Gain in Transcriptional Activity by Primate-specific Coevolution of Melanoma Antigen-A11 and Its Interaction Site in Androgen Receptor*

Male sex development and growth occur in response to high affinity androgen binding to the androgen receptor (AR). In contrast to complete amino acid sequence conservation in the AR DNA and ligand binding domains among mammals, a primate-specific difference in the AR NH2-terminal region that regulates the NH2- and carboxyl-terminal (N/C) interaction enables direct binding to melanoma antigen-A11 (MAGE-11), an AR coregulator that is also primate-specific. Human, mouse, and rat AR share the same NH2-terminal 23FQNLF27 sequence that mediates the androgen-dependent N/C interaction. However, the mouse and rat AR FXXLF motif is flanked by Ala33 that evolved to Val33 in primates. Human AR Val33 was required to interact directly with MAGE-11 and for the inhibitory effect of the AR N/C interaction on activation function 2 that was relieved by MAGE-11. The functional importance of MAGE-11 was indicated by decreased human AR regulation of an androgen-dependent endogenous gene using lentivirus short hairpin RNAs and by the greater transcriptional strength of human compared with mouse AR. MAGE-11 increased progesterone and glucocorticoid receptor activity independently of binding an FXXLF motif by interacting with p300 and p160 coactivators. We conclude that the coevolution of the AR NH2-terminal sequence and MAGE-11 expression among primates provides increased regulatory control over activation domain dominance. Primate-specific expression of MAGE-11 results in greater steroid receptor transcriptional activity through direct interactions with the human AR FXXLF motif region and indirectly through steroid receptor-associated p300 and p160 coactivators.

is a ligand-dependent transcription factor activated by binding testosterone, a major circulating male steroid hormone, or by dihydrotestosterone (DHT), the more potent 5␣-reduced metabolite of testosterone. Evolution of AR among mammals is characterized by complete amino acid sequence conservation in the central DNA binding domain that interacts with androgen response element DNA and in the carboxyl-terminal ligand binding domain that binds androgens with high affinity and specificity (1). Strict sequence conservation in these regions reflects the rigid structural requirements for DNA and hormone binding. In contrast, the AR NH 2 -terminal region, although also required for AR transcriptional activity, is largely unstructured and less well conserved (2). The human AR NH 2terminal region contains activation function 1 (AF1) between amino acid residues 142-337 (3) preceded by a polymorphic CAG-encoded glutamine repeat that expanded during primate evolution (1). Mechanisms involved in AF1 and glutamine repeat function remain to be established. Expansion of the glutamine repeat length to more than 39 residues results in the adult onset muscle wasting disease known as spinal bulbar muscular atrophy (4). A similar glutamine repeat in rat and mouse AR is shifted in position in the NH 2 -terminal region relative to human AR (5).
The mammalian AR NH 2 -terminal region also contains a conserved 23 FQNLF 27 sequence that binds activation function 2 (AF2) in the AR ligand binding domain to mediate the androgen-dependent AR NH 2 -and carboxyl-terminal (N/C) interaction (6,7). An NH 2 -terminal 433 WHTLF 437 WXXLF motif contributes to the human AR N/C interaction and transcriptional activity (6 -8). The functional importance of the intermolecular AR N/C interaction is indicated by a dependence on high affinity androgen binding, its requirement for optimal gene transcription, and inhibition of the N/C interaction by classical AR antagonists (7, 9 -12). There are also naturally occurring human AR single amino acid mutations that cause partial androgen insensitivity and disrupt the AR N/C interaction even though high affinity androgen binding is maintained (13)(14)(15)(16)(17)(18)(19)(20).
The androgen-dependent human AR N/C interaction reduces p160 coactivator binding to AF2 in the ligand binding domain (21), which suggests regulation of activation domain dominance. p160 coactivator binding to AF2 is increased by melanoma antigen-A11 (MAGE-11), an AR coregulator that binds the human AR FXXLF motif region in a manner competitive with the N/C interaction to expose AF2 for increased p160 coactivator recruitment (22). MAGE-11 also interacts with p300 and p160 coactivators (23,24).
Studies in this report address the functional consequences of amino acid sequence differences in the AR NH 2 -terminal region among primates and lower mammals that parallel the evolution of MAGE-11 expression. Our findings suggest that the primate-specific expression of MAGE-11 increases AR transcriptional strength through direct binding to the human AR FXXLF motif region. We provide evidence that MAGE-11 also functions as a more general transcriptional coregulator through interactions with steroid receptor-associated p300 and p160 coactivators.
