Regulation of ALK-1 Signaling by the Nuclear Receptor LXR (cid:1) *

The transforming growth factor (cid:1) (TGF- (cid:1) ) receptor, ALK-1, is expressed specifically on endothelial cells and is essential for angiogenesis, as demonstrated by its targeted deletion in mice and its mutation in the human disease hereditary hemorrhagic telangiectasia. Although ALK-1 and another endothelial-specific TGF- (cid:1) receptor, endoglin, both bind TGF- (cid:1) with identical isoform specificity and form a complex together, neither has been shown to signal in response to TGF- (cid:1) , and the mechanism by which these receptors signal in endothelial cells remains unknown. Here we report the identification of the nuclear receptor liver X receptor (cid:1) (LXR (cid:1) ) as a modulator/mediator of ALK-1 signaling. The cytoplasmic domain of ALK-1 specifically binds to LXR (cid:1) in vitro and in vivo . Expression of activated ALK-1 results in translocation of LXR (cid:1) from the nuclear compartment to the cytoplasmic compartment. The interaction of activated ALK-1 with LXR (cid:1) in the cytoplasmic compartment results in the specific phosphorylation of LXR (cid:1) by ALK-1, primarily on serine residues. LXR (cid:1) subsequently modulates signaling by ALK-1 and the closely related TGF- (cid:1) receptor, ALK-2, as 3–5 days to select for yeast both and prey vectors. replica-plated on His or (cid:1) to assay for interaction as assessed under these conditions. clones tested for ability to and prey dependence and specificity with other baits (lamin, myc, endoglin, ALK-2) to screen out false Clones these screens were sequenced and analyzed. Two-hybrid visible (cid:1) Current are directed at identifying the phosphorylation of LXR and establishing the physiological of this phosphorylation.

The transforming growth factor ␤ (TGF-␤) receptor, ALK-1, is expressed specifically on endothelial cells and is essential for angiogenesis, as demonstrated by its targeted deletion in mice and its mutation in the human disease hereditary hemorrhagic telangiectasia. Although ALK-1 and another endothelial-specific TGF-␤ receptor, endoglin, both bind TGF-␤ with identical isoform specificity and form a complex together, neither has been shown to signal in response to TGF-␤, and the mechanism by which these receptors signal in endothelial cells remains unknown. Here we report the identification of the nuclear receptor liver X receptor ␤ (LXR␤) as a modulator/mediator of ALK-1 signaling. The cytoplasmic domain of ALK-1 specifically binds to LXR␤ in vitro and in vivo. Expression of activated ALK-1 results in translocation of LXR␤ from the nuclear compartment to the cytoplasmic compartment. The interaction of activated ALK-1 with LXR␤ in the cytoplasmic compartment results in the specific phosphorylation of LXR␤ by ALK-1, primarily on serine residues. LXR␤ subsequently modulates signaling by ALK-1 and the closely related TGF-␤ receptor, ALK-2, as demonstrated by specific and potent inhibition of ALK-1-and ALK-2-mediated transcriptional responses, establishing LXR␤ as a potential modulator/mediator of ALK-1/ALK-2 signaling.
Transforming growth factor ␤ (TGF-␤) 1 is a member of a family of dimeric polypeptide growth factors that regulate cellular proliferation and differentiation as well as the processes of embryonic development, wound healing, and angiogenesis in a cell-and context-specific manner (1). Mutations in TGF-␤ receptors or their intracellular signaling molecules have been described in association with tumorigenesis and the human disease hereditary hemorrhagic telangiectasia (HHT). HHT is an autosomal dominant disease in which the primary defect is vascular dysplasia resulting in telangiectasia and arteriovenous malformations. Clinically this disorder results in telangiectases of the skin, recurrent epistaxis, and gastrointestinal bleeding as well as shunting phenomena and neurological sequela because of arteriovenous malformations in the pulmonary and central nervous system (2). The pathological lesions of HHT consist of dilated vessels, lined by a single layer of endothelium attached to a continuous basement membrane, that are thought to form from dilation of postcapillary venules. Genetic linkage studies of families with HHT identified the genes for two receptors in the TGF-␤ family, endoglin and ALK-1 (a type I receptor in the TGF-␤ family) as the genes responsible for hereditary hemorrhagic telangiectasia-1 and -2, respectively (3,4). Endoglin and ALK-1 are expressed specifically on endothelial cells, and endoglin production is increased in proliferating endothelium (5). Mice lacking endoglin or ALK-1 died at embryonic day 10.5-11.5 because of defects in angiogenesis (6 -10). Histochemical studies of the embryos revealed primary defects in the endothelial and smooth muscle cells of the developing blood vessels, confirming an essential role for these TGF-␤ receptors in angiogenesis.
