Proteomic analyses to identify novel therapeutic targets for the treatment of advanced prostate cancer

Abstract

At present there is no cure for advanced prostate cancer once it progresses to an androgen independent stage. Hormonal therapy, radiotherapy, and chemotherapy all have limited durations of efficacy for men diagnosed with androgen independent disease and patients will succumb over a period of several months to two years. The androgen receptor (AR) has been suspected to play an important role in the mechanism of progression to androgen independence. This is because the AR is a transcription factor that ‘normally’ mediates the effects of androgen to regulate expression of genes involved in proliferation and survival of prostate cells. Thus identifying and characterizing the proteins that interact with the AR to facilitate an activated receptor is of critical importance. Proteomic approaches such as isotope-coded affinity tags (ICAT), isotope Tags for Relative and Absolute Quantification (iTRAQ)^TM, Stable Isotope Labeling with Amino acids in Cell culture (SILAC), Tandem Affinity Purification (TAP) of tagged proteins (TAP-tag) and Multidimensional Protein Identification Technology (MudPIT) provide large scale unbiased strategies and have not been previously applied to identify proteins that interact with the AR. Here an example of the power of these proteomic approaches to identify potential therapeutic targets for prostate cancer is provided. Application of MudPIT identified 82 peptides in endogenous complexes immunoprecipitated with the AR from prostate cancer cells. Identification of these novel proteins may ultimately lead to the development of better therapies for the treatment or prevention of advanced prostate cancer.

Advanced prostate cancer

Prostate cancer has a propensity to metastasize to the bone to form predominantly sclerotic or blastic lesions. Approximately 20% of men treated with radical prostatectomy will experience tumour recurrence and 11% of men will already have osseous (bone) metastases at the time of clinical presentation [1]. Most men that succumb to prostate cancer have osseous metastases. The only effective systemic therapy available for metastatic prostate cancer is androgen deprivation by chemical or surgical castration. This therapy is not curative and after an initial response the disease will relapse with the ultimate progression to androgen independence. An early sign of androgen-independence, related to reduced survival, is a rising titer of serum prostate-specific antigen (PSA) after an initial response to androgen deprivation. Currently the mechanism of how prostate cancer escapes hormonal control is unknown. Similarly, how transcription of the androgen-regulated PSA gene escapes regulation by androgens in advanced prostate cancer is also unknown.

Suspected mechanism of hormonal progression to androgen independence

Several theories have been advanced, many of which point to the androgen receptor (AR) as a probable factor in the mechanism of hormonal progression. Data supporting a role for the AR is based upon the following: 1) the AR is expressed in the nuclei of the majority of hormone refractory tumours [2–4]; 2) mutations in the AR can result in hypersensitivity to low castrate levels of androgens or in being activated by steroids other than androgens [5]; 3) amplification of the AR gene has been detected in 20% to 30% of androgen independent tumors [6]; 4) genes normally regulated by androgens such as PSA are re-expressed in androgen-independent disease [7]; 5) the timing and sequence of use of the family of anti-androgens may prolong the time to androgen independence [8]; 6) ligand-independent activation of the AR has been shown to occur in prostate cancer cells maintained in monolayer [9–14]; 7) the AR has been shown to be necessary for the proliferation of androgen independent prostate cancer cells [15]; and 8) increased expression of AR is associated with the development of resistance to antiandrogen therapy [16]. It has also been proposed that there are low levels of androgen remaining in clinical tissues from castrated men that may be sufficient to mediate biological activity [17,18]. Current antiandrogens clinically used target the ligand-binding domain (LBD) of the AR but unfortunately have poor affinity compared to dihydrotestosterone. Thus better drugs that inhibit the AR are of intense interest for the treatment of advanced prostate cancer. Therapeutic strategies targeting the AR proposed include: 1) identifying interacting proteins that are essential for the transcriptional activity of the AR; 2) developing drugs that specifically block interactions with those proteins that enhance AR activity such as co-activators; and 3) developing drugs that specifically stabilize or mimic interactions with repressors of AR activity.

