Amino acid residues required for Gtr1p-Gtr2p complex formation and its interactions with the Ego1p-Ego3p complex and TORC1 components in yeast

Authors


  • Communicated by: Kunihiro Matsumoto

Abstract

The yeast Ras-like GTPases Gtr1p and Gtr2p form a heterodimer, are implicated in the regulation of TOR complex 1 (TORC1) and play pivotal roles in cell growth. Gtr1p and Gtr2p bind Ego1p and Ego3p, which are tethered to the endosomal and vacuolar membranes where TORC1 functions are regulated through a relay of amino acid signaling interactions. The mechanisms by which Gtr1p and Gtr2p activate TORC1 remain obscure. We probed the interactions of the Gtr1p-Gtr2p complex with the Ego1p-Ego3p complex and TORC1 subunits. Mutations in the region (179–220 a.a.) following the nucleotide-binding region of Gtr1p and Gtr2p abrogated their mutual interaction and resulted in a loss in function, suggesting that complex formation between Gtr1p and Gtr2p was indispensable for TORC1 function. A modified yeast two-hybrid assay showed that Gtr1p-Gtr2p complex formation is important for its interaction with the Ego1p-Ego3p complex. GTP-bound Gtr1p interacted with the region containing the HEAT repeats of Kog1p and the C-terminal region of Tco89p. The GTP-bound Gtr2p suppressed a Kog1p mutation. Our findings indicate that the interactions of the Gtr1p-Gtr2p complex with the Ego1p-Ego3p complex and TORC1 components Kog1p and Tco89p play a role in TORC1 function.

Introduction

Guanine nucleotide-binding proteins belong to a superfamily of regulatory GTP hydrolases composed of a large number of proteins, which include Ras-like small G proteins, heterotrimeric G protein subunits and the elongation factors, Tu and G. They have crucial roles in cell growth, proliferation, differentiation and macromolecular trafficking between different intracellular compartments (Exton 1998). In yeast, Ras-like small GTPase proteins include Ras1p, Ypt1p, Cdc42p, Arf1p, Gsp1p, Gtr1p and Gtr2p. These proteins act as molecular switches between an active GTP-bound form and an inactive GDP-bound form for various cellular metabolism and signaling pathways (Bourne et al. 1990; Takai et al. 2001). Gsp1p is a nuclear protein which functions in the nucleocytoplasmic transport of proteins and nucleic acids across the nuclear membrane (Nishimoto 2000; Sazer & Dasso 2000). Gsp1p is activated by Prp20p/Mtr1p, the guanine nucleotide-exchange factor (GEF) (Seki et al. 1996; Nishimoto 2000), and inactivated by Rna1p, the GTPase-activating factor (GAP) (Becker et al. 1995). The Ras-like small GTPases Gtr1p and Gtr2p genetically interact with Prp20p/Mtr1p. A cold-sensitive mutant of GTR1, gtr1-11, was identified as a suppressor of the temperature-sensitive mutant mtr1-2 (Nakashima et al. 1996). Gtr1p and Gtr2p have been suggested to negatively regulate Gsp1-GTPase activity through Yrb2p (Wang et al. 2005). Gtr1p also interacts with Nop8p, a nucleolar protein involved in ribosome biogenesis (Sekiguchi et al. 2004), and Rpc19p, a shared subunit of RNA polymerases I and III (Todaka et al. 2005), suggesting that Gtr1p is involved in RNA synthesis and ribosome maturation.

Unlike other Ras-like small G proteins, Gtr1p can form a homodimer and a heterodimer with Gtr2p (Nakashima et al. 1999). The Gtr1p-Gtr2p complex also forms the EGO complex (EGOC) with Ego1p/Meh1p and Ego3p/Slm4p and localizes to the vacuolar membrane (Dubouloz et al. 2005). Furthermore, the Gtr1p-Gtr2p complex is a regulator of TORC1, a homologue of mammalian mTORC1 (Binda et al. 2009; Valbuena et al. 2012). TORC1, localized in the vacuole and nucleus, is an important regulator of transcription, protein synthesis and autophagy in response to energy status and amino acid supply in yeast (De Virgilio & Loewith 2006; Valbuena et al. 2012; Betz & Hall 2013). TORC1 also plays roles in gene expression and epigenetic regulation (Kim & Chen 2000; Li et al. 2006; Chen et al. 2012). Gtr1p and Gtr2p are also suggested to be involved in epigenetic control of gene expression in yeast (Sekiguchi et al. 2008). Rapamycin binds Fpr1p and inhibits TORC1 activity (Heitman et al. 1991). Caffeine also inhibits TORC1 activity (Reinke et al. 2006). Mutations in genes involved in TORC1 activity could render the mutant yeast cells hypersensitive to rapamycin and caffeine.

Rag A/RRAG A and Rag B/RRAG B are the human homologues of Gtr1p (Schurmann et al. 1995; Hirose et al. 1998). RagC/RRAG C and RagD/RRAG D, the human homologues of Gtr2p, bind to RagA to form a heterodimer (Sekiguchi et al. 2001). The RagA-RagC complex regulates mTOR protein kinase activity through amino acid signaling by controlling its intracellular localization (Sancak et al. 2008). Three protein complexes, Ragulator, GATOR1 and GATOR2, are primarily reported to be involved in the regulation of the Rag complex. Ragulator, consisting of p18/LAMTOR1, p14/LAMTOR2, MP1/LAMTOR3, C7orf59/LAMTOR4 and HBXIP/LAMTOR5, controls the GEF activity of RagA/B, interacts with the heterodimer of the Rag complex (Rag GTPases) and mediates mTORC1 translocation to lysosomal membranes in response to amino acid signals (Sancak et al. 2010; Bar-Peled et al. 2012). GATOR1 (GAP for RagA/B) and GATOR2 (inhibitor of GATOR1) antagonistically transmit amino acid signals to RagA/B (Bar-Peled et al. 2013). The major components of these protein complexes are generally conserved in the budding yeast. For example, the functional counterpart of Ragulator in yeast is encoded by the EGO1 and EGO3 genes (Dubouloz et al. 2005). GATOR1 homologue, Iml1p, Npr2p and Npr3p act as a GAP for Gtr1p (Panchaud et al. 2013). Leucyl-tRNA synthetase functions as a GAP through GTP-to-GDP transition of RagD, leading to the control of mTORC1 activity (Han et al. 2012). In yeast, a leucyl-tRNA synthetase interacts with Gtr1p to sense the cellular leucine level and transmits its signal to TORC1 (Bonfils et al. 2012). Yeast Vam6p, a guanine nucleotide-exchange factor, acts as an activator of Gtr1p and Gtr2p (Binda et al. 2009). Conversely, it was recently reported that GTP charging of Rag GTPases is not altered by cellular amino acid conditions, suggesting additional amino acid-dependent pathway(s) (Oshiro et al. 2014).