ALK-1 and endoglin mutations in HHT result in similar phenotypes: both have similar affinity and specificity for TGF-␤ (with preference for the TGF-␤1 and -3 isoforms) and have been shown to associate with one another, suggesting that both receptors transduce similar signals essential for endothelial cell function (11). However, the nature of the signaling pathway downstream from these receptors is unknown.
To establish the signaling pathway downstream of ALK-1 and endoglin, yeast two-hybrid screens utilizing the cytoplasmic domains of these receptors were performed. Here we report the identification of the nuclear receptor LXR␤ as a protein that specifically binds to the cytoplasmic domain of activated ALK-1. ALK-1 alters LXR␤ cellular localization and phosphorylates LXR␤, primarily on serine residues, and LXR␤ is able to modulate ALK-1 signaling, establishing LXR␤ as a potential member of the ALK-1 signaling pathway.

MATERIALS AND METHODS
Yeast Two-hybrid Screens-The mating system of James, Halladay, and Craig was utilized (12). Briefly, a human lung library (Clontech) was screened with baits composed of the cytoplasmic domain of ALK-1 and mutants of the cytoplasmic domain of ALK-1 (ALK-1-Q200D and ALK-K229R) cloned into pGBD, in-frame with the Gal4 DNA binding domain. Yeast containing bait and the prey library cloned in pGAD, in-frame with the Gal4 activation domain, were mated overnight in YPAD (yeast extract, peptone, adenine, and dextrose) media at 30°C, plated on Trp Ϫ , Leu Ϫ plates and incubated at 30°C for 3-5 days to select for yeast containing both bait and prey vectors. Colonies were replicaplated on His Ϫ or His Ϫ , Ade Ϫ plates to assay for interaction as assessed by growth under these conditions. Interacting clones were tested for ability to repeat, bait, and prey dependence and specificity with other baits (lamin, myc, endoglin, ALK-2) to screen out false positives. Clones that passed these screens were sequenced and further analyzed.
Yeast Two-hybrid Interaction Assay-Appropriate strains of yeast (a strain for bait, ␣ strain for library) were transformed with pGBD control vector or pGBD-ALK-1, -ALK-1Q-D, -ALK-1K-R, -ALK-2, or -END (containing the cytoplasmic domain of these respective proteins) and pGAD-LXR␤ (encoding LXR␤ lacking the first 14 amino acids). These yeast were mated overnight in YPAD media at 30°C, plated on Trp Ϫ , Leu Ϫ plates, and incubated at 30°C for 3-5 days to allow diploid cells to form visible colonies. Colonies were replica-plated on His Ϫ , Ade Ϫ plates to assay for interaction as indicated by colony growth.
GST Affinity Binding Assay-COS-7 or 293T cells expressing HA-or FLAG-tagged LXR␤ or HA-tagged LXR␣ were lysed with 1% Triton X-100 lysis buffer and precleared with glutathione-agarose beads. Equal amounts of cell lysate were incubated with GST fusion proteins of the cytoplasmic domain of ALK-1 (GST-ALK-1), activated ALK-1 (GST-ALK-1Q-D), or GST alone complexed with glutathione-agarose beads. The beads were harvested by centrifugation and washed three times with lysis buffer. Binding proteins were analyzed by SDS-PAGE and Western blot analysis with ␣HA or ␣FLAG antibody as appropriate.
Co-immunoprecipitation Assay-THP-1 macrophage cells were lysed in RIPA lysis buffer with 1 mM phenylmethylsulfonyl fluoride and 0.02% leupeptin. The lysates was precleared by centrifugation, and 500-l aliquots were incubated with polyclonal preimmune serum, polyclonal ␣ALK-1 antibody (made to the cytoplasmic domain of ALK-1, peptide KREPVVAVAAPASSESSST), or ␣LXR antibody (Affinity BioReagents) for 2 h at 4°C. The immune complexes were harvested by addition of Protein A beads. The pellets were washed three times with ice-cold lysis buffer. 50 l of 2ϫ sample buffer was added, and samples were boiled and loaded on a 12.5% SDS-PAGE, transferred to nitrocellulose, and visualized by Western blot analysis with ␣ALK-1 antibody.
Confocal Microscopy-Transfected HEK-293 cells were split onto collagen-coated 35-mm plastic dishes with glass bottoms (MatTek, Ashland, MA) and cultured overnight. GFP images were generated and collected using single line excitation (488 nm) and a 505-nm emission filter with a Zeiss laser scanning confocal microscope (LSM-510).
Phosphorylation Assays-COS-7 cells were transfected with various TGF-␤ receptors and HA-tagged LXR␤ or -␣. After 48 h, the cells were washed in phosphate-free medium and labeled with [ 32 P]orthophosphate (1.0 mCi/ml) for 4 h. Cells were washed with phosphate-buffered saline, lysed with RIPA lysis buffer, immunoprecipitated with ␣HA antibody and Protein G-Sepharose, and analyzed on 10% SDS-PAGE gels; phosphorylation was detected by phosphoimager analysis of the dried gels.
