Understanding variation in complex biological responses, such as those mediated by the p53 tumor suppressor, requires integrating bioinformatics with experimental approaches that reveal fundamental molecular events and their phenotypic consequences at the cellular level. The relationship between the degree of DNA binding and target gene transactivation has been the subject of many studies and there is considerable interest in how human genetic variation in TF binding sites may impact gene expression and disease susceptibility 
. Several recent studies have identified SNPs in p53 REs and evaluated their impact on expression or binding 
. The present results suggest that MAPD can be useful for assessing the magnitude of p53 binding to newly discovered candidate sequences and also for evaluating the impact of SNPs or mutations in those sequences.
Whole genome association studies frequently identify SNPs in gene regulatory regions containing putative TFBSs; assessing TF binding using the approach outlined here is one practical way to evaluate the effects of regulatory region SNPs. One issue in developing good prediction of TFBSs has been the lack of quantitative binding data. We show that integrating binding information from site-directed mutagenesis of a p53 consensus sequence into a computational binding site model improves prediction for a second set of REs.
We used MAPD to investigate sequence-specific binding between p53 protein and a number of known and well-studied REs in the p53 pathway. Various in vitro assays have suggested that p53 has a higher affinity for REs in cell-cycle arrest and DNA repair genes than the REs in apoptosis genes 
. Similar observations were obtained in vivo using a yeast-based system 
, and the present results are in agreement. This is also supported by Riley et al, who report that apoptosis-related p53 REs differ more from consensus binding sequence than cell cycle-related REs. In addition an evolutionary analysis 
indicates that p53 RE sequences in human apoptosis genes tend to diverge considerably from those of rodents—although some weak binding REs do demonstrate strong evolutionary conservation.
We observed a strong correlation between binding to target p53 REs and p53-driven luciferase gene expression in mammalian cells (). We have also seen that binding correlates with expression for the p21 RE when constructed with variable numbers of nucleotides between p53 half-sites 
. These correlations suggest that our in vitro binding assay using p53 in nuclear extracts may be a surrogate for some in vivo transactivation studies. However, further work to evaluate binding to full length native promoter sequence with direct comparisons to chromatin IP and expression is needed 
. The importance of p53 binding has been highlighted by Shaked et al al 
who reported that p53 activation status and RE occupancy differ between tumor cell lines and normal primary cells. Our approach, assessing binding with nuclear extracts, can be used with tumor lines, EBV transformed lymphoblasts or normal primary cells. This will allow quantitative evaluation of the impact of post-translationally modified and activated wt p53 on binding to target gene REs as in , or under various stress and DNA-damaging situations as in Shaked et al 
. Given that p53 mutations are the most common genetic changes in cancer and are a major determinant of treatment outcome 
, the MAPD system may be useful in assessing the impact of mutations on p53 RE interactions (). Since radiation and alkylating agents are primary therapies in cancer treatment, understanding how exposures affect p53 activation and binding could provide insight into tumor biology, carcinogenesis and treatment response 
TF binding to REs is a necessary step in regulating transcriptional networks 
. EMSA and chromatin immunoprecipitation (ChIP) have been essential in advancing the study of gene regulation. However, both approaches are low in sample throughput, labor–intensive, difficult to quantify and have a limited dynamic range. More recently, ChIP combined with microarrays 
or sequencing, and protein binding arrays 
have allowed parallel analyses of TF binding to thousands of sequences, and these techniques are highly effective for qualitative discovery of binding sites. However, currently these approaches have experimental constraints, including relatively low sample throughput, uncharacterized variability, and are expensive. Surface plasmon resonance and fluorescence anisotropy are useful for determining detailed biochemical parameters, but like protein binding arrays are currently restricted to purified proteins. MAPD complements existing approaches for the functional examination of candidate binding sequences because this system permits relative quantification and allows multiple endpoints (up to 100 DNA REs per experiment including internal reference controls). In addition, the use of cell extracts in a rapid, low-cost 96-well format enables complex experimental designs with large numbers of samples and replication. An automated format is possible and screening for inhibitors of binding or compounds that selectively restore binding could be accomplished. A strength of the MAPD approach over other binding methods is the inclusion of an internal reference positive control (for example, p21, a strong binding RE) and a negative control (WRNC, a nonbinding sequence) within the multiplexed assay which provide for quality control and normalization across experiments.
MAPD is a general methodology as demonstrated by detection of not only p53 binding, but also the nuclear hormone receptor ERα binding its RE (Figure S5
). MAPD could be adapted to detect the impact of multiple proteins in a transcription complex or to evaluate multiple binding sites within native or artificially constructed promoters. Ongoing studies are examining the functional evolution of p53 response elements across mammalian species 
, co-evolution of the p53 protein, as well as rules for p53 transactivation 
. Because of MAPD's sensitivity and dynamic range, the impact of SNPs or mutations on binding can be assessed providing a valuable tool in genomics research. We found that subtle changes in binding sequence can have effects that are not captured with existing computational models. Using nuclear extracts from human cells expressing p53 mutants, R175H and S121F, we observe a loss of binding and a change of binding specificity, respectively, consistent with their effect on gene expression. The correlation between binding in this nuclear extract-based in vitro system and transactivation in mammalian cells suggests the important role that sequence-specific p53 interactions play in transcriptional activation. MAPD will facilitate studies of how variation in target binding sites affects binding and fine-tuning the regulation of transcriptional networks as well as the impact of polymorphic variation on master regulatory networks.