifications such as Saccharomyces cerevisiae,
Pichia pastoris, or baculovirus (28–31) would have been used,
the glycosylation pattern obtained would not be identical to
humans. We have shown that 96.5% of the 2413 clones
express a detectable His
6
-tagged fusion protein with an av
-
erage signal intensity of 132.5 units. The sequence analysis of
clones expressing proteins with a relative intensity close to
FIG.4. Serum profiling on the protein microarrays. This com-
posed image is the result of an overlay of an image from a protein
microarray incubated with the control anti-human-IgG antibody (blue
spots) with the image from a separate microarray incubated with the
serum of a diffuse alopecia patient (red spots). The guide dots appear
as white spots. Individual proteins recognized by both the control and
the patient sera appear as purple spots.
A Nonredundant Human Protein Chip
1346 Molecular & Cellular Proteomics 2.12
the background threshold shows that most of them encode
their gene product in the wrong reading frame with respect to
the RGSH
6
epitope. Only a minority of clones exhibiting a low
relative intensity expressed its cDNA insert in the correct
reading frame. Sequence analysis revealed that these clones
encode for either highly structured proteins or large proteins
that may be difficult to express by E. coli (32). Clones showing
a higher relative intensity of the expressed proteins are mainly
in the correct reading frame, but some wrong frame clones
resulting in small 5–8-kDa peptides have also been observed,
indicating that some of the small peptides are not fully de-
graded by E. coli. The use of purified proteins allows certain
discrimination between correct and wrong frame clones.
Sequence analysis of 2303 of the 2413 clones was per-
formed, and the reading frame of 1602 was determined and
gave the following results: 880 (55%) clones have their cDNA
insert in the RGSH
6
-tag corresponding reading frame, includ
-
ing 382 (43%) clones comprising full-length proteins, and 724
TABLE I
Candidate autoantigens detected by sera antibodies from three patients
The table includes the list of proteins identified when the protein chip was screened with serum from three patients, two with rheumatoid
arthritis and one with diffuse alopecia. The table includes the library clone identifier, gene name, database accession number, and the relative
intensity units measured for each of the duplicate clones on the chip.
No.
Clone
MPMGp800
Gene name
Accession
Number
Arthritis
serum43
Arthritis
serum44
Diffuse
alopecia
1 A21618 Autoantigen p69 L21181 37.347 172.012 149.821
48.309 180.647 158.224
2 G02617 Tumor suppressor p33 ING homolog AF044076 38.771 60.704 190.685
43.064 91.959 194.042
3 I17618 cDNA FLJ20427, unknown function AK000434 23.381 103.153 142.685
27.347 128.798 149.877
4 O03614 Unknown 108.744 57.624 34.262
126.624 60.923 26.532
5 B15613 Homolog to mouse synaptotagmin VI NM018800 14.178 123.594 32.764
22.172 146.278 36.381
6 F11617 Diglycerol kinase
XM012056 142.813 61.501 29.240
147.589 85.869 27.069
7 K20617 Ubiquitin B (control clone) NM018955 35.369 52.203 25.680
40.484 72.644 28.148
8 L16612 RAS association domain family 1 XM003273.2 79.750 135.898 39.501
88.095 147.211 50.212
FIG.5. Western immunoblotting of
purified candidate autoantigens. Left
side top, purified candidate autoanti-
gens on Coomassie-stained SDS-PAGE.
Lanes 1– 8, p69; tumor suppressor p33
ING homologue, cDNA FLJ20427, un-
known, mouse synaptotagmin VI homo-
logue, diglycerol kinase
, ubiquitin B
(control clone), and RAS assocation do-
main family. A, Western blot of candi-
date autoantigens probed with serum
from patient with diffuse alopecia. B,
Western blot of candidate autoantigens
probed with serum from patient with al-
opecia areata. C, Western blot of candi-
date autoantigens probed with the sec-
ondary antibody (control).
A Nonredundant Human Protein Chip
Molecular & Cellular Proteomics 2.12 1347
(45%) clones that show a different reading frame to the frame
determined by the RGSH
6
-tag. Using stringent sequence
analysis criteria, no reading frame was determined for 701
clones.