In addition to mTOR, mTORC1 is comprised of several associated proteins: Raptor (the regulatory-associated protein of mTOR), mLst8, PRAS40 (proline-rich Akt substrate of 40 kDa) and DEPTOR (the DEP domain-containing mTOR-interacting protein) (Takahara & Maeda 2013). Yeast TORC1 contains Tor1p (Tor2p) and the associated proteins Kog1p, Lst8p and Tco89p (Reinke et al. 2004). Yeast Tor1p, Kog1 and Lst8p are homologues of mTOR, Raptor and mLst8, respectively. Through the interaction of these associated proteins with mTORC1, mTOR acts as a key modulator of cancer, diabetes and aging in higher eukaryotes including humans [see review in Johnson et al. (2013)].

Studies on the crystal structure of the Gtr1p-Gtr2p complex showed the structural roles of both the N-terminal nucleotide-binding domain and the interacting C-terminal domain for TORC1 activation and amino acid signaling (Gong et al. 2011). In this study, we investigated the interaction of the Gtr1p-Gtr2p complex with its binding partners Ego1p, Ego3p (Ragulator in yeast) and the TORC1 proteins, Tco89p and Kog1p, by biochemical and genetic studies to better understand the mechanisms through which the Gtr1p-Gtr2p complex activates TORC1.

Results

Involvement of the conserved C-terminal domain of Gtr1p and Gtr2p in complex formation

We reported previously that Gtr2p interacted with the N-terminal domain of Gtr1p (1–233 a. a.), but not with Gtr1p modified by the deletion of residues 201–233 in a yeast two-hybrid assay (Wang et al. 2009), indicating that the region important for heterodimer formation may encompass amino acids 201–233 of Gtr1p. As shown in the crystal structure of the Gtr1p-Gtr2p heterodimer (Gong et al. 2011) (Fig. 1A), Leu-207 (7th β-sheet) is in close contact with Leu-193 (7th α-helix) of Gtr1p, suggesting that both conserved Leu residues are important for establishing rigid protein structure via hydrophobic intramolecular interactions and for Gtr1p binding to Gtr2p. An amino acid sequence comparison showed that Gtr2p also has conserved Leu residues (Leu-193 and Leu-207) in the corresponding region (Fig. 1B), suggesting that they are required for Gtr1p-Gtr2p complex formation.

Figure 1.

Mutation at leucine 193 and 207 of Gtr2p hampers the interaction with Gtr1p. (A) Assignment of amino acid residues mutated in this study to the crystal structure of the Gtr1p-Gtr2p complex (PDB, 3R7W), which was visualized with PyMOL software (http://www.pymol.org). The gray and pale gray image is Gtr2p, and the cyan and pale cyan image is Gtr1p. The pale color region is the nucleotide-binding domain. Red sphere, magnesium ion; blue sticks, guanine nucleotides. The amino acid residues substituted in this report are indicated in the figure. (B) Sequence alignment of Gtr1p and Gtr2p showing similar (.) and identical (:) amino acid residues from the 179th to 220th residue. Substituted amino acid residues are underlined. *, amino acid residues were substituted by alanine. (C) The Y190 strain harboring pAS404-GTR1 was transformed with pACT2-GTR2, pACT2-gtr2L193G, pACT2-gtr2L207P, pACT2-gtr2L193GL207P or the pACT2 vector. The β-galactosidase activity of three independent yeast clones was measured as described in the Experimental procedures and shown with the associated standard deviation. The β-galactosidase filter assay was conducted and the results are shown on the right. The blue color represents a positive signal for the interaction. (D) A NBW5Δgtr1gtr2 strain expressing HA-tagged Gtr1p was transformed with either pRS314-MYC-GTR2 (lanes 1, 6), pRS314-MYC-gtr2L193G (lanes 2, 7), pRS314-MYC-gtr2L207P (lanes 3, 8), pRS314-MYC-gtr2L193GL207P (lanes 4, 9) or pRS314 (lanes 5, 10). Yeast cell lysates were used for immunoprecipitation with either an anti-HA-tag or anti-MYC-tag antibodies. Proteins bound to the antibodies were pulled down with protein G-beads and analyzed with Western blotting. The upper panels: the precipitates were run on SDS-polyacrylamide gels and transferred onto polyvinylidene difluoride membranes and blotted with anti-MYC antibody to detect MYC-tagged Gtr2p. The lower panels: the precipitates were blotted with anti-HA antibody to detect HA-tagged Gtr1p. (E) A NBW5Δgtr2 strain was transformed with pRS314-GTR2, pRS314-gtr2L193G, pRS314-gtr2L207P, pRS314-gtr2L193GL207P or pRS314. The transformants were spotted serially on YPD, YPD-caffeine (1 mg/mL) and YPD-rapamycin (20 ng/mL) plates.

To examine whether these conserved Leu residues are involved in Gtr1p-Gtr2p complex formation, mutations were introduced at Leu-193 and/or Leu-207 of Gtr2p and a yeast two-hybrid assay using Gtr1p in pAS404 as bait and various Gtr2p mutant proteins in pACT2 as preys was performed. The resulting strains expressing Gtr1p and either Gtr2-L207Pp or Gtr2-L193GL207Pp did not have β-galactosidase activity (Fig. 1C, rows 3, 4), suggesting that Gtr1p is unable to interact with Gtr2-L207Pp. Thus, Leu-207 of Gtr2p is important for Gtr1p-Gtr2p heterodimer formation. In contrast, both Gtr2p and Gtr2-L193Gp interacted with Gtr1p (Fig. 1C, rows 1, 2). The control, a strain expressing Gtr1p alone, did not show any β-galactosidase activity (Fig. 1C, row 5). To confirm the results, we carried out an in vivo pull-down assay. HA-tagged Gtr1p did not interact with MYC-tagged Gtr2-L207Pp and Gtr2-L193GL207Pp (Fig. 1D lanes 3, 4), whereas Gtr1p interacted with Gtr2-L193Gp although less efficiently (Fig. 1D, lane 2). MYC-tagged Gtr2-L207Pp and Gtr2-L193GL207Pp did not pull down HA-tagged Gtr1p (Fig. 1D, lanes 8, 9), and Gtr2-L193Gp pulled down Gtr1p inefficiently (Fig. 1D, lane 7), suggesting that Leu-193 might also be involved in Gtr1p-Gtr2p complex formation. As a control, Gtr2p pulled down Gtr1p efficiently and vice versa (Fig. 1D, lanes 1, 6). These results suggest that the amino acid residues Leu-193 and Leu-207 are important for Gtr1p-Gtr2p complex formation.

