Identification of Tctex2β, a Novel Dynein Light Chain Family Member That Interacts with Different Transforming Growth Factor-β Receptors*

Endoglin is a membrane-inserted protein that is preferentially synthesized in angiogenic vascular endothelial and smooth muscle cells. Endoglin associates with members of the transforming growth factor-β (TGF-β) receptor family and has been identified as the gene involved in hereditary hemorrhagic telangiectasia. Although endoglin is known to affect cell responses to TGF-β, its mode of action is largely unknown. We performed yeast two-hybrid screening of a human placental cDNA library and isolated a new endoglin-binding partner, a novel 221-amino acid member of the Tctex1/2 family of cytoplasmic dynein light chains named Tctex2β, as the founder of a new Tctex1/2 subfamily. The interaction was localized exclusively to the cytoplasmic domain of endoglin. Reverse transcription-PCR showed expression of Tctex2β in a wide range of tissues, including vascular endothelial and smooth muscle cells, placenta, and testis, as well as in several tumor cell lines. High expression levels were found in human umbilical vein endothelial cells and the large cell lung cancer cell line. Forced expression of Tctex2β had a profound inhibitory effect on TGF-β signaling. Additional Tctex2β-interacting receptors were identified to be the TGF-β type II receptor and most likely beta-glycan, but not ALK5, ALK1, or the bone morphogenetic protein type II receptor. Upon fluorescence tagging, co-localization of Tctex2β and endoglin, as well as Tctex2β, endoglin, and the TGF-β type II receptor, was observed by different microscopy techniques. Our findings link endoglin for the first time to microtubule-based minus end-directed transport machinery, suggesting that some endoglin functions might be regulated and directed by its interaction with the cytoplasmic dynein light chain Tctex2β.

Endoglin (CD105) is a homodimeric transmembrane glycoprotein of ϳ180 kDa that is synthesized to high levels in proliferating endothelial cells in both culture and angiogenic vasculature (1)(2)(3)(4). Cell type-specific expression has further been observed in macrophages (5), T-cells (6), bone marrow cells (7), stromal cells (8), vascular smooth muscle cells (9), and different types of cancer cells (10). Endoglin is classified as a transforming growth factor-␤ (TGF-␤) 5 type III receptor that needs the TGF-␤ type II receptor (T␤RII) to bind to TGF-␤1 and TGF-␤3 (11), in contrast to betaglycan, the second TGF-␤ type III receptor (12). Endoglin has been identified as the gene mutated in hereditary hemorrhagic telangiectasia type 1, an autosomal dominant inherited vascular disorder characterized by multisystemic vascular dysplasia and recurrent hemorrhage (13). Mice embryos homozygously deleted of endoglin are unable to progress beyond embryonic days 10.5-11.5, with failure to form mature blood vessels in the yolk sac, poor vascular smooth muscle development, and arrested endothelial remodeling (14 -16).
TGF-␤ is involved in vascular development and homeostasis with both pro-and anti-angiogenic effects (17,18). Upon ligand binding, T␤RII recruits, phosphorylates, and activates the ALK5 (activin receptor-like kinase-5) type I receptor. Activated ALK5 phosphorylates the receptor Smads (Smad2/Smad3), which are then translocated to the nucleus, where they act as co-transcription factors on specific genes (19). In endothelial cells, a second type I receptor exists, ALK1. ALK5 and ALK1 mediate, in a concentration-dependent manner, TGF-␤ signaling with opposing effects (20,21). Mutations in ALK1 also lead to the vascular disorder hereditary hemorrhagic telangiectasia type 2 (13). ALK1 signaling is propagated through Smad1/ Smad5, promoting vascular endothelial cell proliferation and migration ("activation phase"), whereas ALK5 signaling suppresses endothelial cell proliferation and induces smooth muscle differentiation ("resolution phase") (18). Thus, angiogenesis seems to be regulated in part by an ALK1/ALK5 signaling balance.
Recent findings suggest that endoglin is involved in the finetuning of this balance between TGF-␤/ALK1 and TGF-␤/ ALK5 signaling. Endoglin interacts with T␤RII, ALK5, and ALK1 both in vitro and in vivo (11,22,23). Ectopic expression of endoglin in several cell types inhibits the TGF-␤/ALK5 signaling pathway and counteracts TGF-␤-induced growth inhibition, but promotes TGF-␤/ALK1 signaling (24 -27). Of note is the discrepancy that, although the extracellular domain of endoglin is involved in and required for both pathways, the relatively short cytoplasmic domain of 47 amino acids is required only for TGF-␤/ALK5 (but not TGF-␤/ALK1) signaling (24). Thus, this domain appears to have a key role in regulating the signaling balance in endothelial cells.
The cytoplasmic domain of endoglin is highly conserved in human, rat, and mouse with 99% identity, and there is 71% homology between the corresponding domains in human endoglin and betaglycan. Recent findings have identified the LIM domain proteins ZRP-1 (zyxin-related protein-1) and zyxin as intracellular partners for the cytoplasmic domain of endoglin (28,29). Interaction of ZRP-1 or zyxin with endoglin influences the intracellular distributions of proteins involved in the organization of the actin cytoskeleton or the assembly of focal adhesions, respectively. However, the mechanisms by which endoglin achieves these functions remain unknown.
To gain more understanding of the role of endoglin in TGF-␤ signaling and angiogenesis, we performed yeast two-hybrid screening to identify cellular binding partners for this protein.
We report here the interaction between a novel Tctex1/2 (t-complex testis-expressed protein-1/2)-like protein, designated Tctex2␤, and the cytoplasmic domain of endoglin as well as T␤RII. To the best of our knowledge, this is the first evidence that endoglin can interact with a member of the dynein light chain (DLC) protein family.