Phosphoamino Acid Analysis-In vivo 32 P-labeled LXR␤ was immunoprecipitated with ␣HA antibody, resolved by SDS-PAGE, and transferred to a polyvinylidene difluoride membrane. The polyvinylidene difluoride membrane was washed, and the 32 P-labeled LXR␤ was detected by autoradiography and excised from the membrane. The protein was hydrolyzed to completion by placing in 6 M HCl at 110°C for 1 h, dried down, resuspended, and spotted on a TLC plate along with unlabeled phosphoamino acid standards. The TLC was wet with 66% (pH 1.9, electrophoresis buffer (0.58 M formic acid, 1.36 M acetic acid)): 34% (pH 3.5, electrophoresis buffer (0.5% (v/v) pyridine, 0.87 M acetic acid, 0.5 mM EDTA)) and electrophoresed for 1 h at 1.5 kV. The plate was dried at 50°C, sprayed with 0.25% ninhydrin in acetone, reheated to visualize the standards, and then analyzed by phosphoimager analysis.
Phosphopeptide Mapping-In vivo 32 P-labeled LXR␤ was resolved by SDS-PAGE, detected by autoradiography, and excised from the dried gel. The gel was rehydrated in 50 mM ammonium bicarbonate, pH 7.3, and the proteins were serially extracted by addition of 1% 2-mercaptoethanol and 0.1% SDS until over 60% of labeled protein was extracted by Cerenkov counting. 20 g of Rnase (carrier protein) was added; the mixture was trichloroacetic acid-precipitated, oxidized in cold performic acid, dried, and digested overnight in 50 mM ammonium bicarbonate, pH 8.0. with 10 g of trypsin. The digested peptides were lyophilized, resuspended in pH 1.9 electrophoresis buffer: pH 3.5 electrophoresis buffer (2:1), and spotted on a TLC plate. The TLC was electrophoresed in the first dimension for 20 min at 1.0 kV, then air dried, rotated 90°c ounterclockwise, and separated in the second dimension by chromatography in standard chromatography buffer until the buffer was 2 cm from the top of plate. The plate was dried at 65°C and analyzed by phosphoimager analysis.
Luciferase Reporter Gene Assays-P19 or L6 cells were plated onto 24-well plates at 20,000 cells/ml 1 day prior to transfection. The cells were transiently transfected using FuGene 6 reagent (Roche) with 300 ng/well of total DNA (100 ng of XVent-Lux or pE2.1-Lux, 100 ng of LXR␤ or -␣ expression vectors, and 100 ng of ALK-1 Q-D or -2 Q-D expression vectors). The following morning the cells were washed twice in phosphate-buffered saline and were immediately incubated in ␣-MEM (minimum essential medium) with 0.2% fetal bovine serum for 18 -24 h at 37°C. Afterward, cells were lysed in 100 l of passive lysis buffer (Promega) and 25 l used for luciferase assays (Promega). Data are from experiments done in triplicate and are presented as the mean -fold stimulations (Ϯ S.E.) of the luciferase induced by ALK-1 Q-D or -2 Q-D relative to samples transfected with control vector.

Interaction of LXR␤ with ALK-1 in Vitro and in Vivo-
To elucidate potential downstream signaling molecules for ALK-1, the cytoplasmic domain of ALK-1 was utilized as a bait to screen for interacting proteins with the yeast two-hybrid system of James, Halladay, and Craig (12). Initial screens of a murine embryonic library (day 10.5) and a human lung library with the entire cytoplasmic domain of ALK-1 did not yield any specific clones, whereas screens with the entire cytoplasmic domain of ALK-2 done in parallel revealed numerous clones of FKBP12 as has been previously reported (13) (data not shown). Because the cytoplasmic domain of ALK-1 contains a serine/ threonine protein kinase and constitutively active ALK-1 has been reported to activate BMP-responsive promoters (14,15) and phosphorylate Smad1 and Smad5 (14,15), baits of the cytoplasmic domain of constitutively active ALK-1 (ALK-1 Q200D, mutating glutamine 200 to aspartic acid in the GS domain) and kinase-dead ALK-1 (ALK-1 K229R, mutating lysine 229 to arginine in the ATP binding site) were utilized to rescreen the human lung library. The ALK-1 K229R bait again did not yield any specific clones; however, the screen with the ALK-1 Q200D bait yielded 17 specific clones. Three of these isolates encoded two separate clones, both encompassing nearly the entire open reading frame of the nuclear receptor LXR␤ (Fig. 1A). LXR␤ has been previously identified as an orphan nuclear receptor that forms obligate heterodimers with the retinoid X receptor to bind and regulate promoters with a unique hormone response element (LXRE) (16 -20). LXR␤ contains an amino-terminal PEST domain followed by a DNA binding domain and then a ligand binding domain containing an activation function (AF-2) (Fig. 1A). More recently the oxysterols 24(S), 25-epoxycholesterol and 24(S) hydroxycholesterol, have been identified as potential endogenous ligands for LXR␤ and the closely related nuclear receptor LXR␣ (21,22). In addition, LXR␣ has been established as a sensor of dietary cholesterol, consistent with its restricted expression in tissues rich in lipid metabolism (23). In contrast, the role of LXR␤, which is ubiquitously expressed, remains to be established. Significantly, unlike LXR␣ knockout mice, LXR␤ knockout mice lack a metabolic/lipid metabolism phenotype (24). Association of ALK-1 Q200D with LXR␤ was specific because LXR␤ was not isolated in screens with ALK-1, -1 K229R, or -2 (data not shown). In addition, the cytoplasmic domain of ALK-1 Q200D could interact with LXR␤ in the yeast two-hybrid interaction assay, whereas the cytoplasmic domains of ALK-1, -1 K229R, -2, or endoglin could not (Fig. 1B).
