Cloning and Characterization of a Family of Proteins Associated with Mpl*

Thrombopoietin (TPO) controls the formation of megakaryocytes and platelets from hematopoietic stem cells via activation of the c-Mpl receptor and multiple downstream signal transduction pathways. We used two-hybrid screening to identify new proteins that interacted with the cytoplasmic domain of Mpl, and we found a new family of proteins designated A2D (for Ataxin-2 Domain protein). The A2D are 130-kDa proteins that have three regions similar to those of Ataxin-2, the gene product causing familial type 2 spinocerebellar ataxia. A2D has several isoforms with different C-termi-nal domains, all produced from a single gene by alternative splicing. Northern blotting indicated that the A2D gene is widely expressed in immortalized cell lines and hematopoietic and fetal tissues. A2D proteins were constitutively associated with Mpl in vivo in human hematopoietic UT7 cells. TPO also caused the release of A2D from the activated receptor, and the phosphorylation of A2D on tyrosines residues was dependent on the Mpl C-terminal domain. Finally, A2D bound to the unstimulated erythropoietin receptor, whereas erythropoietin caused dissociation from the erythropoietin receptor, suggesting that A2D proteins are new components of the cytokine signaling system.


From INSERM U363, ICGM, 27 Rue du Faubourg Saint Jacques, 75014 Paris, France
Thrombopoietin (TPO) controls the formation of megakaryocytes and platelets from hematopoietic stem cells via activation of the c-Mpl receptor and multiple downstream signal transduction pathways. We used two-hybrid screening to identify new proteins that interacted with the cytoplasmic domain of Mpl, and we found a new family of proteins designated A2D (for Ataxin-2 Domain protein). The A2D are 130-kDa proteins that have three regions similar to those of Ataxin-2, the gene product causing familial type 2 spinocerebellar ataxia. A2D has several isoforms with different C-terminal domains, all produced from a single gene by alternative splicing. Northern blotting indicated that the A2D gene is widely expressed in immortalized cell lines and hematopoietic and fetal tissues. A2D proteins were constitutively associated with Mpl in vivo in human hematopoietic UT7 cells. TPO also caused the release of A2D from the activated receptor, and the phosphorylation of A2D on tyrosines residues was dependent on the Mpl C-terminal domain. Finally, A2D bound to the unstimulated erythropoietin receptor, whereas erythropoietin caused dissociation from the erythropoietin receptor, suggesting that A2D proteins are new components of the cytokine signaling system. TPO 1 is the primary physiological regulator of megakaryopoiesis (1). TPO acts via Mpl (2), a cytokine receptor originally identified as the cellular homologue of the product of the myeloproliferative leukemia virus (MPLV) oncogene (3). Mpl is found only in tissues that support hematopoiesis, the bone marrow, spleen, and the fetal liver; it is also abundant in CD34ϩ cells and in cells of the megakaryocytic lineage (4). Disruption of the Tpo or Mpl genes in mice leads to the loss of 90% of the megakaryocytes and circulating platelets plus reduced megakaryocyte ploidy (5)(6)(7). These mice have few morphologically and functionally normal platelets, whereas Tpo-deficient mice have fewer than normal hematopoietic stem cells and progenitors of both myeloid and erythroid lineages, indicating that TPO has a pleiotropic range of activities (8,9). Like other members of the cytokine receptor superfamily, Mpl has no intrinsic enzymatic activity. TPO binding induces receptor homodimerization (10) leading to activation of the Janus kinases (Jak)2 and Tyk2 and the stimulation of multiple signaling events. These include the phosphorylation of Mpl tyrosines (11), providing docking sites for proteins containing src homology (SH)2 and phosphotyrosine binding domains. The downstream molecular targets of receptor activation include signal transducers and activators of transcription 1, 3, and 5 (STAT1, STAT3, and STAT5), Jun, Cbl, Vav, Raf-1, mitogen-activated protein kinase (MAPK), phosphatidylinositol 3-kinase, Akt, and the inositol phosphatase SHIP (12)(13)(14)(15)(16). However, eliminating all the tyrosine residues of Mpl does not destroy its capacity to support cell proliferation, suggesting that there are other mechanisms for the transduction of TPO signaling (17,18).
Our understanding of the intracellular signaling pathways activated by TPO is based largely on the heterologous expression of wild type and mutant Mpl genes in immortalized hematopoietic cell lines. These studies suggest that the membraneproximal intracellular domain of Mpl provides the primary proliferative signal requiring activation of Jak/STAT signaling (19). The membrane-distal region is thought to be required for megakaryocyte maturation through sustained activation of MAPK (20) that involves both Ras/Raf1-and Rap1/B-Raf-dependent pathways (21). The importance of MAPK activation in megakaryocytic differentiation was confirmed in normal hematopoietic progenitors isolated from cord blood or bone marrow (22,23). The way these two regions of Mpl help mediate the functions of TPO in vivo was investigated by deleting part of the gene encoding the Mpl C-terminal domain in mice and by comparing their phenotype to that of mice lacking the entire receptor. Mice lacking the membrane-distal half of Mpl intracellular domain had full megakaryocytic lineage maturation including normal platelet counts (24). However, their response to exogenous TPO stimulation was impaired, indicating that the distal region of the receptor determines the quantitative but not a qualitative response to TPO in vivo. Several signaling pathways are affected by this deletion. Shc and Akt are not phosphorylated in platelets from these mice, whereas the phosphorylation of STAT3, STAT5, and MAPK is greatly reduced. However, the truncated receptor still supports cell proliferation and differentiation, suggesting that unidentified signaling pathways may complement the defect due to the distal deletion.
