High-throughput methodologies, such as SAGE, DNA/oligo microarrays, yeast two-hybrid, and RNA-seq that measure abundance and/or dynamic changes of entire mRNA populations in cells or tissues, have had profound impacts on modern biology. However, our ability to monitor dynamic changes in protein products is currently insufficient. For example, the 2D-gel electrophoresis is a traditional approach to profiling protein expression in a cell or tissue; however, it had only limited success, mainly due to its limitation in resolving different proteins (e.g., <4,000) and detection of PTMs. Without using the time-consuming MS/MS approach, the identity of those separated protein spots is not readily known. More importantly, it does not provide any information about the global activity of kinases. One challenging aspect is the fact that the abundance and/or dynamic changes of a given protein often involve various types of protein PTMs, such as phosphorylation, ubiquitylation, SUMOylation, acetylation, and glycosylation. These PTMs can dictate protein activity, function, fate, and subcellular localization. A high-throughput, unbiased, and cost-effective method that allows for global measurement of different types of PTMs in a given cell or tissue could have a huge impact on the study of protein dynamics. Our microarray based approach would specifically work to meet this need.
We view this method as a parallel to the DNA microarray technology. The latter uses in vitro hybridization on a microarray to measure mRNA abundance in cell extracts, whereas ours uses in vitro phosphorylation of specific substrates on a protein microarray to profile kinase activity in cell lysates. Also similar to DNA microarrays, this new method can be further exploited by profiling phosphorylomes at multiple time points and with different cell/tissue types, thus, providing a dynamic global picture of the rapid changes in kinase activity that occur both in cellulo and in vivo. This method is also potentially amenable to the measurement of other PTM enzyme families, such as ubiquitin and SUMO E3 ligases, acetyltransferases, and proteases. Further, the method is capable of examining different sample types (cell extracts, tissue lysates, biological fluids), developmental stages, and physiological states (normal versus diseased or treated versus untreated cells).
We and others have demonstrated that novel biological functions and pathways can be discovered from the use of functional protein microarrays when coupled with sophisticated bioinformatics analyses and in-depth in vivo
characterization. This often occurs when a covalent PTM reaction is studied under highly simplified in vitro
reaction conditions and testing a single PTM enzyme at a time. The successful examples include identification of substrates of protein kinases 
, acetyltransferases 
, and ubiquitin E3 ligases 
. However, these approaches risk the physiologically relevant environment, such as the context of protein complexes, adapter proteins, inhibitors, and subcellular compartmentalization. Also, such methods are not amenable to the characterization of global PTM activity. The use of cell/tissue lysates for profiling kinase activity not only allowed us to examine the dynamic changes in kinase activity as a whole, but also preserved the physiologically relevant microenvironments (i.e., in the context of protein complexes) for the kinases to execute their activity on the functional protein microarrays. Furthermore, the relevant adapters and/or required scaffold proteins are also readily available in these reactions. Indeed, using HGF/c-Met activation in cultured cells and xenograft tissues as a proof-of-principle, we were able to reproducibly recover differentially phosphorylated proteins that are known to be phosphorylated upon activation of the HGF/c-Met signaling pathway. The fidelity of the method was validated by the recovery of a high percentage of known downstream targets associated with HGF-induced cell growth, cell cycle progression, and cell survival, including AKT and MAPK. More importantly, the discovery and confirmation of novel components downstream of the HGF/c-Met signaling pathway demonstrates the usefulness of this method. Our findings underscore the complexity of the signaling events that mediate the phenotypes associated with HGF and tumorigenesis. The ability to monitor the phosphorylation of these proteins simultaneously should provide a valuable technology for network analyses and drug target discovery.
Although the use of cell lysates to discover kinase-substrate relationships has been reported previously 
, our approach allows for a more comprehensive survey of a given proteome. We are aware that this high-throughput approach has its own shortcomings, including loss of compartmentalization and competition from endogenous proteins for the kinases. To enhance the sensitivity of this method, phosphatase and protease inhibitors were added into the lysates upon lysis. While the addition of the former would theoretically increase the auto- and trans-phosphorylation activity of the kinases, such arbitrary effects can be cancelled out when condition-dependent phosphorylation events are selected. Furthermore, additional bioinformatics analyses, as exemplified in this study, helps to exclude obvious false positives. The fact that all the tested novel candidates were confirmed in cell-based assays suggests that this approach is likely to identify differentially phosphorylated proteins in vivo
. Although the antibody arrays currently used are able to reasonably identify phosphorylated kinases in a mixture, it is not our intention to predict upstream kinases of these condition-dependent phosphorylated proteins in this study. Rather we have elucidated the global phosphorylated proteins in our selected system and envision that we will be able to use this technique to predict the corresponding upstream kinases once a more comprehensive and reliable network of human kinase-substrate relationships is available.