To confirm the interaction of ALK-1 with LXR␤, GST fusion proteins of the cytoplasmic domain of ALK-1 and -1 Q200D and an HA-tagged LXR␤ lacking the first 14 amino acids of the coding sequence (corresponding to the longer yeast two-hybrid clone isolated, Fig. 1A) were made. HA-tagged LXR␤ was efficiently expressed in COS-7 cells and was specifically pulled down by GST-ALK-1 Q200D as well as by GST-ALK-1 (Fig. 1C, top panel, lanes 7 and 8) but not by GST alone (Fig. 1C, lane 6). The interaction between ALK-1 and LXR␤ in the GST affinity binding assay was unexpected because this interaction could not be detected in the yeast two-hybrid system (Fig. 1B), suggesting that ALK-1 may be activated in mammalian cell extracts. To further investigate the interaction of ALK-1 and LXR␤, a FLAG-tagged (amino-terminal) full-length LXR␤ was made and utilized in the GST affinity binding assay. FLAGtagged LXR␤ was efficiently expressed in 293T cells and specifically pulled down by GST-ALK-1 Q200D (Fig. 1C, second panel, lane 8) but not by GST-ALK-1 or GST alone (Fig. 1C,  lanes 6 and 7), suggesting that the first 14 amino acids of LXR␤ may regulate the interaction of ALK-1 and LXR␤. Next, the interaction of ALK-1 with the closely related nuclear receptor, LXR␣, was assessed. Neither GST-ALK-1 Q200D nor -ALK-1 was able to pull down LXR␣ (Fig. 1C, top panel, lanes 3 and 4). Taken together these results confirm a specific interaction between ALK-1 and LXR␤ in vitro.
To determine the regions important for the interaction between ALK-1 and LXR␤, we examined the sequences of each protein for potential interaction motifs. This examination revealed an LXXLL motif in the cytoplasmic domain of ALK-1 (amino acids 174 -178) known to mediate interaction of proteins with the AF-2 domain of nuclear receptors (25). To establish whether this LXXLL motif was responsible for mediating interaction of ALK-1 with the AF-2 of LXR␤, the LXXLL motif was mutated in GST-ALK-1 Q200D and the mutant assessed for interaction with full-length LXR␤. Mutation of LXXLL (LG-DLL to LAAAA) of ALK-1 did not alter the ability of ALK-1 Q200D to bind LXR␤ (data not shown). In addition, mutation or deletion of the AF-2 of LXR␤ did not alter the ability of LXR␤ to bind ALK-1 Q200D (Fig. 1C, third panel, lanes 7 and 8, data not shown), confirming that these were not the regions responsible for interaction of ALK-1 and LXR␤. We then analyzed the ability of specific regions of LXR␤ to bind to ALK-1 Q200D in vitro. These studies revealed that the ligand binding domain of LXR␤ (amino acids 286 -461) was unable to bind ALK-1 Q200D in vitro (data not shown), suggesting that the binding motif resides in the first 286 amino acids of LXR␤. Attempts to directly confirm binding of amino acids 1-286 of LXR␤ containing the DNA binding domain were unsuccessful because constructs of LXR␤ (amino acids 1-286) could not be stably expressed (data not shown).
Finally, to determine whether this interaction occurred in vivo and without overexpression of ALK-1 and LXR␤ (i.e. with physiological levels of expression of ALK-1 and LXR␤), we screened cells known to express LXR␤ for expression of ALK-1. This survey revealed that the THP-1 macrophage cell line known to express LXR␤ (26) also expressed ALK-1 (Fig. 1D, lane 2). Therefore, we used the THP-1 cell line to perform co-immunoprecipitation studies with antibodies to LXR␤ and ALK-1. An antibody to LXR␤ was able to co-immunoprecipitate ALK-1 in THP-1 cell lysates (Fig. 1D, lane 3), whereas preimmune serum was not (Fig. 1D, lane 1). Conversely, an antibody to ALK-1 co-immunoprecipitated LXR␤ in THP-1 cell lysates (data not shown). Taken together, these findings strongly suggest a physiological interaction between ALK-1 and LXR␤.