We have performed a yeast two-hybrid screening to identify the proteins that bind to the Mpl intracellular domain and thus find out more about the signaling pathways activated by TPO. We cloned a new family of proteins related to Ataxin-2, Ataxin-2 Domain protein (A2D), which is associated with Mpl.
The gene encoding Ataxin-2, the product of the SCA2 locus, is defective in type 2 spinocerebellar ataxia (25)(26)(27), one of the familial neurodegenerative diseases characterized by unstable CAG repeats, expanded in patients (28,29). These disorders result in progressive loss of neurons from the cerebellum, brain stem nuclei, and spinocerebellar tracts. The only feature shared by the genes causing these diseases is that they have CAG repeats encoding polyglutamines. CAG repeat diseases all have such features as ubiquitination, aggregation of full-length or truncated proteins, and cleavage by caspase (30,31). But just how the expansion of the repeat triggers neuron death remains unclear. Although SCA2 only affects the cerebellum and brain, Ataxin-2 transcripts are found in various tissues. The protein (145 kDa) is located in the cytoplasm and the trans-Golgi network. Several putative functional motifs have been identified in the Ataxin-2 protein (caspase3 cleavage site and clathrin-mediated trans-Golgi signal), but its biological function is unknown.
The present report describes the molecular cloning and characterization of several A2D 130-kDa proteins isoforms, each with different C-terminal domains. We show that the A2D proteins are most abundant in hematopoietic and fetal tissues. A2D proteins are constitutively associated with Mpl in vivo in hematopoietic cells and are released from the receptor upon its activation. TPO triggers the phosphorylation of A2D on tyrosine residues in a way that is dependent on the presence of the Mpl C-terminal domain. We also find that A2D proteins are constitutively associated with the endogenous erythropoietin receptor (EPO-R) and EPO is caused to dissociate from the receptor, suggesting that A2D proteins are new components of cytokine signaling.
Plasmids Construction-The whole human c-Mpl cytoplasmic domain (IC) was cloned by PCR in-frame into the Gal4 binding domain of LexA plasmid to yield LexA-Mpl-IC.
Stable transfections of UT7/Mpl or BaF3/Mpl were prepared by cloning full-length A2D cDNA or N-terminally truncated A2D (starting at amino acid 315) into a modified pCDNA3 vector where the cytomegalovirus promoter was changed to Rous sarcoma virus. The Myc tag was fused to amino acid 1028 of A2D, deleting the C-terminal domains of the various isoforms.
UT7/Mpl cDNA Library-A cDNA library was constructed from the human megakaryoblastic cell line UT7 expressing Mpl in which TPO promotes proliferation and differentiation signals (34). Cells were cultured for 4 days in the presence of TPO, and total RNA was prepared using the TRIzol reagent (Invitrogen). Poly(A) ϩ RNA was obtained with the mRNA isolation system (Invitrogen), and cDNA primed with (dT) [12][13][14][15][16][17][18] was synthesized from 2 g of poly(A) ϩ RNA with the Copy Kit cDNA Synthesis System (Invitrogen), as specified by the manufacturer and ligated to EcoRI linkers. The fractions containing 1-2-kb-long cDNA species were isolated on a sucrose gradient and cloned in the EcoRI site of the pGAD-GH activation domain plasmid (CLONTECH). These were electroporated into XL1-Blue MRFЈ (Invitrogen) to give ϳ10 6 independent colonies. The library was amplified in liquid medium for 4 h, and plasmid DNA was purified with the Mega kit (Qiagen).
Two-hybrid Library Screening and Cloning of Full-length A2D cDNA-Screening was carried out as described previously (35). Saccharomyces cerevisiae strain L40 (MAT␣ trp1 leu2 his3 LYS2::lexA-HIS3 URA3::lexA-lacZ) cells were sequentially transformed with the LexA-Mpl bait plasmid and then with the cDNA library prey plasmids. Double transformants were plated on yeast drop-out medium lacking Trp, Leu, and His and grown for 4 days. The resulting colonies were patched on the same medium and replica-plated on Whatman 40 filters to test for ␤-galactosidase activity. Positive clones were selected, and prey plasmids containing library cDNA inserts were isolated and transformed into bacteria for amplification. The 5 ϫ 10 5 clones screened gave four primary positives. Only one of them, a 1565-bp insert cDNA, M37, interacted specifically with the Mpl cytoplasmic domain in yeast.
The M37 insert was used as a probe to isolate full-length clones from a Jurkat cell line cDNA library cloned into the NotI site of pSPORT plasmid, kindly provided by Dr. L. Maouche-Chretien (INSERM U474). A total of 18 clones with inserts up to 4050 bp were obtained. The cDNAs contained the majority of the coding sequence. The remaining 5Ј sequence was obtained from a human blood cDNA library cloned in ZAP Express TM /EcoRI vector (Stratagene number 938202) that allowed us to clone three complete cDNAs. Both strands of the isolated clones were sequenced on an Applied Biosystem automated DNA sequencer. The complete sequence of A2D-A cDNA corresponding to the longest cDNA can be found in GenBank TM under accession number AJ317970. The 3Ј parts of the cDNAs were heterogeneous, generating isoforms with different C-terminal domains termed A2D-B, -C, -D, and -E (see "Results" and Fig. 4). One of these clones also contained an internal in-frame deletion from nucleotides 633 to 908 that deleted an internal ATG. The M37 clone isolated by two-hybrid screening extended from A2D-D nucleotide 2215 to the poly(A) tail.
