A yeast 2-hybrid analysis of human GTP cyclohydrolase I protein interactions

Yeast strains

The yeast strains used were Saccharomyces cerevisiae Y187 (MATα, ura3–52, his3–200, ade2–101, trp 1–901, leu 2–3, 112, gal4Δ, met⁻,gal80Δ, URA3::Gal1_UAS-GAL1_TATA-lacZ, MEL1) and AH109 (MATa, trp 1–901, leu 2–3, 112, ura3–52, his3–200, gal4Δ, gal80Δ, LYS2::Gal1_UAS-GAL1_TATA-HIS3, MEL1, Gal2_UAS-GAL2_TATA-ADE2,URA3::MEL1_UAS-MEL1_TATA-lacZ). Both strains were purchased from Clontech.

Construction of plasmids

Human full-length Form 1 GCH1 (FL-GCH1), N-terminal domain GCH1 (N-GCH1), core domain GCH1 (C-GCH1), and middle domain GCH1 (MID-GCH1) were amplified by PCR using pCDM8-hGCH1 (supplied by Dr Markus Gutlich, Technical University of Munich) containing full-length human GCH1 as the template and appropriate primer pairs. GCH1 amplicons were digested and inserted into either or both the bait and prey plasmids pGBKT7 and pGADT7 (Clontech). Rat GFRP which is 94% homologous to human GFRP was prepared by PCR using pGEM3Z-rGFRP (supplied by Dr Sheldon Milstien, NIMH) as the template and was cloned into the prey plasmid pGADT7. All constructs were verified by DNA sequencing.

2-hybrid screening of a human brain cDNA library

A concentrated overnight culture of AH109 cells containing the bait plasmid pGBKT7-N-GCH1 was combined in 2X yeast peptone dextrose (YPD) containing kanamycin with Y187 cells pre-transformed with a human brain cDNA library (Clontech, Mountain View, CA, USA). The mating culture was incubated at 30°C for 24 h, harvested by centrifugation and the cell pellet suspended in 0.5X YPD containing kanamycin. The culture was spread on selective dropout media (SD/–His/–Leu/–Trp and SD/–Ade/–His/–Leu/–Trp) and incubated at 30°C until colonies appeared.

Library prey PCR and analysis

Plasmid DNAwas purified from yeast according to the manufacturers protocol (Zymo Research, Orange, CA, USA). A 1: 10 dilution was used as template for PCR reactions that included 0.5 μM of forward and reverse primers (fwd: CTATTCGATGATGAAGATACCCCACCAAACCC; rev: GTGAACTTGCGGGGTTTTTCAGTATCTACGATT), 200 μM dNTP’s and 1 U hot start Taq polymerase. Cycling parameters were 95°C for 12 min, 39 cycles of 95°C for 1 minute, 66°C for 1 minute, 72°C for one minute, and a final extension time at 72°C for 10 min. Amplicons were digested with 10 U of HaeIII in the PCR reaction buffer for 1 h at 37°C and analyzed by agarose gel electrophoresis. Plasmids that produced PCR products with unique restriction patterns were transformed into bacteria, purified and sequenced. Insert DNA sequences were translated in all three frames and the frame producing the longest amino acid sequence was analyzed for nucleotide and/or protein homology using National Center for Bioinformatics (NCBI) Blast programs.

β-galactosidase colony-lift filter assay

Co-transformed colonies were replica-plated on filters and lysed by freeze-thawing. Each replica was then placed colony side up on filter paper saturated with Z-buffer/X-gal solution. Positive β-galactosidase (β-gal) activity was defined by the appearance of blue colonies after 8 h at 30°C.

β-galactosidase liquid assay

Soluble β-gal was assayed as recommended by the supplier (Clontech). Reactions in triplicate were performed on three isolated colonies each of which was used to inoculate 5 mL of SD/–Leu/–Trp media. Cultures were grown overnight at 30°C with shaking until OD600–0.6. Cells were collected by centrifugation, washed, suspended in Z-buffer and lysed by freeze-thawing. β-gal enzyme activity in the lysates was normalized to OD₆₀₀ and expressed as a percentage of β-gal activity in colonies cotransformed at the same time with the positive interaction control pair p53 and SV40T. This normalization process allowed comparisons of soluble β-gal activity to be made across different experiments.