Localization and Co-expression of LXR␤ and ALK-1-Because the nuclear receptor LXR␤ is thought to reside primarily in the nucleus and ALK-1 is a membrane-bound receptor, we explored how these receptors might interact in vivo. To address this issue, we examined the localization of LXR␤ in vivo. We constructed and expressed a GFP fusion protein of LXR␤ in HEK-293 cells with or without ALK-1 Q-D and examined localization of GFP-LXR␤ in live cells by confocal microscopy. In the absence of ALK-1, GFP-LXR␤ was predominately, but not exclusively, expressed in the nucleus in most cells (Fig. 2, A and  B), whereas GFP alone was uniformly distributed between the nucleus and cytoplasm as previously reported (27) (data not shown), and ␤Ϫarrestin 2-GFP was expressed exclusively in the cytoplasm (Fig. 2G). However, in some cells in the absence of ALK-1 (ϳ5%), GFP-LXR␤ was more uniformly distributed between the nucleus and cytoplasm (Fig. 2C). In the presence of ALK-1 Q-D, localization was altered so that GFP-LXR␤ was predominately expressed in the nucleus in only a minority of cells (ϳ20%), (Fig. 2D), whereas in the majority of cells, GFP-LXR␤ was uniformly distributed between the nucleus and cytoplasm (Fig. 2E). In addition, in some cells (ϳ5%), a predominant membrane localization of GFP-LXR␤ was observed (Fig.  2F, arrow), suggesting that ALK-1 Q-D results in a nuclear to cytoplasmic translocation of LXR␤ and that ALK-1 has access to LXR␤ in the cytoplasm.
To investigate the potential physiological relevance of the interaction of ALK-1 and LXR␤, we surveyed the expression patterns of ALK-1 and LXR␤ by performing virtual Northern blot analysis of SAGE and EST databases and by examining published Northern blot analysis of ALK-1 (28 -30) and LXR␤ (31). These surveys confirmed that LXR␤ is widely expressed, whereas ALK-1 has a more limited pattern of expression, primarily in vascular tissues (Table I). Significantly, ALK-1 and LXR␤ were co-expressed in 94% of tissues, including in endothelial cells, where ALK-1 is specifically expressed.
LXR␤ as a Substrate for the ALK-1 Kinase-Because LXR␤ specifically interacts with activated ALK-1 (ALK-1 Q-D), we explored whether LXR␤ was a substrate for the ALK-1 serine/threonine kinase. HA-tagged LXR␤ was expressed in COS-7 cells in the presence and absence of either wild type, activated, or kinase-dead ALK-1. The cells were labeled with orthophosphate, and LXR␤ was immunoprecipitated with ␣HA antibody. Activated ALK-1 was able to specifically phosphorylate LXR␤ (Fig. 3A, lane 4), whereas ALK-1 (Fig. 3A,  lane 3) and ALK-1 K-R (Fig. 3A, lane 5) were not. To characterize the phosphorylation of LXR␤ by activated ALK-1, LXR␤ phosphorylated in vivo by activated ALK-1 was subjected to phosphoamino acid analysis. In vivo phosphorylated LXR␤ was primarily phosphorylated on serine residues, with a trace amount of threonine phosphorylation and no detectable tyrosine phosphorylation (ratio of pSer:pThr:pTyr of 12: 1:0), (Fig. 3B). To identify the number of phosphorylation sites, phosphopeptide mapping was performed. LXR␤ was phosphorylated in vivo on at least two distinct sites, with one major phosphopeptide (Fig. 3C).
