Protein Microarray Technology
Over the past decade, microarray technology has revolutionized the study of gene expression. Arrays of nucleic acids, comprising either single-stranded oligonucleotides or double-stranded PCR products, have been used to measure the abundance of thousands of transcripts in cell or tissue samples. The technology is well suited to large-scale, system-wide investigations for two reasons: (1) it enables many different samples to be interrogated simultaneously in a rapid and economical fashion; and (2) it enables such experiments to be performed hundreds or even thousands of times, with different cells under different conditions. These two features of the technology apply equally well to the system-wide study of protein function.
We have previously developed and reported methods to fabricate microarrays of purified proteins at high spatial density on chemically-derivatized glass microscope slides (abstract). In order to perform functional assays on the arrayed proteins, we developed surfaces that permit covalent attachment of proteins in multiple orientations, preserve the folded and active states of the immobilized proteins, and exhibit low non-specific binding properties. In proof-of-concept experiments, we demonstrated that microarrays of proteins can be probed with fluorescently labeled molecules (proteins, peptides, or small organic compounds) to identify stable interactions (Fig. 1), and that the selectivity of the proteins for their cognate ligands is preserved on the arrays. These experiments demonstrated that protein microarray technology provides a robust way to study protein function in a rapid, economical, and system-wide fashion.
We are now applying this technology to the comprehensive and quantiative study of whole families of protein interaction domains (SH2/PTB domains and PDZ domains). These projects involve probing microarrays of 150-300 proteins with hundreds of fluorescent peptides or proteins. In order to reduce variation introduced at the probing and washing steps, as well as to facilitate processing thousands of arrays, we developed a way to fabricate protein microarrays in separate wells of 96- and 384-well microtiter plates (Fig. 2). To maximize the number of spots in each well, we array our proteins onto microtiter plate-sized pieces of aldehyde-derivatized glass, custom manufactured by Erie Scientific (Fig. 2A). To do this, we use a contact-printing microarrayer from TeleChem International (the NanoPrint Microarrayer). We then attach the glass to the bottom of a bottomless microtiter plate, using an intervening silicone gasket custom manufactured by Grace Bio-Labs (Fig. 2B). The gasket forms a watertight seal between adjacent arrays and enables us to wash the arrays with a microtiter plate washer. The arrays are then visualized with a Tecan LS400 scanner.
Figure 2. Strategy to fabricate protein microarrays in microtiter plates. (A) Piezoelectric microarrayer, depositing proteins on an aldehyde-derivatized glass plate. (B) 96 different domains are deposited in duplicate in each well and the same array is fabricated 96 times on a single piece of glass. The glass slab is then fixed to the bottom of a bottomless microtiter plate using an intervening silicone gasket.
Although probing a protein array with a ligand of interest highlights a subset of specific interactions, the resulting information can be misleading. We have learned not to rely solely on the fluorescent intensity of individual spots. Some interactions show up brightly while others are less obvious. This, we presume, is because some proteins lose very little activity when immobilized while others are less well behaved. Importantly, the intensity of the spot does not always correlate with the strength of the interaction. A truly valuable protein interaction network should indicate not just connections between proteins, but also the strength of those connections. To address this challenge, we probe our arrays with eight different concentrations of each peptide, ranging from 10 nM to 5 µM (Fig. 3). We then quantify the fluorescence at each spot. Assuming that the system reaches equilibrium during the incubation step, the mean fluorescence of duplicate spots on our arrays (Fobs) can be described by equation 1,
where Fmax is the maximum fluorescence at saturation, [pep] is the total peptide concentration, and KD is the apparent equilibrium dissociation constant. For each fluorescent ligand, we fit separate binding curves for each of the proteins on the array. Since nonspecific binding increases linearly with ligand concentration, while specific binding saturates, we score as “specific” those interactions that fit well to equation 1 (R2>0.9) and exhibit a KD below 2 µM with an Fmax that is at least two-fold higher than the mean fluorescence of control spots.
Figure 3. Microarrays of human SH2 and PTB domains in a 96-well microtiter plate. A small amount of Cy5-BSA was added to each protein, giving rise to the red image on the left. The arrays were probed with eight concentrations of each of eleven 5(6)-TAMRA-labeled phosphopeptides, giving rise to the green image on the right.
To assess the accuracy of our method, we measured KD’s for eight domain-peptide interactions using surface plasmon resonance (SPR). For every interaction, the free energy of binding (deltaG) calculated from our high-throughput analysis matched within 5% the value measured by SPR. Overall, the deltaG measured by each method agreed with an R2 correlation coefficient of 0.87.
To see quantitative networks that we have uncovered using this approach, click here.