We have demonstrated the characterization of antibody
specificity on high-density protein microarrays, as well as their
potential application in profiling the antibodies repertoire in
sera from patients with autoimmune diseases. In order to
minimize all surface-related effects that might cause nonspe-
cific binding or unspecific folding of the deposited proteins on
conventional slides, we worked with a highly aqueous gel.
This also allows the addition of reagents that may increase
binding or change the surrounding environment to a more
physiological state. A related three-dimensional surface was
developed by Mirzabekov and coworkers, who have used a
gel photo or persulfate-induced copolymerization technique
to produce functional protein microarrays on polyacrylamide
gel pads (33). Furthermore, an additional advantage of the
three-dimensional structure of the gel matrix, when compared
with conventional two-dimensional slides, is the higher pro-
tein binding capacity and their high detection sensitivity (34).
Other methods to generate protein microarrays involve pro-
tein immobilization on a flat, two-dimensional surface by ei-
ther covalent coupling to a cross-linker attached to the sur-
face (35–37) or noncovalent interaction to an immobilized
biomolecule such as biotin (38). Recently, oriented immobili-
zation of proteins either on chlorinated glass slides (39) via
affinity tags (40) or by biotinylation of capture molecules and
their immobilization on streptavidin coated supports (41) have
been described.
For the generation of antibody arrays, which consist of
immobilized antibodies used to detect different proteins, for
example in body fluids or tissue extracts, it is increasingly
important to analyze these antibodies for their specificity
against a large number of different proteins. We have ana-
lyzed and compared the specificity of two mouse monoclonal
antibodies, anti-GAPDH and anti-HSP90

, respectively, using
our protein microarrays. It has been show that both antibodies
recognize their antigen with high specificity, but additional
proteins were also detected, indicating a certain cross-reac-
tivity. Both antibodies were designed for their use in Western
immunoblotting and, therefore, recognize linear epitopes.
However, because refolding of immobilized, denatured pro-
teins was shown to occur in a single step using Tris-buffered
saline (42), which is also used here in the binding assay, a
partial refolding of the immobilized proteins during incubation
cannot be excluded. Therefore, the possibility remains that
the resulting structural epitopes might mimic the antigenic con-
formation. In respect to antibodies immobilized on antibody
arrays, best results will be obtained using antibodies, which
were previously characterized. In addition, these results show
that this Uniclone set is indeed nonredundant, because the
antibodies characterized here recognize housekeeping genes,
which were highly redundant in the original cDNA library (14).
In the proof of principle experiment to profile the antibody
repertoire in serum, we chose sera from two different patients
with rheumatoid arthritis and one serum from a patient show-
ing characteristic symptoms of alopecia. Protein p69 was
recognized by the sera from patients with different autoim-
mune diseases (arthritis and alopecia). Interestingly, p69 is
also known to be a candidate autoimmune target in type 1
diabetes as the recombinant protein was recognized by au-
toantibodies and T cells from diabetic children (43). These
results suggest that this protein may be involved in a more
general way in autoimmune diseases. The microarray data
was confirmed by Western immunoblot analysis of equal
amounts of the purified proteins (Fig. 5) with the diffuse alo-
pecia patient sera. This serum detected specifically the au-
toantigen p69 and the tumor suppressor p33 ING homolo-
gous protein, whereas the gene product of cDNA FLJ20427
was not detected on the Western immunoblot (Fig. 5A). This
might be due to changes in the conformation after SDS treat-
ment or to effects of the immobilization either on the chip or
during Western immunoblot procedure. Additionally, weaker
signals were obtained for the homologue of mouse synapto-
tagmin VI and the protein belonging to the RAS association
domain family I. Both of these proteins were also recognized
by the serum of the patient with characteristics specific for
alopecia areata, whereas autoantigen p69 and tumor suppres-
sor p33 ING homologous protein were not detected, consistent
with the clinical diagnosis of different types of hair loss (Fig. 5B).
This suggests that protein arrays containing a specific subset of
proteins may be used for subtyping of diseases. A more in-
depth analysis of the selected autoimmune diseases, including
the characterization of sera from large patient and control co-
horts selected after clinical diagnosis, is currently under way.
