The conversion of gene array technology into protein arrays that can identify and quantify proteins is fraught with problems. There are distinct advantages for those that succeed but a series of technical challenges will have to be overcome before ‘off-the-shelf’ proteins arrays become a reality. Protein arrays, particularly antibody arrays, can overcome the sensitivity problems of 2DGE proteomics and simultaneously offer the prospect of massive parallel processing of large numbers of proteins (Lal et al., 2002). However, unlike the PCR reaction for nucleic acids, proteins cannot be amplified.
Advantages of protein arrays
As a strategy, it is a better to analyse expressed proteins rather than the mRNAs that encode them. The advantages include the ability to detect PTMs and the certainty that a protein has been expressed in response to an altered gene expression. Several years ago we reported (
Bennetts et al., 1986) there was little or no correspondence between the amounts of mRNAs (determined from Northern blots) of α-cardiac and α-skeletal muscle actin isoforms and the total quantity of translated actin determined from one-dimensional PAGE gels. The same is true for cardiac myosin heavy chain expression, namely changes observed in myosin heavy chain mRNA did not correspond to changes at the protein level (
Coumans et al., 1997). Although this is clearly not true for all myocardial proteins, the correspondence of changes in gene arrays and proteomics in the analyses of eukaryotic cells can be quite poor (
La Naour et al., 2001).
The main challenge for protein array technology is the identification and quantification of proteins selectively retained on the arrays. This largely depends on ligand affinity.
Binding affinity and specificity
At present, specificity is mainly achieved using monoclonal antibodies. These, rather than polyclonals, are used because they can be reliably produced and have definable (although usually not precisely defined) epitopes. Their binding affinities need to be high, preferably in the picomolar range, but most monoclonal antibodies have lower affinities. The exceptional affinity of biotin for streptavidin has made these binding partners a popular choice for downstream detection.
Binding specificity also must be considered. Monoclonal antibodies are not necessarily the obvious choice for a capture device. It is true that monoclonal antibodies have highly specific epitopes but these epitopes are not necessarily unique. In fact, cross reactions with unexpected proteins can and do occur. Some years ago, we defined the epitope of a monoclonal antibody to myosin light chains. It turned out to be an unusual PTM, trimethyl alanine (Boey et al., 1992). No other contractile protein has such a modification and we believed it to be unique. It was not. Another group also had reported an antibody directed against a nuclear protein which was almost certainly had the same epitope. Others have gone so far as to say that most monoclonal antibodies have some, limited degree of cross-reactivity.
An important issue is the preservation of activity of the capturing antibody. The chemical nature of the solid support is critical for the preservation of protein structure (Mitchell, 2002). More than 50% of antibodies are thought to denature on solid planar surfaces such as glass, and considerable effort has been spent on ameliorating the forces that distort protein structure.
Finding and arraying the right antibody
Thousands of monoclonal antibodies are available for use in protein arrays and the number increases daily. Search engines are quite good at finding commercial sources (e.g.
www.abcam.com) but research laboratories generate many more antibodies. Even when an antibody is directed against a protein, more than one clone may be available, each with a different affinity and epitope.
Once the right antibody has been identified, its use in an array is normally very economical. 50–500 pl can be delivered to an array surface, so it follows that the purchase of a single 0.5 ml sample of a monoclonal antibody will normally be sufficient to dispense onto very large numbers of arrays.
A reasonable estimate of the number of proteins expressed in a cardiomyocytes is about 104, and gene arrays tell us that about 10% of these differ in non-diseased and heart failure (Barrans et al., 2002). It is wise to employ more than one clone for those proteins whose expression is critically elevated (or depressed) and of course isotype and negative controls should be present in the array. At an average purchase price of about $200 per antibody, clearly the development of a comprehensive protein array for myocardial proteins associated with human heart failure can be costly. The cost of each antibody in an array will be small (
Surface denaturation of antibodies can be overcome by either capturing them by their Fc portions, or by immobilizing Fab fragments by their free cysteinyls that face away from the antigen-combining surface. Accelr8 Technology (www.accelr8.com) and Prolinx (www.prolinx.com) have developed novel three-dimensional coatings for solid supports that may reduce denaturation while increasing ligand densities. They also increase the cost.