Androgen receptor structure

The AR belongs to the steroid hormone receptor family. Members of this family are composed of four domains: the N-terminus domain (NTD), the LBD, the DNA-binding domain (DBD) and the hinge region [19]. The NTD contains the transcriptional activation sites (AF-1 and AF-5) that are essential for AR transcriptional activity [20]. Deletions of these regions result in loss of receptor transcriptional activity [21]. In an inactivated state the NTD is unfolded. Folding of this domain is stabilized through protein-protein interactions [22,23]. The DBD is the most conserved region among all the members of the steroid receptor family. It contains amino acids sequences that allow specific contacts with the DNA, androgen response elements (AREs), as well as protein-protein interactions essential for dimerization [24]. The LBD is required for the specific binding of ligands and subsequent activation of the receptor. Deletion of the LBD yields a constitutively active receptor capable of initiating transcription which is consistent with this domain having a regulatory role [25]. Mutations in the LBD have been identified from clinical samples and can cause destabilization of AR conformation or loss of specificity in ligand-binding (promiscuity). The hinge region contains a nuclear localization signal required for AR translocation to the nucleus.

Activation of the androgen receptor

The AR mediates the effects of androgens in the cell. In the absence of androgens, the non-activated AR resides predominantly in the cytoplasm in an unfolded state [26]. Binding of ligand (androgen) results in an activation step such that the AR can effectively bind to its respective ARE. Addition of androgens results in a ligand-induced change in transformation of the AR that may include the following of which each is a potential drug target (Fig 1): 1) dissociation of heat-shock proteins which may act as a cytoplasmic anchor or inhibit DNA interactions; 2) translocation from the cytoplasm to the nucleus; 3) a change in AR conformation required for DNA binding and protein-protein interactions; 4) post-translational modification (e. g. phosphorylation or acetylation); 5) dimerization and AR-DNA interaction; and 6) assembly of protein complex (protein-protein interactions) on the DNA. Fundamental to AR transcriptional activity is its antiparallel dimerization involving interaction between its NTD and the LBD. Specific sequences involved in the interaction are the FXXLF (²³FGNLF²⁷) core sequence in the NTD that interacts with ⁴³³WXXLF⁴³⁷ in the LBD [27, 28]. The role of some coregulators appears to involve stabilizing (coactivator) or interfering (co-repressor) with this interaction while other proteins alter post-transcriptional modifications or cellular localization [29]. Inhibition of interaction between the NTD and LBD is a potential drug target as well as inhibitors that block one or more of the above mechanisms of transformation.

Figure 1. Activation of the androgen receptor and potential drug targets.

The AR can be activated by binding androgens (DHT) or in the absence of androgens via alternative pathways involving IL-6 and growth factors (GF) that interact with receptors in the plasma membrane. Inhibition of MAPK activity can block activation of the AR by both androgen and IL-6 or PKA by an unknown mechanism. The AR is a phosphoprotein and translocates to the nucleus. NCoR and SMRT are co-repressors that inhibit AR transcriptional activity. In the nucleus the AR dimerizes and interacts with coactivators (e.g., SRC-1) and the basal transcriptional machinery (BTM) and proteins that alter histone acetylation (e.g., CBP and SWI/SNF) to enhance transcription of genes containing AREs such as PSA and genes involved in proliferation and progression.

Proteins interacting with AR

Interaction of the AR with specific proteins is required for its successful transformation from the cytoplasmic inactive state to its nuclear transcriptionally active form. Thus a battery of proteins, each with a different function is required for each step in the transformation process. For example the AR will have to interact with proteins involved in protein trafficking which would regulate AR activation through controlling its translocation to the nucleus [30]. The AR may also be regulated by degradation [31] or in its formation of a complex with the transcriptional machinery. Coregulators can promote AR transcription [32] by acting as bridging proteins with the transcription factors or operating in multi-complexes [33]. Proteins can also inhibit transcriptional activity of the AR [34]. Inappropriate transcriptional derepression can play a role in the development of genetic diseases [35]. Posttranslational modifications of the receptors such as phosphorylation are involved in the process. Thus the search for proteins that interact with the AR is the major focus in the quest for development of new therapies for the treatment of advanced prostate cancer. A database of proteins that interact with the AR can be found at www.androgendb.mcgill.ca. These proteins include transcription factors, kinases, coactivators, corepressors, chaperons, and enzymes such as ubiquitin protein ligases. Unfortunately most all of these interactions have not been validated with the endogenous proteins or in physiologically relevant cells lines. Forced constitutive expression typically leads to overexpression of ectopic proteins beyond “normal” physiological levels and loss of regulation that may be dependent upon phase of cell cycle or other cellular states. This may result in aberrant interactions and/or altered cellular localization leading to false positives. Overexpression of kinases can cause loss of specificity of substrate phosphorylation. Protein-protein interactions may be cell-specific thereby emphasizing the requirement for analysis in relevant cell lines.