Cloning Genomic Sequences-Genomic DNA surrounding position 261 in cDNAs was cloned by PCR with 1 g of genomic DNA from UT7 and Jurkat cell lines with the following oligonucleotides: sense, 5Ј-GCAACAGGCCGTGGCCCGT (positions 225-243), and antisense, 3Ј-GAAGGCATCTTGGCGCCGCAG-5Ј (positions 390 -408) with the Advantage-GC Genomic PCR kit (CLONTECH) under the following conditions: 1 cycle at 94°C for 5 min; 5 cycles at 94°C for 5 s, 68°C for 2 min; 5 cycles at 94°C for 5 s, 65°C for 30 s, 68°C for 2 min; 15 cycles at 94°C for 5 s, 60°C for 30 s, 68°C for 2 min; and 1 cycle at 68°C for 7 min. The concentration of GC-Melt was 1 M. The PCR products were cloned in the pCR 2.1-TOPO vector with the PCR 2.1-TOPO cloning kit (Invitrogen), and the nucleotide sequence of 4 clones of each cell line was determined.
Northern Blot Analysis-A human multiple tissue Northern blot (CLONTECH) containing ϳ1 g of poly(A) ϩ RNA from each tissue (hematopoietic, fetal, and adult) was hybridized according to the instructions of the manufacturer using a 1.5-kb EcoRI insert from the M37 clone as probe. Northern blotting of human and murine immortalized cell lines was performed using 10 g of total RNA from each cell line.
Chromosomal Localization-Human metaphase cells for fluorescence in situ hybridization were prepared from phytohemagglutininstimulated lymphocytes by using the thymidine synchronization and bromodeoxyuridine release technique for banding. The A2D cDNA probe was biotinylated by the nick translation procedure (Invitrogen) and used for fluorescence in situ hybridization, which was performed as described previously (36).
Antibodies-Anti-A2D antibodies were obtained by cloning the cDNA coding for A2D amino acids 927-1057 in-frame into the expression vector pGex-4T2 (Amersham Biosciences). The resulting recombinant protein was purified by affinity chromatography with glutathione-Sepharose TM 4B beads and used to immunize rabbits. A monoclonal antibody M1 (Eastman Kodak Co.), specific for the FLAG epitope tag sequence, was used to detect Mpl. Anti-phosphotyrosine monoclonal antibodies 4G10 and PY72 were provided by Dr. Drucker (Portland, OR) and Dr. Sefton (La Jolla, CA). Monoclonal antibody against Myc tag (9E10) was purchased from Santa Cruz Biotechnology, Inc., and anti-Jak2 antiserum was from Upstate Biotechnology, Inc. (catalog number 06-255). The anti-EPO-R antiserum used for immunoprecipitation was provided by Dr. Mayeux (INSERM U363, ICGM), and antibodies used for immunoblotting were purchased from Santa-Cruz (sc-695).
Cell Culture, Stimulations, and Transfections-Growth factor-dependent human megakaryoblastic UT7 cells expressing murine FLAG-Mpl were cultured in ␣-minimum essential medium supplemented with 10% fetal calf serum and either 2.5 ng per ml of granulocyte-macrophage colony-stimulating factor (GM-CSF) or 2 units/ml EPO. Before TPO stimulation, cells were washed twice in serum-free medium and suspended in ␣-minimum essential medium supplemented with 0.4% bovine serum albumin. Stimulations were performed by adding either 100 ng per ml recombinant Hu-PEG-MGDF or 10 nM TPO mimetic peptide. For EPO stimulations, exponentially growing UT7 cells were washed and incubated for 16 -18 h in Iscove's modified Dulbecco's medium supplemented with 0.4% bovine serum albumin and 20 g of iron transferrin under serum-free conditions. The cells were stimulated with 10 units/ml EPO at 37°C. UT7/Mpl cells were stably transfected by electroporation at 960 microfarads and 250 V using a GenePulser (Bio-Rad).
Immunoprecipitation and Immunoblotting-Cells were washed with phosphate-buffered saline (PBS) and lysed on ice in lysis buffer: 50 mM NaCl, 50 mM Tris, pH 7.5, 1% Brij 98 (or 30 mM octyl ␤-D-thioglucopyranoside), 1 mM Na 3 VO 4 , 1 mM phenylmethylsulfonyl fluoride, 10 g/ml leupeptin, 10 g/ml aprotinin, and 1 g/ml pepstatin. Insoluble membrane fragments were removed by centrifugation at 150,000 ϫ g for 45 min at 4°C. For immunoprecipitation, lysates were incubated with preimmune serum bound to protein G-Sepharose beads for 30 min and then incubated with specific antibodies for 1 h at 4°C. Protein G-Sepharose was added, and the incubation was continued for 20 min at 4°C. Samples were washed once with lysis buffer and three times with wash buffer: 50 mM NaCl, 50 mM Tris, pH 7.5, 0.1% Brij 98 (or 3 mM octyl ␤-D-thioglucopyranoside). Precipitates were dissolved by boiling in SDS sample buffer and resolved by SDS-PAGE. Immunoblots were prepared as described previously (21).
