Morris Plains, NJ, 07950
Capturing Proteins Using
microarrays are enabling researchers to perform
high-throughput proteomic analyses
disagree on the exact number of meaningful genes in the
human genome, but proteins clearly far outnumber genes,
perhaps by a factor of ten or twenty, especially after
considering post-translational modifications. When all
the possibilities are tallied, distinct proteins number
between half a million and one million.
the image to
Schematic view of
the procedure to screen protein-protein
interactions using antibodyArray. (Source:
proteomic equivalents of gene microarrays, antibody
microarrays enable parallel, multiplexed, miniaturized,
high-throughput proteomic analysis that includes
individual protein detection, protein expression
profiling, and interactions between proteins and drugs,
ligands, antibodies, or other proteins.
microarrays use immobilized antibodies affixed to glass,
membranes, microplate wells, mass spectrometer plates,
silicon, plastic, or other surfaces to capture less
abundant proteins of interest from complex biological
fluids and tissues. Chip substrates depend on the
desired detection and antibody immobilization methods.
In the early days of peptide microarrays, developers
used silicon chips because flat, high-quality silicon
was less expensive than glass with similar properties.
Hypromatrix Inc., Worcester, Mass., has tried several
different substrates but settled on nitrocellulose
membranes and glass substrates coated with
nitrocellulose membranes for antibody immobilization.
According to company president John Hou, nitrocellulose
coupled with proprietary chemistry keeps antibodies in
their active state and makes spots easier to detect.
Applications: wherever you
Complementing the Binding Capacity
With only about 500 antibodies commercially
available, covering the entire human proteome with
antibody microarrays won't happen any time soon.
Still, companies continue to chip away at the
proteome with innovative antibody-generating
technologies. Epitomics Inc., South San Francisco,
has developed a suite of antibody-generating
methods. TargetMAb generates rabbit monoclonal
antibodies (MAbs) for proteins of all classes
directly from DNA plasmids. MultiMab multiplexes
immunization of many antigens to create antibodies
several at a time. Through its PathoMAb
technology, Epitomics immunizes rabbits directly
with disease tissues or cells, generating
thousands of MAbs against both cell surface
molecules and intracellular targets. Using a
subtractive method to compare antibodies generated
in this manner from those in normal tissues,
scientists identify disease-specific antibodies
Some companies have found a way
to complement the binding capacity of antibodies,
or to eliminate antibodies from their protein
recognition chips altogether.
ProteinChip Biology system, from Ciphergen
Biosystems Inc., Fremont, Calif., consists of
instrumentation, ProteinChip Arrays, and software
for data collection and analysis, including
comparison of the presence and abundance of
proteins in crude samples. Like other protein
analysis chips, ProteinChip arrays can probe
individual proteins, DNA-protein interactions, and
protein-protein associations. Ciphergen also has a
ProteinChip interface for its QStar instruments,
allowing data collection in single MS and MS-MS
mode directly from the ProteinChip arrays.
By incorporating built-in chromatography
chemistries, ProteinChip does not rely solely on
antibody-protein or other high-affinity
interactions. For example, Ciphergen offers arrays
containing weak cation exchange, strong anion
exchange, metal affinity capture, reverse phase,
hydrophobic interaction, normal-phase silicate,
and inert gold surfaces.
Inc., South San Francisco, Calif., eliminates the
"antibody limitation" by creating all the
molecular recognition surfaces it needs through a
type of molecular imprinting technology known as
ProteinPrint. Aspira first creates a peptide
corresponding to a signature sequence in the
target protein. Next, polymerizable monomers
self-assemble around the peptide and are
cross-linked in place. The peptide is then
removed, leaving behind an artificial molecular
recognition pocket that is complementary to the
peptide in both shape and chemical functionality.
When the polymer is exposed to a sample of
denatured protein, target proteins bind to the
recognition pockets. Washing, staining, and
detection are carried out as with traditional
antibody detection systems.
offers several advantages over traditional
antibodies. Scientists can create recognition
sites without knowing the full target sequence,
without purchasing or creating antibodies.
Typically, Aspira targets unique COOH-terminal
seven-peptide sequences for denatured proteins, or
an internal sequence for active proteins. On the
minus side, the targeted region must somehow be
accessible to the recognition site, whereas
natural antibodies take care of that detail.
Whereas molecular imprinting is faster and less
expensive than raising and purifying antibodies,
it does take time.
Colo., uses photoaptamers—modified,
single-stranded DNA or RNA which, like antibodies,
bind proteins with the affinity and specificity
normally associated with antibodies. Photoaptamers
differ from conventional aptamers in that
thymidine is substituted with brominated
deoxyuridine, which gives the molecules the
additional capability of cross-linking to target
proteins under the influence of ultraviolet light,
hence, the "photo" in "photoaptamer." Like
antibodies, photoaptamers can be fabricated into
arrays and used for protein capture.