EXPERIMENTAL PROCEDURES
Cell Culture-Cell media and fetal bovine serum were purchased from Invitrogen and PAA Laboratories. HEK293, NIH3T3, HeLa, COS-1, HepG2, and mink lung epithelial cells (Mv1Lu) were maintained in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, penicillin/ streptomycin, and L-glutamine. Primary human dermal microvascular endothelial cells were obtained from PromoCell (Heidelberg, Germany) and cultured in endothelial cell basal medium MV with SupplementPack MV. All cell lines were cultured at 37°C in a humidified 5% CO 2 environment.
For the mammalian two-hybrid assay (Clontech), the coding sequence of the endoglin cytoplasmic domain was cloned into vector pM, resulting in pM-Endo. The whole coding sequence of Tctex2␤ was cloned into pVP16, yielding pVP16-Tctex2␤. For expression in mammalian cells and immunoprecipitation assays, the full coding sequence of Tctex2␤ was fused to the N-terminal His 6 tag of the pcDNA4/HisMax vector (Invitrogen), resulting in His Tctex2␤. Human endoglin and HA-tagged T␤RII (T␤RII HA ) are in the pCMV5 vector (22,30). For fluorescence microscopy studies, human endoglin was fused to the N-terminal HcRed sequence of the pHcRed-N1/1 vector (a kind gift from Dr. Peter March, Faculty of Life Sciences, University of Manchester) or pECFP (Clontech), resulting in Endo HcRed or Endo ECFP , respectively. The full-length coding sequence of Tctex2␤ was fused to the C-terminal enhanced green fluorescent protein (EGFP) sequence of the pEGFP-C1 vector (Clontech), resulting in EGFP Tctex2␤. The correct sequences and reading frames of all constructs derived from PCR products were verified by DNA sequencing.
Yeast Two-hybrid Screening and Growth Assay-A yeast strain (PJ69-4A) based on the Gal4 system was used (32). Yeast media were purchased from Clontech. Yeast transformations were performed using a standard lithium acetate method as recommended in the Matchmaker yeast protocol handbook (Clontech). For library screening, a sequential transformation was performed with first the bait plasmid (pGBT9-Endo cyto ) and then with 50 g of human placental library cDNA (cloned into the pACT2 vector). The co-transformants were streaked onto appropriate selective media plates. To determine the strength of interactions, high stringency synthetic medium lacking histidine and adenine but containing X-gal was also used. Interacting clones were selected by their abilities to grow or turn blue on appropriate selective plates. Plasmid DNA from positive yeast clones was rescued into bacterial strain KC8 (Clontech) and sequenced using a poly(T) sequencing primer. The resultant sequences were checked for similarity to known transcripts in the nucleotide sequence data bases using the BLAST algorithm. For yeast two-hybrid growth assays, yeast cells were sequentially transformed first with endoglin, beta-glycan, or different type I/II receptor constructs and then with pACT2-Tctex2␤. Phenotypes of yeast co-transformants on selective media were tested.
Mammalian Two-hybrid Assay-HEK293 cells were seeded 1 night before transfection at an initial density of 1.5 ϫ 10 5 cells/well in 6-well plates in Dulbecco's modified Eagle's medium plus 10% fetal bovine serum. Prior to transfection, cells were switched to medium containing 5% fetal bovine serum to reduce the endogenous level of alkaline phosphatase in the serum. Upon reaching 50 -60% confluence, cells were transfected with the pG5SEAP reporter construct (0.3 g) and a specific combination of pM-Endo (1.5 g)/pVP16-Tctex2␤ (1.5 g) or pM (1.5 g)/pVP16 (1.5 g) as a negative control using the CalPhos TM mammalian transfection kit (Clontech). 48 h after transfection, the cell culture medium was collected and subjected to secreted alkaline phosphatase activity assay using the Great EscAPe kit (Clontech). Relative light units were measured using a Berthold Lumat LB 9507 luminometer and expressed as the means Ϯ S.E. Values from mock-transfected samples served as the basal level of alkaline phosphatase activity in the culture medium. Transfection experiments were carried out in triplicate and repeated three times.
Immunoprecipitation and Western Blotting-HEK293 cells were transfected with appropriate combinations of pCMV5, His Tctex2␤, endoglin, and T␤RII HA as indicated. 48 h later, cells were lysed with lysis buffer (50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 0.5% Nonidet P-40, 50 mM NaF, 1 mM phenylmethylsulfonyl fluoride, and Sigma protease inhibitor mixture) at 4°C for 30 min. The cell lysates were precleared with 20 l of protein G-Sepharose (Sigma) for 1 h before they were immunoprecipitated overnight at 4°C with 50 l of protein G-Sepharose and the indicated antibodies. The immunocomplexes were then washed three times with 500 l of lysis buffer and resolved in 20 l of Laemmli buffer. The resulting immunoprecipitated sample and 20 l of each total cell lysate were separated on SDS-8% (for endoglin and T␤RII HA ) and 12% (for His Tctex2␤) polyacrylamide gels and blotted onto nitrocellulose membranes (Bio-Rad) for immunoblotting with the indicated antibodies. Immunodetection were performed with the ECL Western blotting detection kit (Amersham Biosciences) or with the Odyssey infrared imaging system (LI-COR Biosciences).
5Ј-Rapid Amplification of cDNA Ends (RACE) and Reverse Transcription (RT)-PCR-For cloning of Tctex2␤ cDNA, 5Ј-RACE was performed with the SMART TM RACE kit (Clontech) using human placental and testis total RNAs (Clontech). RACE-PCR products were cloned into the pCR2.1-TA cloning vector (Invitrogen) and then sequenced using forward and reverse M13 sequencing primers. The Tctex2␤ gene-specific primer used for 5Ј-RACE was 5Ј-TTGTAGCGTGGCGGGC-TGAGCTCG-3Ј.
