MicroRNAs (miRNAs) are ~22 nucleotide non-coding RNAs that regulate protein production by pairing to appropriate complementary stretches in mRNAs 
. Hundreds of miRNAs are encoded in the human genome, with an estimated 30% of mRNAs possessing conserved miRNA binding sites, suggesting that miRNA-based regulation is an integral component of the global gene expression program 
. The importance and functional range of miRNAs is evident from their widespread occurrence and the diverse and often strong phenotypes and disease states associated with mutation or altered expression of miRNAs 
. miRNAs function through formation of a ribonucleoprotein complex termed the RNA-induced silencing complex (RISC) 
. In humans, RISC is minimally composed of a guide miRNA bound to an Argonaute protein (Ago 1, 2, 3 or 4), along with Dicer and the HIV transactivating response binding protein (TRBP) 
. Experiments in mice and human cell lines show that Ago2 is the central RISC component, capable of cleaving target mRNA when there is perfect miRNA:mRNA complementarity 
. However, most miRNA:mRNA interactions in metazoans have imperfect complementarity 
, and it is likely that an overwhelming majority of miRNA targets are not cleaved by Ago2. In many cases it is likely that miRNAs repress translation and/or promote decay of their mRNA targets 
A combination of experimental and computational approaches has begun to elucidate how mRNA targets are specifically recognized by miRNAs. From this large body of work, several salient features of target recognition have emerged. First, it is likely that most miRNA target sites are located in 3′-untranslated regions (UTRs) of mRNAs 
. Sites in coding sequences and, in at least one instance, 5′-UTR can also lead to decreased protein levels, although they do so less efficiently than sites in 3′-UTRs 
. Second, a stretch of six to eight nucleotides near the 5′-end of the miRNA, the “seed region”, are particularly important for miRNA function 
. Their importance is underscored by the fact that the complementary regions are among the most evolutionarily conserved regions in mRNA targets and in some instances the seed match alone appears sufficient to confer recognition 
The observation that miRNAs cause decreases in the abundance of at least some mRNA targets provides a powerful strategy for determining what features in mRNA and miRNA sequences contribute to specificity 
. Recently, Lim et al.
found that transfection of each of two miRNAs, heart-specific miR-1 and brain/kidney-specific miR-124, into HeLa cells led to decreases in abundance of at least 96 and 174 mRNAs respectively, many of which were likely to be direct targets as inferred from the enrichment of seed matches in their 3′-UTRs (~90% had 6mer seed matches) 
. The observation that many of these targets had conserved seed matches in their 3′-UTRs and that overexpression of the miRNA induced a muscle-like or brain-like gene expression program, respectively, suggested many of the apparent targets were physiological, even though miR-1 and miR-124 are not normally present in HeLa cells. In addition to the 3′-UTR sites, the authors found evidence for some targeting to sites in coding sequences. This miRNA overexpression/microarray approach was subsequently expanded to 11 miRNAs and used to identify additional features in mRNAs that contribute to changes in target mRNA levels 
. These data provided the basis for a model for the effectiveness of each seed match site in 3′-UTRs of mRNAs for ~450 miRNAs (TargetScan 4.0). Other miRNA target prediction methods are based on limited experimental data and theoretical considerations (e.g.
mRNA secondary structure surrounding predicted sites), but only limited functional data are available to test their performance 
One limitation of current approaches is that targets are often inferred from changes in mRNA abundance; however, miRNA-induced decreases in protein levels can only partially be accounted for by changes in mRNA levels, consistent with the view that miRNAs affect both translation and mRNA decay 
. In addition, identifying targets by altering miRNA expression and measuring changes in mRNA levels returns no information on which targets might be the most important in carrying out the actual biological processes (e.g.
cellular differentiation) and is limited to the study of the altered miRNA. Conservation is commonly used as a filter to identify likely targets, but many functional sites are not conserved and many conserved sites do not seem to be functional 
. Although useful, the existing methods may be capturing an incomplete and possibly biased subset of miRNA targets.
A direct experimental method to identify miRNA targets that does not rely on any specific mechanism of regulation, conservation, or the altered expression of specific miRNAs is required to fully explore the suite of miRNA targets. Here we describe a simple method that provides quantitative information about which mRNAs are being regulated by miRNAs in a cell population. We express affinity-tagged Ago2 in Human Embryonic Kidney (HEK) 293T cells, immunopurify the resulting tagged Ago2 complexes, and identify the associated mRNAs and miRNAs using DNA microarrays. This Ago2 immunopurification (IP)/microarray approach allows miRNA targets to be comprehensively identified in an unbiased fashion, and provides a method for comprehensively assessing the regulation of mRNAs by RISC. In addition, mRNA targets of particular miRNAs can be identified by comparing the Ago2 IP/microarray profiles of cells expressing a particular miRNA to the Ago2 IP/microarray profiles of untreated cells.