Deleting nonessential components specific to TORC1, but not TORC2, renders cells hypersensitive to caffeine and rapamycin (Reinke et al. 2006). Because Gtr1p and Gtr2p are regulators of TORC1 activity, loss-of-function mutations in these nonessential genes result in hypersensitivity to caffeine and rapamycin (Dubouloz et al. 2005). Thus, we used caffeine and rapamycin treatment to monitor the functions of Gtr1p and Gtr2p. The NBW5Δgtr2 strain was transformed with gtr2 mutant plasmids. gtr2L207P and gtr2L193GL207P did not complement Δgtr2 mutant on YPD-caffeine and YPD-rapamycin plates (Fig. 1E, middle and right panels), implying that complex formation between Gtr1p and Gtr2p is indispensable for their function. It should be noted that gtr2L193G complemented Δgtr2 mutant on an YPD-caffeine plate, although it did not complement Δgtr2 mutant completely on an YPD-rapamycin plate, suggesting that Leu-193 of Gtr2p might be important for the activation of TORC1. GTR2 complemented the Δgtr2 mutant, whereas the pRS314 vector alone did not complement the Δgtr2 mutant. Similarly, mutations were introduced into Gtr1p at Leu-193 and Leu-207. Gtr1-L193GL207Pp did not interact with Gtr2p (Fig. 2A, row 1) and did not complement the Δgtr1 mutant (Fig. 2B). As a positive control, GTR1 complemented the Δgtr1 mutant. As a negative control, the pRS316 vector alone did not complement the Δgtr1 mutant.

Figure 2.

Mutational analysis of Gtr1p indicates amino acid residues important for its function. (A) A yeast two-hybrid assay to test the interaction of Gtr1p with Gtr2p. The β-galactosidase assay was performed as described in Fig. 1C. (B) A NBW5Δgtr1 strain was transformed with pRS316-GTR1, pRS316-gtr1L193GL207P or pRS316. The transformants were serially diluted and spotted onto YPD and YPD-caffeine (1 mg/mL) plates. (C) A NBW5Δgtr1 strain was transformed with pAS404-GTR1, pAS404-gtr1L182AI183AP184A, pAS404-gtr1F196A, pAS404-gtr1I205AI206A, pAS404-gtr1F208A, pAS404-gtr1L214AV215AI216A or pAS404. The transformants were serially diluted and spotted onto YPD, YPD-caffeine (1 mg/mL) and YPD-rapamycin (20 ng/mL) plates.

Next, we investigated whether other amino acid residues might be important for Gtr1p-Gtr2p complex formation within 179–220 a.a. of Gtr1p (indicated in Fig. 1A and B), because a sequence comparison showed that this region of Gtr1p contains identical or similar amino acid residues to the corresponding region of Gtr2p (Fig. 1B). Mutations were introduced into the amino acid residues of Gtr1p shown by asterisks in Fig. 1B. The triple mutants Gtr1-L182AI183AP184Ap and Gtr1-L214AV215AI216Ap interacted less with Gtr2p (Fig. 2A, rows 4, 6) and lost their function (Fig. 2C), whereas Gtr1-F196Ap and Gtr1-F208Ap were functionally equivalent to Gtr1p and Gtr1-I205AI206Ap was less functional (Fig. 2C). Because Leu-182, Ile-183, Pro-184, Leu-214, Val-215 and Ile-216 of Gtr1p localize in the region adjacent to the nucleotide-binding domain (Fig. 1A), the 179–220 a.a. region might be important for maintaining the structural integrity of Gtr1p. The control, the pAS404 vector, did not suppress the mutation. These results suggest that the hydrophobic residues described above in the 179–220 a.a. region of Gtr1p were also important for the interaction with Gtr2p. Taken together, the results suggest that the interaction between Gtr1p and Gtr2p is indispensable for their function (Figs 1, 2).

Genetic interaction between GTR1 and GTR2 in their nucleotide-bound states

To investigate the genetic interaction between Gtr1p and Gtr2p when they are bound to guanine nucleotides, suppression analyses were performed using mutant strains gtr1S20L (gtr1-11: mutation at the 20th serine to leucine) and gtr2Q66L (mutation at the 66th glutamine to leucine). GTP-bound GTR1, gtr1Q65L, and GDP-bound GTR2, gtr2S23N, are active, whereas GDP-bound GTR1, gtr1S20N or gtr1S20L (gtr1-11), and GTP-bound GTR2, gtr2Q66L, are less active (Wang et al. 2009). The gtr1S20L mutation was significantly suppressed by over-expression of gtr2Q66L (Fig. 3A), suggesting that Gtr2-Q66Lp has an activity. As a negative control, gtr1S20L did not rescue the caffeine-sensitive phenotype of the gtr1S20L strain. We then examined whether the caffeine-sensitive phenotype of the gtr2Q66L strain was suppressed by the gtr1 mutants. The gtr2Q66L strain grew very slowly on the caffeine plate. As shown in Fig. 3B, gtr1Q65L (GTP-bound form) efficiently suppressed the gtr2Q66L mutation, whereas gtr1S20N (GDP-bound form) and vector alone did not suppress the gtr2Q66L mutation. These genetic results indicated that mutants of Gtr1p and Gtr2p were suppressed by the GTP-bound form mutant of Gtr2p and Gtr1p, respectively.

Figure 3.

Genetic interaction between GTR1 and GTR2 in their nucleotide-bound states. (A) A NBW5gtr1S20L (gtr1-11) strain was transformed with YEplac112-GTR2, YEplac112-gtr2S23N, YEplac112-gtr2Q66L, YEplac195-GTR1, YEplac195-gtr1S20L or YEplac112. The transformants were serially diluted and spotted onto YPD and YPD+caffeine (0.75 mg/mL) plates. (B) A NBW5Δgtr2 strain expressing Gtr2-Q66Lp was transformed with YEplac195-gtr1Q65L, YEplac195-gtr1S20N or YEplac195. The transformants were serially diluted and spotted onto YPD and YPD+caffeine (0.75 mg/mL) plates. (C) Yeast Y190 strains harboring Gal4 DNA-binding domain fusions with either gtr1S20L, gtr1Q65L, gtr2S23N or gtr2Q66L were streaked on nitrocellulose membranes, and a β-galactosidase filter assay was performed as shown in Fig. 1C.