For RT-PCR, total RNAs from primary human umbilical vein endothelial cells (HUVECs) and HMEC-1, HepG2, large cell lung cancer (LCLC), and T47D cells were isolated as described previously (31). Total RNA from smooth muscle cells was generously supplied by Dr. Judith Gillard (Department of Biophysics, University of Manchester). To exclude the possibility of genomic DNA contamination, total RNA was digested with RNase-free DNase I (Roche Applied Science, Mannheim, Germany) according to the manufacturer's instructions. Subsequently, 1 g of each total RNA was reverse-transcribed in a 50-l volume with a mixture of oligo(dT) oligonucleotides and (dN) 6 random hexamers following the Fermentas RevertAid H Minus Moloney murine leukemia virus protocol. In parallel, negative control samples lacked Moloney murine leukemia virus. 2 l of the resulting first-strand cDNA was PCR-amplified with appropriate primer pairs using a ReadyMix TM PCR kit (Sigma). Samples were first held at 96°C for 5 min and then cycled 30 times at 94°C for 30 s, 58°C for 30 s, and 72°C for 1 min, followed by an additional 7 min at 72°C. PCR products were visualized by electrophoresis on 1.5% agarose gel containing ethidium bromide. Primer pairs were based on human sequences (given here from 5Ј to 3Ј): Tctex2␤, ATGGCCAGCAG-GCCTCTGCC (forward) and TCACTCGCAGTAGAGCCCGT (reverse); and glyceraldehyde-3-phosphate dehydrogenase, TCA-ACGGATTTGGTCGTAT (forward) and ATGAGTCCTTCCA-CGATAC (reverse).
Fluorescence Microscopy and Immunofluorescence-For fluorescence microscopy, HeLa cells were seeded on coverslips and transfected as indicated. 1 day after transfection, cells were fixed with 4% paraformaldehyde on ice, mounted on Mowiol (Dabco), and used for digital imaging.
Two-photon microscope images were obtained with the TriMScope 2-photon system for multiphoton microscopy (LaVision BioTec, Bielefeld, Germany) with a Zeiss Axiovert 200 fluorescence microscope equipped with a 63ϫ oil objective. This system was upgraded for two-photon live-time imaging with a Chamelon XR laser (Coherent Inc.). Two-photon excitation was at 800 nm, and emission for GFP was at 530 nm and for CFP at 480 nm. Images were analyzed using ImSpector Version 3.0 software. Images obtained with the Zeiss ApoTome system were acquired with an inverted wide-field microscope (Zeiss Axiovert 200 M and Axiovision Version 05.2005 software) equipped with a 75-watt xenon lamp (Hamamatsu Photonics), a 12-bit CCD camera (Zeiss AxioCam HRm), a 63ϫ oil objective (Plan Neofluar, 1.25 numerical aperture), and appropriate filter sets for the different fluorophores (CFP, excitation at 436/20 nm, beam splitter at 455 nm, and emission at 480/40 nm; yellow fluorescent protein, excitation at 500/20 nm, beam splitter at 515 nm, and emission at 535/30 nm; and DsRed, excitation at 565/30 nm, beam splitter 585 nm, and emission at 620/60 nm).
Immunofluorescence was done as described (31). In brief, HeLa cells transfected with endoglin were seeded on coverslips, fixed, permeabilized, and incubated with a 1:100 dilution of the endoglin-specific monoclonal antibody SN6 (Serotec). Subsequently, cells were incubated with a TRITC-labeled anti-mouse secondary antibody (Molecular Probes). After several washing steps, cells were mounted on Mowiol and used for fluorescence microscopy.
For confocal microscopy, 1 night before transfection, NIH3T3 or HEK293 cells were seeded on polylysine-coated coverslips in 6-well plates at a density of 0.8 ϫ 10 5 cells/well. Cells were transiently transfected with combinations of different constructs ( EGFP Tctex2␤, Endo HcRed , or vectors alone as controls). After an additional 48 h, cells were fixed with 4% paraformaldehyde on ice. Coverslips were mounted on Mowiol and subjected to fluorescence analysis using a Zeiss confocal microscope. Cells were imaged using LSM 510 software, and the Multi-track function was chosen to collect potentially overlapping emissions separately. Green and red fluorophores were excited using the argon (488 nm) and helium/neon (543 nm) visible lasers, respectively.
Luciferase Reporter Assay-The luciferase assay was performed as described previously (31). In brief, Mv1Lu or HepG2 cells were transiently transfected with (CAGA) 12 -luciferase (an ALK5 signaling-responsive reporter construct) (33) together with His Tctex2␤ or endoglin constructs as indicated. The pEYFP-N1 vector (Clontech) was always included to serve as an internal control for transfection efficiency. Luciferase activity was assayed 16 h after incubation with 4 ng/ml TGF-␤3 or TGF-␤1. All experiments were performed in triplicate and repeated at least three times.
Proteasome Inhibitor Assay-The proteasome inhibitors MG132, lactacystin, and proteasome inhibitor I (proteasome inhibitor set I, catalog no. 539164, Calbiochem) were used. Inhibitors were suspended in Me 2 SO at 5 mM. HEK293 or COS-1 cells were seeded in 6-well plates. Cells were mocktransfected; transfected with His Tctex2␤, endoglin, or T␤RII HA alone; or cotransfected with His Tctex2␤ and endoglin or with His Tctex2␤ and T␤RII HA using the cationic transfection reagent jetPEI TM (Polyplus Transfection) according to the manufacturer's instructions. The next day, cells were either left untreated or were incubated overnight for 16 h with a combination of 1 l/ml each inhibitor or equal amounts of Me 2 SO. After incubation, cells were lysed in 400 l, and the protein concentration of each lysate was determined using the BCA TM protein assay (Pierce) according to the manufacturer's instructions. Subsequently, equal amounts of protein (30 g) were separated by SDS-PAGE and blotted onto a nitrocellulose membrane (Macherey-Nagel). After incubation of the membrane with specific primary and secondary antibodies as indicated, expressed proteins were visualized with the Odyssey infrared imaging system.