GST-fusion protein purification

Full-length hGCH1 was cloned into the pGEX4T1 vector, expressed in Escherichia Coli strain BL21 and purified from inclusion bodies using a protocol modified from Frangioni and Neel (1993). For purification, GST or GST-FL-GCH1 was combined with glutathione-agarose, washed repeatedly and gel-bound protein eluted with glutathione. GST and GST-FL-GCH1 were dialyzed and protein concentrations determined by the Bradford Assay. Sodium dodecyl sulphate polyacrylamide gel electrophoresis (SDS-PAGE) showed purity of each to be greater than 95%.

In vitro binding assay

300 ng of GST or GST-FL-GCH1 protein was combined with 10–20 000 c.p.m. of in vitro translated [³⁵S]methionine-labeled protein, 200 μL of Tris-buffered saline (TBS) containing 0.1% Tween-20 (TBStw), and 100 μg/mL acetylated bovine serum albumin (BSA). Samples were rotated at 4°C for 4–8 h. Glutathione-agarose was blocked in TBStw containing 1 mg/mL acetylated BSA and 0.5 μL of packed beads in 100 μL of TBStw was added to each sample and incubated, rotating at 4°C for 1 h. Samples were centrifuged, beads washed with TBStw, transferred to a fresh tube, washed again and bound [³⁵S]-labeled proteins eluted by heating at 70°C for 10 min in NuPAGE sample buffer. Proteins were separated on NuPAGE 10% acrylamide gels, transferred to nitrocellulose, the membrane sprayed with Enhance (PerkinElmer Life Sciences, Wellesly, MA, USA) and placed against film at −80°C.

GCH1 enzyme activity assay

100 μL of yeast cell extract was combined with 100 μL of reaction buffer (0.l M Tris HCl, pH 8.0, 2 mM GTP, 0.3 M KCl, 10% glycerol, 1 mM EDTA) and incubated at 37°C for up to 3 h. The reaction product dihydroneopterin triphosphate was oxidized, de-phosphorylated and proteins precipitated as previously described (Kapatos et al. 1983). Clarified supernatants were filtered and Derythro-neopterin content determined by HPLC with fluorescence detection. Enzyme activity was expressed as pmoles neopterin/h/mg protein.

GCH1 assembly

The yeast 2-hybrid assay system requires that one protein (bait) be fused to the GAL4 DNA-binding domain (DBD) while the other protein (prey) is fused to the GAL4 activation domain (AD). If the bait and prey proteins interact, the GAL4 DBD and AD fusion domains are brought into close proximity, reconstituting GAL4 function and activating transcription of multiple reporter genes. Based upon published functional and structural studies of GCH1, we fused four forms of human GCH1 in frame via their N-termini to either the GAL4 DBD or AD (Fig. 1c). Full-length-GCH1 (FL-GCH1) includes amino acids 1–250 (Fig. 1c), contains all secondary features of GCH1 (Fig. 1b) and is the most abundant form of human GCH1 found in tissues (Togari et al. 1998; Hirayama et al. 2001). N-terminal domain GCH1 (N-GCH1) is truncated at the C-terminus to contain amino acids 1–96 (Fig. 1c) and includes structural features h1, h2 and h3 (Fig. 1b). Core domain GCH1 (C-GCH1) is truncated at the N-terminus to contain amino acids 42–250 (Fig. 1c) and is minus extended h1 (Fig. 1b). Middle domain GCH1 (MID-GCH1) is truncated at both the N- and C-termini to contain amino acids 42–236 (Fig. 1c) and is minus h1 and h6 (Fig. 1b).