Previously, we showed that Gtr2p, but not Gtr1p, exhibited transcription activating activity as a Gal4DBD-Gtr2p fusion protein in yeast cells (Sekiguchi et al. 2008). Herein, we found that Gtr2-Q66Lp, but not Gtr2-S23Np, exhibited significant transcription activating activity (Fig. 3C), suggesting that Gtr2-Q66Lp has a distinct activity which was not found in Gtr2-S23Np. Gtr1-Q65Lp and Gtr1-S20Lp did not show the transcription activating activity.

Interaction of the Gtr1p-Gtr2p complex with Ego1p and Ego3p

Gtr1p and Gtr2p interact with Ego1p and Ego3p to form EGOC (Dubouloz et al. 2005; Wang et al. 2009; Zhang et al. 2012), and thus, we investigated in detail how the Gtr1p-Gtr2p heterodimer interacts with the Ego1p-Ego3p heterodimer. First, we examined whether Gtr1p interacts with Ego1p, Ego3p or both (Fig. 4A). We confirmed that Gtr1p and Gtr2p interacted with Ego1p and Ego3p (Fig. 4A, rows 1–9) as previously reported (Binda et al. 2009). When we examined the Y190 strain using the yeast two-hybrid assay, the results indicated that Gtr1p interacted with both Ego1p and Ego3p (Fig. 4A, rows 1, 2) and Gtr2p interacted with both Ego1p and Ego3p (Fig. 4A, rows 10, 11). Gtr1p and Gtr2p strongly interacted with Ego1p, but weakly interacted with Ego3p (compare rows 1/2, 4/5, 7/8, 10/11). GTR1, GTR2 or EGO3 of the Y190 strain was deleted to test the protein interactions in the absence of each gene product as described in the Experimental Procedures. In the absence of Gtr2p, Gtr1p minimally interacted with Ego1p and Ego3p (Fig. 4A, rows 14, 15). Consistent with this, in the absence of Gtr1p, Gtr2-S23Np only minimally interacted with both Ego1p and Ego3p (Fig. 4A, rows 21, 22). In the absence of Ego3p, Gtr1p minimally interacted with Ego1p (Fig. 4A, row 17), whereas Gtr1p and Gtr2p were strongly interacting in Δgtr2, Δego3 and Δgtr1 (Fig. 4A, rows 13, 18, 20). These results suggest that both Gtr1p-Gtr2p and Ego1p-Ego3p complex formations are required for the efficient interaction of the Gtr1p-Gtr2p complex with the Ego1p-Ego3p complex. To confirm this, pull-down experiments were performed as shown in Fig. 4B. In the absence of Gtr1p, Gtr2p failed to interact with Ego3p (Fig. 4B, upper panel, lane 7), whereas in the presence of Gtr1p (wild-type strain), Gtr2p interacted with Ego3p (Fig. 4B, upper panel, lane 3). GST protein failed to interact with Ego3p in WT and Δgtr1 strains (Fig. 4B, upper panels, lanes 4, 8). The equivalent amount of MYC-Gtr2p was loaded onto the gel (Fig. 4B, upper panels, lanes 1, 2, 5, 6). These results suggested that lack of any of the four proteins, Gtr1p, Gtr2p, Ego1p and Ego3p, resulted in decreased interactions between the Gtr1p-Gtr2p and Ego1p-Ego3p complex proteins. Thus, it is likely that both Gtr1p-Gtr2p and Ego1p-Ego3p complex formations are required for efficient mutual interactions.

Figure 4.

Gtr1p-Gtr2p and Ego1p-Ego3p complex formation was important for their mutual interaction. (A) Two-hybrid yeast assays to test the interaction of Gtr1p and Gtr2-S23Np with Ego1p, Ego3p, Gtr1p and Gtr2p. Y190Δgtr1, Y190Δgtr2 and Y190Δego3 were constructed by deleting GTR1, GTR2 or EGO3 as described in the Experimental Procedures. Gtr2-S23Np was used for Gtr2p because Gtr2-S23Np is an active form and Gtr2p activates transcription and cannot be used (Sekiguchi et al. 2008). The β-galactosidase assay was performed and the results are shown in the right columns as described in Fig. 1C. (B) Ego3p failed to interact with Gtr2p in the absence of Gtr1p. A NBW5 strain expressing MYC-tagged Gtr2p was transformed with pEGKG-EGO3 (lanes 1, 3) and pEGKG (lanes 2, 4). A NBW5Δgtr1 strain expressing MYC-tagged Gtr2p was transformed with pEGKG-EGO3 (lanes 5, 7) and pEGKG (lanes 6, 8). Yeast cell lysates were prepared as described in 'Experimental procedures' and mixed with the glutathione Sepharose beads. Proteins bound to the beads were analyzed using Western blotting. The upper panels: Proteins bound to the beads were transferred onto the nylon membrane and blotted with the anti-MYC antibody. The lower panels: Proteins bound to the beads were transferred onto the nylon membrane and blotted with the anti-GST antibody. Arrowheads indicate the position of the proteins. (C) The GTP-bound Gtr2p (Gtr2-Q66Lp) and the GDP-bound Gtr2p (Gtr2-S23Np) were bound to Ego1p and Ego3p, and vice versa. Yeast strains expressing MYC-Gtr2p, MYC-Gtr2-Q66Lp or MYC-Gtr2-S23Np and T7-tagged Ego1p or T7-tagged Ego3p were constructed as described in Materials and Methods. Left two panels showed that Gtr2p proteins were pulled down with an anti-MYC antibody, followed by blotting with either an anti-T7 antibody or the anti-MYC antibody (lanes 1–6). Right two panels showed that either Ego1p or Ego3p protein was pulled down with the anti-T7 antibody, followed by blotting with either the anti-T7 antibody or the anti-MYC antibody (lanes 7–12).