ALK-1 has been proposed to form a signaling complex with the type II TGF-␤ receptor (T␤RII) and endoglin in endothelial   3. ALK-1 phosphorylates LXR␤ on more than one serine/ threonine residue in vivo. A, COS-7 cells expressing ALK-1 or ALK-1 mutants and full-length HA-tagged LXR␤ as indicated were metabolically labeled with [ 32 P]orthophosphate, lysed, and immunoprecipitated with ␣HA antibody. Immunoprecipitates were analyzed by SDS-PAGE and phosphoimager analysis (upper panel). The position of radiolabeled LXR␤ is indicated by an arrow. Portions of the total cell lysates (TL) were analyzed by Western blot to control for expression of the indicated proteins (lower panel). B, 32 P-labeled LXR␤ (phosphorylated in vivo by ALK-1 Q-D) was subjected to phosphoamino acid analysis using onedimensional thin layer electrophoresis and analyzed by phosphoimager analysis. The positions of co-electrophoresed unlabelled phosphoamino acids are indicated (pS, serine; pT, threonine, pY, tyrosine). C, tryptic digests of gel-purified in vivo 32 P-labeled LXR␤ were resolved in two dimensions with electrophoresis in pH 1.9 buffer: pH 3.5 buffer (2:1) and chromatography in standard chromatography buffer and analyzed by phosphoimager analysis. The origin (Ori) is indicated by the ϩ. Data shown are representative of two to three independent experiments. cells (11). To investigate the specificity of phosphorylation of LXR␤ by activated ALK-1, we investigated the ability of wild type ALK-1 to phosphorylate LXR␤ in the presence of T␤RII or T␤RII and endoglin in the presence and absence of TGF-␤1. Although activated ALK-1 phosphorylated LXR␤ (Fig. 4A, lane  2), wild type ALK-1 could not (Fig. 4A, lane 1) even in the presence of T␤RII (Fig. 4A, lane 3), T␤RII and TGF-␤1 (Fig. 4A,  lane 4), or T␤RII, endoglin, and TGF-␤1 (Fig. 4A, lane 5), suggesting that an uncharacterized ligand stimulates ALK-1 to phosphorylate LXR␤ in vivo. Because T␤RII was also HAtagged (like LXR␤), autophosphorylated T␤RII was immunoprecipitated in these experiments as well (Fig. 4A, lanes 3-5), confirming that T␤RII was expressed and catalytically active. T␤RII also co-immunoprecipitated ALK-1 as has been previously reported (11), confirming expression of ALK-1 in these studies. Interestingly, consistent with a recent report (32), endoglin expression reduced autophosphorylation of T␤RII and also reduced phosphorylation of ALK-1 (Fig. 4A, compare lane  5 with lane 4).
We then investigated the ability of other type I TGF-␤ superfamily receptors to phosphorylate LXR␤. Constitutively active ALK-2 (Fig. 4B, ALK-2 Q-D, lane 4) or constitutively active ALK-5 (Fig. 4B, ALK-5 T204D, lane 5) were unable to phosphorylate LXR␤ under the same expression conditions that resulted in phosphorylation of LXR␤ by activated ALK-1 (Fig. 4B,  lane 3), confirming the specificity of this phosphorylation by activated ALK-1. Finally, to investigate the specificity of this phosphorylation we analyzed the ability of activated ALK-1 to phosphorylate LXR␣ in vivo. Activated ALK-1 was unable to phosphorylate LXR␣ in vivo (Fig. 4C, lane 3) under conditions it was able to phosphorylate LXR␤ (Fig. 4C, lane 5). The specific phosphorylation of LXR␤ (but not LXR␣) by ALK-1 in vivo strongly supports a physiologically relevant interaction between ALK-1 and LXR␤.
Effect of LXR␤ on the ALK-1 Signaling Pathway-Although the signaling pathway downstream of ALK-1 has not been elucidated, ALK-1 Q200D has been demonstrated to activate BMP-responsive promoters, including XVent2 (15). Therefore we assessed the ability of LXR␤ to affect ALK-1 Q200D-mediated Vent2 promoter-driven luciferase activity. ALK-1 Q200D was able to induce Vent2 promoter-driven luciferase activity (Fig. 5A), and LXR␤ was able to significantly reduce the ability of ALK-1 Q200D to induce Vent2 promoter-driven luciferase activity (Fig. 5A). To establish the specificity of this effect, we analyzed the effect of LXR␤ on ALK-1, ALK-2, or constitutively active ALK-2-mediated Vent2 promoter-driven luciferase activity. In addition to ALK-1 Q200D, only constitutively active ALK-2 was able to induce Vent2 promoter-driven luciferase activity, and LXR␤ was able to potently inhibit the ability of constitutively active ALK-2 to perform this function as well (Fig. 5B). In contrast, LXR␤ had little effect on the Vent2 promoter alone (Fig. 5A) or the PAI-1 based promoter, pE2.1, alone or in the presence of ALK-1 Q200D (Fig. 5C) that the effects of LXR␤ were specific. To further assess the specificity of LXR␤ effects on the Vent2 promoter, we examined the effects of LXR␣. Under identical conditions, LXR␣ was unable to significantly affect either ALK-1 Q200D (Fig. 5D) or constitutively active ALK-2 (Fig. 5E)-mediated induction of Vent2 promoter-driven luciferase activity. Taken together, these results suggest that LXR␤ may be a specific modulator of ALK-1 and -2 signaling. DISCUSSION ALK-1 and endoglin are two endothelial