In summary, using our protein microarrays consisting of a
large number of human proteins, the specificity and cross-
reactivity of antibodies can be characterized. The detected
proteins can easily be identified by mass spectrometry or
DNA sequencing, as they are derived from clones of an ar-
rayed cDNA expression library. This approach can also be
applied to determine the binding specificity of antibodies,
which were previously unknown, or to profile the binding of
mixtures of antibodies, such as are found in sera of patients
with autoimmune diseases.
Acknowledgments—We thank Prof. Dr. Ulrike Blume-Peytavi and
Dr. Dorothee Hagemann for providing the patient sera.
* This work was funded by Bundesministerium fu¨ r Bildung und
Forschung (BioFuture 0311018 and DHGPII 0311323) and the Max-
Planck-Gesellschaft. D. J. C acknowledges funding from the Health
Education Authority and Science Foundation Ireland, Dublin, Ireland.
The costs of publication of this article were defrayed in part by the
payment of page charges. This article must therefore be hereby
marked “advertisement” in accordance with 18 U.S.C. Section 1734
solely to indicate this fact.
** To whom correspondence should be addressed: Centre for Hu-
man Proteomics, Royal College of Surgeons in Ireland, 123 St. Ste-
A Nonredundant Human Protein Chip
1348 Molecular & Cellular Proteomics 2.12
phen’s Green, Dublin 2, Ireland. Tel.: 353-(1)-402-8550; Fax: 353-(1)-
402-2453; E-mail: chp@rcsi.ie.
REFERENCES
1. Epstein, C. B., and Butow, R. A. (2000) Microarray technology—Enhanced
versatility, persistent challenge. Curr. Opin. Biotechnol. 11, 36 –41
2. Lockhart, D. J., and Winzeler, E. A. (2000) Genomics, gene expression and
DNA arrays. Nature 405, 827– 836
3. Blohm, D. H., and Guiseppi-Elie, A. (2001) New developments in microarray
technology. Curr. Opin. Biotechnol. 12, 41– 47
4. Meier-Ewert, S., Maier, E., Ahmadi, A., Curtis, J., and Lehrach, H. (1993) An
automated approach to generating expressed sequence catalogues.
Nature 361, 375–376
5. Lennon, G. G., and Lehrach, H. (1991) Hybridization analysis of arrayed
cDNA libraries. Trends Genetics 7, 314 –317
6. Schena, M., Heller, R. A., Theriault, T. P., Konrad, K., Lachenmeier, E., and
Davis, R. W. (1998) Microarrays: Biotechnology’s discovery platform for
functional genomics. Trends Biotechnol. 16, 301–306
7. Eickhoff, H., Schuchhardt, J., Ivanov, I., Meier-Ewert, S., O’Brien, J., Malik,
A., Tandon, N., Wolski, E. W., Rohlfs, E., Nyarsik, L., Reinhardt, R.,
Nietfeld, W., and Lehrach, H. (2000) Tissue gene expression profiling
using arrayed normalized cDNA libraries. Genome Res. 10, 1230 –1240
8. Walter, G., Bu¨ ssow, K., Cahill, D., Lueking, A., and Lehrach, H. (2000)
Protein arrays for gene expression and molecular interaction screening.
Curr. Opin. Microbiol. 3, 298 –302
9. Cahill, D. et al (2001) in Proteomics (Pennington, S. and Dunn, M., eds.) pp.
1–20, BIOS Scientific Publishers, Oxford
10. Pohlner, K., Portig, I., Pankuweit, S., Lottspeich, F., and Maisch, B. (1997)
Identification of mitochondrial antigens recognized by antibodies in sera
of patients with idiopathic dilated cardiomyopathy by two-dimensional
gel electrophoresis and protein sequencing. Am. J. Cardiol. 80,
1040 –1045
11. Latif, N., Baker, C. S., Dunn, M. J., Rose, M. L., Brady, P., and Yacoub,
M. H. (1993) Frequency and specificity of antiheart antibodies in patients
with dilated cardiomyopathy detected using SDS-PAGE and western
blotting. J. Am. Coll. Cardiol. 22, 1378 –1384
12. Greiner, J., Ringhoffer, M., Simikopinko, O., Szmaragowska, A., Huebsch,
S., Maurer, U., Bergmann, L., and Schmitt, M. (2000) Simultaneous
expression of different immunogenic antigens in acute myeloid leukemia.
