Small glutamine-rich tetratricopeptide repeat-containing protein is composed of three structural units with distinct functions

Archives of Biochemistry and Biophysics 435 (2005) 253–263 www.elsevier.com/locate/yabbi

Small glutamine-rich tetratricopeptide repeat-containing protein is composed of three structural units with distinct functions Shen-Ting Liou, Chung Wang¤ Institute of Molecular Biology, Academia Sinica, Taipei, Taiwan, ROC Received 17 September 2004, and in revised form 18 December 2004 Available online 5 January 2005

Abstract Previously, we identiWed the human small glutamine-rich tetratricopeptide repeat-containing protein (SGT) as a co-chaperone. The tetratricopeptide repeat (TPR) domain in SGT is responsible for interacting with Hsc70. In this study, we demonstrated that the TPR domain of SGT also interacted with Hsp90. Moreover, we investigated the functional signiWcance of regions of SGT outside the TPR domain. Evidently, the N-terminal domain of SGT is necessary and suYcient for its self-association; and, SGT may be a dimer elongated in shape. The C-terminal glutamine-rich region has the capacity to interact with short peptide segments composed of consecutive non-polar amino acids. The C-terminal fragment of SGT indeed plays a role in the association of SGT with in vitro translated rat type 1 glucose transporter, an integral membrane protein folded in a non-physiological state. Moreover, in the presence of SGT, the degradation of the transporter in reticulocyte lysates is inhibited. Taking together, SGT can be separated into three structural units with distinct functions.  2004 Elsevier Inc. All rights reserved. Keywords: Tetratricopeptide repeats; SGT; Molecular chaperone; Protein domain

Direct physical interaction of proteins with other molecules plays a pivotal role in biological processes. Frequently, proteins are composed of diVerent structural domains, and each domain possesses a speciWc function. For instance, the tetratricopeptide repeat (TPR)1 domain found in many proteins is believed to mediate protein–protein interaction [1,2]. Each TPR domain is composed of several 34-amino acid repeat motifs in tandem. While TPR motifs have a relatively low conservation in primary sequences, they generally form a

¤

Corresponding author. Fax: +8862 2782 6085. E-mail address: [email protected] (C. Wang). 1 Abbreviations used: TPR, tetratricopeptide repeat; Hsc70, 70-kDa heat shock cognate protein; Hsp90, 90-kDa heat shock protein; SGT, small glutamine-rich TPR-containing protein; Tpr1, tetratricopeptide repeat protein 1; CHIP, C-terminus of hsc70-interacting protein; Hop, hsc70/hsp90-organizing protein; GST, glutathione S-transferase; GLUT1, type 1 glucose transporter. 0003-9861/$ - see front matter  2004 Elsevier Inc. All rights reserved. doi:10.1016/j.abb.2004.12.020

conserved antiparallel pair of helices with equal length, and a minimum of three repeats appears to be required for forming a functional domain [2]. The structures of several TPR domains have been determined by X-ray crystallography [3–8]. From these X-ray structures, it is clear that adjacent TPR motifs are packed together in parallel arrangement generating a superhelical structure with a groove on the surface. This groove can be used for intramolecular interaction [5] or for interacting with other cellular proteins [4]. Considerable amount of evidence supports the view that the 70-kDa heat shock protein (Hsp70) and the 90kDa heat shock protein (Hsp90) are highly conserved molecular chaperones in the cytosol of eukaryotes [9– 11]. They participate in a number of cellular events, including protein folding and assembly of protein complexes. It is also clear that both Hsp90 and Hsp70 must work with other cellular proteins to exert their function. A group of proteins containing TPR domains have been

254

S.-T. Liou, C. Wang / Archives of Biochemistry and Biophysics 435 (2005) 253–263

shown to interact with Hsp90 [12] or with Hsc70 [13]. Hop is one of the best characterized members in this group. It serves as an adaptor protein and has the capacity to interact with both Hsp70 and Hsp90 [14]. Hop contains three distinctive but homologous TPR domains. The N-terminal three TPR motifs form a domain (TPR1) interacting only with the C-terminal PTIEEVD sequence of Hsp70, whereas the second TPR domain (TPR2A) interacts with the MEEVD sequence at the C-terminus of Hsp90 [4]. The interaction of these two TPR domains with Hsp70 and Hsp90 is speciWc. On the other hand, single TPR domain with three motifs within CHIP [15] and probably Tom70 [16] are capable of interacting non-discriminately with both Hsp70 and Hsp90. The small glutamine-rich TPR-containing protein (SGT) [17], also known as viral protein U (Vpu)-binding protein [18], was identiWed as a protein interacting with NS1 protein of parvovirus, and with Vpu of human immunodeWciency virus. Structurally, the TPR domain is in the middle of the protein. The glutamine-rich (Qrich) fragment (ca. 55 amino acids) is located at the Cterminal region of the protein, and there are 13 glutamines scattered in this fragment. SGT appears to be located in both cytosol and nucleus, and appears to be post-translationally modiWed [17]. Since NS1 has been shown to accumulate in nucleus and is involved in viral DNA replication, SGT was thought to play a role in regulating gene expression [19]. More recently, Winnefeld et al. [20] have shown that reduction in the level of SGT results in a mitotic arrest of cultured cells. The result implies that SGT plays a role in cell division. Moreover, SGT, Hsc70, and the cysteine string protein (a group III DnaJ protein) have been demonstrated to form trimeric protein complexes, indicating that they may function as a synaptic chaperone machine [21]. A brain-speciWc homologue of SGT (
SGT) with the same binding capacity as SGT was recently identiWed [22]. In addition, SGT has been shown to interact with growth hormone receptor [23]. In this interaction, the Wrst TPR motif is responsible for interacting with the ubiquitin-dependent endocytosis motif of the receptor. SGT therefore could have diVerent functions in cells depending on its subcellular location and the interacting partners. Evidence has also been presented suggesting that SGT is a co-chaperone. For instance, SGT and several molecular chaperones were co-immunoprecipitated with the intracellular
-amyloid peptide [24]. However, it was not clear whether or not SGT interacted directly with the
-amyloid peptide. We previously have demonstrated that the TPR domain in SGT interacts with Hsc70 [13], and that SGT may inhibit Hsc70-dependent in vitro refolding of luciferase [25]. Herein, we show that the TPR domain of SGT also interact with Hsp90. Since the TPR domain in SGT only comprises 35% of the protein mass, it is likely that the remaining 65% of SGT may

have additional activities via interacting with other proteins. We demonstrate in this report that the N-terminal fragment of SGT is responsible for the dimerization of the protein, and that the C-terminal Q-rich region has the capacity to interact with hydrophobic amino acid segments within polypeptides. Therefore, it raises the possibility that SGT itself may be a molecular chaperone.