The GTP-bound Gtr1p (Gtr1-Q65Lp) (Fig. 4A, rows 7, 8) and the GDP-bound Gtr1p (Gtr1-S20Np) (Fig. 4A, rows 4, 5) interacted with Ego1p and Ego3p in the yeast two-hybrid assay. Moreover, the GTP-bound Gtr2p (Gtr2-Q66Lp) and the GDP-bound Gtr2p (Gtr2-S23Np) bound to Ego1p and Ego3p, and vice versa in yeast extracts, showing that Gtr2p, Gtr2-Q66Lp and Gtr2-S23Np interacted with Ego1p (Fig. 4C, lanes 1–3, 7–9) and Ego3p (Fig. 4C, lanes 4–6, 10–12). The interaction between Gtr1/2 and Ego1/3 does not seem to be dependent on the guanine nucleotide-bound states.

Interaction of Gtr1p and Gtr2p with Kog1p in the region containing HEAT repeats

The Gtr1p-Gtr2p complex has been implicated in regulating TORC1 function in yeast. To understand how the Gtr1p-Gtr2p complex regulates TORC1 function, the Gtr1p-Gtr2p complex-interacting subunit(s) of TORC1 needs to be identified. Thus, a yeast two-hybrid assay was performed using GTR1 as bait and either TOR1, KOG1, LST8 or TCO89 as prey. The Gtr1p interacted with Kog1p and Tco89p (Figs 5A, 6C), but not Lst8p and Tor1p (results not shown). We confirmed a previous report that Kog1p and Raptor, a mammalian orthologue of Kog1p, interact with the Gtr1p and Rag complex, respectively (Sancak et al. 2008; Binda et al. 2009). A yeast two-hybrid assay showed that the C-terminal region (661–1156 a.a.) of Kog1p interacted with Gtr1p (Fig. 5A, rows 2, 3), but only minimally with Gtr2p (Fig. 5A, rows 7, 13). Even in the absence of Gtr2p, Kog1p (661–1156 a.a.) interacted with Gtr1p (Fig. 5A, row 9). Consistently, GST-Kog1-(661-1156)p pulled down recombinant Gtr1p and Gtr2p in vitro (Fig. 5B, lanes 2, 3), and Kog1p exhibited slightly higher affinity to Gtr2p than to Gtr1p (Fig. 5B, lane 1). We confirmed that similar amounts of Gtr1p and Gtr2p were present (results not shown). Gtr1p and Gtr2p could interact with Kog1p in the region containing the 3rd (777–814 a.a.) and 4th (888–925 a.a.) HEAT repeats (Fig. 5C), which form rod-like helical structures and might be involved in protein–protein interactions. To confirm this, various mutant constructs in the 3rd (777–814 a.a.) HEAT repeat were constructed (Fig. 5D). The yeast two-hybrid assay showed that the conserved amino acid residues in the 3rd HEAT repeat of Kog1p (Val797Arg798) were important for the interaction with Gtr1p. The GTP-bound Gtr1p (Gtr1-Q65Lp) (Fig. 5A, row 5), but not the GDP-bound Gtr1p (Gtr1-S20Np) (Fig. 5A, row 4), interacted with Kog1p. The GTP-bound Gtr2p (Gtr2-Q66Lp) and the GDP-bound Gtr2p (Gtr2-S23Np) were bound to Kog1p (Fig. 5B, lanes 4, 5). Even in the absence of Ego3p, Tor1p and Tco89p, Gtr1p interacted with Kog1-(661-1156)p (Fig. 5A, rows 10–12). We also examined the genetic interactions between GTR1, GTR2 and KOG1. The kog1ts mutant (YKK410) did not grow in the presence of caffeine or a low concentration of rapamycin. gtr2Q66L suppressed the kog1 mutation in the presence of caffeine (Fig. 5E, F). Gtr1-L193GL207Pp did not interact with Kog1p (Fig. 5A, row 8), suggesting that Leu-193 and Leu-207 of Gtr1p might also be required for the interaction of Gtr1p with Kog1p. The above results indicated that Gtr1p and Gtr2p interact directly with the region containing the HEAT repeats of Kog1p and could regulate TORC1 function.

Figure 5.

Interaction of the HEAT domain of Kog1p with the GTP-bound form of Gtr1p. (A) Determination of the interaction region of Kog1p with Gtr1p and Gtr2p by yeast two-hybrid assay. (B) Association of Kog1-(661-1156)p with Gtr1p in vitro. Cell lysates from E. coli BL21 strain transformed with pET28a-GTR1, pET28a-GTR2, pET28a-gtr2Q66L or pET28a-gtr2S23N were prepared and mixed with either GST-Kog1-(661-1156)p beads (lanes 1–5) or GST beads (lanes 6–10) as described above the panels. Beads were pulled down and subjected to SDS-PAGE followed by Western blotting with the anti-T7 antibody (Gtr1p, Gtr2p) (lanes 1–10) or the anti-GST antibody (lanes 11, 12). (C) Schematic representation of the motifs of Kog1p and the region of deletion clones used in Fig. 5A is shown. RNC (Raptor N-terminal conserved domain), HEAT and WD repeats are depicted. (D) Mutations in the HEAT repeat of Kog1p abolished the interaction to Gtr1p. Either amino acid residues PV (793, 794), VR (797, 798) or PVPEVR (793–798) of Kog1p was exchanged with EL to construct Kog1-(661-1156)p containing either kog1pv-el, kog1vr-el or Δpvvr-el mutations in pACT2 vectors. Comparison between yeast Kog1p and human raptor amino acid residues is shown with mutated amino acid residues (underlined). A yeast two-hybrid β-galactosidase filter assay was performed as shown in Fig. 1C. (E) Suppression analysis of a kog1 mutation by guanine nucleotide-bound mutants of GTR1 or GTR2. The kog1 mutant yeast strain (YYK410) was transformed with GTR1 or GTR2 mutant plasmids and serially diluted and spotted onto YPD and YPD+caffeine (1 mg/mL) plates. (F) Suppression of the kog1 mutation by the gtr2Q66L mutation. The kog1 mutant yeast strain was transformed with either the GTR2 or gtr2Q66L plasmid and serially diluted and spotted onto YPD, YPD+caffeine (1 mg/mL) and YPD+rapamycin (20 ng/mL) plates.

Figure 6.