Medical diagnostics are an obvious
commercial opportunity for antibody microarrays, but to
be useful in this context, chips are limited to a few
key analytes and must incorporate built-in sample
preparation and readout. For example, the Triage Cardiac
microarray from Biosite Diagnostics Inc., San Diego,
which is used in hospitals and other emergency medical
settings, tests for just three cardiac markers. Biosite
also is working on an antibody-based stroke diagnostic
microarray that will test five stroke-related protein
A more esoteric, albeit further-off
medical application, termed "personal proteomics," is
more in line with the sophistication that developers
envision for antibody microarrays. Like personal
genomics or personal medicine, personal proteomics seeks
to sub-define an illness according to specific molecular
characteristics—in this case, protein expression. For
example personal genomics seeks to differentiate
diagnostically similar tumors according to the genes
expressed by those tissues. The idea is that genomically
distinct tumors may be associated with radically
different prognoses and require different therapies.
Personal proteomics takes this idea a step further:
identical-looking tumors exhibiting similar genomics are
differentiated on the basis of their protein expression
The other significant application of
antibody arrays is research proteomics, which stresses
research-based protein function analysis, determination
of enzyme activity, antibody cross-reactivity studies,
selection of proteins from phage or ribosome display
libraries, and epitope mapping.
AntibodyArray chip is used to study protein-protein
interactions, post-translational modifications, and
protein expression patterns. "Our chips are available
for less than $1,000, take 6 hours to process, and give
many fewer false positives because they don't create
interactions, they just detect them," says Hou.
Hypromatrix offers three antibody arrays. Its
400-antibody signal transduction array, says Hou,
"covers a lot of ground and can be used as a starting
point for protein expression analysis." Hypromatrix also
makes an apoptosis array containing 150 antibodies, and
a cell-cycle array with 60 antibodies. Custom arrays
using any combination of the 400 well-characterized
antibodies are also available.
According to Dev
Baines, of Prometic Life Sciences, Montreal, Quebec,
Canada, today's antibody microarrays are too specific to
be used to mine all proteins of interest in a sample.
That is why Prometic focuses on group-specific chemical
affinity ligands, rather than antibodies, for protein
capture. High-abundance proteins such as albumin are
first removed chromatographically. Then low-abundance
proteins of interest are captured by Prometic's
synthetic chemical ligands, which are less specific than
antibodies but have the potential for picking out many
more diverse proteins than any collection of antibodies.
"Instead of focusing on active site or docking
mechanisms, our affinity technology targets the 500 or
so unique folds that occur in almost every protein.
Essentially what we're doing is to scale down what we've
done successfully with chromatography columns to
array-sized devices, which allows parallel processing.
Proteins collected in this fashion are then subjected to
Although clinical proteomics
is the ultimate goal of many antibody microarray
efforts, not everyone believes chips are the answer to
devilishly complex proteomic puzzles. "Clinical
proteomics is strictly comparative," says Lisa Bradbury,
senior scientist at Ciphergen Biosystems Inc., Fremont,
Calif., who contrasts chip-based protein profiling with
more conventional 2D proteomics. "If you use an antibody
array to profile late-stage versus early-stage colon
cancer, you may miss important proteins because you
can't cover everything. Proteomic methods are unbiased
because they are not limited by selected antibodies. The
end point is to define distinct populations through
differential protein expression. Those differences don't
necessarily need to tie in with disease mechanisms, but
they must at least be signature differences."
(above) Scanning electron
micrograph of lower part of microarray device
(ultrasonic welding of a clear upper plastic part
creates a capillary space that contains the sample
fluid) illustrating the reaction chamber where
plasma contacts labeled antibody reagents.
(above) The Zyomyx Human
Cytokine Biochip, with 30 biologically relevant
cytokines, is the first in a series of protein
profiling biochips designed by Zyomyx to
facilitate highly sensitive and comprehensive,
multiplex protein expression profiling using
minimal sample amounts.
drug development, protein chips will benefit companies
looking for subsets of "responders" for clinical trials.
After commercialization of proteomic-based discoveries,
antibody chips will help determine which patients will
benefit from those drug therapies. In every case,
treatment options are limited by what is available.
"This revolution isn't going to happen overnight," says
Larry Cohen, CEO of Zyomyx Inc., Hayward, Calif., "but
in the long term, it's quite possible."
claim to fame is its activity-retaining protein
immobilization technologies, which combine
nanotechnology and biochemistry. In February 2003, the
company launched its Protein Profiling Biochip system
and its first protein biochip product, the Zyomyx Human
Cytokine Biochip containing 30 cytokines."