Without making any a priori distinctions between the formation of GCH1 dimers or decamers, our initial studies set out to determine whether the yeast 2-hybrid assay can resolve the interaction of full length GCH1 (FL-GCH1) monomers. Yeast cultures were co-transformed with FL-GCH1 fused to the DBD or to the AD (Fig. 1c). Colony (data not shown) and liquid β-gal assays revealed a very robust interaction of FL-GCH1-DBD with FL-GCH1-AD (Fig. 2a) essentially equivalent to the positive interacting control pair p53 and SV40T. This interaction was specific in that no signal was detected between FL-GCH1-DBD and the negative control protein SV40T-AD (Fig. 2b). The three GCH1 deletion-DBD fusion constructs were next individually transformed with FL-GCH1-AD. Results from this analysis showed that the interaction of C-GCH1 and MID-GCH1 with FL-GCH1 was no different than the interaction of FL-GCH1 with itself, indicating that these two truncated forms of GCH1 retain the structural integrity necessary to associate with FL-GCH1 monomers (Fig. 2a). In contrast, N-GCH1 had no activity in this assay and from this we conclude that in isolation h1–h3 cannot interact with FL-GCH1 monomers (Fig. 2a).

In order to interpret these observations within the structural framework of GCH1 developed through crystallography, it was important to determine the level of oligomerization achieved by GCH1 monomers in the yeast 2-hybrid assay. Because biochemical studies indicate that only the GCH1 decamer is catalytically active (Yim and Brown 1976), the question of whether GCH1 fusion proteins have the capacity to form decamers might be determined by measuring GCH1 enzyme activity in protein extracts prepared from yeast transformed with FL-GCH1-DBD, C-GCH1-DBD and MID-GCH1-DBD. The results of this experiment showed that although C-GCH1 was not as catalytically active as FL-GCH1, both generate GCH1 enzyme activity well above the endogenous activity of non-transfected yeast, indicating that FL-GCH1 and C-GHC1 fusion proteins form functional decamers (Table 1). In contrast, MID-GCH1 had no detectable enzyme activity in this assay, even though this construct retains all amino acid residues known to be involved in formation of the GCH1 active site and interacts strongly with FL-GCH1 in the 2-hybrid assay (Fig. 2a). These results suggest that the MID-GCH1 construct cannot oligomerize beyond the complexity of the dimer (Fig. 1a).

GCH1–GFRP interaction

Although FL-GCH1 and C-GCH1 expressed in yeast form decamers it remained a possibility that this enzyme activity represents a pool of GCH1 that is not tethered to yeast genomic DNA via the fusion partner and therefore not involved in the 2–hybrid interaction. To test this, the interaction of GCH1 with its accessory protein GTP Cyclohydrolase I Feedback Regulatory Protein (GFRP) was investigated. This was carried out because biochemical (Yoneyama and Hatakeyama 1998) and structural (Maita et al. 2002) studies have conclusively shown that only the GCH1 decamer is capable of interacting with the GFRP pentamer. Thus for the 2-hybrid assay to detect the GCH1–GFRP interaction presumably requires not only that FL-GCH1 bound to genomic DNA via the fusion partner exist as a decamer but that GFRP also assemble as a pentamer. Despite these constraints, the colony-lift filter β-gal assay proved that the interaction between FL-GCH1 and GFRP was readily detectable and specific in that GFRP did not interact with the negative control protein p53 (Fig. 2b). The liquid β-gal assay measured the strength of the FL–GCH1 and GFRP interaction at approximately 10% of FL-GCH1 with itself (Fig. 2c). Reversal of DBD and AD fusions for GCH1 and GFRP produced identical results (data not shown) further demonstrating the specificity of the GCH1–GFRP interaction. When interpreted along with the enzyme activity data, these observations support our conclusion that FL-GCH1 is assembled as a decamer on genomic GAL4 DNA binding elements. Moreover, because the dissociation constant for GCH1 and GFRP is estimated to be in the nM range (Milstien et al. 1996), the binding of GCH1 and GFRP sets a scale by which the strength of other GCH1 binding partner interactions can be compared.

Based upon these results, we next set out to use GFRP as a tool to examine the state of assembly of the three truncated forms of GCH1. We could find no evidence that N-GCH1 and GFRP interact in the 2-hybrid assay (Fig. 2c). Nonetheless, the absence of N-terminal amino acids 1–42 in C-GCH1 decreased the strength of the GCH1–GFRP interaction by approximately 80%, indicating a significant role for extended h1 in the stabilization of the GCH1-GFRP complex (Fig. 2c). Further truncation of C-terminal amino acids 237–250 completely eliminated GFRP binding to GCH1, even though MID-GCH1 retains all structural components reportedly necessary for the GCH1–GFRP interaction to proceed (Fig. 2c) (Maita et al. 2002). These differences between GCH1 constructs cannot be accounted for by differences in protein expression levels because, as shown in Fig. 2(a), C-GCH1 and MID-GCH1 interact equally with FL-GCH1. Furthermore, N-GCH1 functions quite effectively as bait in a screening assay (see below). Based upon the overall subunit configuration of the crystal GCH1-GFRP complex, these results support our contention that FL-GCH1 and C-GCH1 assemble as decamers, whereas MID-GCH1 subunits form dimers unable to interact further.