Subcellular Fractionations and Immunofluorescence-Cells were washed with cold PBS and suspended in ice-cold buffer (210 mM mannitol, 70 mM sucrose 5 mM Tris, pH 7.5, 1 mM EDTA, supplemented with protease inhibitors). The cells were allowed to stand for 15 min and homogenized by four passes through a ball-bearing homogenizer (cell cracker EMBL) with a clearance of 10 m. Nuclei were collected by centrifugation at 500 ϫ g for 5 min at 4°C and washed several times with PBS. They were then suspended in SDS sample buffer. The supernatant containing the membrane and cytosolic proteins was centrifuged at 10,000 ϫ g for 30 min. The pellet containing the heavy membranes was washed 3 times with PBS and suspended in lysis buffer (1% Nonidet P-40, 50 mM Tris, pH 7.5, 137 mM NaCl). The supernatant was centrifuged at 100,000 ϫ g for 60 min to give the light membranes in the pellet and the cytosolic fraction in the supernatant. The protein concentration of each fraction was determined by the BCA Protein Assay Reagent A (Pierce).
Immunofluorescence experiments were carried out on UT7 cells fixed with 3.7% formaldehyde in PBS and permeabilized with 0.1% Triton X-100. The permeabilized cells were incubated with anti-A2D antibodies, followed by fluorescein isothiocyanate-conjugated anti-rabbit secondary antibody and examined under a Zeiss microscope.

Molecular
Cloning of A2D cDNAs-We performed a yeast two-hybrid screening using the entire cytoplasmic domain of Mpl fused to LexA as a bait to identify new mediators involved in TPO signaling. We constructed a human cDNA library from the human megakaryoblastic cell line UT7 expressing Mpl in which TPO promotes both proliferation and differentiation (see "Experimental Procedures"). One positive clone (M37) was isolated by screening 5 ϫ 10 5 independent transformants. To validate this result and identify the domain of Mpl involved in this interaction, we tested in the yeast two-hybrid system the binding of M37 to cytoplasmic deletion mutants of Mpl (Fig. 1). Removal of the 24 C-terminal amino acids of Mpl (⌬98 -121) prevented the interaction with M37, whereas this 24-amino acid fragment alone interacted actively with M37. M37 also did not interact with Lex-Lamin or Lex-Ras V12 showing that the M37/Mpl interaction was specific.
We assessed the expression of M37 transcripts by Northern blotting on various human hematopoietic tissues. Hybridization with the M37 probe revealed a set of transcripts with two major forms, 4 and 4.5 kb (see below and Fig. 4). M37 was also abundantly expressed in purified megakaryocytic CD41ϩ cells, whose proliferation and maturation are controlled by TPO.
M37 is an incomplete cDNA since it contained a 1.5-kb insert including a poly(A) tail. The M37 probe was used to screen cDNA libraries from human circulating blood cells and Jurkat cell line and to isolate full-length cDNA. A total of 21 independent clones were purified and analyzed. The largest was 4380 nucleotides long, very close to the size of the transcript observed by Northern analysis. DNA sequencing followed by sequence alignment with the FASTA program revealed a match with the cDNA of the human Ataxin-2-like protein A2LG (Gen-Bank TM accession number AF034373). 2 There were several discrepancies between our cDNAs and the A2LG sequence. (i) The 5Ј 148 nucleotides of the A2LG published sequence were not present in our cDNAs. This sequence corresponds to repeated sequences in the human genome, as shown by Gen-Bank TM data base analysis. (ii) 6 point mutations did not change the amino acid sequence, whereas 4 insertions of single nucleotides in our clones changed the reading frame in two regions of the cDNA as follows. 1) Insertion of a C at nucleotide 261 led to a translation initiation different from that of A2LG protein. Therefore, A2D and A2LG proteins have different Nterminal domains upstream of amino acid 32 of A2D. 2) Insertions of single nucleotides at positions 1454, 1465, and 1508 modified 18 amino acids of the A2LG published sequence. We cloned the corresponding exon from the genomic DNAs of UT7 and Jurkat cells (see "Experimental Procedures") to verify the insertion at position 261. This insertion was found in the genomic clones and was also present in 5 overlapping cDNAs from a human circulating blood cell library taken from 4 adult males. A polymorphism at this position is therefore very unlikely. We confirmed our nucleotide sequence several times on both strands, and we designated this sequence as A2D-A for Ataxin-2 Domain Protein. This sequence has been registered in the GenBank TM data base under accession number AJ317970. We believe that A2LG and A2D-A are produced by transcription of the same gene.
The sequence of the A2D-A cDNA revealed an open reading frame extending from nucleotide 1 to 3396. The first methionine-encoding sequence at position 169 is preceded by a consensus sequence for initiating translation of a 1076-amino acid protein ( Fig. 2A). Although the sequence upstream of nucleotide 169 has no stop codon, the entire encoded protein sequence is probably in A2D-A since in vitro transcription/translation generated a 130-kDa protein, very close to the size of the A2D protein in UT7 cell lysates (see below Fig. 5B). The A2D-A sequence also contained 2 additional in-frame ATGs (amino acids 183 and 315) flanked by nucleotides that fit the Kozak consensus that could be alternative translation initiation sites.