DNA Transfection and Gene Expression Studies-Mammalian two-hybrid assays were performed in HeLa cells (5 ϫ 10 4 / well) (26) in 12-well plates transfected using FuGENE 6 (Roche Applied Science), expression vector DNA, and 0.1 g/well 5XGAL4Luc reporter gene. The day after transfection, cells were transferred to serum-free medium and incubated for 24 h in the absence and presence of DHT. AR transcription assays in HeLa cells utilized wild-type and mutant pCMV-hAR and pCMV-mAR and 0.1 or 0.25 g/well PSA-Enh-Luc, BH500-Luc that contains the Ϫ426 to ϩ28 rat probasin promoter provided by Robert J. Matusik (30,31), or MMTV-Luc. Cells were harvested in 0.25 ml of lysis buffer containing 1% Triton X-100, 2 mM EDTA, and 25 mM Tris phosphate, pH 7.8. Transcription assays in CV1 cells (4 ϫ 10 5 /6-cm dish) were performed using calcium phosphate precipitation (21). Immediately after transfection and 24 h later, cells were incubated for 24 h in serumfree phenol red-free medium in the absence and presence of DHT and harvested in 0.5 ml of lysis buffer. Luciferase activity was measured using an automated Lumistar Galaxy luminometer (BMG Labtech), and values (mean Ϯ S.E.) are representative of at least three independent experiments. The siRNA oligonucleotide experiments were performed in HeLa cells (2 ϫ 10 5 /well) in 6-well plates and in COS cells (2 ϫ 10 6 cells/10-cm dish) using p300 siRNAs, nonspecific siRNA-3 (Dharmacon RNA Technologies), and Lipofectamine 2000 (Invitrogen) in the absence of antibiotics (24).
Genome Analysis of MAGE Families and Two-hybrid Screens-The human MAGE-11 amino acid sequence was compared with the NCBI BLAST Assembled RefSeq Database for human, Rhesus macaque, mouse and rat genomes. We obtained no experimental evidence for human MAGE-11 isoform B indicated by the human genome database.
The Clontech Mate and Plate yeast two-hybrid screen was performed by preparing mouse testis complementary DNA (cDNA) pool using SMART cDNA synthesis technology, recombination with Matchmaker prey vector pGADT7-Rec, and transformation in yeast strain Y187. Two independent screens of 2 ϫ 10 8 colonies of a mouse testis library were performed using GAL-mAR-(4 -52) or GAL-mAR-(2-285) as bait in yeast strain Y2HGold. Selection parameters included resistance to aureobasidin A, X-␣-Gal blue colony formation, and growth in the absence of histidine or adenine. Screening using GAL-mAR-(4 -52) as bait yielded 11 positive independent clones compared with 17 positive independent clones using GAL-mAR-(2-285) as bait. None of the identified protein interactions with mouse AR were FXXLF motif-dependent.
HeLa-AR-PSA-Luc-A6 cell growth assays were performed by plating 5000 cells/well of 24-well plates in 0.5 ml of medium containing 10% fetal bovine serum. Cells were incubated for increasing times and assayed using the cell counting kit (Dojindo Laboratories). WST-8 reagent (0.02 ml) was added to 0.2 ml of serum-free medium and incubated for 2.5 h at 37°C. Optical density measurements were determined at 485 nm.

Species-specific AR FXXLF Motif Interactions-
The AR NH 2terminal FXXLF motif sequence 23 FQNLF 27 is conserved among mammals with the hydrophobic character maintained throughout lower vertebrates including fish ( Fig. 1A) (7,35). However, the sequence flanking the AR FXXLF motif differs with Ala 33 in mouse, rat, and other lower vertebrates evolved to Val 33 in human, chimpanzee, Rhesus macaque, lemur, and other primates (Fig. 1, A and B). Human AR Val 33 is in a predicted ␣-helical region that extends from the FXXLF motif that interacts with AF2 in the ligand binding domain bound to a high affinity androgen. AR FXXLF motif binding to AF2 competes with AR FXXLF motif binding to MAGE-11, a human AR coregulator (22,36).