N-terminal amino acids 1–42 include non-conserved poly proline domains

As already mentioned, the N-terminal amino acid sequence of GCH1 shows a high degree of species diversity. Alignment of the human, chimpanzee, macaque, mouse and rat N-terminal 1–42 amino acids reveals a number of significant differences, the most obvious being that the three primate forms have a nine amino acid N-terminal extension (Fig. 1d). Except for a single amino acid difference at position 8, the human and chimpanzee sequences are completely homologous and contain three consensus motifs (PXXP) for binding to SRC Homology 3 (SH3) domains; P₂₇PRP₃₀, P₃₂SRP₃₅ and P₃₉PRP₄₂. These include two overlapping ligands, P₂₃ER-DPPRP₃₀ and R₂₉PGPSRP_35, that have been shown to bind specifically to Type-I SH3 domains (Cohen et al. 1995). Two substitutions at P₃₉ and P₂₃ interrupt P₃₉PRP₄₂ and P₂₃ER-DPPRP₃₀ of the macaque N-terminal 42 amino acid sequence, which otherwise is completely homologous to the chimpanzee protein. In contrast, the mouse and rat N-termini have single amino acid substitutions at positions equivalent to P₂₇, P₃₂ and P₃₉ in the primate and lack all three PXXP domains.

Identification of putative binding partners that interact through the GCH1 N-terminal domain

N-terminal amino acids 1–96 were chosen as the bait to search for GCH1 protein binding partners because in the intact protein this protruding extension of h1–h3 offers binding sites for SH3 domain-containing proteins that are away from the main body of the protein. Screening of a human brain cDNA library containing 5 × 10⁶ independent clones with an average insert size of 1.5 kb with N-GCH1 yielded 38 unique prey selected under the most stringent growth and interaction conditions (– ade/–leu/–tryp/–his). From these 38 prey, nine were selected based upon their ability to interact with FL-GCH1 and thus to recognize the N-terminal 96 amino acids within the context of the GCH1 decamer. In order to demonstrate the requirement of extended h1 for these interactions, we screened these nine prey against C-GCH1, which lacks amino acids 1–42 and the SH3-ligand sequences.

This screen selected for seven clones that interact with N-GCH1 and FL-GCH1 but not with C-GCH1 and therefore require amino acids 1–42 for binding (Table 2). These seven clones were sequenced; open reading frames established and full-length proteins identified. Grouped according to the strength of interactions (Figs 3a and b) these seven include: Activator of Heat Shock 90 kDa Protein (Aha1; BC007398); Rearranged L-Myc Fusion protein (RLF; XM_029839); 71 kDa Heat Shock Cognate Protein (Hsc70; Y00371); Specificity Protein 3 (Sp3; AY070137); Cannabinoid Receptor CB1 Interacting Protein 1 (CB1-IP1; AAN41658); Rab26 (Rab26; BC007681) and the β1 subunit of Na⁺/K⁺ ATPase Transporting Protein (ATPase; XM_005493). Based upon liquid β-gal assays, the interaction of FL-GCH1 with each of these new binding partners was significantly greater than the negative control pair lamin/SV40T, and ranged from 10% to 0.3% of the activity of the established GCH1 binding partner GFRP. Table 2 also shows that these binding partners can be grouped into two broad functional categories: membrane-related (Aha1, Hsc70, Rab27, Na⁺/K⁺ ATPase, CB1-IP1) and gene transcription (Sp3, RLF).