Our data indicate that the A2D-A protein is very rich in proline (15%) and serine (11%) and contains 27 tyrosines. The protein is basic except for a highly acidic region (residues 169 -325). The sequence contains no obvious functional do- main, but it has three substantial regions similar to those of Ataxin-2. The first (amino acids 208 -352) is 60% similar to Ataxin-2 (Fig. 3A) and is also conserved in a putative protein of Drosophila (GenBank TM accession number AE003708). The second (amino acids 651-668) is 100% similar to Ataxin-2 and homologous to a short region of the antiproliferative transducer of ErbB-2 (TOB1) protein (Fig. 3B). The last (amino acids 895-941) is 60% similar to Ataxin-2 (Fig. 3C). The N-terminal portion of A2D also contains 8 interrupted glutamines replacing the polyglutamine tract of Ataxin-2.
A heterogeneity at the 3Ј end of the cloned cDNAs, compared with A2D-A, generated proteins with different C-terminal domains (Fig. 2B). They contained internal deletions affecting the reading frame as follows: 4 clones that lacked nucleotides 3308 -3875 were designated as A2D-B (GenBank TM accession number AJ317971); one clone A2D-C lacked nucleotides 3308 -3457 (GenBank TM accession number AJ317972); 8 clones A2D-D lacked nucleotides 3253-3851 (GenBank TM accession number AJ317973); and 4 clones A2D-E lacked nucleotides 3253-3874 (GenBank TM accession number AJ317974). The predicted amino acid sequences of the C-terminal domains of these isoforms are indicated in Fig. 2A. A cDNA library prepared from purified human megakaryocytes was similarly heterogeneous (data not shown). We derived a probe S (Fig. 2C), from nucleotides 3302-3840, which discriminated between these transcripts in Northern blots, revealing only the longest 4.5-kb transcript corresponding to A2D-A.
Mapping of the human gene by fluorescence in situ hybridization with the A2D cDNA probe revealed two loci on chromosome 16p11 and 7p21 (data not shown). By screening a human genomic library with the same probe, we cloned part of the A2D gene, including exons corresponding to cDNA nucleotides 1100 -3700, and a pseudogene that had been located by others on chromosome 7p21 (GenBank TM accession number AC005014). 3 We did not isolate any related gene, so we presume that the A2D gene is unique and located on chromosome 16. Moreover, sequence analyses indicated that the different transcripts A2D-A to -E were issued from this single gene by alternative splicing. In addition, the A2D and ATAXIN-2 genes 3 E. Ryan and C. Wilson, unpublished results. have a similar exon-intron structures (37), suggesting that they came from a common ancestor (data not shown).
Tissue Distribution of A2D Transcripts-As described above, the A2D mRNAs are found in hematopoietic tissues, with the highest expression in thymus, lymph nodes, and spleen. A2D transcripts are also abundant in fetal tissues, particularly in the fetal kidney (Fig. 4). The only adult organs that contains abundant A2D mRNAs is the testis. Transcripts were also weakly detected in the pancreas and placenta. The 4.5-kb isoform was the major transcript observed in fetal tissues, spleen, and lymph nodes, whereas the shorter form was predominantly detected in lymphocytes. A2D transcripts were also highly expressed in most immortalized human and murine hematopoietic cell lines; these also contained additional mRNAs ϳ3.7 kb long.
A2D Proteins Associate with Mpl in Vivo-The DNA sequence of the M37 clone corresponding to the 3Ј-half of the A2D cDNAs predicted two overlapping reading frames (ORF1 and ORF2), one of 347 codons and the other of 408 codons. We raised rabbit polyclonal antisera against putative proteins from ORF1 in-frame with the Gal4 activation domain and ORF2 corresponding to the A2D reading frame to identify proteins encoded by this cDNA. These antibodies were tested both on in vitro translation products and on UT7 lysates. They showed that ORF1 in-frame with the Gal4 activation domain in the yeast vector pGAD was probably non-coding, because no protein was found in UT7 lysates (data not shown). By contrast, the antiserum raised against ORF2 specifically precipitated a major form of ϳ130-kDa from UT7 lysates that was not precipitated by preimmune serum. This serum was designated as anti-A2D antibody (Fig. 5A). The 130-kDa band probably includes several isoforms with different C-terminal domains. Two additional bands of 120 and 100 kDa were detected by immunoblotting (Fig. 5C). Ribosomal frameshifting, which is frequent in yeast (38), may have resulted in the translation of a portion of the protein encoded by ORF2 in fusion with the Gal4 activation domain. We therefore looked to see if the 130-kDa protein interacted with Mpl in vivo.