Materials and methods Engineering the plasmids containing the C-terminal region of Hsp90 and Hsc70 To construct the plasmids for Hsp90, we ampliWed the C-terminal fragment of Hsp90 by polymerase chain reaction (PCR) using primers A and B (Table 1). Human liver cDNA library in PACT2 (Clontech) was used as template. Primer A is identical to the coding sequence of hsp90 (amino acids #621–625) with an extra NdeI site at the 5⬘-end, and primer B contains an additional XhoI site and is complementary to the coding sequence of the last Wve residues of Hsp90. The PCR products were ligated into pGEM-Teasy vector (Promega), resulting in plasmid pC90/Te. The insert was then isolated after treating the pC90/Te with restriction enzymes NdeI and EcoRI, and was ligated with the pGBKT7 vector (Clontech) previously digested with these two enzymes. (The EcoRI site on the vector was downstream of the 3⬘-end of the cDNA fragment.) In addition, to generate the mutant of Hsp90 without the C-terminus, we carried out PCR ampliWcation with primers A and C (Table 1), and we used pC90/Te as the template. Primer C is complementary to the coding sequence of Hsp90 but with the last four amino acids (EEVD) deleted and contains an XhoI site and a stop codon. The products were ligated into pGEM-Teasy vector resulting in pC90(EEVD)/Te, and the insert was subsequently cloned into pGBKT7 vector as described. These two pGBKT7 plasmids were used for the yeast two-hybrid assays. The EcoRI–XhoI cDNA fragments were excised from pC90/Te and pC90(EEVD)/Te, respectively. (Another EcoRI site was on the pGEM-Teasy vector located Table 1 Nucleotide sequences of the primers used in this study Primers

Nucleotide sequences

A B C D E F G H I

5⬘-CATATGATGGCCAAGAAACAC 5⬘-CTCGAGTTAATCCACTTCTTCCAT 5⬘-CTCGAGTTACATGCGAGAGGCATC 5⬘-CATATGTACAAAGCTGAGGATGAG 5⬘-GGATCCTTAATCGACCTCTTC 5⬘-CATATGAGCACTAGACTAAAGGAG 5⬘-CTCGAGTTATTTTAGTCTTTCATTACG 5⬘-CCCGGGACCAGCCCCTCGCAG 5⬘-AAGCTTCCCCTCTCTCAGGC

S.-T. Liou, C. Wang / Archives of Biochemistry and Biophysics 435 (2005) 253–263

upstream of the 5⬘-end of the cDNA fragment.) They were inserted into pGEX-4T-3 vector (Amersham–Pharmacia) using these two restriction sites. These plasmids were used to express the GST fusion proteins. To engineer the C-terminal fragment of Hsc70, we ampliWed the corresponding cDNA using primers D and E (Table 1). Primer D, with an NdeI site at the 5⬘-end, is identical to the coding sequence of Hsc70 (amino acids #525–530), and Primer E is complementary to the coding sequence (amino acids #643 to stop codon) with an additional BamHI site. Plasmid pCt-30/R [26] was used as the template. The products were inserted into the pET-15b vector (Novagen) using the NdeI and BamHI restriction sites, resulting in plasmid pCt-12/R. Subsequently, the XhoI–HindIII fragment containing the cDNA was isolated and ligated with pGEX-KG [27] with these two sites. The XhoI site (amino acids #542–543) previously was introduced into pCt-30/R [26], and the HindIII site was on the pET-15b vector. These plasmids obtained were used to express the GST fusion proteins. Co-immunoprecipitation of Hsp90 and Hsc70 with SGT Rat liver was dissected, minced with a pair of scissors, and homogenized in 10 volumes of homogenization buVer (0.22 M sorbitol, 75 mM sucrose, 10 mM Hepes– KOH, 1 mM EDTA, and 0.5 mM pefabloc, pH 7.4) and the homogenate was centrifuged at 1000g for 10 min. The supernatant was collected and centrifuged at 17,000g for 10 min. Subsequently, the supernatant was subjected to high speed centrifugation (150,000g) for an hour. The Wnal supernatant obtained (liver cytosol) was precipitated with (NH)2SO4 at 80% saturation. The precipitant was then dialyzed with ice-cold phosphatebuVered saline (PBS) for 4 h, clariWed by centrifugation, and incubated with anti-SGT antibodies that had been cross-linked to protein A–Sepharose. After washing with PBS, the proteins bound to the resin were eluted with 20 mM HCl, neutralized with Tris, and used for immunoblotting analysis with antibodies prepared against SGT, Hsp90, and Hsc70. Expression of the TPR domains Expression and puriWcation of the TPR domain for human SGT (hSGT; SGT in [22]) was described previously [13]. To engineer the expression plasmids for the TPR domain of Tpr1 [13,28], we ampliWed the corresponding cDNA using primers F and G (Table 1); and pACT2-TPR1 [13] was used as the template. Primer F is identical to the coding sequence (amino acids #116–121) with an extra NdeI site at the 5⬘-end, and primer G is complementary to the coding sequence (amino acids #244–239) with an additional stop codon and an XhoI site. The PCR products were ligated to the pGEM-T vector (Promega) and sequenced. The insert was then