Species-specific differences in the AR N/C interaction and AR FXXLF motif binding to MAGE-11 were investigated for human, rat, and mouse AR in mammalian two-hybrid assays. The androgen-dependent N/C interaction between the AR FXXLF motif and AF2 in the ligand binding domain (6) was assayed in the absence and presence of DHT using GAL-hAR-(658 -919) in which the GAL4 DNA binding domain was expressed as a fusion protein with the AR ligand binding domain conserved in human, mouse, and rat (5).
When assayed using VP-hAR-(1-660) that contains human AR NH 2 -terminal and DNA binding domains, the V33A mutation decreased the human AR N/C interaction and eliminated the interaction with MAGE-11 (Fig. 2D). Conversely, VP-mAR- Dependence of the AR N/C interaction on Val 33 was investigated further by expressing corresponding hAR-(1-503) and FIGURE 1. Vertebrate evolution of AR FXXLF motif region and model of AR, MAGE-11, TIF2, and p300 interactions. A, human AR amino acid residues 16 -34 (Homo sapiens; GenBank accession no. P10275) are compared with corresponding AR regions from chimpanzee (Pan troglodytes; GenBank accession no. NP001009012), crab-eating macaque (Macaca fascicularis; GenBank accession no. AAC73050), collared brown lemur (Eulemur fulvus collaris; GenBank accession no. U94178), house mouse (Mus musculus; GenBank accession no. NP038504.1), Norway rat (Rattus norvegicus; GenBank accession no. NP036634), African clawed frog (Xenopus laevis; GenBank accession no. NP001084353), goldfish (Carassius auratus; GenBank accession no. AAM09278), Japanese eel (Anguilla japonica; GenBank accession no. BAA75464; AR-␣), and rainbow trout (Oncorhynchus mykiss; GenBank accession no. NP001117656; AR-␣). Highlighted in red are FXXLF motif residues required for the AR N/C interaction or interaction with MAGE-11. B, human AR NH 2 -terminal FXXLF motif residues 23-33 (hydrophobic residues underlined) precede AF1. High affinity androgen binding in the human AR ligand binding domain (LBD) initiates FXXLF motif binding to AF2 in the N/C interaction and restricts p160 coactivator LXXLL motif binding to AF2. Val 33 flanking the AR FXXLF motif in humans and other primates was required for the N/C interaction inhibition of AF2 and a direct interaction with MAGE-11. AR in lower mammals such as rat and mouse has Ala 33 instead of Val 33 . Ala 33 was not compatible with a direct interaction with MAGE-11. In agreement with MAGE family genome sequence analysis that MAGE-11 is primate-specific, MAGE-11 or another FXXLF motif-interacting protein was not identified in a mouse testis expression library using the mouse AR FXXLF motif region as bait. A direct interaction between human AR and MAGE-11 provides a primate-specific gain in function. MAGE-11 increases activity of other steroid hormone receptors in an FXXLF motif-independent manner through interactions with receptor-associated p300 and p160 coactivators. DBD, DNA binding domain.
The results suggest that the evolutionary change from alanine to valine at AR residue 33 in the primate lineage strengthens the androgen-dependent AR N/C interaction in primates. The same transition mutation to Val 33 also enables human AR to interact directly with MAGE-11.
Greater Transcriptional Activity of Human AR-The ability of MAGE-11 to increase human AR transcriptional activity (22)(23)(24)27) together with the inability of mouse AR to interact with MAGE-11 (Fig. 2) and the apparent absence of MAGE-11 in mouse (see Fig. 6) suggested that MAGE-11 may contribute to species-specific differences in AR transcriptional strength. Transcriptional activities of human and mouse AR were compared in HeLa cells that express endogenous MAGE-11 (Fig.  3A, top panel, lane 1) and in CV1 cells with low levels of MAGE-11 mRNA (37) and no detectable MAGE-11 protein (Fig. 3A, top panel, lane 2). Transcriptional activity was determined using the human PSA and rat probasin enhancer-luciferase reporter genes that depend on the AR N/C interaction for maximal activation by human AR (7). In agreement with an androgen-dependent N/C interaction for human and mouse AR (Fig. 2), both receptors were stabilized in the presence of 10 nM DHT (Fig. 3A, lower panel) (21,38).