Common methods used to determine protein-protein interactions Yeast two-hybrid

There are a number of methods commonly used to investigate protein-protein interactions. One powerful approach is the yeast two-hybrid system. The yeast two-hybrid system was first described in 1991. The technique is based on the construction of a gene fusion. The protein under investigation (the bait) is fused to the DNA-binding domain of GAL4. A library of genes fused with a transactivation domain is used in order to screen for a possible partner (the prey). Identification of an interaction is achieved when a functional transcription factor is generated [36]. Isolation of positive yeast clones is obtained employing nutritional markers and enzymatic reporters. Variations of this technique are related to the nature of the transcription factor and the reporter gene used. Strong transcription activation domains are suitable to investigate weak protein-protein interactions but, on the other hand, increase the detection of “false positives”. The strain of yeast is also an important variable in obtaining results. A high number of copies of the upstream activating sequence, to which DNA can bind, will increase the sensitivity of the test but also the number of unspecific positive results. More recent two-hybrid systems employ a third selection marker to reduce false positives. These positive signals are due to different reasons: components under investigation can be intrinsically sticky or proteins directly interact with the promoters [37]. Overexpression of proteins in yeast can lead to modifications in yeast permeability and growth medium-specific differences in toxicity. Proteins are expressed differently depending on the strain of yeast such that the same vector may be present in varying numbers in different colonies which may vary with modification in protocols used by different laboratories. This makes the comparison of data obtained in various laboratories difficult [38]. With current estimates of 47–91% for false-positive rates in genome-wide screens [39], strategies to reduce false positives have been proposed that can include the simultaneous use of different reporter genes in order to minimize casual activations due to a particular promoter. Using two different sequential screenings with different host strains containing different GAL4 promoters may aid in reducing false positives. A procedure that employs a classical transformation method followed by a second screening employing a yeast genetic mating technique can isolate “true” interactions [40]. Protein-proteins interactions may also be missed by the screen to produce “False negatives”. For example, ORFs used are products of PCR and mutations in some populations can cause misincorporation in PCR products. In addition, some proteins may require cell-specific posttranslational modifications (eg., phosphorylation and/or acetylation) or processing (cleavage) prior to interaction [41].

Mammalian Two-Hybrid assay

The Yeast Two-Hybrid System is very effective for investigating protein-protein interactions. Therefore systems to detect protein interactions in mammalian cells have been developed [42]. The principle beneath this assay is the same as the one employed in yeast. A bait protein of interest is expressed as a fusion protein with the GAL4 DNA binding domain and the target protein is expressed as a fusion protein containing the herpes simplex virus VP16 activation domain. These two domains are both part of a transcriptional activator (the GAL4 protein). Therefore, if the two proteins, expressed in mammalian cells, interact, the transcriptional activator is reconstituted. This leads to the activation of a reporter gene usually containing the chloramphenicol acetyl transferase (CAT) or the luciferase gene, whose activity is measured in cells extracts. In comparison to yeast, this method displays two important advantages. First, it allows the detection of interactions that depend on posttranscriptional modifications, that may not occur in yeast. Second, it allows monitoring modifications in the interaction in response to cell stimulation [43]. This system is frequently used to confirm the interaction of known proteins. Once an interaction between two proteins has been identified in a yeast-two hybrid screening, the mammalian two-hybrid assay can be used in combination with deletion or site-directed mutagenesis to identify the specific domains and/or residues required for the interaction. This system was used to show that the FXXLF motif in AR is the region that mediates specific interactions with coregulators. ARA70, ARA55 and ARA54 were previously selected with the yeast two-hybrid system and also showed the presence of FXXLF motifs. The mammalian two hybrid system provided evidence of androgen-dependent interactions of this motif with the AR LBD involving specific amino acid residues [44]. Mutations in AR disrupted interactions between the LBD and coregulator protein TIF2, as well disrupted dimerization between the AR LBD and NTD [45].