Cell Migration in the Wound Assay-HeLa cells were seeded in 6-well plates and either singly transfected with 3 g of mock DNA or 1.5 g of His Tctex2␤, endoglin, or T␤RII supplemented with 1.5 g of mock DNA or doubly transfected with 1.5 g of His Tctex2␤ and 1.5 g of endoglin or T␤RII using jetPEI TM . The next day, cells were wounded with a 20 -200-l tip. Cells were washed once with phosphate-buffered saline, and fresh growth medium was added. Subsequently, the wounded cell layers were photographed at four different positions. After 48 h, the same positions were photographed again to document cell migration.
Cell Proliferation Assay-To assess cell proliferation under fixed cell culture conditions, the increase in cell numbers was measured after different time points by staining the cell nuclei with crystal violet as described previously (34). In brief, 5 ϫ 10 5 primary human dermal microvascular endothelial cells were transfected with 3 g of mock DNA or His Tctex2␤ by nucleofection with a Nucleofector (Amaxa Biosystems, Cologne, Germany) according to the manufacturer's instructions using the S-005 nucleofection program. After nucleofection, cells were plated in 6-well plates to recover for 1 day. Subsequently, cells were trypsinized and seeded in a 96-well plate at a density of 2000 cells/well. After 4 h of adhesion, cells were incubated for 24 and 48 h with 0.5 ng/ml TGF-␤1 or 4 ng/ml TGF-␤1 or were left untreated. After the different time points, cells were fixed with 5.5% glutaraldehyde and subsequently washed three times with double-distilled H 2 O and then air-dried for 1 h. A 0.1% crystal violet solution (100 l/well; pH 4.5) was added for 20 min. Cells were washed three times with double-distilled H 2 O and air-dried again. Finally, 10% acidic acid (100 l/well) was added, and extinction was measured at 595 nm. An aliquot of cells assayed after 4 h of adhesion served as a reference at time 0.

Identification of Tctex2␤, a Novel Tctex1/2 Family Protein, as an Interacting Partner for the Cytoplasmic Domain of Endoglin-
To identify new intracellular binding partners for endoglin, the yeast Gal4 two-hybrid system was used (32). A human placental cDNA library was screened with the cytoplasmic domain of endoglin (Endo cyto ) as bait and cloned into the Gal4 DNA-binding domain vector pGBT9. A number of clones strongly and reproducibly interacting with endoglin were identified as determined by their abilities to grow in the absence of adenine and histidine and to activate the lacZ reporter. Among the identified clones was ZRP-1, confirming a previous report (29). However, the strongest interaction was identified in a clone provisionally named N22.1 with an ϳ650-bp insert. When sequenced, it was found to correspond to a hypothetical protein (GenBank TM accession number XM_291623) with homology to mouse Tctex1. The N22.1 clone is only a partial sequence of this predicted protein, corresponding to the C-terminal 182amino acid region (residues 40 -221).
A computational expression sequence tag (EST) "walk" was performed in the EST data base. The only two EST entries overlapping with N22.1 correspond to its 3Ј-and 5Ј-ends, respectively. The 3Ј-EST, from human lung epithelial cells (699 bp; GenBank TM accession number CA314455), covered the whole N22.1 insert sequence, whereas the 5Ј-EST, from human placenta (726 bp; GenBank TM accession number CF994778) extended an additional 626 bp upstream of the insert sequence.
To determine and isolate the full-length N22.1 cDNA, 5Ј-RACE with human placental and testis RNAs was performed. The cloned and sequenced RACE products confirmed the 5Ј-EST sequence from our computational analysis. Taken together, the putative complete N22.1 cDNA was determined to be 1276 bp long, consisting of a single open reading frame of 666 bp, a 5Ј-untranslated region of 509 bp, and a 3Ј-untranslated region of 101 bp containing a consensus polyadenylation signal (aataaa) located 15 bases upstream of the poly(A) tail.
Because of the homology to the Tctex1/2 protein family and additional analyses (see below), N22.1 was renamed and is henceforth referred to as Tctex2␤. The complete nucleotide sequence and its deduced amino acid sequence are shown in Fig. 1A and have been deposited in the GenBank TM Data Bank with accession number DQ132441. An additional BLASTn search with the whole cDNA mapped this gene to human chromosome 1p34.1, with two exons and one intron of 108 bp. The existence of the intron was confirmed by RT-PCR with primer pairs located in exons 1 and 2 flanking the intron (data not shown).
Bioinformatic Analysis of the Tctex2␤ Protein-A series of software-based analyses were performed to provide clues about the structural features of the Tctex2␤ protein. The Compute pI/Mw program showed that Tctex2␤ has a theoretical pI of 9.87 and an expected molecular mass of 23 kDa. In amino acid composition, it is ϳ10% serine and has a proline-rich N terminus (20% prolines of the N-terminal 114 amino acids). Motif searching by PROSITE (35) showed four consensus protein kinase C phosphorylation sites and two casein kinase II phosphorylation sites. Secondary structure prediction by PSIPRED (36) showed the layout of the helixstrand structure, in which the N-terminal section formed several ␣-helices, whereas in the C terminus, four ␤-strands were predicted. This predicted structure agrees very well with the known secondary structure of the Tctex1 protein (37,38).