cell-specific receptors that are both essential for angiogenesis, and mutation of either receptor results in the human disease hereditary hemorrhagic telangiectasia (33). For these reasons there remains significant interest in the mechanisms by which these receptors function and/or signal. Earlier studies have established that both of these receptors bind TGF-␤ ligand with identical isoform specificity (11,34), both bind other TGF-␤ receptors including T␤RII, and both form a complex with each other (35). Despite this knowledge, the signaling pathway downstream from these receptors has remained undefined. Several factors account for this. First, despite binding TGF-␤, neither ALK-1 nor endoglin has been shown to signal in a TGF-␤ ligand-dependent fashion (28,36). Second, both endoglin and ALK-1 can bind other ligands, with endoglin binding Activin A, BMP-2, and BMP-7 along with their respective ligand binding receptors (35) and ALK-1 binding Activin A along with the type II activin receptor as well as an uncharacterized ligand in serum (11,28). Finally, when examined, effects of ALK-1 and endoglin have not been consistent with endoglin overexpression either decreasing (37) or increasing (32) TGF-␤ signaling and constitutively activated ALK-1 resulting in Smad1, Smad5, and Smad8 phosphorylation (14,15), activating BMP-responsive promoters including XVent2 (15) and TLX2 (14) and inhibiting the TGF-␤-responsive promoter, PAI-1 (11). Here we present evidence that the cytoplasmic domain of activated ALK-1 specifically binds the nuclear receptor LXR␤ as demonstrated by 1) the isolation of two distinct, nearly full-length clones of LXR␤ (but not LXR␣) in a yeast two-hybrid screen with the cytoplasmic domain of activated ALK-1 as a bait (Fig. 1A); 2) the specific interaction of the cytoplasmic domain of activated ALK-1 with LXR␤ in a yeast two-hybrid interaction assay (Fig.  1B); 3) the specific interaction of the cytoplasmic domain of ALK-1 with LXR␤ (but not LXR␣) in GST pull-down assays (Fig. 1C); and 4) the specific interaction of full-length endogenous ALK-1 with endogenous LXR␤ in co-immunoprecipitation studies (Fig. 1D). This interaction appears to be of physiological significance because 1) it results in the translocation of LXR␤ from the nuclear compartment to the cytoplasmic compartment (Fig. 2); 2) it results in the specific phosphorylation of LXR␤ (but not LXR␣) by ALK-1 (Figs. 3 and 4); and 3) LXR␤ (but not LXR␣) is able to specifically modulate a known cellular function of ALK-1, namely activation of XVent2 promoter-driven luciferase activity (Fig. 5). Taken together, these studies serve to establish LXR␤ as a potential mediator/modulator of ALK-1 signaling. Because the physiological ligand and signaling pathway downstream of ALK-1 has not been elucidated, we cannot determine the precise role that LXR␤ plays in the ALK-1 signaling pathway at this time. However, because LXR␤ is a transcription factor and has effects at the transcriptional level on at least one known target of ALK-1, LXR␤ may function by inhibiting the ALK-1 signaling pathway, either by directly inhibiting binding of other transcription factors or by recruiting transcriptional repressors. Studies are currently underway to evaluate these hypotheses.
ALK-2 is the type I receptor in the TGF-␤ superfamily most closely related to ALK-1 and shares many properties with ALK-1. Similar to ALK-1, ALK-2 binds the type II activin receptor, binds Activin A along with the type II activin receptor, and can activate a PAI-1-based reporter gene when coexpressed with the type II activin receptor (28). However, Xenopus animal cap explant assays suggest that ALK-2 does not function as an activin receptor in vivo (38), and subsequent studies have suggested that ALK-2 acts as a BMP receptor. For example, ALK-2 can bind BMP-7 when expressed with the type II activin receptor (14). In addition, constitutively activated ALK-2, like ALK-1, can phosphorylate Smad1, Smad5, and Smad8 (14,39) and can activate BMP-responsive promoters including Xvent2 (15) and TLX2 (14). Finally, ALK-2, like ALK-1, can complex with T␤RII and bind TGF-␤1 along with T␤RII (40,41). Thus, the current finding that the ALK-1 interacting protein, LXR␤, can mediate ALK-1 signaling as well as ALK-2 signaling is not entirely unexpected. One substantial difference between ALK-1 and -2 is that, in contrast to ALK-1, ALK-2 is ubiquitously expressed (40), similar to LXR␤ (31), suggesting the LXR␤ may play a larger role in mediated/modulating ALK-2 signaling in vivo. We have demonstrated that ALK-2 does not bind LXR␤ (Fig. 1B) and activated ALK-2 does not significantly phosphorylate LXR␤ (Fig. 4B), suggesting that, at least in the case of ALK-2 signaling, phosphorylation of LXR␤ by ALK-2 may not be necessary for the effects of LXR␤ on ALK-2 signaling. The role and mechanism of action of LXR␤ in modulating ALK-2 signaling is currently under further investigation.