Glutathione S-transferase (GST) -pulldown

The GST pull-down assay has been employed to identify interactions as well as validate results obtained from the yeast two-hybrid system or other approaches. The bait is expressed as a GST fusion protein usually produced in E. coli and adhered onto a glutathione-agarose column. Cellular extracts can be run on the column to identify proteins in the extract that bind to the bait. Proteins interacting with the GST-fusion protein can subsequently be identified by mass spectrometry. Alternatively if the assay is being used to validate interactions obtained from another screen such as the yeast two hybrid, both bait and prey may be applied as either a recombinant fusion protein or the prey may be overexpressed in cells and then applied. Critical components of this assay are the selection and formation of the bait protein (the purified protein can be tagged or subcloned) and if the interaction under study is stable or transient. Several weaknesses of the assay should be kept in mind. In the case of mixing together two recombinant proteins at relatively high concentrations there is little physiological relevance since the proteins may just be sticky. Nor does this approach consider that the proteins may not colocalize in the cell and thereby never come into proximity with one another. A false-positive rate of 61% and false-negative rate of 38% were reported for this approach [46]. The elution of the complex from the support is also a determinant point of control. Usually the complex is eluted with SDS-PAGE sample buffer. This method leads to denaturation of protein and, consequently, of the complex, dilution of the sample and elution from the matrix of unspecific bound proteins. It is important that proteins maintain a correctly folded conformation similar to that found in vivo. The AR NTD which is a major therapeutic target [11, 47] has been reported to not be folded correctly in vitro thereby raising the possibility for aberrant interactions to occur [48].

Co-immunoprecipitation of interacting proteins

The optimal method to confirm a protein-protein interaction is co-immunoprecipitation of interacting endogenous proteins within relevant cells. Here an antibody is used to immunoprecipitate the bait protein with its interacting partners from cells. However, this assay is dependent upon the availability of “good” antibodies that are specific and target an epitope that is not masked by interacting proteins. Levels of expression of the interacting proteins and the affinity of interaction are major determinants. Few endogenous interactions have been reported for prostate cancer. The first was for SRC-1 interaction with the AR NTD in response to both ligand and IL-6 [13]. Since then Daxx, Dax1, SMRT and NCoR have been identified from endogenous complexes with the AR in prostate cells [49–52]. To circumvent the general lack of good antibodies and levels of expression in cells, transfection experiments have been utilized to overexpress one or both of the two interacting proteins. Overexpression of proteins in the cells, however, can lead to modifications in the physiological environment within the cell. Transient overexpression of glucocorticoid receptor (GR) without hormone treatment in HeLa cells causes the accumulation of non-hormone-binding receptor forms that bind DNA in vitro. These forms negatively interfere with constitutive active mutants of GR converting the receptor from an activator to an inhibitor [53].

Proteomic approaches to examining protein-protein interactions

Proteomics addresses the relative abundance of the protein product, post-translational modifications, subcellular localization, turnover, interaction with other proteins, and functional aspects, all of which cannot be addressed by RNA analysis. In terms of levels of expression of genes, a poor correlation of less than 0.5 has been determined between mRNA and protein levels. This is due to differences in the rates of degradation of individual mRNAs and proteins, and because many proteins are modified after they have been translated. Also, one mRNA transcript can give rise to more than one protein. It has been estimated that there may be as many as 24% more proteins than genes in the organism Mycoplasma genitalium [54]. In humans there could be at least three times more proteins than genes. Post-translational modification of proteins is highly important for biological processes and the propagation of cellular signaling pathways. Of these modifications, phosphorylation is known to play a critical role. Verification of a gene product by proteomic methods is an important step in genomic annotation providing key information about true levels of expression, post-translational modifications, protein-protein interactions, and intracellular localization of gene products. These unbiased approaches may identify multiple pools of the protein of interest with subsequent differing interacting partners rather than only those proteins that interact in the nucleus (general two-hybrid systems) that depend on transcriptional read-out. In prostate cancer research, to date large-scale proteomics approaches have only been applied for the discovery of new biomarkers or to characterize changes in the proteome of prostate cells in response to androgens.

2D-PAGE

An established approach for differential protein-protein interactions involves combining co-immunoprecipitation with two dimensional electrophoresis (2D PAGE). 2D-PAGE encompasses separation or proteins, in-gel digestion of excised spots followed by mass spectrometric analysis of the peptide fragment patterns to provide a rapid method for protein identification. Additional information regarding post-translational modification status can also be obtained using tandem mass spectrometry (MS) with different levels of sensitivity, throughput and protein sequence coverage depending on instrument specifications and configuration. This approach provides greater resolution and separation of isoforms and proteins differing in posttranslational modifications that may not be resolved into discrete protein bands when using 1D gels. Weaknesses of 2D PAGE technology include the relatively large amount of protein required for detection of resolved spots as well as the difficulties of resolving very basic and very hydrophobic molecules. Proteins with very high molecular charges or very low molecular mass also will not separate on gels. However, improvements have been made in protein separation, lysis solutions, and detection methods for 2D PAGE techniques [55]. To date, there are no reports of application of 2D-PAGE followed by MS analysis of protein-protein interactions with the AR. The proteomes of human prostate cancer specimens and cells maintained in culture in response to androgens has been investigated using 2D gels followed by MS analysis [56–60]