By BLASTp and EST data base searches, Tctex2␤ homologs have been identified in several different mammalian species, including mouse (70% identity; GenBank TM accession number BC092499), pig (80% identity; Gen-Bank TM accession number AJ973122), rat (74% identity; Gen-Bank TM accession number AJ973123), and chimpanzee (98% identity; GenBank TM accession number XM_513130). An alignment of amino acid sequences using ClustalW (39) revealed the homology between Tctex2␤ and members of the Tctex1/2 protein family, especially in the C-terminal regions (Fig. 1B). For example, Tctex2␤ shares 27% identity with mouse Tctex2, 18% identity with human Tctex1, and 26% identity with a human Tctex2b homolog (GenBank TM accession number BC021177). The unrooted tree generated by PhyloDraw (40) shows the phylogenesis of Tctex2␤ (Fig. 1C). Of the 15 proteins we aligned, human Tctex2␤ and its several close homologs clearly form a distinct subdivision (the Tctex2␤ group) different from the other three subfamilies, i.e. the Tctex1 group (including mouse, rat, human, and Chlamydomonas Tctex1 as well as human RP3), Tctex2 group (including mouse, rat, and human Tctex2) and Tctex2b group (including Chlamydomonas Tctex2b and a human testis Tctex2b homolog).
Confirming the Specific Interaction between Endoglin and Tctex2␤-The interaction between endoglin and full-length Tctex2␤ was reconfirmed in yeast by sequential transformation, first with different endoglin constructs as indicated in Table 1 or the pGBT9 empty vector and subsequently with pACT2-Tctex2␤. For this purpose, full-length wild-type endoglin (Endo wt ) and the endoglin-coding sequence lacking the cytoplasmic part (Endo ⌬cyto ) were cloned into pGBT9. As summarized in Table 1 (first four rows), yeast cells co-transformed with either Endo wt /Tctex2␤ or Endo cyto /Tctex2␤ not only were able to grow on His/Ade-deficient selection medium, but also could turn blue because of activation of the lacZ reporter. However, neither the Tctex2␤/pGBT9 nor Endo ⌬cyto / Tctex2␤ co-transformant was able to grow on selective medium. These results strongly suggest that Tctex2␤ specifically interacts with the cytoplasmic domain of endoglin.
The interaction between endoglin and Tctex2␤ was also confirmed using a mammalian two-hybrid assay in HEK293 cells in which the cytoplasmic domain of endoglin was fused to the DNA-binding domain of the pM vector , whereas Tctex2␤ was fused to the activation domain of the pVP16 vector. A third vector, pG5SEAP, encoding the secreted alkaline phosphatase gene, served as the reporter, the induction of which is caused by interaction of pM and pVP16 fusion proteins. Cotransfection of the three constructs, pMendoglin, pVP16-Tctex2␤, and pG5SEAP, resulted in a 4.1fold induction of secreted alkaline phosphatase activity compared with the negative control samples (p Ͻ 0.01). This indicates that the endoglin/Tctex2␤ interaction also happens in mammalian cells.
HEK293 cells were transiently transfected with the pCMV5 vector (mock), His Tctex2␤, or Endo wt alone or with His Tctex2␤/Endo wt together as indicated in Fig. 2. Western blotting of the total cell lysate using anti-His antibody showed a single specific band (ϳ28 kDa), which corresponded to the 221-amino acid Tctex2␤ polypeptide. Endoglin showed two bands, with the lower molecular mass band corresponding to the non-fully processed form. As shown in Fig. 2A, Tctex2␤ was coprecipitated by anti-endoglin antibody from the lysates of His Tctex2␤/Endo wt cotransfectants. Furthermore, we observed that HEK293 cells expressed low amounts of endogenous endoglin. Thus, we transfected HEK293 cells with His Tctex2␤ alone and precipitated the endogenous endoglin protein from cell lysates. Subsequent Western blotting demonstrated that the His-tagged Tctex2␤ protein was coprecipitated (Fig. 2B). These experiments further confirm the specific in vivo interaction between endoglin and Tctex2␤.
Expression Profiling of Tctex2␤-To investigate the potential physiological relevance of the novel Tctex2␤ protein, the expression pattern of Tctex2␤ was studied by RT-PCR analysis with glyceraldehyde-3-phosphate dehydrogenase as an internal control. Given that the whole coding sequence of Tctex2␤ is within one exon, genome-free cDNAs were prepared. This was verified by the fact that non-reverse transcriptase-treated samples gave negative PCR results (Fig. 3, lanes 2). Tctex2␤ gene expression could be seen in primary vascular endothelial cells (HUVECs), an endothelial cell line (HMEC-1) and smooth muscle cells, all of which are supposed to be key cell types of endoglin function. Expression was also detected in other tissues and cell types, including human placenta and testis, as well as in several tumor cell lines such as LCLC, HepG2 (hepatocellular carcinoma cells), and T47D (breast cancer cells). Different levels of expression were observed, with the highest expression in LCLC cells and HUVECs; lower expression in HepG2, HMEC-1, and smooth muscle cells; and weakest expression in T47D cells, placenta, and testis.