255

excised, and cloned into the pET-15b vector using the NdeI and XhoI sites. The resulting plasmid was used for expressing the recombinant protein. In vitro pull-down with GST fusion proteins The pull-down assays were performed by following the methods of Liu et al. [13]. BrieXy, the C-terminal fragments of Hsp90 and Hsc70 fused to GST were puriWed, and the puriWed proteins were dialyzed against PBS. The proteins (20 g) were then incubated with glutathione–Sepharose (Pharmacia Biotech; 20 l) at 4 °C overnight. After washing with PBS to remove unbound protein, the resins were incubated with the TPR domains of SGT and Tpr1 (20 g) at room temperature for 2 h. At the end of incubation, the resin was Wrst washed with PBS, and the proteins bound were eluted with 50 mM glutathione (pH 7.5). The eluates were analyzed with SDS gel electrophoresis, and the gels were subsequently stained with Coomassie brilliant blue for visualization. Generation of plasmid containing SGT and its deletion mutants To engineer the SGT plasmids without its Wrst 87 residues, we replaced the SacI–XhoI fragment in phSGT(TPR)/15b [13] with the C-terminal SacI–XhoI fragment of phSGT/15b [13], resulting in plasmids pSGT(N)/15b. The cDNA insert in the plasmid phSGT(N)/15b was then excised with NdeI and XhoI digestion, ligated into the pAS2-1 vector (Clontech) using the same restriction sites, and resulted in plasmid pAS-SGT(N) for the yeast two-hybrid assays. To obtain the plasmid for the N-terminal fragment of SGT, plasmid phSGT/15b [13] was Wrst digested with SacI. The product was treated with T4 DNA polymerase to remove the 3⬘-overhang. It was then digested with XhoI, treated with Klenow enzyme in the presence of the four nucleotides, and was self-ligated. The resulting plasmid, Sgt(N)/15b, was used for protein expression. Here, nine extra amino acid residues (VERSMKRRY) were added at the C-terminus of the polypeptides. To generate the cDNA fragment for the C-terminal Qrich region, we performed PCR using plasmid hSGT/15b [13] as the template. The two primers were H and I (Table 1). Primer H is identical to the coding sequence of SGT (amino acids #299–301) except that a C to G substitution was made to introduce a SmaI site. Primer I is complementary to the 3⬘-non-coding sequence (»90 nucleotides downstream the stop codon) with an additional HindIII site. The PCR products were cloned into the pGEMTeasy vector and the insert was excised from the plasmid using restriction enzymes SmaI and PstI. (The PstI site is on the vector.) The isolated cDNA fragment was ligated with the pAS2-1 vector using the two restriction sites. To generate the correct reading frame for the Q-rich region,

256

S.-T. Liou, C. Wang / Archives of Biochemistry and Biophysics 435 (2005) 253–263

the plasmid obtained then was digested with EcoRI, treated with Klenow enzyme in the presence of the four nucleotides, and then self-ligated. The resulting plasmid, pAS-SGT(Q), was used for two-hybrid analysis. Subsequently, pAS-SGT(Q) was doubly digested with restriction enzymes NdeI and HindIII. The fragment containing the Q-rich region was isolated and ligated with the pET15b vector previously treated with the same enzymes. The resulting plasmid, SGT(Q)/15b, was used to express the Q-rich region of SGT in Escherichia coli.

equilibrated with buVer A (75 mM KCl, 25 mM Tris, pH 7.5). The mixtures were incubated at room temperature for 45 min with constant mixing, and the resin then was washed with buVer A. Subsequently, the protein bound was eluted with 100 mM imidazole (pH 7.5), and was displayed with SDS gel electrophoresis. The dried gels were Xuorographed and the amount of [35S]GLUT1 was quantiWed as described in the previous paragraph.

Protein cross-linking

The stability of GLUT1 translated in the reticulocyte lysate system was determined by a pulse-chase analysis. BrieXy, GLUT1 translations were carried out with or without SGT proteins in a total volume of 50 l. At the end of translation, non-radioactive methionine was added to the mixtures to a Wnal concentration of 20 mM, and the incubations were allowed to continue at 30 °C. After various time periods, an aliquot (2.5 l) of the sample was withdrawn and was boiled in SDS sample buVer. The samples were subsequently displayed with SDS gel electrophoresis for Xuorography. The amount of 35Slabeled radioactivity in GLUT1 was then quantiWed as described in the previous paragraph.

To cross-link the proteins, 3 g of either recombinant SGT or the N-terminal fragment of SGT in 100 l of PBS was mixed at room temperature with 1 l of glutaraldehyde solution (0.8 M). At various time points, the mixtures were quenched by adding 10 l of 1 M glycine (pH 9.0). Subsequently, the proteins were boiled in SDS sample buVer for gel electrophoresis, and then were subjected to immunoblotting with mouse antibodies prepared against the (his)6-tag (Serotec). In vitro translation of GLUT1 The cDNA of rat GLUT1 was isolated by screening a rat cDNA library (Clontech) using a GLUT1 cDNA fragment [29] as probe, and a full-length cDNA obtained then was inserted into the EcoRI site of the pBluescript vector (Stratagene). The plasmid was suitable for in vitro transcription of GLUT1 mRNA with T7 polymerase. Translation of GLUT1 was performed using TNT T7 coupled reticulocyte lysate system (Promega) by the procedures recommended by the manufacturer. In most cases, either SGT or its deletion mutants were added to the lysates at a Wnal concentration of 4 M, and luciferase cDNA (provided by the manufacturer) was used as a control. At the end of translation reaction, one-tenth of the products was used for SDS gel electrophoresis and for Xuorography with X-ray Wlms. To quantify the relative level of translation, the protein bands (either GLUT1 or luciferase) were excised from the dried gels and placed into scintillation vials. The proteins were eluted from the gel slices by incubating with 0.4 ml of 0.5% SDS in 25 mM Tris (pH 8.0) at room temperature overnight. The 35S-labeled radioactivity was then measured with a scintillation counter. For comparison, the amount of GLUT1 or luciferase without SGT proteins was taken as 100%.