Isotope Coded Affinity Tag (ICAT) and isotope Tags for Relative and Absolute Quantification (iTRAQ)^TM

ICAT technology is a powerfull approach that allows quantitative and identification of protein differences between different cellular states. ICAT reagents consist of three functional elements which include a thiol-specific reactive group, an ethylene glycol linker group that occurs in a “heavy” or “light” state and a biotin tag [61]. The cysteine side chains in complex mixtures of proteins from two different cell states are labelled using the heavy (deuterated or C13) ICAT reagent for one cell state and the light (non-deuterated or C12) ICAT reagent for proteins in the second cell state. The two labeled mixtures are then combined, proteolytically digested and run on an avidin column to pull out only the labeled peptides via the biotin tag. This reduces the complexity of the sample prior to analysis by a combination of nanoscale liquid chromatography (LC)-MS/MS [61]. ICAT has the power to quantitatively identify proteins including acidic or basic proteins, membrane proteins, low copy number proteins, and high molecular weight proteins [62]. Weaknesses of an ICAT MS approach include the following. Firstly, proteins that do not contain cysteine residues, or proteins that do not contain cysteine residues on tractable peptides upon proteolytic digestion, will not be detected due to the nature of the labeling procedure [63]. The cysteine-based ICAT tags will generally not yield information on changes in the proteome based on post-translational modifications such as phosphorylation. Secondly, avidin columns used for affinity separation of labeled peptides may also present difficulties associated with low capacity, non-specific binding and irreversible binding [63]. Thirdly, high sample complexity and the data-acquisition rate of the mass spectrometer used may limit coverage of differentially expressed proteins [63]. However, advances in sample fractionation at both the protein and peptide level, improved data acquisition schemes and optimized ICAT tags are currently being implicated to further overcome some of the difficulties associated with ICAT [64]. Co-immunoprecipitation of proteins followed by analysis using ICAT LC-MS/MS technology has successfully identified differences in: the composition of subunits of 20S proteosomes [65]; proteins interacting with transcription factors and large polymerase II preinitiation complex [66, 67]; and the mSin3 corepressor complex [68]. iTRAQ is similar in concept to ICAT. The power of iTRAQ is that it allows 4 samples to be analysed at a time and is thus more cost-effective. Depending upon how the analysis is set-up, this can allow duplicates in a single analysis. After combining the iTRAQ Reagent-labeled sample peptide digests, high resolution LC fractionation is performed to separate the peptides for LC-MS/MS analysis and ultimate protein identification and quantitation. The peptides are labeled on lysine residues and N-terminus with cleavable iTRAQ reagent (multiplex isobaric tags) to produce MS/MS signature ions with the relative peak area corresponding to the proportion of the labeled peptides. The labeling strategy means that most proteins are labeled and some peptides may contain multiple labels. Thus this approach differs from ICAT by not reducing the complexity of the sample through labeling strategies that enable MS analysis of the labeled peptides only. The peptides are automatically identified and quantified using Pro QUANT Software (Applied Biosystems, USA). Caveats using iTRAC for quantifying changes in protein-protein interactions are the same as for ICAT and involve the label reaction occurring after the co-immunoprecipitation and cost of label. So far there are no reports using iTRAC and only few reports that have applied ICAT in the prostate cancer field [62, 69–73]. However none of these reports have applied ICAT or iTRAQ approaches to examine protein-protein interactions in prostate cancer cells.

Stable Isotope Labeling with Amino acids in Cell culture (SILAC)

An alternative approach to ICAT and iTRACTM is SILAC. SILAC incorporates a tag in vivo into proteins for relative quantitation by mass spectrometry. SILAC has been successfully used to quantitate relative protein abundance [74], protein-protein interactions [75], and phosphorylation [76, 77]. Two cell populations are grown in identical culture conditions with the exception that one culture medium contains a “heavy” form while the other contains a “light” form of a particular amino acid, such as ¹³C- and ¹²C-labeled lysine, respectively. After 5 passages of the cells, the isotope label is 100% incorporated into newly synthesized protein to replace the naturally occurring amino acid [78]. Replacement of the naturally occurring amino acids with the isotope-containing amino acids does not alter the “normal” behaviour of the protein [78]. Arginine and lysine are the preferred heavy isotopes because trypsin cleaves after these residues thereby allowing every peptide ending with an arginine or lysine to be quantitated and compared to the “light state”. Labeling with both arginine and lysine isotopes provides greater coverage of the protein to increase confidence of identification. This powerful approach has been successfully applied to identify protein-protein interactions with EGF, GLUT4 and yeast 26 S proteasome complex [75, 79, 80] but to date has not been applied with AR or any of the steroid receptors.