Androgen-dependent human AR transactivation of PSA-Enh-Luc (Fig. 3B) and probasin BH500-Luc (Fig. 3C) exceeded that of mouse AR in HeLa cells in the presence of 0.1 nM DHT. The activity of human and mouse AR F26A,F27A FXXLF motif mutants decreased to the greatest extent at the lower concentrations of DHT. It was noteworthy that the transcriptional strength of hAR-L26A,F27A in which the FXXLF motif binding site for MAGE-11 was disrupted was similar to that of wild-type mouse AR (Fig. 3, B and C). In CV1 cells that essentially lack MAGE-11 (Fig. 3A), human and mouse AR transactivation of PSA was similar and FXXLF motif-dependent (Fig. 3D).
The results suggest that maximal human and mouse AR transactivation requires the AR N/C interaction. The greater transcriptional strength of human AR relative to mouse in cells expressing MAGE-11 was consistent with the coactivator effects of MAGE-11 interaction with the human AR FXXLF motif region.
The requirement for MAGE-11 in androgen-dependent human AR transactivation was investigated by quantitative RT-PCR of endogenous PSA in LAPC-4 prostate cancer cells that have MAGE-11 mRNA levels ϳ10-fold greater than HeLa cells, ϳ100-fold greater than Ishikawa human endometrial cancer cells, and ϳ1000-fold greater than normal human prostate cells (37). As seen with the HeLa-AR cells (Fig. 4), lentivirus AR shRNA-5 and MAGE-11 shRNA-947 were most effective in silencing AR and MAGE-11 expression in LAPC-4 cells, respectively, based on immunoblot (Fig. 5A, lanes 4 and 6) and RT-PCR analyses of MAGE-11 mRNA (Fig. 5B). Human AR shRNA-5 and MAGE-11 shRNA-947 decreased androgen-dependent human AR transactivation of PSA to an extent not seen with the partial silencing of MAGE-11 (Fig. 5C). Specificity of inhibition was suggested by androgen-dependent up-regulation of PSA without lentivirus or with control lentivirus that did not alter AR or MAGE-11 levels (Fig. 5, A-C). MAGE-11  shRNA-947 did not alter AR protein levels (Fig. 5A, lane 4), but knockdown of AR decreased MAGE-11 mRNA (Fig. 5B). In agreement with the low levels of MAGE-11 in most cells and the requirement for maximal silencing of MAGE-11 to inhibit AR activity, the results suggest that MAGE-11 is a low abundance but essential human AR coregulator.
Absence of MAGE-11 Homologue in Mice-To determine whether less evolved mammals express MAGE-11 or a related protein that interacts with the AR FXXLF motif, genome-wide sequence comparisons (Fig. 6) and two-hybrid screens were performed. Comparison of human, Rhesus macaque (Macaca mulatta), mouse, and rat MAGE gene families suggests divergent evolution between the lower mammals and primates (Fig.  6, A-D). The 429-amino acid human and Rhesus macaque MAGE-11 proteins represented by the top black bar (Fig. 6, A  and B) were absent in the mouse and rat genomes (Fig. 6, C and  D). Human and Rhesus macaque MAGE-11 have 93% amino acid sequence homology with shared regions important for coregulator function. Conserved functional domains include nuclear localization residues 18 -23, Ser 174 mitogen-activated protein kinase phosphorylation site, 185 MXXIF 189 interaction motif for p300, Lys 240 and Lys 245 monoubiquitinylation sites required to bind the human AR FXXLF motif, 260 FPEIF 264 FXXIF interaction motif for TIF2, and Thr 360 Chk1 cell cycle checkpoint kinase phosphorylation site in the F-box-like 329 -369 residues that bind the human AR FXXLF motif (23, 24, 27) (Fig. 6E, highlighted in red). The 7% sequence divergence between human and Rhesus macaque MAGE-11 outside these functional domains is consistent with continuing evolution of the MAGE gene family in primates.
With MAGE-11 absent in the mouse and rat genomes, there remained the possibility of a related AR FXXLF motif-interacting protein in mouse. This idea gained some support from the extensive sequence conservation in the MAGE homology domain region that interacts with the human AR FXXLF motif. Yeast two-hybrid screens were performed using a mouse testis library. Mouse GAL-mAR-(4 -52) bait vector contained Ala 33 instead of Val 33 in GAL-hAR-(4 -52) used to identify MAGE-11 in a human testis library (22). A second mouse testis library screen was performed using GAL-mAR-(2-285) that contained additional mouse AR NH 2 -terminal sequence.