Turn-over of the in vitro translated GLUT1

Other methods The yeast two-hybrid assays were performed as described previously [13]. Gel Wltration chromatography was carried out essentially as described previously [13]. BrieXy, 100 g of proteins in buVer B (75 mM KCl, 25 mM Tris, pH 7.0) was applied to a Superdex 200 HR column (1 £ 30 cm) in equilibrium with the same buVer. The column was eluted with buVer B at a Xow rate of 0.5 ml/min and the absorbance at 280 nm was monitored. The Stokes’ radius of puriWed SGT and its N-terminal fragment was determined by dynamic light scattering with a DynaPro-MS800 spectrophotometer (Protein Solutions). Here, the proteins eluted from Superdex 200 HR column were pooled together, concentrated with a centrifugal Wlter device (Amicon), and dialyzed with PBS. They were then used for dynamic light scattering measurements with protein concentrations ranging from 1 to 4 mg/ml. To prepare the antibodies, puriWed recombinant SGT and GST fused with the C-terminal fragment of Hsp90 were injected subcutaneously into rabbits. Sera collected after the fourth boost were used in this study. The Tricine–SDS gel electrophoresis was performed by following the procedures of Schagger and von Jagow [30].

Precipitation of SGT-associated GLUT1 Results To verify if GLUT1 might be associated with SGT, we carried out in vitro translation of GLUT1 in the presence of SGT (4 M). Aliquots of the translation products (45 l) were incubated with Talon resin (20 l; Clontech)

Previously, we demonstrated that the TPR domain of SGT interacted with Hsc70 [13]. Since other investigators have shown that domains with three TPR motifs in

S.-T. Liou, C. Wang / Archives of Biochemistry and Biophysics 435 (2005) 253–263

CHIP [15] and Tom70 [16] interact with both Hsp90 and Hsp70, we therefore investigated if the TPR domain of SGT may also interact with Hsp90. Tpr1 [28], another protein containing a TPR domain with three repeats known to interact with Hsc70 [13], also was included in this investigation. We Wrst utilized the yeast two-hybrid analysis to assess if SGT and Tpr1 might interact with the C-terminal region of Hsp90. The result showed that both SGT and Tpr1 were capable of interacting with Hsp90 (Fig. 1A). Since the EEVD motif at the C-terminus of Hsp90 is responsible for interacting with the TPR domains of Hop [4], we then determined if this motif in Hsp90 may also play a role in interacting with SGT and Tpr1. The deletion mutant therefore was prepared and used for the yeast two-hybrid analysis. Indeed, deleting the EEVD sequence abolished the interaction with SGT and Tpr1 (Fig. 1A), indicating that the C-terminus of Hsp90 was necessary for interacting with the TPR domains of SGT and Tpr1. We next investigated if SGT interacted with full-length Hsp90. Thus, the rat liver cytosol was prepared and immunoprecipitation was performed using antibodies against SGT. Subsequently, we examined if Hsp90 and Hsc70 might be co-precipitated with SGT. The results shown in Fig. 1B demonstrated that a fraction of Hsp90 and Hsc70 was co-immunoprecipitated with SGT, implying that SGT had the capacity to form complexes with both Hsp90 and Hsc70. We then considered if the C-terminal fragment of Hsp90 might be associated in vitro with the puriWed TPR domains of SGT and Tpr1. The TPR domains of Tpr1 and SGT Wrst were expressed in bacteria. The puriWed polypeptides were mixed with the C-terminal fragments of Hsp90 and Hsc70 fused to GST for pull-down assays. As expected, both TPR domains were brought down together with the C-terminal domain of Hsc70 (Fig. 1C, lanes 2 and 5), conWrming that they interacted with Hsc70. Moreover, these two TPR domains also interacted physically with the C-terminal fragment of Hsp90 (Fig. 1C, lanes 1 and 4). Under the same experimental conditions, the fusion protein without the C-terminal EEVD motif, however, failed to pull-down these two TPR domains (data not shown). Thus, it was clear that the TPR domain of SGT interacted non-discriminately with both Hsc70 and Hsp90. While the TPR domain of SGT had the capacity to interact with the C-terminus of Hsc70 and Hsp90, the question, however, remained unanswered as to whether or not the portion of SGT outside the TPR domain might interact with other proteins. To identify putative SGT-interacting proteins, we carried out yeast twohybrid screening with human SGT as a bait, and we obtained 35 independent positive clones derived from 20 diVerent cDNAs. Among these clones, SGT itself appeared four times, indicating that SGT could form oligomers. This conclusion was in agreement with the previous suggestion that SGT may be a dimer [17]. In

257

Fig. 1. Interaction of SGT and Tpr1 with Hsp90 and Hsc70. (A) The yeast two-hybrid assays: the C-terminal fragments of Hsp90 (from amino acids 621 to the C-terminus) and the corresponding mutant without the last four residues (EEVD) were in pGBKT7. Full-length SGT and Tpr1 were in pACT2; and the shaded regions represent the TPR motifs in the primary sequences of these proteins. The plasmids of interest were transformed into yeast strain Y190 for the two-hybrid assays. The
-galactosidase activity was determined by Wlter assays, and the blue Wlters are shown here. Lamin C was used as a negative control. (B) Co-immunoprecipitation of Hsp90 and Hsc70 with SGT: liver cytosol (10 mg protein/ml) in a volume of 200 l was incubated with anti-SGT antibodies previously coupled to protein A–Sepharose (200 l resin). The resin was washed with PBS, and the protein bound was eluted twice with 200 l acid each. The eluates were neutralized and were then subjected to SDS gel electrophoresis for Western blotting analysis with antibodies against SGT, Hsc70, and Hsp90. Lane 1 is 1/50 of the cytosol; lane 2 represents 1/5 of the Wnal 200 l PBS used for washing the resin before acid elution; and lanes 3 and 4 are 1/5 of the two acid eluates. (C) In vitro interaction of TPR domains with Hsp90 and Hsc70: the C-terminal fragments of Hsp90 and Hsc70 fused with GST were puriWed, and were allowed to bind to glutathione–Sepharose. The resin was mixed with the TPR domain of Tpr1 (lanes 1 and 2) and SGT (lanes 4 and 5). PuriWed GST incubated with Tpr1 (lane 3) and SGT (lane 6) were used as controls. After washing with PBS, the proteins bound to the resin were eluted with 50 mM of glutathione and analyzed by SDS gel electrophoresis. A Coomassie brilliant blue-stained gel is shown. The locations on the gel for the TPR domains of SGT (asterisk) and Tpr1 (circle) are indicated.

order to identify the domain in SGT responsible for selfassociation, we prepared deletion mutants of SGT and then performed a series of yeast two-hybrid analysis. The

258

S.-T. Liou, C. Wang / Archives of Biochemistry and Biophysics 435 (2005) 253–263

Fig. 2. Elution proWles of SGT and the N-terminal fragment of SGT. PuriWed SGT and its N-terminal fragment were subjected to gel Wltration analysis using a Superdex 200HR column as described in Materials and methods. The elution positions for aldolase (158 kDa) and BSA (67 kDa) were 12.2 and 14 ml, respectively.