We used a coprecipitation approach in the UT7/Mpl cells that proliferate in response to GM-CSF or EPO and differentiate to a megakaryocytic phenotype in response to TPO. The UT7/Mpl cells were deprived of growth factor and stimulated (or not) with TPO at 37°C for 20 min. The tagged FLAG-Mpl receptor was precipitated with anti-FLAG antibody. Immunoblotting with anti-A2D antibodies revealed A2D proteins in the immunoprecipitate from cells deprived of growth factor (Fig. 6A). By contrast, A2D did not coprecipitate with Mpl upon TPO stimulation. Reprobing the membrane with anti-FLAG antibody showed that the same amounts of Mpl were precipitated from unstimulated and stimulated UT7/Mpl cells. Similar amounts of A2D proteins were detected in total lysates of unstimulated and stimulated cells. A control with parental UT7 cells that does not express Mpl confirmed the specificity of this result. Thus, A2D proteins are constitutively associated with inactive Mpl in vivo and dissociate after ligand stimulation.
In agreement with these results, immunofluorescence anal- FIG. 4. Northern blot analyses of A2D. Human multiple tissues Northern blot derived from hematopoietic, fetal, and adult organs (CLONTECH; each lane contains 1 g of poly(A) ϩ RNA), and a Northern blot of total RNA from megakaryocytic progenitors (CD41 ϩ ) was hybridized with the M37 1.5-kb insert probe. CD41 ϩ cells were cultured from umbilical cord blood cells as described previously (22). mRNA sizes (in kb) are indicated on the left. Northern blots of human and murine cell lines were hybridized with the same probe. Each lane contains 10 g of total RNA. The murine hematopoietic cell lines were NS1 (plasmocytoma), BW (T lymphoma), WEHI, J774 (monocytic cells), P815, Reds (mast origin), IW32, and Friend (Friend-induced erythroleukemia). The human hematopoietic cell lines were U937, Raji, and UT7.

FIG. 3. Alignment of the amino acid sequence of A2D and related proteins.
A, A2D-A amino acids (aa) 208 -351 are 65% homologous to Ataxin-2 amino acids 351-493 (GenBank TM accession number HSU70323). This region is also present in a putative protein of Drosophila (GenBank TM accession number AE003708). Identical residues are highlighted in dark shading, and similar amino acids are in light shading. B, A2D-A amino acids 651-667 are 100% similar to Ataxin-2 amino acids 904 -922. This region is also present in the protein TOB1 (Swiss-Prot accession number P50616). C, A2D amino acids 895-940 are 60% similar to Ataxin-2 amino acids 1172-1216. ysis of UT7/Mpl cells with anti-A2D antibodies revealed staining at the cell periphery (data not shown). In addition, subcellular protein fractionation of UT7/Mpl cells (Fig. 7) indicated that A2D proteins were mainly expressed in the light membranes and to a lesser extend in the heavy membranes corresponding to endoplasmic reticulum/Golgi fraction. Weak bands were also detected in the cytosolic and nuclear fractions. Fractions were identified by the presence of the caspase3 in the cytosol and of Mpl in the light membranes.
TPO Induces a Transient Tyrosine Phosphorylation of A2D That Is Dependent on the Mpl C-terminal Domain-Cytokines induce tyrosine phosphorylation of molecular targets involved in signal transduction. As A2D proteins are rich in tyrosine residues and associate with Mpl, we measured the capacity of TPO to stimulate the phosphorylation of A2D proteins. Starved UT7/Mpl cells were stimulated or not with TPO for various times, and their lysates were immunoprecipitated with anti-A2D or preimmune antibodies and then resolved by SDS-PAGE. Immunoblotting using anti-phosphotyrosine antibody (Fig. 8A) showed that a 130-kDa tyrosine-phosphorylated protein was specifically precipitated by anti-A2D antibodies after TPO stimulation. Stimulation also resulted in the detection of a nonspecific 110-kDa band by preimmune serum. The phosphorylation of the p130 in TPO-stimulated cells was detectable by 2 min and then decreased after 20 min. Reprobing with anti-A2D antibody demonstrated that this 130-kDa phosphorylated protein comigrated with A2D. We also detected other A2D-associated proteins whose tyrosines were phosphorylated; a 40-kDa protein (Fig. 8A), which has not yet been identified, was strongly and stably phosphorylated after TPO stimulation; and phosphorylated Mpl was also detected when the membranes were overexposed (Fig. 8B). The overall pattern of phosphorylation after TPO stimulation is shown on the left panel of Fig. 8A.
We confirmed that the phosphorylated 130-kDa protein corresponded to A2D, using UT7/Mpl cells expressing a Myc-tagged A2D protein lacking its N-terminal domain (amino acids 1-314 deleted). This protein has a molecular mass of 100 kDa that is easily distinguished from the endogenous 130-kDa phosphorylated protein. Cells were stimulated or not with TPO and immunoprecipitated with anti-Myc antibody. Immunoblotting with anti-phosphotyrosine antibodies revealed a 100-kDa tyrosine-phosphorylated protein after TPO stimulation (Fig. 9). Reprobing with anti-Myc antibody demonstrated that the phosphorylated protein was the Myc-tagged A2D protein.