Interaction of GCH1 and Aha1

A GST pull-down strategy was used next to test the binding of GST-GCH1 to the seven in vitro translated [³⁵S]-labeled prey inserts. Translation of each clone into ³⁵S-labeled protein was verified by SDS-PAGE (data not shown). Of the seven proteins tested only the interaction between GST-GCH1 and Aha1 was detected above background binding to GST alone (Fig. 4). The amount of Aha1 bound in this assay was low, most probably because the interaction between Aha1 and GCH1 is weak (~10% of GFRP-GCH1) and very sensitive to the washes required to eliminate background binding.

GCH1 Assembly

GCH1 is a homodecamer assembled from five dimers arranged into a ring-like structure that has the appearance of two pentamers aligned face-to-face (Fig. 1a). The spontaneous association of monomers to form dimers and the salt and substrate induced association of dimers to form decamers (Yim and Brown 1976) suggests that distinct protein domains are involved in dimer and decamer assembly. This concept is supported by crystal structures of the GCH1 decamer which have been interpreted to indicate dimers being formed through a four-helix bundle composed of h2 and h3 from two monomers, with no role played by h1. In agreement, we show here that extended h1 is not required for subunit dimerization. We also report, however, that N-GCH1, which contains h1, h2 and h3, does not interact with FL-GCH1. This suggests that the interaction of h2 and h3 may not be sufficient for subunit dimerization, which may also require orientation and stabilization through the interaction of the anti-parallel β-sheets from each monomer. Alternatively, when isolated from the main body of the protein, h2 and h3 may not be able to orient so as to interact with full length GCH1 monomers. The latter possibility seems unlikely, however, given that the N-GCH1 construct was used to select for binding partners that were subsequently found to interact with FL-GCH1 through the N-terminal domain.

Analysis of crystal structures indicates that five GCH1 dimers associate into one decamer through the formation of a hydrogen-bonded ladder between b1 of one monomer within a dimer with b4 of another monomer within an adjacent dimer. Nonetheless, based upon truncation of C-terminal amino acids 236–250 our studies indicate that h6, located in the pore region in close proximity to h6 of adjacent monomers outside of the dimer, also serves to stabilize the decamer. In support of this, size exclusion analysis of the bacterial enzyme has shown that residues Lys₁₃₆ and Arg₁₃₉ of h4 and Glu₁₅₂ of h5 are also critically important for decamer formation (Lee et al. 2002). That these three residues are involved in the binding of GTP within the active site cavity may be related to the ability of GTP to induce the conversion of dimers to the decamer (Yim and Brown 1976). Similarly, Arg₁₈₄ located in h5 of the human enzyme has also been shown to be involved in GCH1 oligomerization (Suzuki et al. 1999). Taken together these observations support the concept that h4, h5 and h6 are either directly involved in decamer formation from dimers or serve to stabilize the decamer by positioning b1 and b4 of adjacent dimers.

The GCH1–GFRP interaction

The yeast 2-hybrid system was also used to study GCH1 domains required for the GCH1–GFRP interaction. We previously hypothesized that the N-terminal region of GCH1 was involved in this interaction (Swick et al. 2002). Using crystals of GCH1 complexed with GFRP it was subsequently demonstrated, however, that this interaction is primarily due to van der Waals forces and a salt bridge formed between Arg₁₁ of GFRP with Glu₂₃₆ of GCH1 (Bader et al. 2001; Maita et al. 2002). Asp₄₂ of GFRP is also thought to form direct and water-mediated hydrogen bonds with Glu₂₃₆ and Arg₂₃₅ of GCH1. Our current findings show that while not absolutely required for binding to GFRP, the GCH1 N-terminal 42 amino acids are very much involved in this interaction, suggesting that extended h1 may stabilize this protein-protein complex.