SGT and the N-terminal fragment would be 143 and 93 kDa, respectively. Thus, the molecular sizes obtained from gel Wltration and dynamic light scattering were in good agreement. The results supported the view that there might be four subunits in each oligomer formed by SGT or its N-terminal fragment, assuming that the oligomers were spherical in shape. On the other hand, it is also possible that the shape of SGT might be signiWcantly deviated from spherical. Consequently, a trimer or a dimer could have a Stokes’ radius much larger than that predicted on the basis of a spherical shape. Therefore, we carried out cross-linking experiments to determine the number of subunits in each molecule. The results clearly showed that both SGT and its N-terminal fragments were likely to be dimers (Fig. 3). If indeed the molecules were dimers, they had to be non-spherical in shape. In any case, the N-terminal fragment of SGT was necessary and suYcient for its self-association.

results indicated that the N-terminal fragment of SGT was necessary for the self-association of SGT (data not shown). Identical result was recently obtained by Tobaben et al. [22]. The question then arose as to whether the N-terminal fragment alone was suYcient for the self-association. Therefore, we puriWed the polypeptides and used gel Wltration chromatography to determine the sizes of the protein molecules. As shown in Fig. 2, puriWed SGT eluted predominantly as a single peak, and that the Nterminal fragment of SGT also appeared as a single peak with an elution position close to that of SGT. Both SGT and its N-terminal fragment were eluted between those of aldolase (158 kDa) and BSA (67 kDa) with apparent molecular masses corresponding to tetramers instead of dimers. Using dynamic light scattering, we further veriWed if these assignments of the apparent sizes of the proteins were correct. The values obtained for the Stokes’ radius of SGT and the N-terminal fragment were 5 and 4.1 nm, respectively (Table 2). Assuming that the proteins were spherical, the predicted molecular masses for Table 2 Hydrodynamic radius of SGT and its N-terminal fragment

SGT (35.8K) N (16.0K)

Radius (nm)

M (kDa)

% Mass

5.0 § 0 4.1 § 0.1

143 § 1 93 § 4

99.8 § 0.1 100 § 0

The molecular weights of monomeric SGT and its N-terminal fragment are 35.8K and 16.0K, respectively. Stokes’ radius was determined by dynamic light scattering, and the average of four determinations was shown here. The molecular mass (M) was predicted with the assumption that the molecule was spherical. The peak distributions for SGT and its N-terminal fragment were ranging from 13.9 to 20.6% and 12.9 to 22.3%, respectively; and % mass represents the percent of total mass input in the peaks.

Fig. 3. Cross-linking of the polypeptides. Recombinant SGT and the Nterminal fragment of SGT were puriWed with immobilized metal aYnity chromatography using Ni2+–NTA resin. The puriWed proteins were cross-linked with glutaraldehyde for various time intervals as indicated. The reactions were quenched with glycine, and the proteins were displayed with SDS gel electrophoresis for Western blotting analysis. The reaction of quencher added immediately after the addition of glutaraldehyde was taken as zero time point. The result of a blot reacting with antibodies against (his)6-tag is given here, and pre-stained molecular weight markers (Sigma; SDS 7B) were used. The apparent masses for the cross-linking products of the N-terminal fragment and SGT were 34 and 100 kDa, respectively. The data support that the molecules are dimers, although the apparent mass of the SGT product is signiWcantly larger than that calculated for a dimer. This discrepancy may be resulted from cross-linking and reaction with glutaraldehyde.

S.-T. Liou, C. Wang / Archives of Biochemistry and Biophysics 435 (2005) 253–263

259

Besides SGT itself, most of other SGT-interacting clones obtained from the yeast two-hybrid screening were either prematurely terminated polypeptides or proteins with incorrect open reading frames. In particular, 12 out of the 35 positive clones were derived from eight diVerent cDNAs ligated to the activation domain of GAL4 (GAL4AD) in opposite orientation. Once translated, the poly(T), the complementary of poly(A) tail of mRNA, in these clones encodes polyphenylalanine. These clones thus produced fusion proteins containing a peptide segment with consecutive phenylalanine residues located at the C-terminus of GAL4AD. To determine if SGT might interact with these peptides, we took advantage that the poly(T) in certain clones can easily be separated from the rest of the cNDAs by restriction enzyme digestion. Consequently, one could obtain plasmid essentially containing poly(T) fused with the GAL4AD. One of such clones was used for yeast two-hybrid assays. From the results shown in Fig. 4, SGT indeed interacted with the peptide containing polyphenylalanine. We then identiWed the region of SGT responsible for this interaction. Various deletion mutants of SGT in the pAS2-1 vector therefore were engineered and utilized for the yeast two-hybrid analysis. The results indicated that the C-terminal Q-rich region in SGT was necessary and suYcient for interacting with the poly(phe)-containing segment (Fig. 4). Examination of the amino acid sequences of the rest of the SGT-interacting clones revealed that most of them also contained segments of six or more consecutive non-polar residues (data not shown). However, for a protein in native state, the hydrophobic segments frequently are buried. Conceivably, SGT might interact with the non-polar segment in proteins that were not in their native structure. Therefore, we next considered

whether or not SGT might interact with rat type 1 glucose transporter (GLUT1) translated in reticulocyte lysates. Rat GLUT1 contains 10 segments with six or more consecutive non-polar residues [31]. Moreover, when GLUT1 is translated in reticulocyte lysate system, its transmembrane regions would be exposed and resulted in non-physiological folding. We puriWed from bacteria the recombinant SGT with (his)6-tag at the Nterminus (Fig. 5, lane 1). We then performed in vitro translation of GLUT1 in the presence of SGT. Interestingly, addition of SGT in the translation mixtures brought about a twofold increase in the production of GLUT1 (Fig. 6). We next used the in vitro translated GLUT1 to determine whether it might be associated with SGT. Thus, the translation products with or without SGT were mixed with Talon resin that has the capacity to bind the (his)6-tag containing proteins. The protein bound to the resin was subsequently eluted for further analysis. As shown in Fig. 7A (lane 2) and Fig. 7B, 40% of the total GLUT1 input were pull-down if SGT was present in the translation mixtures, suggesting that GLUT1 might be associated with SGT.