Tctex2␤ Inhibits TGF-␤-induced ALK5 Signaling-Considering the role of endoglin in modulating cell responses to TGF-␤, we next investigated whether Tctex2␤ is also involved in the TGF-␤ signaling pathway and what the functional implications of the interaction between endoglin and Tctex2␤ are. The (CAGA) 12 -luciferase reporter has been widely used as a specific reporter for TGF-␤/ALK5 signaling specifically responding to Smad3/Smad4 activity (33). Mv1Lu cells were used to test the role of Tctex2␤ and endoglin in TGF-␤ signal- DECEMBER  ing. Cells transfected with (CAGA) 12 -luciferase responded with a Ͼ11-fold increase in reporter activity after TGF-␤3 treatment (Fig. 4A). Overexpression of endoglin decreased ligand-induced reporter activity by 33% as expected. In contrast to the previously identified DLC km23/mLC7-1, which enhances and supports TGF-␤ signaling responses (41), Tctex2␤ overexpression caused an even stronger down-regulation of signaling compared with endoglin, with 45% inhibition of ligand-induced reporter activity. However, cotransfection of endoglin and Tctex2␤ did not alter the level of inhibition as for Tctex2␤ alone, suggesting that there is no synergistic inhibitory effect for endoglin and Tctex2␤. Similar results were obtained with HepG2 cells. Here again, overexpression of Tctex2␤ inhibited TGF-␤1-induced signaling (Fig. 4B). This suggests that the inhibitory activity of Tctex2␤ on signaling is less likely cell typespecific, but a more general feature of Tctex2␤ function.

Analysis of the Interaction between Tctex2␤ and Other Members of the TGF-␤ Receptor
Family-We were greatly interested in the possibility that Tctex2␤ might interact with other members of the TGF-␤ receptor family. In view of this, the cytoplasmic domains of different type I receptors (ALK5, wild-type ALK1, and constitutively active ALK1), type II receptors (activin type IIA receptor (ActRIIA), T␤RII, and bone morphogenetic protein type II receptor (BMPRII)), and the betaglycan type III receptor were cloned into the pGBT9 vector and tested for their ability to interact with pACT2-Tctex2␤ in a yeast twohybrid system. Of the different receptors, T␤RII, betaglycan, and ActRIIA could also associate with Tctex2␤, as inferred by colony growth on His/Ade-deficient plates, whereas BMPRII, ALK5, ALK1, and constitutively active ALK1 failed to interact with Tctex2␤ (Table 1, fifth through eleventh rows). Nevertheless, it is worth noting that, as judged by the inability of T␤RII/ Tctex2␤ or ActRIIA/Tctex2␤ co-transformants to turn blue on

Summary of the interactions between Tctex2␤ and different members of the TGF-␤ receptor family
Yeast two-hybrid growth assays were performed by sequential transformation of yeast cells, first with appropriate domains of different TGF-␤ receptors fused to the DNA-binding domain (BD) of pGBT9 and then with full-length Tctex2␤ fused to the activation domain (AD) of pACT2. The pGBT9 or pACT2 vector was used as an appropriate negative control. The presence and strength of interactions between the indicated proteins were judged by the abilities of yeast co-transformants to grow on His/Ade-deficient plates or to turn blue on His/Ade-deficient X-gal plates. Ϫ, no interaction; ϩ, interaction; blue, strong interaction.

Endoglin Binding to a Dynein Light Chain Protein
high stringency plates, it seemed that these interactions were less strong than the endoglin/Tctex2␤ or betaglycan/Tctex2␤ interaction. Association between T␤RII and Tctex2␤ was further confirmed by co-immunoprecipitation studies in HEK293 cells (Fig. 5), whereas interaction between ActRIIA and Tctex2␤ could not be detected by co-immunoprecipitation (data not shown). Subcellular Distribution of Tctex2␤ and Co-localization Studies with Endoglin and T␤RII-HeLa, HEK293, and NIH3T3 cells were transiently transfected with EGFP Tctex2␤ in the presence and absence of Endo ECFP or Endo HcRed and then subjected to two-photon microscopy and confocal fluorescence microscopy. Microscopy analysis revealed that Tctex2␤ was distributed throughout the cytoplasm and nucleus (Fig. 6, A-C). Cells cotransfected with endoglin and Tctex2␤ showed clear co-localization of Tctex2␤ and endoglin (Fig. 6A). Similar results were also found in HEK293 and NIH3T3 cells (data not shown). HeLa cells transfected with EGFP Tctex2␤ alone clearly showed that Tctex2␤ localized to vesicles and microtubules and was highly abundant in the nucleus, but not in the nucleoli (Fig. 6, B and C).
Next, we analyzed in more detail the co-localization and intracellular distribution of Tctex2␤, endoglin, and T␤RII using HeLa cells cotransfected with all three constructs. Fluorescence microscope images were taken using the ApoTome  12 luciferase reporter construct together with His Tctex2␤, endoglin, or a combination of the two. Cells were incubated with or without TGF-␤3 (4 ng/ml) for 16 h, after which luciferase activities were measured. B, HepG2 cells were transfected as described for Mv1Lu cells and incubated with or without TGF-␤1 (4 ng/ml) for 16 h, after which luciferase activities were measured. All transfection experiments were performed in triplicate and repeated at least three times, and the results shown are the average -fold changes. S.D. values are indicated by error bars, and all experiments were normalized for enhanced yellow fluorescent protein fluorescence intensity. *, p Ͻ 0.05 compared with the mock transfection group; **, p Ͻ 0.01.  . Subcellular distribution of Tctex2␤ and co-localization with endoglin. A, HeLa cells were cotransfected with EGFP Tctex2␤ and Endo ECFP . Subcellular localization of EGFP Tctex2␤ and Endo ECFP is shown in green and red (false-colored), respectively. Co-localization of Tctex2␤ and endoglin is represented by yellow in the overlaid images. Images were taken by two-photon microscopy with the TriMScope 2-photon system for multiphoton microscopy. B, shown is a series of Z-stack images of HeLa cells transfected with EGFP Tctex2␤. Images were taken with the TriMScope 2-photon system as described for A. C, shown is a composite image of Z-stack images from B.
imaging system, which produces images of confocal microscopy quality and properties. Endoglin and T␤RII both localized to vesicles and the plasma membrane and clearly co-localized, although there was not a complete vesicular overlap (Fig. 7). Tctex2␤ was, as seen before, distributed throughout the cytoplasm as well as in the nucleus and was associated with vesicles and what we think are microtubule structures. Co-localization of Tctex2␤, endoglin, and T␤RII was seen at the cell membrane forming the junction (Fig. 7, white arrowheads) between two cells and in cellular protrusions and therefore maybe areas of cell adhesion. Surprisingly, there was almost no co-localization of Tctex2␤, endoglin, and T␤RII in vesicles. Vesicles containing endoglin and T␤RII were in general larger than and distinct from those associated with Tctex2␤, although these data at least suggest that Tctex2␤, endoglin, and T␤RII can be present in one protein complex.