The yeast two-hybrid assay demonstrated that the M37/Mpl interaction is dependent on the Mpl C-terminal amino acids 98 -121. There was no coprecipitation of A2D with Mpl in UT7 cells expressing a truncated form of Mpl (⌬34) deleted from amino acids 70 -121 (data not shown). We used UT7 cells expressing Mpl-⌬34 to determine whether Mpl binding was required for TPO-induced A2D tyrosine phosphorylation. This cell line contained amounts of Mpl similar to those in UT7 cells were incubated or not with TPO for 20 min at 37°C and lysed in octyl ␤-D-thioglucopyranoside 30 mM. Anti-FLAG immunoprecipitates proteins (2 ϫ 10 7 cell equivalents) and whole cell lysates were analyzed by SDS-PAGE and anti-A2D immunoblotting. The membrane was reprobed with anti-FLAG antibody. B, association of A2D with EPO-R in UT7 cells. EPO-starved UT7 cells were incubated or not with 10 units/ml EPO for 10 min at 37°C and then lysed in 30 mM octyl ␤-D-thioglucopyranoside. Cell lysates (2 ϫ 10 7 cell eq) were immunoprecipitated (IP) with anti-EPO-R antiserum. Precipitated proteins and whole cell lysates were analyzed by SDS-PAGE and immunoblotted with anti-A2D antibodies. The same membrane was reprobed with anti-EPO-R antibodies. Reciprocally (bottom line), cell lysates were immunoprecipitated with anti-A2D antibodies or preimmune serum (pi). Immunoprecipitated and whole lysates were analyzed by SDS-PAGE and immunoblotting (IB) with anti-EPO-R. expressing wild type Mpl (WT) (Fig. 10A). Cells expressing WT-Mpl and truncated Mpl-⌬34 were stimulated with TPO for 5 min, and their lysates were immunoprecipitated with anti-A2D antibodies and loaded on SDS-PAGE. Immunoblotting with anti-phosphotyrosine (Fig. 10B) indicated that A2D was phosphorylated in cells expressing WT but not in UT7/Mpl-⌬34 cells. Reprobing the membrane with anti-A2D antibodies showed that the same amounts of A2D protein were precipitated in both cell lines. Because the tyrosines of Mpl-⌬34 are not phosphorylated after TPO stimulation and because Mpl-⌬34 lacks the tyrosine residues 112 and 117, the major sites of receptor phosphorylation (17), receptor activation was verified by assessing Jak2 activation. Immunoprecipitation with anti-Jak2 antibodies, followed by immunoblotting with anti-phosphotyrosine antibody, demonstrated similar phosphorylation of Jak2 in both cell lines upon stimulation (Fig. 10C). Hence, Mpl ⌬34 is functional; thus, the need for the C-terminal part of Mpl cytoplasmic domain for A2D recruitment and tyrosine phosphorylation agrees with the two-hybrid results involving this region for A2D binding.
A2D Proteins Associate with EPO-R in Vivo-As A2D proteins are mainly present in the hematopoietic cells of many lineages, we determined whether they also interacted with another receptor belonging to the cytokine receptor family. We investigated the ability of A2D to bind to the EPO-R, as Mpl and EPO-R have structural and functional similarities. We used UT7 cells for their high surface EPO-R expression and their proliferation in response to EPO. Cells were starved for 16 h and then stimulated or not with EPO. The EPO-R was immunoprecipitated, and the immunoprecipitates were immunoblotted with anti-A2D antibodies. A2D proteins were found associated with the EPO-R in unstimulated UT7 cells (Fig. 6B). The amounts of A2D coprecipitated with the EPO-R decreased markedly after 10 min of EPO stimulation, whereas the amounts of receptor remained constant, as shown by reprobing the membrane with anti-EPO-R antibodies. Reciprocally, a 66-kDa protein that reacted with anti-EPO-R antibodies was coprecipitated with A2D from unstimulated UT7 cells, and adding EPO decreased the amounts of EPO-R bound to A2D proteins (Fig. 6B). This indicates that A2D is constitutively associated with EPO-R and dissociates after ligand stimulation, as is the case with Mpl. Lysates were immunoprecipitated with anti-A2D or preimmune antibodies and resolved on SDS-PAGE. Immunoprecipitates were immunoblotted with anti-phosphotyrosine (PTYR) antibodies. The membrane was reprobed with anti-A2D antibodies. C, TPO stimulation of UT7/Mpl WT and UT7/Mpl⌬34 cell lines was verified by Jak2 activation. The lysates were incubated with anti-Jak2 antibodies, and the immunoprecipitates were immunoblotted with anti-phosphotyrosine antibodies. The membrane was reprobed with anti-Jak2 antibodies.

DISCUSSION
The yeast two-hybrid system has proved to be a powerful tool for identifying proteins directly associated with the receptors for tumor necrosis factor, IL-1, macrophage colony-stimulating factor, and FAS ligand. These are downstream molecular targets of receptor activation (39,40). By contrast, only a few reports have described the cloning of proteins interacting with a member of the cytokine receptor superfamily. A proteinarginine methyltransferase, PRMT1, was cloned by binding to the IFNAR1 chain of the interferon-␣/␤ receptor and shown to regulate STAT1 function (41,42). Similarly, a clone corresponding to RACK1, an Src kinase, and protein kinase Cassociated protein, was also isolated from a yeast two-hybrid screen with the common ␤-chain of the IL-5/IL-3/GM-CSF receptor (43). More recently, syntenin was shown to interact with the ␣-subunit of the IL-5 receptor and to be critical for activation of the transcription factor Sox4 by IL-5 (44).