Fig. 4. Interaction of SGT with A polyphenylalanine-containing peptide. Human SGT and its deletion mutants as indicated were in pAS21. The interacting partner in pACT2 was a cDNA fragment containing predominantly poly(A) sequence in opposite orientation. Once translated, it produces amino acids FFFFFLFIQHLL fused at the C-terminus of GAL4 activation domain. The plasmids were transformed into yeast strain Y190 for the two-hybrid assays. The activities of
-galactosidase were determined by color assays, and the blue Wlters are shown here.

Fig. 5. PuriWed recombinant polypeptides. The recombinant SGT (lane 1), SGT without the C-terminal region, SGT(Q) (lane 2), the TPR domain (lane 3), and the Q-rich fragment (lane 4) were expressed in bacteria, and were puriWed using Ni2+–NTA resin. The puriWed proteins were subjected to Tricine–SDS gel electrophoresis. A Coomassie blue-stained gel is shown. Molecular weight markers are carbonic anhydrase (29,000), soybean trypsin inhibitor (20,000), horse heart myoglobin (17,000), and fragments of myoglobin after CNBr cleavage (10,600; 8200; and 6200), respectively.

260

S.-T. Liou, C. Wang / Archives of Biochemistry and Biophysics 435 (2005) 253–263

Fig. 6. The eVect of SGT on translation of GLUT1. (A) In vitro translation of GLUT1 and luciferase: translation in vitro of GLUT1 and luciferase was carried out as described in Materials and methods. Lane 1 was without the addition of any SGT protein. Lanes 2, 3, 4, and 5 were with SGT, SGT(Q), TPR domain, and Q-rich fragment, respectively. (B) QuantiWcation of the relative level of translation: the 35Slabeled radioactivity in each protein band was quantiWed. The amount of radioactivity in the translation products without SGT protein was taken as 100% and the relative level of translation was calculated accordingly. The average § SD of four separate translations is shown.

We then asked if the Q-rich region in SGT might be important for this interaction. We Wrst puriWed SGT without the Q-region, SGT(Q), and its C-terminal Qrich fragment (Fig. 5, lanes 2 and 4). Addition of SGT(Q) in reticulocyte lysates also resulted in a twofold increase in GLUT1 translation. On the other hand, the C-terminal Q-rich fragment of SGT brought about a 70% reduction in GLTU1 translation (Fig. 6). However, these were not general phenomena, since SGT and its fragments did not have such an eVect on the translation of luciferase (Fig. 6). In any case, we assessed if GLUT1 might be associated with these SGT fragments. The results shown in Fig. 7 demonstrated that a fraction of GLUT1 did associate with SGT(Q), but the level of association was reduced to one-fourth of that with SGT (Figs. 7A, lane 3 and B). Thus, the Q-rich fragment in SGT appeared to play a role on the association with GLUT1. However, this fragment by itself did not bring down any GLUT1 (Figs. 7A, lane 5 and B), despite the

Fig. 7. Binding of the in vitro translated GLUT1 to SGT. (A) Association of GLUT1 with SGT: translation of GLUT1 was performed in the absence or presence of SGT proteins, and 90% of the products were mixed with Talon resin. The proteins bound to the resin then were eluted with 100 mM imidazole, analyzed by SDS gel electrophoresis, and followed by Xuorography. Lane 1 was without SGT proteins; lanes 2, 3, 4, and 5 were with SGT, SGT(Q), TPR domain, and Q-rich fragment, respectively. (B) QuantiWcation of the amount GLUT1 bound to the Talon resin: in each case, the amount of GLUT1 bound to the resin was quantiWed, and it was divided by the total amount GLUT1 loaded to the resin. No GLUT1 was bound to the resin in the absence of SGT proteins.

fact that the Q-rich region was suYcient for interacting with the hydrophobic peptide segments on the yeast two-hybrid assay (Fig. 4). One possible explanation was that the aYnity of the Q-rich region for GLUT1 was relatively weak and the resulting complexes were dissociated during the pull-down processes. However, the newly synthesized GLUT1 may bind to Hsc70 that is known to exist in reticulocyte lysates. Since Hsc70 has the capacity to interact with the TPR domain of SGT, it raised the possibility that a fraction of GLUT1 was associated with the SGT proteins through Hsc70. If so, the TPR domain alone would produce a similar, if not identical, result because it is responsible for interacting with Hsc70. Therefore, we next puriWed the TPR domain (Fig. 5, lane 3), performed in vitro translation (Fig. 6), and quantiWed the association of GLUT1 with the TPR domain. As shown in Fig. 7A (lane 4), a fraction of GLUT1 was pull-down by the TPR domain, but the amount of the protein appeared greatly reduced. Nevertheless, this apparent low level of association might be, at least partially, originated from lower amount of GLUT1 was used in this experiment (see Fig. 6). Indeed, after quantiWcation, it became evident that identical level of the total GLUT1 input (ca. 10%) was associated with both SGT(Q) and the TPR

S.-T. Liou, C. Wang / Archives of Biochemistry and Biophysics 435 (2005) 253–263

domain (Fig. 7B). The association of GLUT1 with the TPR domain and SGT(Q) conceivably was mediated by Hsc70. However, since the amount of GLUT1 associated with SGT was four times of those associated with SGT(Q) or with the TPR domain (Fig. 7B), it is likely that majority of the GLUT1 might form complexes with SGT through direct physical interaction. And, the Q-rich region is likely to be involved in this interaction. The last question considered in this study was why SGT brought about an apparent increase in GLUT1 translation (Fig. 6). One of the possible explanations was that SGT might aVect the stability of GLUT1 translated in reticulocyte lysate system. Therefore, we performed pulse-chase experiments and examined the degradation of GLUT1 as a function of time. As shown in Figs. 8A and B, in the absence of SGT, the level of [35S]GLUT1 decreased monotonically with time and the half-life was about 60 min. However, in the presence of SGT, the amount of [35S]GLUT1 remained largely unchanged (Figs. 8A and B). These results supported the notion