None of the 28 positive mouse AR-interacting proteins were FXXLF motif-dependent or recapitulated the transcription enhancing effects of MAGE-11 seen with human AR. Results from MAGE gene family comparisons and two-hybrid screens provide evidence that the coregulator function of MAGE-11 is primate-specific.
Transcriptional Synergy between MAGE-11 and Receptor-associated p300 and p160 Coactivators-MAGE-11 binds TIF2, a coactivator that interacts with AF2 in the AR ligand binding domain, and p300 (23,24), an acetyltransferase that interacts with the NH 2 -terminal region of AR and other steroid receptors (39) (Fig. 1B). This raised the possibility that MAGE-11 has transcriptional effects independent of binding an FXXLF motif by interacting with receptor-associated coactivators. Initial support for this came from the coimmunoprecipitation of FLAG-MAGE in a complex with human and mouse AR (Fig.  7A), human GR (Fig. 7B), and human PR-B (Fig. 7C) expressed in cells treated with receptor agonists and EGF. The experimental conditions were similar to the optimal interaction conditions for nuclear and cytoplasmic MAGE-11 interaction with liganded and unliganded human AR, respectively (22,27). MAGE-11 also increased the transcriptional activity of mouse AR and the L26A,F27A FXXLF motif mutant and was most effective with mAR-A33V (Fig. 8A). The hAR-L26A,F27A FXXLF motif mutation that disrupts the N/C interaction decreased activity, whereas hAR-V33A with an N/C interaction retained activity that increased with MAGE-11. MAGE-11 functioned synergistically with TIF2 and mouse AR (Fig. 8B) and increased human PR-B and GR activity with TIF2 and p300 (Fig. 8, C and D).
Human and Mouse AR AF2 Activity-The inhibitory effect of the human AR N/C interaction on AF2 activity can be relieved either by mutating the AR FXXLF motif to FXXAA or by MAGE-11 binding to the human AR FXXLF motif (21,23). The inability of mouse AR to interact directly with MAGE-11 raised the possibility that mouse AR AF2 activity may be inhibited to a greater extent than human AR by the N/C interaction. On the other hand, a weaker mouse AR N/C interaction (Figs. 2E and 8A) may not inhibit AF2 activity. To test this further, the effect of the AR N/C interaction on human and mouse AR AF2 activation by TIF2 was investigated using hAR⌬120 -472, an AF1 deletion mutant (Fig. 9A) that depends on AF2 for transcriptional activity (23), and the corresponding mAR⌬101-452 mouse mutant. In agreement with previous studies (23), transactivation of MMTV-Luc by hAR⌬120 -472 required the expression of both TIF2 and MAGE-11 (Fig. 10A). The inhibitory effect of the human AR N/C interaction on AF2 activity that was overcome by MAGE-11 was also indicated by TIF2 activation of the hAR⌬120 -472-L26A,F27A FXXLF motif mutant in the absence of MAGE-11. Dependence on TIF2 LXXLL motif binding to AF2 was shown by loss of transcriptional activity by the K720A charge clamp mutant. However, hAR⌬120 -472-V33A in which Val 33 was changed to Ala 33 to mimic mouse AR was inactive with and without TIF2 or MAGE-11. This suggests that the human AR N/C interaction inhibition of AF2 activity in hAR⌬120 -472-V33A could not be rescued by MAGE-11 consistent with the requirement for AR Val 33 to interact directly with MAGE-11.
In contrast, the corresponding mouse mAR⌬101-452 and L26A,F27A FXXLF motif mutants were strongly activated by TIF2 with or without the expression of MAGE-11 (Fig. 10B). When Ala 33 was changed to Val 33 in mAR⌬101-452-A33V to mimic human AR, transcriptional activity required TIF2 and MAGE-11. Partial retention of activity by the mAR⌬101-452-K700A charge clamp mutant suggested that mouse AR may be activated to some extent by TIF2 outside the mouse AR AF2 site. However, similar to human AR, transcriptional activity was lost by combining mutations in mAR⌬101-452-A33V,K700A.