We have now identified a single positive clone, M37, from a two-hybrid screen using the cytoplasmic domain of Mpl as bait. However, the M37 coding sequence was not in-frame with the Gal4 activation domain in the vector pGAD. Nonetheless, we have demonstrated that the encoded protein is constitutively associated with Mpl suggesting that a ribosomal frameshifting allowed the interaction in yeast. We used this clone to characterize a new family of 130-kDa proteins (A2D) that are expressed in immortalized cell lines and hematopoietic and fetal tissues. The predicted amino acid sequence of A2D contains 3 regions of homology with Ataxin-2, the product of the causative gene for the familial SCA2 disease (type 2 spinocerebellar ataxia) whose biological function is presently unknown. A2D and ATAXIN-2 genes have several common features; in addition to a similar genomic organization, their cDNAs both contain several long overlapping ORFs. One of them in A2D contained 347 codons for which we could not find any translation product. The main ORF of A2D starts at nucleotide 169 (ATG 1 ) and is preceded by an upstream sequence that contains 78% GC and no stop codon. The same observation has been described for the ATAXIN-2 gene (26). To determine the position of the transcription initiation site, we performed a 5Ј-rapid amplification of cDNA ends from UT7 RNA. This method did not reveal additional upstream sequences suggesting that our cDNAs contain the complete coding sequences. We characterized several cDNAs generated by alternative splicing of the single A2D gene encoding five A2D isoforms with different C-terminal domains. We also identified shorter proteins of 120 and 100 kDa (p120 and p100) that were not recognized by antibodies raised against the A2D N-terminal domain (data not shown) suggesting they could be issued from internal in-frame ATGs (ATG 183 and ATG 315 ). Indeed, GC boxes upstream of these internal ATGs could be a part of another promoter region. These data correlate well with the finding that an alternative physiological promoter has been found in a CpG island of the ATAXIN-2 gene within the first exon containing the CAG repeats. This shorter transcript may allow the synthesis of a protein using the internal in-frame ATG (45).
A2D binds to cytokine receptors Mpl and EPO-R. A2D is associated with these unoccupied receptors in vivo, and stimulation with TPO or EPO causes the rapid dissociation of A2D from the activated receptor. A similar constitutive binding and release following activation occurs for heterotrimeric G proteins that dissociate from receptors containing seven membrane-spanning domains (46) and also from the EPO-R (47), allowing activation of target effectors. Consistent with this, TPO stimulates the rapid and transient tyrosine phosphorylation of A2D in UT7/Mpl cells. This depends on Mpl binding, suggesting that A2D proteins function as signaling molecules.
We used the yeast two-hybrid assay to show that the Mpl cytoplasmic amino acids 98 -121 are needed for M37 binding. These data were confirmed by coimmunoprecipitation experiments in UT7 cells producing the wild type or a truncated form of Mpl. The Mpl C-terminal region contains tyrosine residue 112 that is both critical for receptor phosphorylation and needed for the TPO-dependent activation of STAT3, Shc, SHIP, and Gab proteins. This domain is involved in TPO-induced proliferation, because UT7 cells expressing the Mpl-⌬4 mutant (lacking amino acids 98 -121) do not grow in the presence of TPO (48). We were unable to stably transfect UT7/Mpl cells or the murine IL-3-dependent BaF3/Mpl cells with an A2D expression plasmid in which A2D was fused to a C-terminal Myc epitope (data not shown). The UT7/Mpl/A2D-Myc clones survived for 1 month and proliferated very slowly under GM-CSF, whereas BaF3/Mpl cells quickly stopped expressing exogenous A2D-Myc. The A2D amino acid sequence contains a short domain, conserved in the TOB protein, cloned as a protein associated with the c-ErbB2 receptor (49). The gene encoding TOB belongs to the anti-proliferative BTG1 gene family, and its product blocks the growth of NIH3T3 cells. Therefore, the toxicity of overproduced A2D for hematopoietic cell lines suggests that A2D proteins deliver growth inhibitory signals.
A2D proteins may also function as scaffold proteins. If so, their overproduction would sequester downstream signaling molecules involved in proliferation. Indeed, the sequence of A2D proteins contains two tandem repeats (between amino acids 677 and 693) of consensus sequences ((R/K)SXSXP) for binding to 14-3-3 proteins that interact with many key signaling molecules and regulate signal transduction events (50). A2D proteins also contain proline-rich regions with several PXXP motifs that could bind to SH3 domains. The tyrosine residue 349 of A2D in the sequence YXXL corresponds to the consensus sequence for binding to the Shc-SH2 domain. A2D proteins are also rich in serine residues, some of them in a potential target sequence for MAPK phosphorylation. However, our preliminary experiments have shown no association between A2D and Shc, MAPK, or 14-33 by coimmunoprecipitation or pull-down experiments. Likewise A2D had no effect on MAPK activation (data not shown). We still need to determine whether A2D proteins are adaptor molecules or initiate new signaling pathways.
In conclusion, we have identified a new family of proteins that are probably involved in TPO signaling. A single cell line produces proteins that have different C-terminal domains or lack their N-terminal region (p120 and p100) and are encoded by a single gene. A2D proteins are abundant in all hematopoietic lineages, indicating that they may be involved in other cytokine signaling systems. They also associate with the EPO-R, dissociate from the activated receptor, and are tyrosine-phosphorylated by EPO (data not shown). Finally, A2D proteins may be encoded by a gene belonging to a larger family because we detected a 160-kDa protein having epitopes similar to A2D proteins. The synthesis of this protein is greatly increased during megakaryocytic differentiation. 4