Fig. 8. Stability of GLUT1 in reticulocyte lysates. (A) Pulse-chase of GLUT1 with or without SGT: GLUT1 was translated in the absence or presence of SGT using the reticulocyte lysate system with [35S]methionine, and was chased by adding an excessive amount of non-radioactive methionine. At various time points indicated, aliquots were withdrawn for SDS gel electrophoresis, and the gels were dried for Xuorography with X-ray Wlms. A set of Xuorograms is shown here. (B) The relative amount of [35S]GLUT1 as a function of time: the GLUT1 pulse-chase experiments were performed in the presence of SGT and its fragments. The results were Wrst analyzed with SDS gel electrophoresis and Xuorography. Subsequently, the radioactivity of [35S]GLUT1 at various chasing time points was quantiWed, and the amount of GLUT1 at time D 0 was taken as 100%. The relative amount of [35S]GLUT1 at other time points was calculated accordingly and was plotted. The average of two determinations is shown.

261

that, in reticulocyte lysates, SGT might prevent GLUT1 from degradation. Consequently, the amount of GLUT1 was increased. We then investigated the eVect of various SGT fragments on the degradation of in vitro translated GLUT1. From Fig. 8B, SGT(Q) also inhibited the degradation of GLUT1. On the other hand, the Q-rich fragment appeared to facilitate the turn-over of GLUT1, and the half-life of GLUT1 was reduced to 40 min. Therefore, the variation in the GLUT1 levels shown in Fig. 6 correlated with the eVect of SGT proteins on GLUT1 stability (Fig. 8B).

Discussion Herein, we demonstrated that SGT is composed of three structural units with distinct activities. In the Wrst place, the N-terminal region is necessary and suYcient for the dimerization of SGT. This conclusion is in agreement with the previous report that this region is responsible for its self-association [22]. To determine the oligomeric state of SGT, we Wrst estimated the size of SGT using gel Wltration chromatography and dynamic light scattering. The results suggest that SGT would be a tetramer, if it is spherical in shape. On the other hand, the results of cross-linking experiments clearly indicate that SGT is a dimer. This discrepancy may be resulted from the assumption that the molecules were spherical. Since such an assumption could be incorrect, therefore, it is likely that SGT is a dimer. If so, it should be elongated in shape. The second conclusion of this study is that the TPR domain of SGT also interacts with Hsp90. It was previously shown that the TPR domain of SGT interacts with Hsc70 [13]. Our results and those of others on CHIP [15] and Tom70p [16] indicate that several TPR domains with three motifs may interact non-discriminately with both Hsp70 and Hsp90. The situation of Hop, in which its TPR domains interact speciWcally with Hsp70 or with Hsp90, might not be common. In contrary to our conclusion, Angeletti et al. [32], however, concluded that the TPR domain of SGT by itself was not suYcient for interacting with Hsp90. The reason for the disagreement appears to be originated from the deWnition of the TPR domain. While they deWned the TPR domain as three repeats only, we included an additional 15 amino acids at the C-terminus of the third TPR motif. These C-terminal residues could form an -helix, which may be essential for the stabilization of the TPR domain structure [2]. We previously showed that, in Hsc70, the tripeptide PTI preceding the EEVD motif is required for the interaction with SGT [25]. It is likely that the hydrophobic residue preceding the EEVD motif in Hsp90 is also needed for interacting with the TPR domain of SGT, although it needs to be proven. Interestingly, a dimeric SGT has two TPR domains per molecule. It therefore has the

262

S.-T. Liou, C. Wang / Archives of Biochemistry and Biophysics 435 (2005) 253–263

potential to interact simultaneously with Hsc70 and Hsp90. The situation could be similar to that of Hop, although SGT in principle could also interact with two Hsp90 or two Hsc70. In any case, the interplay among Hsc70, Hsp90, and SGT and the signiWcance of these interactions remain to be elucidated. The third conclusion of this study is that SGT interacts with polypeptides containing consecutive non-polar amino acid residues. This conclusion is consistent with the previous observation that SGT may be associated with
-amyloid peptides [24], which are known to contain peptides with consecutive hydrophobic residues [33]. More recently, Wang et al. [34] have shown that SGT interacted with the N-terminal signal peptide of myostatin, which also contains consecutive hydrophobic residues [35]. In our yeast two-hybrid screening, we also identiWed heptaglobin, a secretory protein with a hydrophobic signal peptide, as an SGT-interacting protein (data not shown). We demonstrate here that SGT interacts with GLUT1 translated in reticulocyte lysates. However, neither the N-terminal nor the C-terminal tails of GLUT1 aVects the interaction of GLUT1 with SGT (data not shown). While it remains unproven, the result suggests that SGT might interact with the trans-membrane segments of GLUT1. Nevertheless, if GLUT1 was in native state, these segments would be hidden from interacting with SGT, since they are buried in the plasma membrane. The results support the view that only the non-native GLUT1 interacts with SGT. In agreement with this notion, the
-amyloid peptides are non-native. Similarly, the signal peptide of myostatin is not included in the well folded native protein. However, with or without SGT, a similar amount of GLUT1 in reticulocyte lysates (ca. 70%) was in the pellets after centrifugation (data not shown), suggesting that binding of SGT does not aVect the aggregation of non-native GLUT1. Nevertheless, our results indicate that SGT recognizes unfolded hydrophobic segments with broad sequence speciWcity. Thus, SGT might possess chaperoning capability, albeit this hypothesis needs to be proven. It is also intriguing that SGT appears to increase the level of the in vitro translated GLUT1. However, in the presence of microsomes, such an eVect was not observed (data not shown). In this case, GLUT1 is co-translationally inserted into the membrane and folded into native conformation. Moreover, our pulse-chase analysis suggests that the apparent increase in GLUT1 translation by SGT was due to a decrease in GLUT1 degradation, although it is not entirely ruled out that SGT might also aVect the translation of GLUT1 in the reticulocyte lysate system. It is unclear how SGT inhibits non-physiologically folded GLUT1 from degradation. Evidently, Q-rich fragment of SGT was not necessary for this protection, since SGT(Q) exerted a similar eVect. It raises the possibility that SGT and/or SGT(Q) might interact with the protein degradation machinery in the

reticulocyte lysate system to prevent non-native GLUT1 from being degraded. Further studies are needed to verify this hypothesis. In summary, SGT is a dimer that is non-spherical in shape. The TPR domain of SGT interacts with both Hsc70 and Hsp90. Moreover, the C-terminal Q-rich region of SGT has the capacity to interact with polypeptides containing consecutive non-polar amino acid residues. Indeed, SGT interacted with GLUT1 in its non-native state. SGT thus has the potential to be a molecular chaperone.