We noted that hAR⌬120 -472 migrated as a double band on immunoblots, whereas mAR⌬101-452 was a single band (Fig.  10, A and B, top insets). In agreement with previous studies showing that human AR is phosphorylated on Ser 94 (41,42), the hAR⌬120 -472-S94A mutant migrated as a single band similar to wild-type mouse AR. However, the inhibitory effect of the human AR N/C interaction on AF2 activity not seen with mouse AR could not be attributed to differences in Ser 94 phosphorylation. Human AR and mouse AR share the same NH 2terminal EDGSPQAH sequence that contains human AR Ser 94 ,Pro 95 (underlined), which is Ser 74 ,Pro 75 in mouse AR. Human AR Ser 94 is shifted in position relative to mouse AR by glutamine repeat residues 58 -78 (5). In addition, hAR⌬120 -472-S94A required MAGE-11 for AF2 activation by TIF2 similar to hAR⌬120 -472 (Fig. 10A). When the glutamine repeat length of hAR⌬120 -472 was reduced from 24 to 5 residues to mimic mouse AR, the double band migration was lost (Fig. 10A, bottom inset), but TIF2 and MAGE-11 were required for transactivation (data not shown).
The results suggest fundamental differences in the regulation of human and mouse AR transcriptional activity. Val 33 in  human AR resulted in a stronger N/C interaction and inhibition of AF2 that was released by the competitive binding of MAGE-11. Ala 33 in mouse AR caused a weaker AR N/C interaction that did not require MAGE-11 for AF2 activity in an animal model that does not express MAGE-11.

DISUCSSION
Convergent Evolution of AR NH 2 -terminal Sequence and MAGE-11-Our findings suggest species-specific differences in AR transactivation mediated by the effects of the AR N/C interaction and MAGE-11. The androgen-dependent AR N/C  interaction between the NH 2 -terminal FXXLF motif and AF2 hydrophobic binding surface of the ligand binding domain appears to occur in all mammals although to a greater extent in primates due to the evolutionary transition mutation from AR Ala 33 to Val 33 . Evidence for the androgen-dependent AR N/C interaction among all mammals was supported by androgendependent stabilization of human and mouse AR, an interaction that slows the dissociation rate of bound androgen from human AR (6,10). Previous studies confirmed a rat AR N/C interaction that increased activity in association with transcriptional effects of p160 coactivators (43). However, the evolutionary change from alanine to valine at residue 33 flanking the 23 FQNLF 27 motif in a predicted extended ␣-helical region not only strengthens the N/C interaction in primates but enables human AR to interact directly with MAGE-11, a coregulator that coevolved in the primate lineage. Studies using AR AF1 deletion mutants that depend on AF2 for activity showed that human AR Val 33 contributes to the inhibitory effect of the human AR N/C interaction on AF2 activation by TIF2 that is relieved by MAGE-11. Convergent evolution of the AR NH 2 -terminal FXXLF motif flanking sequence with the expression of MAGE-11 in primates together with the ability of MAGE-11 to increase human AR transcriptional activity and the dual functions of the human AR FXXLF motif compared with mouse AR suggests that human and nonhuman primates have acquired increased regulatory control over AR activation domains in association with a primate-specific gain in function (Fig. 11). Our previously proposed model for an evolutionary transition from AF2 to AF1 among different steroid hormone receptors (36) appears to apply also to the evolution of mammalian AR where primate AR acquired the ability to regulate AF2 activity. The functional importance of MAGE-11 in human AR transactivation of androgen-responsive genes was supported by the effects of lentivirus shRNA silencing MAGE-11. The results suggest that primates evolved increased regulatory control over AR activation domain usage to increase transcriptional strength.
Species-specific AR Function-Complete amino acid sequence conservation in the AR DNA and ligand binding domains throughout mammalian evolution preserves the structural constraints necessary for high affinity androgen and DNA binding. In contrast is the unstructured AR NH 2 -terminal region, which is also important for AR function but less well conserved. Human and mouse AR NH 2 -terminal regions share ϳ84% amino acid sequence homology with the most highly conserved regions at human AR amino acid residues 1-53, 234 -247, and 360 -429 (1,35). Human AR residues 1-53 contain the FXXLF motif that mediates the N/C interaction (6,10) and serves as the binding site for MAGE-11 (22). This region precedes the polymorphic glutamine repeat that expanded during primate evolution to an average length of 20 -22 residues (1). Human AR glutamine repeat residues 58 -78 are shifted in mouse AR to residues 174 -193 (1,5,25). Expansion to more than 39-glutamine repeat length in human AR causes spinal bulbar muscular atrophy or Kennedy disease, an adult onset muscle wasting disease associated with an AR gain in function that requires the N/C interaction (4,44). Human AR amino acid residues 234 -247 are within the binding region for CHIP (carboxyl terminus of the Hsp70-interacting protein), an E3 ubiquitin ligase that targets AR for degradation (35). Human AR residues 360 -429 were implicated in transactivation (45).