Exp. Hematol. 28, 1413–1422
13. Cahill, D. (2000) Protein arrays: A high-throughput solution for proteomics
research? in Proteomics: Current Trends Supplement (Blackstock, W.,
and Mann, M., eds) pp. 47–51, Elsevier Science, Ltd., London
14. Bu¨ ssow, K., Cahill, D., Nietfeld, W., Bancroft, D., Scherzinger, E., Lehrach,
H., and Walter, G. (1998) A method for global protein expression and
antibody screening on high-density filters of an arrayed cDNA library.
Nucleic Acids Res. 26, 5007–5008
15. Lehrach, H., Bancroft, D., and Maier, E. (1997) Robotics, computing, and
biology. Interdisciplinary Sci. Rev. 22, 37– 44
16. Lueking, A., Horn, M., Eickhoff, H., Bussow, K., Lehrach, H., and Walter, G.
(1999) Protein microarrays for gene expression and antibody screening.
Anal. Biochem. 270, 103–111
17. Bu¨ ssow, K., Nordhoff, E., Lu¨ bbert, C., Lehrach, H., and Walter, G. (2000) A
human cDNA library for high-throughput protein expression screening.
Genomics 65, 1– 8
18. Schmidt, F., Lueking, A., Nordhoff, E., Gobom, J., Klose, J., Seitz, H.,
Egelhofer, V., Eickhoff, H., Lehrach, H., and Cahill, D. J. (2002) Genera-
tion of minimal protein identifiers of proteins from two-dimensional gels
and recombinant proteins. Electrophoresis 23, 621– 625
19. Herwig, R., Schmitt, A. O., Steinfath, M., O’Brien, J., Seidel, H., Meier-
Ewert, S., Lehrach, H., and Radelof, U. (2000) Information theoretical
probe selection for hybridisation experiments. Bioinformatics 16, 890 – 898
20. Yang, Y. H., Dudoit, S., Luu, P., and Speed, T. P. (2000) Technical Report,
University of California, Berkley, CA
21. Schuchhardt, J., Beule, D., Malik, A., Wolski, E., Eickhoff, H., Lehrach, H.,
and Herzel, H. (2000) Normalization strategies for cDNA microarrays.
Nucleic Acids Res. 28, E47
22. Hochuli, E., Dobeli, H., and Schacher, A. (1987) New metal chelate adsorb-
ent selective for proteins and peptides containing neighboring histidine
residues. J. Chromatogr. 411, 177–184
23. Borrebaeck, C. A. (2000) Antibodies in diagnostics—From immunoassays
to protein chips. Immunol. Today 21, 379 –382
24. Robinson, W. H., DiGennaro, C., Hueber, W., Haab, B. B., Kamachi, M.,
Dean, E. J., Fournel, S., Fong, D., Genovese, M. C., de Vegvar, H. E.,
Skriner, K., Hirschberg, D. L., Morris, R. I., Muller, S., Pruijn, G. J., van
Venrooij, W. J., Smolen, J. S., Brown, P. O., Steinman, L., and Utz, P. J.
(2002) Autoantigen microarrays for multiplex characterization of autoan-
tibody responses. Nat. Med. 8, 295–301
25. Altschul, S. F., Madden, T. L., Schaffer, A. A., Zhang, J., Zhang, Z., Miller,
W., and Lipman, D. J. (1997) Gapped BLAST and PSI-BLAST: a new
generation of protein database search programs. Nucleic Acids Res. 25,
3389 –3402
26. MacBeath, G., and Schreiber, S. L. (2000) Printing proteins as microarrays
for high-throughput function determination. Science 289, 1760 –1763
27. Haab, B. B., Dunham, M. J., and Brown, P. O. (2001) Protein microarrays for
highly parallel detection and quantitation of specific proteins and anti-
bodies in complex solutions. Genome Biology 2,
research0004.0001– 0004.0013
28. Holz, C., Lueking, A., Bovekamp, L., Gutjahr, C., Bolotina, N., Lehrach, H.,
and Cahill, D. J. (2001) A human cDNA expression library in yeast
enriched for open reading frames. Genome Res. 11, 1730 –1735
29. Lueking, A., Holz, C., Gotthold, C., Lehrach, H., and Cahill, D. (2000) A
system for dual protein expression in Pichia pastoris and Escherichia coli.