The results of our initial yeast two-hybrid screen with the cytoplasmic domain of activated ALK-1 were somewhat unexpected because they suggested an interaction with a nuclear receptor, LXR␤, thought to be directly targeted to the nucleus. Our initial results supported an interaction in vivo, because ALK-1 could phosphorylate LXR␤ in vivo and LXR␤ could effect ALK-1 signaling in vivo. In addition, receptors in the TGF-␤ superfamily have been demonstrated to internalize, and some studies have suggested that internalization may be essential for TGF-␤ signaling (42,43). Furthermore, in a least one report, a type I TGF-␤ superfamily receptor, ALK-5, has been demonstrated to have a nuclear presence (44), and members of the nuclear receptor family have been demonstrated to have a presence in the cytoplasm (45). Thus, an interaction between LXR␤ and ALK-1 could conceivably occur in the nucleus or cytoplasm. Our findings that LXR␤ is expressed in both the nucleus and cytoplasm and that activated ALK-1 is able to induce translocation of LXR␤ from the nucleus and cytoplasm strongly support a physiological interaction between ALK-1 and LXR␤ and suggest that this interaction occurs in the cytoplasm.
LXR␣ and -␤ were initially identified in low-stringency screens of human and rat liver cDNA libraries, respectively, (20,46) as well as by degenerate PCR of cDNA from an osteoblastic cell line (16). LXR␣ and -␤ are highly conserved isoforms with 77% identity in their DNA and ligand binding domains. Accordingly, both LXR␣ and -␤ bind DNA as heterodimers of retenoid X receptor and preferentially bind to hormone response elements (LXREs) that consist of two hexanucleotide repeats (AGGTCA) separated by four nucleotides (DR-4) (46). In addition, both LXR␣ and -␤ are activated by specific oxysterols including 24(S),25-epoxycholesterol, 22(R)-hydroxycholesterol, and 24(S)-hydroxycholesterol (21,22). Despite these common properties, LXR␣ and -␤ appear to have distinct functions in vivo. LXR␣ is expressed in a tissue-specific manner, primarily in tissues rich in lipid metabolism such as adipose tissue, the liver, kidney, and intestine, whereas LXR␤ is ubiquitously expressed (31). In addition, LXR␣ has been established as a sensor of dietary cholesterol (23), whereas LXR␤ knockout mice lack a metabolic/lipid metabolism phenotype (24). Finally, although all LXR target genes encode proteins that have major roles in controlling cholesterol and/or fatty acid metabolism, in every case LXR␣ either exclusively regulates their expression or has significantly higher activity than LXR␤ in activating these genes (23). These results highlight the need to define specific functions for LXR␤. The present studies suggest that one of the functions of LXR␤ may be to regulate TGF-␤ superfamily signaling, particularly for the ALK-1 and -2 receptors. Clarification of the role of LXR␤ in the ALK-1 pathway awaits further elucidation of this pathway.
Nuclear receptors are classically activated by binding their respective lipophilic ligands. However, most nuclear receptors are phosphorylated on serine and threonine residues, and their level of phosphorylation generally increases upon ligand binding (47). This phosphorylation has been demonstrated to regulate DNA binding, transcriptional activation, and/or the stability of these receptors. In addition, ligand-independent activation of nuclear receptors by phosphorylation has been reported (48). This has been best documented for the estrogen receptor (ER): (1) it can be activated by growth factors (epidermal and insulin-like growth factors) or activators of protein kinase A and protein kinase C; 2) ligand-independent activation is associated with altered phosphorylation of ER; and 3) mutation of mitogen-activated protein (MAP) kinase phosphorylation sites on ER is able to abrogate activation by epidermal growth factor (49). Kinases which have been demonstrated to phosphorylate nuclear receptors include cyclin-dependent kinases, MAP kinases, casein kinase II, protein kinase A, and protein kinase C (47). The current studies identify ALK-1 as a specific kinase for LXR␤ but not for LXR␣. Because LXR␣ is not active in modulating ALK-1 signaling, these results suggest that this phosphorylation event is critical for regulating LXR␤ effects. Current studies are directed at identifying the site of phosphorylation of LXR␤ by ALK-1 and establishing the physiological relevance of this phosphorylation.
Oxysterols have well defined roles in regulating cholesterol, fatty acid, triglyceride, and sphingolipid metabolism (50). In addition, oxysterols have been implicated in regulating a diverse array of cellular processes, including calcium metabolism (51), mitochondrial function, apoptosis, and signal transduction (50). In terms of vascular biology, oxysterols have been implicated in the process of atherosclerosis (52) and have been demonstrated to induce apoptosis in endothelial cells (53) and vascular smooth muscle cells (54). Whether oxysterols, through their effects on LXR␤, have effects on ALK-1 signaling in endothelial cells and vascular smooth muscle cells, both of which have a prominent role in angiogenesis and the pathogenesis of HHT, remains to be determined.