Interaction of the C-terminal domain of Tco89p with the GTP-bound form of Gtr1p. (A) Determination of the interaction region of Tco89p with Gtr1p using a yeast two-hybrid assay. We found that Saccharomyces cerevisiae Tco89p has evolutionarily conserved regions at the N-terminus and C-terminus. These conserved regions are conN (the N-terminal conserved region) and conC (the C-terminal conserved region). The amino acid sequence of Saccharomyces cerevisiae Tco89p is shown, and the conN and conC regions are underlined. (B) Association of Tco89-(401-800)p with Gtr1p or Gtr2p in vitro. Cell lysates from E. coli BL21 strain transformed with pET28a-GTR1 or pET28a-GTR2 were prepared and mixed with GST-Tco89-(401-800)p beads as shown in Fig. 5B (lanes 1–5). Beads were pulled down and subjected to SDS-PAGE followed by Western blotting with the anti-T7 antibody (Gtr1p, Gtr2p) (upper panels, lanes 1–5) or the anti-GST antibody (GST-Tco89-(401–800)p) (lower panel, lane 6). (C) Yeast two-hybrid analysis to detect interactions between Gtr1p, Gtr2p and Tco89p as described in the Experimental Procedures. (D) The Δtco89 yeast strain (Y12072) was transformed with tco89 mutant plasmids and serially diluted and spotted onto either a YPD, YPD+rapamycin (20 ng/mL) or YPD+caffeine (1 mg/mL) plate.

Interaction of Gtr1p with the C-terminal region of Tco89p was dependent upon GTP binding

Next, we tested the interaction between Gtr1p and Tco89p (Fig. 6). The yeast two-hybrid assay showed that Gtr1p interacted with Tco89p, a subunit of TORC1 (Fig. 6C, row 1). Thus, we examined the significance of the interaction between Gtr1p and Tco89p. The interacting region of Tco89p with Gtr1p was determined by using deletion clones of Tco89p. Gtr1p interacted with the C-terminal half of Tco89p, Tco89-(401-800)p (Fig. 6A, upper panel and Fig. 6C, row 3). Consistently, Gtr1p interacted with Tco89-(401-800)p in vitro (Fig. 6B, lane 2), whereas Gtr2p interacted with Tco89-(401-800)p weakly in the presence of both Gtr1p and Gtr2p (Fig. 6B, lane 1), suggesting that Gtr1p could be the main binding partner of Tco89p. By comparing the amino acid sequence of yeast Tco89p with Tco89 homologues of other species, we found two evolutionarily conserved regions (conN and conC) as shown in Fig. 6A. Although the deletion of the conN region (31–47 a.a.) from Tco89p did not influence the interaction of Gtr1p with Tco89p (Fig. 6C, row 4), the deletion of the conC region (589–607 a.a.) from Tco89p resulted in the decreased interaction of Gtr1p with Tco89p (Fig. 6A, 6C, rows 5, 11). This suggests that the conC region of Tco89p might be involved in the interaction with Gtr1p. Tco89-(401-800)p had higher β-galactosidase activity than full-length Tco89p (Fig. 6C, compare rows 1, 3), indicating that the N-terminal region (1–400 a.a.) of Tco89p might affect the interaction of Tco89p with Gtr1p. In the absence of either Gtr2p or Ego3p, the interaction between Gtr1p and Tco89p was observed, suggesting that the interaction of Gtr1p with Tco89p could occur in the absence of Gtr2p or Ego3p (Fig. 6C, rows 6, 7). In the absence of Tor1p, the mode of interaction of Gtr1p with Tco89p was similar to that in WT (Fig. 6C, rows 8–13). GTP-bound Gtr1p (Gtr1-Q65Lp) interacted with Tco89p (Fig. 6C, row 14), whereas GDP-bound Gtr1p (Gtr1-S20Np) did not interact with Tco89p (Fig. 6C, row 15). The GDP-bound Gtr2p (Gtr2-S23Np) interacted very weakly with Tco89p (Fig. 6C, row 16). The GTP-bound Gtr2p (Gtr2-Q66Lp) and the GDP-bound Gtr2p (Gtr2-S23Np) were bound to Tco89p in vitro (Fig. 6B, lanes 4, 5).

We examined whether the conC region of Tco89p would be indispensable for Tco89p function by introducing tco89 mutant plasmids into the Δtco89 strain. Whereas TCO89 and tco89ΔconN complemented the Δtco89 mutation, tco89ΔconC failed to complement the Δtco89 mutation in the presence of rapamycin and caffeine (Fig. 6D). This suggests that the conC region of Tco89p might be essential for its function through its interaction with Gtr1p.

Discussion

The results described herein confirmed that Leu-193 and Leu-207 of Gtr1p and Gtr2p are important for Gtr1p-Gtr2p complex formation and function by detailed analysis of various point mutants of Gtr1p and Gtr2p. The GTP-bound form of Gtr2p (Q66L), which is regarded as a loss-of-function mutant, may also have a positive function. The GDP- and GTP-bound forms of Gtr2p were shown to efficiently interact with Ego1p and Ego3p in vivo. Both the GTP- and GDP-bound forms of Gtr2p interacted with Kog1p and Tco89p. The HEAT repeats domain of Kog1p and the C-terminal region of Tco89p were found to be important for the interactions with Gtr1p and Gtr2p.

We previously showed that the C-terminal region of Gtr1p is required for the interaction with Gtr2p (Wang et al. 2009). Mutational analysis of Gtr1p and Gtr2p indicated that the conserved residues Leu-193 and Leu-207 of Gtr1p and Gtr2p were important for complex formation. Gong et al. identified the intermolecular interaction regions to be α8, β9, β10 of Gtr1p and β7, α8, β9, β10 of Gtr2p (Gong et al. 2011). As shown in Fig. 1A, hydrophobic interactions between Leu-207 (7th β-sheet) (7β) and Leu-193 (7th α-helix) (7α) in both Gtr1p and Gtr2p might occur, helping to induce protein folding to form the Gtr1p-Gtr2p complex and interactions with other proteins. Consistently, we showed that point mutations in Leu-193 and Leu-207 caused loss of function (Figs 1E, 2B). Gtr2-L193Gp did not complement the rapamycin-sensitive phenotype of Δgtr2. This might be caused by inefficient interaction of Gtr2-L193Gp with Gtr1p. It is also possible that Leu-193 and Leu-207 of Gtr1p and Gtr2p might be important for the Gtr1p-Gtr2p interaction in the presence of other interacting proteins such as Ego1p and Ego3p. As a similar example, kinesin motor protein changes its conformation in the presence of microtubules and ATP (Rice et al. 1999).