Protein Expr. Purif. 20, 372–378
30. Albala, J. S., Franke, K., McConnell, I. R., Pak, K. L., Folta, P. A., Rubinfeld,
B., Davies, A. H., Lennon, G. G., and Clark, R. (2000) From genes to
proteins: High-throughput expression and purification of the. J. Cell.
Biochem. 80, 187–191
31. Lueking, A., Horn, S., Lehrach, H., and Cahill, D. J. (2003) A dual-expression
vector allowing expression in E. coli and P. pastoris, including new
modifications. Methods Mol. Biol. 205, 31– 42
32. Makrides, S. C. (1996) Strategies for achieving high-level expression of
genes in Escherichia coli. Microbiol. Rev. 60, 512–538
33. Arenkov, P., Kukhtin, A., Gemmell, A., Voloshchuk, S., Chupeeva, V., and
Mirzabekov, A. (2000) Protein microchips: Use for immunoassay and
enzymatic reactions. Anal. Biochem. 278, 123–131
34. Angenendt, P., Glo¨ kler, J., Murphy, D., Lehrach, H., and Cahill, D. J. (2002)
Towards antibody microarrays: A comparison of current microarray sup-
port materials. Anal. Biochem. 309, 253–260
35. Jones, V. W., Kenseth, J. R., Porter, M. D., Mosher, C. L., and Henderson,
E. (1998) Microminiaturized immunoassays using atomic force micros-
copy and compositionally patterned antigen arrays. Anal. Chem. 70,
1233–1241
36. Mendoza, L. G., McQuary, P., Mongan, A., Gangadharan, R., Brignac, S.,
and Eggers, M. (1999) High-throughput microarray-based enzyme-linked
immunosorbent assay (ELISA). BioTechniques 27, 778 –780, 782–776,
788
37. Mooney, J. F., Hunt, A. J., McIntosh, J. R., Liberko, C. A., Walba, D. M.,
Rogers, C. T. (1996) Patterning of functional antibodies and other pro-
teins by photolithography of silane monolayers. Proc. Natl. Acad. Sci.
U. S. A. 93, 12287–12291
38. Yang, Z., Frey, W., Oliver, T., and Chilkoti, A. (2000) Light-activated affinity
micropatterning of proteins on self-assembled monolayers on gold.
Langmuir 16, 1751–1758
39. Hergenrother, P. J., Depew, K. M., and Schreiber, S. L. (2000) Small-
molecule microarrays: Covalent attachment and screening of alcohol-
containing small molecules on glass slides. J. Am. Chem. Soc. 122,
7849 –7850
40. Zhu, H., and Snyder, M. (2001) Protein arrays and microarrays. Curr. Opin.
Chem. Biol. 5, 81–91
41. Peluso, P., Wilson, D. S., Do, D., Tran, H., Venkatasubbaiah, M., Quincy, D.,
Heidecker, B., Poindexter, K., Tolani, N., Phelan, M., Witte, K., Jung,
L. S., Wagner, P., and Nock, S. (2003) Optimizing antibody immobiliza-
tion strategies for the construction of protein microarrays. Anal. Bio-
chem. 312, 113–124
42. Holzinger, A., Phillips, K. S., and Weaver, T. E. Single-step purification/
solubilization of recombinant proteins: Application to surfactant protein
B. BioTechniques 20, 804 – 806, 808
43. Miyazaki, I., Gaedigk, R., Hui, M. F., Cheung, R. K., Morkowski, J., Rajotte,
R. V., and Dosch, H. M. (1994) Cloning of human and rat p69 cDNA, a
candidate autoimmune target in type 1 diabetes. Biochim. Biophys. Acta
1227, 101–104
A Nonredundant Human Protein Chip
Molecular & Cellular Proteomics 2.12 1349