Amino acid sequence conservation in the AR DNA binding domain throughout evolution suggests that interaction sites in chromatin may be shared across mammalian species. However, recent studies have shown surprisingly low sequence conservation for the majority of mammalian AR DNA binding sites associated with androgen-regulated genes (46). This suggests that species-specific differences in AR gene regulation may be attributed to differences in coregulatory proteins such as MAGE-11.
Genome-wide sequence analysis and two-hybrid screens showed that the AR coregulator function of MAGE-11 is primate-specific. This is consistent with differences in the rat and mouse AR FXXLF motif flanking sequence where Ala 33 in mouse AR is not compatible with direct binding to MAGE-11. However, absence of an FXXLF motif that binds MAGE-11 in mouse AR or other steroid receptors did not rule out transcriptional effects. MAGE-11 coexpressed with mouse AR or human PR-B or GR was present in a complex presumably through its interaction with p300 and p160 coactivators. In agreement with this, MAGE-11 increased steroid receptor transcriptional activity also by an FXXLF motif-independent mechanism. Evidence that MAGE-11 can increase steroid receptor activity through an association with p300 and p160 coactivators broad- FIGURE 11. Model of MAGE-11 interaction with steroid receptors. MAGE-11 F-box residues interact with the extended human AR FXXLFXXVXXV motif sequence 23 FQNLFQSVREV 33 that ends with Val 33 . Both hAR and mAR undergo the androgen-dependent N/C interaction. However, human AR Val 33 that is Ala 33 in mouse AR is required for a direct interaction with MAGE-11 and for the AR N/C interaction inhibition of AF2 activity that is relieved by MAGE-11. MAGE-11 MXXIF motif interaction with p300 (24) enhances transcriptional activity from the NH 2 -teminal regions of human AR, PR-B, and GR. MAGE-11 FXXIF motif interaction with p160 coactivators (23) enhances AF2 activity from the human AR, PR-B, and GR carboxyl-terminal ligand binding domains (LBD). Although mouse AR is activated by p300 and p160 coactivators, MAGE-11 is not present in mouse, and the mouse AR NH 2terminal FXXLFXXVXXA motif sequence would not interact directly with MAGE-11. Human and nonhuman primate AR transcriptional activity is regulated by the primate-specific coevolution of MAGE-11 and the AR NH 2 -terminal FXXLFXXVXXV interaction site for MAGE-11. DBD, DNA binding domain. ens its potential impact in human reproductive physiology. MAGE-11 interaction with receptor-associated p300 and p160 coactivators may have been the evolutionary progenitor for the primate-specific direct interaction with human AR (Fig. 11).
Species-specific differences in AR activity have been reported previously. Rat AR activation of MMTV-chloramphenicol acetyltransferase reporter gene in CV1 cells was less than that of human AR with differences attributed to the AR NH 2terminal region (47). However, we did not detect MAGE-11 protein in CV1 cells, and mRNA levels were low (37). This suggests that additional mechanisms involving the NH 2 -terminal region increase human AR transcriptional strength. Attempts to mimic human AR in a humanized mouse model made use of a human AR NH 2 -terminal region 5Ј SmaI site fragment that begins at human AR amino acid residue 37 so that the chimera retained mouse Ala 33 (48). Even with the absence of MAGE-11 in mouse, the unregulated p160 coactivator access to AF2 that characterizes mouse AR would be expected to be retained in the presence of AR Ala 33 .