Acknowledgments We thank Yun-Yee Chang for the yeast two-hybrid screening and Shin-Jen Wu for assistance. We are also grateful to Dr. Hsou-min Li for comments on the manuscript. The work is supported by grants from National Science Council (NSC-91-2311-B-001-108) and from Academia Sinica.

References [1] J.R. Lamb, S. Tugendreich, P. Hieter, Trends Biochem. Sci. 20 (1995) 257–259. [2] L.D. D’Andrea, L. Regan, Trends Biochem. Sci. 28 (2003) 655– 662. [3] A.K. Das, P.T.W. Cohen, D. Barford, EMBO J. 17 (1998) 1192–1199. [4] C. ScheuXer, A. Brinker, G. Bourenkov, S. Pegoraro, L. Moroder, H. Bartunik, F.U. Hartl, I. MoareW, Cell 101 (2000) 199–210. [5] K. Lapouge, S.J.M. Smith, P.A. Walker, S.J. Gamblin, S.J. Smerdon, K. Rittinger, Mol. Cell 6 (2000) 899–907. [6] G.J. Gatto Jr., B.V. Geisbrecht, S.J. Gould, J.M. Berg, Nat. Struct. Biol. 7 (2000) 1091–1095. [7] P. Taylor, J. Dornan, A. Carrello, R.F. Minchin, T. Ratajczak, M.D. Wilkinshaw, Structure (London) 9 (2001) 431–438. [8] C.R. Sinars, J. Cheung-Flynn, R.A. Rimerman, J.G. Scammell, D.F. Smith, J. Clardy, Proc. Natl. Acad. Sci. USA 100 (2003) 868–873. [9] B. Bukau, A.L. Horwich, Cell 92 (1998) 351–366. [10] A.J. Caplan, Trends Cell Biol. 9 (1999) 262–268. [11] J.C. Young, I. MoareW, F.U. Hartl, J. Cell Biol. 154 (2001) 267– 274. [12] J.C. Young, W.M.J. Obermann, F.U. Hartl, J. Biol. Chem. 273 (1998) 18007–18010. [13] F.-H. Liu, S.-J. Wu, S.-M. Hu, C.D. Hsiao, C. Wang, J. Biol. Chem. 274 (1999) 34425–34432. [14] S. Chen, D.F. Smith, J. Biol. Chem. 273 (1998) 35194–35200. [15] P. Connell, C.A. Ballinger, J. Jiang, Y. Wu, L.J. Thompson, J. Hofeld, C. Patterson, Nat. Cell Biol. 3 (2001) 93–96. [16] J.C. Young, N.J. Hoogenraad, F.U. Hartle, Cell 112 (2003) 41–50. [17] C. Cziepluch, E. Kordes, R. Poirey, A. Grewenig, J. Rommelaere, J.-C. Jauniaux, J. Virol. 72 (1998) 4149–4156. [18] M.A. Callahan, M.A. Handley, Y.H. Lee, K.J. Talbot, J.W. Harper, A.T. Panganiban, J. Virol. 72 (1988) 5189–5197. [19] C. Cziepluch, S. Lampel, A. Grewenig, C. Grund, P. Lichter, J. Rommelaere, J. Virol. 74 (2000) 4807–4815. [20] M. Winnefeld, J. Rommelaere, C. Cziepluch, Exp. Cell Res. 293 (2004) 43–57. [21] S. Tobaben, P. Thakur, R. Fernandez-Chacon, T.C. Sudhof, J. Rettig, B. Stahl, Neuron 31 (2001) 987–999.

S.-T. Liou, C. Wang / Archives of Biochemistry and Biophysics 435 (2005) 253–263 [22] S. Tobaben, F. Varoqueaux, N. Brose, B. Stahl, G. Meyer, J. Biol. Chem. 278 (2003) 38376–38383. [23] J.A. Schantl, M. Roza, A.P. DeJong, G.J. Strous, Biochem. J. 373 (2003) 855–863. [24] V. Fonte, V. Kapulkin, A. Taft, A. Fluet, D. Friedman, C.D. Link, Proc. Natl. Acad. Sci. USA 99 (2002) 9439–9444. [25] S.-J. Wu, F.-H. Liu, S.-M. Hu, C. Wang, Biochem. J. 359 (2001) 419–426. [26] S.-M. Hu, C. Wang, Arch. Biochem. Biophys. 332 (1996) 163–169. [27] K.-L. Guan, J.E. Dixon, Anal. Biochem. 192 (1991) 262–267. [28] A.E. Murthy, A. Bernards, D. Church, J. Wasmuth, J.F. Gusella, DNA Cell Biol. 15 (1996) 727–735.

263

[29] C. Wang, S.-M. Hu, Biochem. Biophys. Res. Commun. 177 (1991) 1095–1100. [30] H. Schagger, G. von Jagow, Anal. Biochem. 166 (1987) 368–379. [31] M.J. Birnbaum, H.C. Haspel, O.M. Rosen, Proc. Natl. Acad. Sci. USA 83 (1986) 5784–5788. [32] P.C. Angeletti, D. Walker, A.T. Panganiban, Cell Stress Chaperones 7 (2002) 258–268. [33] C.D. Link, Proc. Natl. Acad. Sci. USA 92 (1995) 9368–9372. [34] H. Wang, Q. Zhang, D. Zhu, Biochem. Biophys. Res. Commun. 311 (2003) 877–883. [35] A.C. McPherron, S.J. Lee, Proc. Natl. Acad. Sci. USA 94 (1997) 12457–12461.