Commercially available antibody arrays (BD Biosciences, Amersham) are arrayed on modified glass and at present contain only limited numbers (about 250–500) of antibodies. Like the early gene arrays, they will rapidly progress into more extensive devices while maintaining a unit cost of a few hundred dollars. Several companies now offer custom antibody arraying services but the range of antibodies from these companies is currently limited and tends to be specialised (e.g. for chemokines and cytokine receptors).
Initially, our choice of antibodies for a human cardiac protein array will be based on those suggested by the gene arrays (200–300 genes) and those identified by 2DGE (about 70 proteins). This number will be manageable in the start-up phase and can be expanded later.
Blocking indiscriminate binding
When a solid support is used to immobilize an antibody, the unoccupied surfaces generally must be blocked to prevent attachment of non-specific proteins. Usually a universal protein such as powdered skim milk can be used as effectively as more costly proteins like bovine serum albumin. No blocking protein will be ideal but skim milk is particularly problematic where biotin–streptavidin complexes are used in downstream detection because of the presence of biotin.
A better solution is to avoid the blocking process all together. One approach is to control the hydrophobilicy–hydrophilicy of the capturing surface. Hydrophobic surfaces bind most proteins better than hydrophilic surfaces but separation on this basis alone is unlikely to avoid the need to block. Two companies mentioned above have developed surfaces that do not seem to require the blocking step. Accelr8 Technology uses a cross-linked matrix combined with the selective capture of ligands via streptavidin-linked surfaces (to capture of biotinylated proteins), or via thiol- or amine-reactive surfaces to covalently attach proteins. Prolinx Inc. uses polymer ‘brushes’ containing phenyl(di)boronic acid that cross-reacts with proteins pre-modified with salicylhydroxamic acid.
Detection of bound proteins
Antibodies can be used to detect ligands captured by primary antibodies in what amounts to a solid state ELISA assay. There are at least two problems with this approach. Firstly, the detecting antibody must not have the same epitope as the capture antibody. Also, the antibody must have a sufficiently high affinity to detect low abundance components in the array. This can be achieved by labelling the second antibody with a fluorescent probe (these are often commercially available) but this approach can run into sensitivity problems. Clearly the assay is limited by the weakest binding affinity.
Another approach is to biotinylate the second antibody and then detect the biotin moiety with a streptavidin conjugated to a visible probe like HRP or a fluorescent label. Streptavidin is now available linked to magnetic microbeads (www.miltenyibiotec.com) or to streptavidin-conjugated Q-Dots™(www.qdots.com). With these, levels of detection can be substantially lowered since single Qdots and microbeads can be unambiguously identified and quantified. None of these solutions compare with detection by mass spectrometry (see below).
Advantages of protein arrays over 2DGE gels
A major driver of protein array technology has been the short-comings of two-dimensional gel electrophoresis. We have already reviewed the substantial contribution of
2DGE gels to our understanding of the protein changes associated with heart failure in both man and animal models of human heart failure. The principal bottleneck with
2DGE is the identification of the proteins where more than one protein may be present per spot. MALDI mass spectrometry greatly improved this problem, and narrow range first dimension p
I range strips also helped.
New generation seven tesla Fourier transform ion cyclotron resonance mass spectrometers are now available that can identify protein isoforms that differ by only a few Daltons (Lee et al., 2002). These instruments may be expensive but they bring immense power to the problem of identifying protein differences between failing and non-failing hearts. They may even overcome a fundamental failing in the analysis of the proteome, namely absolute rather than relative quantitative evaluation of individual proteins.
Answers to all your Biology Questions