No Detectable Influence of Tctex2␤ on Cell Proliferation and Migration-To analyze what effects Tctex2␤ might have on cellular functions, we investigated its influence on cell proliferation. Given that Tctex2␤ interacts with endoglin and that endoglin is highly expressed in endothelial cells, primary human dermal microvascular endothelial cells were used. Cells were mock-transfected or transfected with Tctex2␤ and treated with low (0.5 ng/ml) and high (4 ng/ml) amounts of TGF-␤1 for 24 and 48 h. This experiment was repeated three times, but there was no significant difference in the proliferation rate of Tctex2␤and mock-transfected cells (data not shown).
Next, we tested the effect of Tctex2␤ on cell migration. For this purpose, HeLa cells were mocktransfected or transfected with Tctex2␤ alone or in combination with endoglin or T␤RII or with all three together. Subsequently, cells were wounded, and cell migration was analyzed after 48 h. No obvious difference could be observed between Tctex2␤and non-Tc-tex2␤-transfected cells.
Ectopic Tctex2␤ Expression Increases Endoglin and T␤RII Protein Amounts-Over the course of our analyses, we observed that the amounts of endoglin or T␤RII from Tctex2␤ cotransfections detected by Western blotting varied compared with single transfections. Thus, we speculated that Tctex2␤ might influence the proteasomal degradation of endoglin and T␤RII. To test this hypothesis, HEK293 cells were transfected with either the different expression constructs alone or with endoglin or T␤RII in combination with Tctex2␤. Transfected cells were then treated for 16 h with either proteasome inhibitors or Me 2 SO or were left untreated. After cell lysis, equal amounts of proteins were analyzed by SDS-PAGE and Western blotting for the amounts of endoglin, T␤RII, and Tctex2␤. Coexpression of endoglin or T␤RII with Tctex2␤ clearly resulted in an increased amount of endoglin as well as T␤RII (Fig. 8A). This stabilizing effect in the presence of Tctex2␤ was equal to or even higher than that in the proteasome inhibitor-treated single transfectants. The same result was also seen in COS-1 cells coexpressing endoglin and Tctex2␤ (Fig. 8B).

DISCUSSION
In this study, we have reported a new DLC protein named Tctex2␤ that interacts with the TGF-␤ type III receptor endoglin as well as T␤RII. Sequence alignment, phylogenetic analysis, and secondary structure prediction clearly placed Tctex2␤ in the Tctex1/2 family of DLCs, a family that was first identified in mouse testis and later found in the cytoplasmic dyneins of many tissues (42,43). On the basis of our bioinformatic analyses, we are confident of the existence of a new Tctex subfamily (Tctex2␤) formed by human Tctex2␤ and its several close homologs in chimpanzee, pig, mouse, and rat (Fig. 1C). Cytoplasmic dyneins are involved in a variety of intracellular motile processes, including mitosis, maintenance of the Golgi apparatus, and trafficking of membranous vesicles and other intracel- FIGURE 7. Subcellular distribution and co-localization of Tctex2␤, endoglin, and T␤RII. To analyze whether Tctex2␤, endoglin, and T␤RII co-localize, HeLa cells were triply transfected with EGFP Tctex2␤, T␤RII ECFP , and endoglin expression constructs. Subcellular localization of EGFP Tctex2␤ and T␤RII ECFP is shown in green and blue, respectively. Subcellular localization of endoglin is shown in red by immunofluorescence using antiendoglin monoclonal antibody SN6 and a TRITC-labeled secondary antibody. Co-localization of Tctex2␤, endoglin, and T␤RII appeared to be in the cell membrane, marked in the overlay image by the white arrowhead as well as in the overlay image for Tctex2␤ and endoglin. The overlay image for endoglin and T␤RII demonstrates co-localization in the cell membrane and in vesicles distinct from vesicles associated with Tctex2␤. Images were taken with the ApoTome imaging system. DECEMBER 1, 2006 • VOLUME 281 • NUMBER 48 lular particles. These movements are required for the spatial organization of the cytoplasm and, as a consequence, are crucial for many processes such as cell division, embryonic development, and regulation of signaling (44,45). DLCs bind to the cargo and therefore determine cargo specificity. Three distinct families of cytoplasmic DLCs have been identified and designated LC8, LC7/Roadblock, and Tctex1/2. Accumulating evidence shows that DLCs can interact with different, functionally unrelated proteins, facilitating dynein association with specific cargoes. For example, known binding partners for Tctex1 include rhodopsin, Doc2, Fyn kinase, the Trk receptor, CD5, and CD155 (46,47).