CpG dinucleotides are hot spots for mutation that account for approximately one-third of disease-causing germ line mutations (49,50) and a significant proportion of human AR mutations that cause the androgen insensitivity syndrome (51). CpG dinucleotides can become methylated, and 5-methylcytosine can undergo spontaneous deamination that results in a C3 T transition mutation in genomic DNA. Our findings suggest that mutations at CpG dinucleotides also contribute to the evolution of new gene sequences. Ala 33 in mouse and rat AR is coded by GCG, which in primates changed to GTG coding for Val 33 . Rat and mouse AR GCG codon 33 is followed by A in the sequence GCGA, and CGA was preferentially associated with mutations in the retinoblastoma gene (50). Methylation and deamination of CpG in AR codon 33 in the less evolved mammals may have caused the evolutionary transition from Ala 33 to Val 33 needed for the AR N/C interaction inhibition of AF2 activity and the binding to MAGE-11 that releases that inhibition in primates. MAGE-11 may have evolved to regulate AR transcriptional activity in part by relieving the inhibitory effects of the AR N/C interaction associated with the AR Ala 33 to Val 33 mutation.
Species-specific MAGE Gene Families-The rapid evolution of the MAGE gene family among mammals results from gene duplication and retrotransposition (52). The MAGE-11 gene is part of a MAGE-A subfamily expressed at Xq28 on the human X chromosome and contains three short unique 5Ј coding exons that precede a major, more highly conserved 3Ј coding exon (22,53). A two-hybrid interaction screen of a human testis library using the human AR FXXLF motif region as bait identified MAGE-11 as a human AR coregulator (22). A similar screen of a mouse testis library using the corresponding mouse AR FXXLF motif region as bait suggested the absence of a related AR FXXLF motif-interacting protein in mouse. This finding was consistent with genome-wide sequence analysis of MAGE gene families that no other mouse MAGE family protein corresponds to human MAGE-11. It was also consistent with the absence of an inhibitory effect of the mouse AR N/C interaction on AF2 activity. The results support the absence of MAGE-11 in mammals outside the primate lineage and a primate-specific AR gain in function that increases transcriptional strength of AR and other steroid hor-mone receptors. The absence of MAGE-11 in lower mammals suggests its coevolution with the AR FXXLF motif flanking sequence and increased regulatory control of AR activation domain dominance through evolution.
The ability of MAGE-11 to increase human AR transcriptional activity places it among a growing list of AR coactivators (54). However, the precise mechanisms that underlie its coregulator function remain to be established. MAGE-11 is a multifunctional protein that interacts with p300 and p160 coactivators (23,24) and undergoes post-translational modification that includes phosphorylation by mitogen-activated protein kinase and cell cycle checkpoint kinase Chk1 and monoubiquitinylation at lysine residues required to interact with the human AR FXXLF motif (23,27). A role for MAGE-11 in androgen-dependent cell cycle regulation is suggested by a number of MAGE family proteins that influence cell cycle progression and apoptosis in humans (55). A MAGE-A subfamily binding partner is a RING E3 ubiquitin ligase whose activity increases in association with MAGE proteins (56). Some MAGE-A proteins inhibit p53 tumor suppressor activity during tumor development (57). MAGE-11 was shown to be an interaction partner and inhibitor of the major hypoxia-inducible factor HIF-1␣-hydroxylating enzyme prolyl hydroxylase 2 and stabilizes HIF-1␣ in a potential tumor-associated regulatory mechanism (58). It remains to be established whether these functions are important for MAGE-11 activity as a steroid receptor coregulator.
The ability of MAGE-11 to increase transcriptional strength of human AR through a direct interaction with the AR FXXLF motif has important implications in normal human reproductive physiology and cancer. MAGE-11 mRNA levels increase more than 50-fold in the mid-secretory glandular epithelium of human endometrium during the window of receptivity for implantation, suggesting a role in human embryo implantation or survival (34). MAGE-11 mRNA levels are acutely up-regulated by cyclic AMP, a second messenger signaling response to the preovulatory luteinizing hormone surge. However, the physiological effects of AR and MAGE-11 in human endometrium during the window of implantation remain to be established. MAGE-11 is coexpressed with AR in human ovarian granulosa cells 3 where it may function in oocyte survival (34). MAGE-11 mRNA levels increase more than 100-fold in a subset of patients with castration-recurrent prostate cancer undergoing androgen deprivation therapy. The increase in MAGE-11 during prostate cancer growth results from progressive hypomethylation of a CpG island at the transcription start site (37). Increased levels of MAGE-11 that increase human AR signaling during prostate cancer progression may promote prostate cancer cell survival (59). In normal men, increased AR signaling by MAGE-11 may improve physical strength and survival, and in women, it may be important for reproductive success.