All arrays are not created equal
Genomics has progressed much more rapidly than
proteomics because complementary gene-gene interactions
are more predictable and easier to bring about than
including affinity for other proteins, depends on the
molecule's primary, secondary, and tertiary structure;
for genes, primary structure is usually enough to retain
activity. Although gene hybridization interactions are
strong and specific, antibody-protein interactions vary
greatly and suffer from unpredictable cross-reactivity.
Antibodies are difficult to make and are available for
only a small number of proteins, whereas automated
oligonucleotide synthesizers can create gene probes by
the dozens. Finally, genes are easily amplified using
polymerase chain reaction (PCR) or other techniques, but
no such method exists for proteins.
Reproducible, reliable protein immobilization
was worked out during microarrays' "peptide days"
through processes that are now quite familiar: attach a
spacer molecule to the substrate and add the antibody of
interest via a covalent sulfhydril, amine, or hydroxyl
group linkage. Antibody arrays still require robotic
manufacture if high reproducibility is desired. Typical
spot creation methods include contact printing,
piezoelectric spotting, and photolithography.
Even though they employ standard immobilization
and spacer chemistry, antibody arrays are much more
difficult to fabricate and operate than gene chips, says
Bruce Kimmel, director of protein therapeutics at
Diversa Corp., San Diego. "Regardless of the platform,
you need to be able to do a sandwich assay, which
requires a capture agent and detection reagents.
Calibrating such a system is extraordinarily difficult
when you have hundreds of different capture events
occurring on the same chip. Binding reagents need to
work in the same buffers and have the same washing
Unlike DNA affixed to microarrays,
currently there is no antibody collection that
represents the entire proteome. So which antibodies go
into arrays? "Choosing them is part science, part art,"
says Hou of Hypromatrix. "At the very least, you want
proteins that give you the best opportunity to capture
what you're looking for. One approach to that is to use
arrays of well-studied proteins related to whatever your
experiment is trying to show. And of course, you can't
use an antibody that doesn't exist or which is not
available at the right purity."
proteins: comparing apples and oranges
the image to
antibody arrays to DNA arrays is an "apples and oranges"
proposition. Gene chips are able to measure
transcriptional mRNA profiles in multiplexed, parallel
fashion because DNA is homogeneous and gene probes are
easily synthesized or purchased in several useful
formats. "You could say that no great technological
breakthroughs were required for gene arrays to progress
to where they are today," says Gunars Valkirs, vice
president of discovery at Biosite Diagnostics Inc. "The
same is not true for antibody arrays. Antibodies exist
for only a small fraction of the proteome. Even if
10,000 antibodies were available, the cost of producing
a 10,000- protein array would be prohibitive."
Density is another huge digression point between
DNA and antibody chips, says Valkirs. "With current
technology, antibody arrays can't be multiplexed
effectively beyond a certain point, say 100 assays per
chip, because of serious signal-to-noise issues. It's
very easy to optimize one assay on a chip, or even 10.
For 100 assays, however, you need to add labeled
antibodies for each one of those tests, which means at
least 100 times the noise. That wouldn't be so bad
except you may still only be looking for a couple of
proteins that require high sensitivity, in which case
the signal-to-noise [ratio] becomes unacceptable."
Valkirs also noted that maintaining interassay
and interchip precision and specificity is difficult
with antibody arrays. "Unless something remarkable
happens on the detection side, specifically label-free
detection that does not require a second labeled
antibody, antibody arrays will be limited in the number
of simultaneous assays they can provide."
Inch wide, mile deep
outnumber genes by such a large margin, and with
relatively few readily available antibodies, proteomics
by necessity covers a much narrower slice of its domain
than does genomics. However, it also cuts a lot deeper,
says Larry Cohen of Zyomyx. "You really only need to
make two measurements to know what's going on with a
gene: its sequence and when it generates its messenger
RNA. Those measurements are not trivial, but that's
basically all you need. It's not that simple with
proteins. Because proteins are responsible for how cells
function, you'd like to measure their structures,
interactions, and expression levels, preferably in
highly parallel fashion."
microarrays only investigate a tiny fraction of the
proteome, the information they give is still valuable.
"DNA arrays were on the market long before we knew the
entire genome sequence," says Cohen, "but they were
still exciting and worthwhile tools. Although protein
arrays only investigate a subset of the proteome, they
remain highly valuable for a variety of applications."
The question, "Why use protein arrays?" is
related to the deeper question, "Why proteomics?" The
point of medical genomics is to discover the immutable
biochemical factors governing disease. Ultimately,
however, scientists come around again to proteins
because those are the agents responsible for what goes
on in the cell. Not coincidentally, proteins are also
drug discovery's principal target. "You simply can't get
protein concentrations, or even confirm a protein's
existence, from DNA," says Cohen. "Even mRNA only really
infers what's going on by giving some idea of the
potential for synthesizing a protein. The only way to be
sure of proteomic status is to measure proteins
Angelo DePalma, PhD
DePalma is a freelance writer based in Newton,