A new GCH1 protein binding partner

Modern proteomics has shown that new functions for an old protein can often be inferred from the biological activity of that protein’s binding partners. One strategy for identifying proteins involved in unconventional roles for GCH1 and BH4 might therefore be to search for proteins that interact with GCH1. GCH1 is already known to have a number of protein partners. Mammalian GCH1 recruits GFRP (Harada et al. 1993) and possibly rabphilin-3A (Imazumi et al. 1994) while certain N-terminal extended isoforms of fruit fly GCH1 bind tyrosine hydroxylase (TH) and catecholamines-up (CATSUP) (Krishnakumar et al. 2000; O’Donnell et al. 2002). In all but one case, these interactions are refinements on established biochemical pathways and presumably not related to unconventional roles for GHC1 and BH4: GFRP confers on GCH1 sensitivity to l-phenylalanine, BH4 (Yoneyama and Hatakeyama 1998) and DAHP, the only known inhibitor of GCH1 (Xie et al. 1998); TH requires BH4 for activity and so a physical association of TH with GCH1 brings together a BH4-requiring and a BH4 biosynthetic enzyme; CATSUP is a zinc transporter and GCH1 is a zinc requiring enzyme (Auerbach et al. 2000; O’Donnell et al. 2002). In contrast, the interaction of GCH1 with rabphilin-3A suggests a novel role for GCH1 and BH4 in vesicular trafficking that may be related to the reported effects of BH4 on neurotransmitter release (Koshimura et al. 1990; Mataga et al. 1991; Imazumi et al. 1994) or G-protein signaling (Ziegler and Gütlich 1996). None of these reported GCH1 partners were isolated here from the human brain library, possibly because N-GCH1 was used as the bait or these interactions are dependent upon post-translational modification of GCH1 or the binding partner.

As pointed out by others, the GCH1 N-terminal region is heterogeneous across species (Maier et al. 1995; Nar et al. 1995; Krishnakumar et al. 2000), suggesting that this domain is involved in functions unrelated to catalysis, such as the binding of accessory proteins. Indeed, our analysis shows that the N-termini of human and chimpanzee GCH1 contain three consensus motifs capable of binding to SH3 domains, including two ligands that are bound specifically by Type-I SH3 domains (Cohen and Ren 1995). These domains are partially conserved in the macaque sequence yet, are completely absent from the rat and mouse proteins. Unlike the highly conserved catalytic core, it would appear that the N-terminal extension of h1 continued to evolve in the primate lineage up to 6 million years ago, possibly as a ligand for binding SH3 domain-containing proteins.

It is of interest that five of the seven GCH1 binding partners are associated with membranes while the other two partners are found in the nucleus, which suggests possible connections to the vectorial secretion of BH4 (Schaffner et al. 1994) and the nuclear localization of GCH1 (Elzaouk et al. 2004). Nonetheless, of the seven only Aha1 was validated in the GST pull-down assay. The reason for this is unclear, but likely involves the low affinity of these proteins for GCH1 and the washing protocol. Alignment of the Aha1 GCH1 binding domain with nostrin and SRC, two proteins that contain canonical SH3 binding sites, reveals some similarities but no identity within the conserved clusters of aromatic residues and surrounding charged amino acids that rather loosely define an SH3 domain. The absence of complete SH3 homology may also be responsible in part for the low level of interaction Aha1 shows with GCH1 in the yeast 2-hybrid assay, which is approximately one-tenth of that of the known binding partner GFRP.

Aha1 is a membrane trafficking protein that has recently been characterized as an activator of human heat shock protein 90 (Hsp90) (Panaretou et al. 2002). Hsp90 is the core component of a chaperone complex that has been shown to associate with endothelial nitric oxide synthase (eNOS) and is recruited to eNOS by factors that stimulate the production of nitric oxide (García-Cardeña et al. 1998). The N-terminal region of Aha1 interacts with the central domain of Hsp90 and stimulates Hsp90 ATPase activity (Lotz et al. 2003). The sequence of Aha1 isolated here based upon its interaction with GCH1 contained only the C-terminus and is outside of the N-terminal domain required for interaction with Hsp90. Since GCH1 regulates the biosynthesis of BH4 and BH4 is required by and not normally saturating for eNOS, the interaction of GCH1 with Aha1 may recruit GCH1 into the eNOS/Hsp90 complex so as to support local changes in nitric oxide production by endothelial cells. Loss of GCH1 expression and BH4 deficiency has been implicated in the decline in endothelial cell eNOS activity and nitric oxide production associated with diabetes (Meininger et al. 2000), hypertension (Hong et al. 2001:; Zheng et al. 2003) and atherosclerosis (Laursen et al. 2001). Although the physiological relevance of the Aha1–GCH1 interaction requires further study, the interaction of GCH1 with Aha1 discovered here might play an important role in normal vascular endothelial cell function.