It was unexpected that gtr2Q66L was able to complement gtr1S20L and kog1ts mutations. That is, the GTP-bound form of Gtr2p (gtr2Q66L), which is thought to be an inactive form of the GTPase, suppresses the caffeine sensitivity of gtr1S20L and kog1ts (Figs 3, 5). It should be noted that Gtr1-S20Lp possesses some activity because it suppressed the prp20ts mutation (Nakashima et al. 1999), and Kog1tsp also has some activity at a permissive temperature (Nakashima et al. 2008). To explain these phenotypes, we assumed the following: (i) Gtr2-Q66Lp may have weaker activity than Gtr2p, and over-expression of Gtr2-Q66Lp might yield sufficient activity to suppress the effects of these mutations; (ii) Gtr2-Q66Lp might use a suitable structure that enables its binding partners, Gtr1-S20Lp and/or Kog1tsp, to exhibit high activities via some unknown mechanism; (iii) Gtr2-Q66Lp might have a novel activity in the nucleus. Herein, we showed that Gtr2-Q66Lp had transcription activating activity (Fig. 3C). TORC1 regulates gene expression in yeast (Cardenas et al. 1999) and in nuclei (Li et al. 2006). Because Gtr2p is localized in the nucleus in Δgtr1 yeast cells (Nakashima et al. 1999) and has transcription activating activity (Sekiguchi et al. 2008), we suspect that regulation of gene expression in response to nutrients by TORC1 might involve Gtr2p.

The interaction of the Gtr1p-Gtr2p complex with the Ego1p-Ego3p complex could be important for TORC1 function because deletion of one of these proteins resulted in rapamycin hypersensitivity. To show this, we used a modified yeast two-hybrid assay system using Y190 strains that lack one of the complex proteins. This modified two-hybrid system allowed us to identify the interactions in the complex formation of both Gtr1p-Gtr2p and Ego1p-Ego3p as shown in Fig. 4. Deletion of one of these proteins abolished the interaction of these two complexes. Thus, the presence of all four proteins, Gtr1p, Gtr2p, Ego1p and Ego3p, was necessary to achieve maximum binding capacity. It is not fully understood why the Gtr1p-Gtr2p complex together with the Ego1p-Ego3p complex is required for TORC1 function in yeast. One possibility is that the Ego1p-Ego3p complex might activate the Gtr1p-Gtr2p complex. Activation and inactivation of the Gtr1p-Gtr2p complex are regulated by GTP-GDP exchange of both proteins in a reciprocal manner. Therefore, the Ego1p-Ego3p complex might act as a GEF (Ragulator) of Gtr1p, which activates the Gtr1p-Gtr2p complex for TORC1 activation (Fig. 7A). Another possibility is that the Gtr1p-Gtr2p heterodimer may be tightly tethered to the vacuole membrane by binding the Ego1p-Ego3p complex to form a tetramer to activate vacuolar-localized TORC1.

Figure 7.

Model for the interaction of Gtr1p and Gtr2p with Ego1p, Ego3p and Tco89p and Kog1p. (A) Proposed mode of interaction of the Gtr1p-Gtr2p complex with Ego1p and Ego3p. The Gtr1p-Gtr2p complex interacted with the Ego1p-Ego3p complex. Subtracting Ego1p from the four protein complex might cause the loss of interactions of Ego3p with the Gtr1p-Gtr2p complex. (B) Proposed mode of interaction of the Gtr1p-Gtr2p complex with the TORC1 components Kog1p, Tco89p and Tor1p. GTP-bound Gtr1p interacted with Gtr2p, Tco89p and Kog1p, which are associated with Tor1p. GDP-bound Gtr1p did not interact with Tco89p and Kog1p.

Kog1p is a yeast orthologue of mammalian Raptor, which is a component of mTORC1. Rag GTPases recruit mTORC1 to the surface of the lysosome through Raptor (Bar-Peled et al. 2012). The region containing the HEAT repeats of Kog1p might play an important role in complex formation between the Gtr1p-Gtr2p complex and Tor1p in a GTP-GDP-dependent manner (Fig. 7B). It is also intriguing that the interaction of the Gtr1p-Gtr2p complex with the region containing the HEAT repeats of Kog1p resembles the HEAT repeats of importin proteins that associate with RanGTPase (Conti et al. 2006). Because Gtr1p and Gtr2p are genetically related to the yeast Ran/Gsp1-GTPase cycle (Nakashima et al. 1999), they may share similar mechanisms of action for protein transport.

Tco89p was identified as a novel component of TORC1 in yeast and was localized to the region proximal to the plasma membrane, but unlike other TORC1 components, Tco89p can also be found at the vacuolar membrane (Reinke et al. 2004). We showed that Tco89p is strongly associated with Gtr1p in vitro, suggesting that the vacuolar membrane-localized Tco89p might be associated with the Gtr1p-Gtr2p complex through binding with the Ego1p-Ego3p complex. It should be noted that Tco89p might favor Gtr1p over Gtr2p (Fig. 6B). Conversely, Kog1p might favor Gtr2p over Gtr1p (Fig. 5B). A tight association of the Gtr1p-Gtr2p complex with TORC1 might be assured through the interaction of Gtr1p and Gtr2p to Tco89p and Kog1p, respectively.

Consistent with a previous report (Binda et al. 2009), our present results showed that GDP-bound Gtr1p did not bind Tco89p or Kog1p, whereas GTP-bound Gtr1p did bind Tco89p and Kog1p. Gtr1p might be involved in the localization of TORC1 by releasing Kog1p and/or Tco89p from the vacuole upon inactivation by a GAP such as Iml1p. In contrast, both GDP-bound and GTP-bound Gtr2p bound Tco89p and Kog1p. Taken together, it is likely that the nucleotide-bound form of Gtr1p could be a determinant for the interaction of the Gtr1p-Gtr2p complex to Tco89p and/or Kog1p (Fig. 7B).