Endoglin Binding to a Dynein Light Chain Protein
The existence of a gene family often indicates indispensable but also differing functions of its members in different tissues. Expression profiling of human Tctex2␤ revealed that Tctex2␤ mRNA is present in primary vascular endothelial and smooth muscle cells, where endoglin is preferentially expressed. Thus, it is possible that Tctex2␤ might play a role in angiogenesis or other endoglin-related functions in the vasculature. Other than that, a range of Tctex2␤ expression levels were found in placenta, testis, and several tumor cell lines. Even though the Tctex2␤ expression levels varied among the different cell types tested, the data suggest a ubiquitous expression profile. The highest expression was seen in LCLC cells and HUVECs, whereas the Tctex2␤ mRNA level in testis was relatively low. In contrast, both Tctex1 and Tctex2 are enriched in testis, whereas Tctex1 expression in lung tissue is differentially regulated, with greatly decreased levels in adult lung. Tctex2 is more restricted to testis, liver, and fetal thymus, with only very low levels detectable in fetal and adult lungs (43). These first expression data for Tctex2␤ in comparison with Tctex1 and Tctex2 suggest that this family of DLCs might have overlapping but also diverging cell type-specific functions.
Endoglin is a new member of the TGF-␤ receptor family identified to interact with a DLC. BMPRII and the bone morphogenetic protein type IA (ALK3) and IB (ALK6) receptors have been reported to bind to Tctex1 (48). In vitro kinase assays demonstrated that BMPRII phosphorylates Tctex1, which is abolished by BMPRII mutations causing the human disorder primary pulmonary hypertension. T␤RII has been shown to interact with and phosphorylate a member of the LC7/Roadblock group of DLCs, km23/mLC7-1 (41), upon TGF-␤ treatment. Overexpression of mLC7-1 induces specific TGF-␤ responses, including JNK (c-Jun N-terminal kinase) activation, c-Jun phosphorylation, and mink lung epithelial cell growth inhibition. Furthermore, TGF-␤ induces the recruitment of mLC7-1 to the dynein intermediate chain. In addition, it has been demonstrated that mLC7-1 is required for TGF-␤-induced activation of the p3TP-Lux promoter reporter (49).
We also investigated the role of Tctex2␤ in TGF-␤ signaling and found that overexpression of Tctex2␤ in mink lung epithelial and HepG2 cells severely inhibited TGF-␤-induced (CAGA) 12 -luciferase reporter activity by Ͼ40% (Fig. 4, A and  B). This inhibitory effect was even stronger than that seen for endoglin. However, coexpression of Tctex2␤ and endoglin demonstrated no synergistic effect of further reduced reporter activity. At the moment, we do not know whether Tctex2␤ is upstream or downstream of endoglin in the signaling cascade. However, because Tctex2␤ is a DLC, and DLCs are generally involved in signaling by mediating the transport of receptor signaling complexes in a retrograde fashion along the microtubules toward the nucleus, it is conceivable that Tctex2␤ operates downstream of endoglin. Our results seem to be contradictory to the up-regulation of TGF-␤ signaling by mLC7-1; however, the (CAGA) 12 -luciferase reporter becomes activated by the Smad3/Smad4 pathway, whereas p3TP-Lux reporter induction depends on an intact MAPK (mitogen-activated protein kinase) pathway (49). Thus, it is possible that the functions of Tctex2␤ in Smad and non-Smad signaling pathways are different. In general, this raises the possibility that the function of DLCs might be pathway-specific.
Aside from endoglin, we have demonstrated that T␤RII can associate with Tctex2␤, too, and that the three proteins form a multimeric complex (Fig. 7). Fluorescence microscope images showed that the main co-localization of the multimeric Tctex2␤-endoglin-T␤RII complex occurred at the cell membrane, whereas in vesicles, only rarely did we observe the three proteins together or Tctex2␤ with endoglin or with T␤RII. In contrast, vesicular co-localization of endoglin with T␤RII was clearly visible. However, these vesicles appeared bloated com- pared with the smaller sized vesicles associated with Tctex2␤. Endoglin and T␤RII are transmembrane proteins transported from the Golgi toward the cell membrane, an anterograde movement. Only during ligand-induced signaling or protein recycling/proteasomal degradation would we expect to see membrane receptors in retrograde moving vesicles. Tctex2␤ is a DLC protein family member. These proteins are thought to be involved in retrograde protein transport. This suggests that the observed endosomal co-localization of endoglin and T␤RII is most likely in vesicles on their way to the cell membrane and that Tctex2␤ is, under the conditions used in this study, not involved in receptor internalization and hence receptor recycling/proteasomal degradation of endoglin and T␤RII. The current hypothesis is that Tctex2␤ may block signaling by preventing receptor internalization, which could lead to an increased retention time of the receptors at the cell membrane. This is supported by our finding that coexpression of Tctex2␤ with endoglin or T␤RII leads to a more or less equivalent increase in the two receptors comparable with that in endoglin-or T␤RIItransfected cells treated with proteasome inhibitors. In addition, these data also suggest that membrane-localized endoglin and T␤RII are constantly degraded by the proteasome system and replaced by newly synthesized endoglin and T␤RII. Apart from that, because the cytoplasmic domain of T␤RII, which interacts with Tctex2␤, has kinase activity, it would be very interesting to test whether Tctex2␤ might be phosphorylated by the T␤RII kinase domain and whether the phosphorylated Tctex2␤ is involved in determination of cargo specificity or other signaling events.
In summary, we have identified a new DLC protein (Tctex2␤) as a novel binding partner for endoglin and T␤RII. Tctex2␤ shows an unexpected role as a TGF-␤ signaling inhibitor. By fluorescence microscopy, co-localization of Tctex2␤, endoglin, and T␤RII could be detected at the cell membrane, but was hardly seen in vesicles. Furthermore, coexpression of Tctex2␤ with endoglin or T␤RII appears to stabilize the two receptors, therefore preventing proteasomal degradation of endoglin and T␤RII. This prompted us to speculate that Tctex2␤ might increase the retention time of endoglin and T␤RII at the cell surface by blocking internalization, which would also lead to inhibition of ligand-induced signaling. Further studies are now necessary to investigate this hypothesis.