Tandem Affinity Purification (TAP) of tagged proteins (TAP-tag) and Multidimensional Protein Identification Technology (MudPIT)

The TAP-tag approach for purification of native protein complexes [81] followed by MS analysis provides a powerful approach for identifying components of protein complexes. A detailed protocol can be found at www-db.embl-heidelberg.de. This approach involves creation of fusion proteins of the target protein that will subsequently be immunoprecipitated for analysis. The expression vector for the tagged protein can be transfected into cells and subsequently the fusion protein and associated proteins are recovered from the cell lysates using an antibody to the tag. The protein complexes can then be run on SDS-PAGE to resolve the proteins for subsequent MS analysis and identification. Alternatively the immunocomplexes can be enzymatically cleaved (trypsin), followed by multidimensional LC-MS/MS analysis. The later approach is termed TAP-MudPIT [82] and has been used to identify 117 proteins interacting with 14-3-3 sigma in vivo in human embryonic kidney 293T cells [83]. TAP-tag has been used predominantly in yeast and has not been previously reported as a strategy for identifying protein-protein interactions with steroid receptors. We have recently employed MudPIT [84] to identify novel protein-protein interactions of endogenous complexes of the AR in human LNCaP prostate cancer cells treated with synthetic androgen (R1881) (Fig 2A). Using two washes of differing stringencies, we identified 82 peptides in these experiments including peptides of the AR. Proteins known to interact with the AR were detected such as BRCA1 [85] and SRCAP [86]. Of these candidates, 24 proteins were considered for further validation based upon low expression levels as determined from SAGE data, predicted nuclear localization, and known or predicted function. The novel proteins identified have various predicted functions including activators and repressors of transcription, ATPase, DNA repair, chromatin, signal transduction pathways, RNA binding, and unknown function (Fig 2B). Most proteins contained one or more nuclear receptor box (NR box contianing an LXXLL motif where L = leucine and X = any amino acid) found in proteins that interact with nuclear receptors [87]. These results indicate that power of this unbiased approach to isolate and identify proteins interacting with different pools of the AR. However, further experiments are required to validate the interactions using any of the above assays. In addition, it will be essential to show biological effect of the interaction thereby necessitating additional approaches.

Figure 2. Identification of proteins that interact with the AR using MudPIT.

A, Schemata showing the experimental approach to identify proteins in endogenous complexes immunoprecipitated with the AR from prostate cancer cells treated with androgen. The protein complex immunoprecipitated using an antibody to the AR are digested with trypsin. The resulting peptides are separated using multidimensional liquid chromatography (MDLC) and analyzed by MS/MS. Spectra are correlated to proteins using MASCOT database searches. From two separate experiments, a total of 82 peptides were identified. B, The predicted function of 24 proteins immunoprecipiated with the AR.

Conclusions

To date there are no effective therapies available once prostate cancer becomes androgen independent. Generally the patient will succumb to his disease within 2 years after initial rises in serum PSA levels (biochemical failure) that signify the emergence of androgen independent disease. The AR has been suggested to play a predominant role in this terminal stage. Current strategies for developing new drugs for the treatment of advanced prostate cancer are aimed at blocking AR activity through improved antiandrogens or by targeting proteins that interact with the AR to inhibit its transcriptional activity. Historically the experimental approaches used to identify interacting proteins have involved two-hybrid systems, immunoprecipitation, and pull-down assays. Each of these approaches has strengths and weaknesses that necessitate validation experiments to confirm true interactions. Surprisingly, no large scale approaches have been applied previously to identify interacting proteins with any of the steroid receptors including the AR. Immunoprecipitation of the endogenous AR and interacting proteins followed by MudPIT identified novel proteins not previously reported to interact with the AR. Further work will be required to determine if any of these proteins would serve as a good therapeutic target for advanced prostate cancer. Thus large scale proteomic approaches provide the capability for unbiased identification of novel protein-protein interactions.