Experimental procedures

Strains and media

Saccharomyces cerevisiae strains used in this study are listed in Table S1 in Supporting Information. The NN7-3B strain (Nakashima et al. 1999) and the YYK410 strain (Nakashima et al. 2008) were used for suppression analysis. NBW5, NBW5Δgtr1 and NBW5Δgtr2 strains were described previously (Nakashima et al. 1999). Yeast strains were grown in YPD (2% glucose, 2% peptone and 1% yeast extract), SD-Ura (2% glucose, 0.67% yeast nitrogen base without amino acids, supplemented with amino acids required for the growth of auxotrophic strains except for uracil). The yeast two-hybrid assay (Chien et al. 1991) was performed using the Saccharomyces cerevisiae Y190 strain to test protein interactions in vivo. Transformation of Saccharomyces cerevisiae was carried out by a lithium acetate method using DMSO (Hill et al. 1991). Gene-knockout strains were constructed by replacing the target gene with the kanamycin-resistant gene (Guldener et al. 1996). Briefly, we amplified the kanamycin gene with primers carrying the sequence tag of the target gene. The amplified DNA was introduced into the yeast strains using the lithium acetate–polyethylene glycol method (Ito et al. 1983). Kanamycin-resistant colonies were examined by polymerase chain reaction (PCR) if the target gene was deleted. To further confirm the target genes were deleted, each deletion strain was transformed with wild-type genes to examine whether transformed strains exhibited the wild-type phenotype. Similarly, yeast cell lines expressing T7-tagged-Ego1p or Ego3p were generated by introducing DNAs containing the C-terminal end of Ego1p or Ego3p, T7-tag and the kanamycin-resistant gene as described (Funakoshi & Hochstrasser 2009). The E. coli strains DH5α and BL21 were used for protein over-expression and purification.

Yeast two-hybrid assay

The Y190 strain was transformed with pAS404-conjugated genes, and the transformants were selected on a SD-Trp plate (2% agar, 2% glucose, 0.67% yeast nitrogen base without amino acids, supplemented with all essential amino acids except for tryptophan). Then, the transformants were transformed with pACT2-conjugated genes and were selected on a SD-Leu, -Trp plate (SD-L-T) (2% agar, 2% glucose, 0.67% yeast nitrogen base without amino acids, supplemented with all essential amino acids except for leucine and tryptophan). The isolated transformants were re-plated on the SD-L-T plate and incubated at 30 °C for 2–7 days. The β-galactosidase liquid and filter assays were performed using the chromogenic substrate o-nitrophenol-β-D-galactoside as described previously (Nakashima et al. 1999; Sekiguchi et al. 2001). One unit of β-D-galactosidase is defined as the amount of enzyme liberating 1 nM of o-nitrophenol in 1 min. A blue color representing a positive signal appeared within 30 min at 30 °C as shown in Fig. 1.

Construction of plasmids

The plasmids used in this study are listed in Table S2 in Supporting Information. Each construct was checked by sequencing using an ABI PRISM® 3100 Genetic Analyzer (Applied Biosystems, Foster City, CA, USA). The oligonucleotides used in this study are listed in Table S3 in Supporting Information.

Purification of proteins in E. coli and the in vitro binding assay using GST-fusion proteins

Escherichia coli DH5α individually transformed with pGEX-KG-kog1-(661-1156), pGEX-KG-tco89-(401-800) and pGEX-KG vector was grown in 100 mL of LB medium. Escherichia coli BL21 individually transformed with pET28a-GTR1 or pET28a-GTR2 vector was grown in 100 mL of LB medium. The culture was induced with 0.2 mm isopropyl β-thiogalactoside and was grown at 26 °C for 4 h. The cells were collected by centrifugation and resuspended in 30 mL of lysis buffer (50 mm Tris (pH 7.5), 150 mm NaCl, 2.5 mm MgCl2, 10% glycerol, 0.5% NP-40, 1 mm PMSF, 0.1 μg/mL aprotinin and 1 mM DTT) and sonicated for 5 min three times on ice (Sonicator™, Heat System-Ultrasonics Inc. Plainview, NY, USA), with a microtip, 40% cycle and output 4. The lysate was centrifuged at 10 000 × g for 30 min at 4 °C. The supernatants (100 μL of each sample) were mixed for 30 min at 4 °C and incubated with 30 μL of 50% slurry (v/v) glutathione Sepharose 4B (GE healthcare, Piscataway, NJ, USA) for 30 min at 4 °C with rotation. Proteins bound to glutathione Sepharose-4B beads were washed three times with 500 μl of the binding assay buffer (50 mm Tris-Cl (pH 7.4), 1 mm EDTA, 150 mm NaCl, 0.1% NP-40, 1 mm PMSF, 0.1 μg/mL aprotinin and 1 mm DTT) and then were subjected to SDS-PAGE, transferred onto a nylon filter and identified by Western blotting using antibodies.

Mouse antihemagglutinin (HA) (Cat. no. MMS-101P) was purchased from CRP, Inc. (Cumberland, VA, USA). Mouse anti-c-MYC (9E10) antibody (Cat. No. sc40) and anti-GST antibody (Cat. No. sc138) were purchased from Santa Cruz Biotechnology (Santa Cruz, CA, USA). Mouse anti-T7 tag antibody was purchased from Merck Biosciences Novagen (Darmstadt, Germany). Rabbit anti-T7 tag antibody was purchased from MBL Inc. (Nagoya, Japan). Horseradish peroxidase-conjugated goat anti-mouse IgG was purchased from GE Healthcare. Protein G–horseradish peroxidase conjugate was purchased from Prozyme Inc. (Hayward, CA, USA).

In vitro binding assay in yeast

The yeast cell lysate was prepared by suspending the cells in S buffer (50 mm potassium phosphate buffer, pH 6.5, 120 mm NaCl, 1 mm MgCl2, 0.1% Triton-X-100, 10% glycerol, 1 mm 2-mercaptoethanol, 0.2 mm PMSF) and then disrupting the yeast cells with glass beads by vortexing two times for 2 min. The resultant extract was diluted to 600 μL with S buffer. The anti-MYC, anti-HA or anti-T7 antibody was mixed with 200 μL of extract and incubated for 30 min at 4 °C with rotation; then, the protein G-beads were added to the mixture, and it was further rotated for 30 min at 4 °C as shown in Fig. 1. The glutathione Sepharose beads were mixed with 200 μL of extract and incubated for 30 min at 4 °C with rotation at 4 °C as shown in Fig. 4. Then, the beads were washed five times with 500 μL of S buffer and resuspended with 30 μL of SDS-PAGE sample buffer. All the bound proteins prepared in the 20 μL SDS-PAGE sample buffer were loaded on a SDS-polyacrylamide gel.

Acknowledgements

We thank Drs T. Nishimoto and N. Nakashima for providing us with samples. We thank Ms F. Sekiguchi for technical assistance. We appreciate the technical support from the Research Support Center, Graduate School of Medical Sciences, Kyushu University. This work was supported by JSPS KAKENHI Grant Numbers 21570007 and 24570160.

Ancillary