Interactions between Conserved Domains within Homodimers in the BIG1, BIG2, and GBF1 Arf Guanine Nucleotide Exchange Factors*

Guanine nucleotide exchange factors carrying a Sec7 domain (ArfGEFs) activate the small GTP-binding protein Arf, a major regulator of membrane remodeling and protein trafficking in eukaryotic cells. Only two of the seven subfamilies of ArfGEFs (GBF and BIG) are found in all eukaryotes. In addition to the Sec7 domain, which catalyzes GDP/GTP exchange on Arf, the GBF and BIG ArfGEFs have five common homology domains. Very little is known about the functions of these noncatalytic domains, but it is likely that they serve to integrate upstream signals that define the conditions of Arf activation. Here we describe interactions between two conserved domains upstream of the Sec7 domain (DCB and HUS) that determine the architecture of the N-terminal regions of the GBF and BIG ArfGEFs using a combination of biochemical, yeast two-hybrid, and cellular assays. Our data demonstrate a strong interaction between DCB domains within GBF1, BIG1, and BIG2 to maintain homodimers and an interaction between DCB and HUS domains within each homodimer. The DCB/HUS interaction is mediated by the HUS box, the most conserved motif in large ArfGEFs after the Sec7 domain. In support of the in vitro data, we show that both the DCB and the HUS domains are necessary for GBF1 dimerization in mammalian cells and that the DCB domain is essential for yeast viability. We propose that the dimeric DCB-HUS structural unit exists in all members of the GBF and BIG ArfGEF groups and in the related Mon2p family and probably serves an important regulatory role in Arf activation.

tion has been assigned. It was originally identified in plant GNOM, where it was shown to be capable of dimerization in yeast two-hybrid and in vitro pull-down assays (16). Little is known about the other domains, except for an almost invariant 5-residue motif in the HUS domain, the HUS box (15,17) (Fig.  1B), which is essential for aspects of Golgi traffic in yeast (18).
Here we combine biochemical, yeast two-hybrid, and cellular assays to analyze the domain architecture and interdomain interactions of the GBF and BIG groups of large ArfGEFs. Our data demonstrate the existence of two distinct interactions involving the DCB domain of the mammalian large ArfGEFs: homodimerization via a DCB/DCB interaction (generalizing previous results from plants) and a novel DCB/HUS interaction depending on the HUS box. We propose that the DCB/DCB and DCB/HUS interactions define a common structure in all members of the BIG and GBF groups of ArfGEFs, which probably also exists in the related eukaryotic Mon2p family.

EXPERIMENTAL PROCEDURES
Expression of Recombinant DCB BIG1 , DCB BIG2 , and DCB-HUS-Sec7 BIG1 -The DCB domain of human BIG1 (DCB BIG1 , residues 2-224) was introduced into pET28a (Novagen) modified to remove the thrombin cleavage site and to include alternative restriction sites (KpnI and AgeI). The E221K mutation was introduced into DCB BIG1 by PCR using the QuikChange site-directed mutagenesis kit (Stratagene). Both wild type and mutant DCB BIG1 were expressed in the Rosetta(DE3)pLysS Escherichia coli strain (Merck KGaA). DCB BIG1 was purified on a Ni 2ϩ -nitrilotriacetic acid affinity column (GE Healthcare) followed by precipitation in ammonium sulfate to 70% saturation and gel filtration on a Superdex 75 column (GE Healthcare). The DCB domain of human BIG2 (DCB BIG2 , residues 2-224) was cloned, expressed, and purified as described for DCB BIG1 .
A construct spanning the DCB, HUS, and Sec7 domains of human BIG1 (DCB-HUS-Sec7 BIG1 , residues 2-888) was introduced into pFastBac HTA vector (Invitrogen) using EcoRI and KpnI restriction sites. Sf21 cells infected by baculoviruses harboring this construct were used to express DCB-HUS-Sec7 BIG1 . The recombinant protein was purified on a Ni 2ϩ -nitrilotriacetic acid affinity column followed by a desalting column and a gel filtration Superdex 200 column (GE Healthcare). Limited proteolysis was performed with 10 units of thrombin (Amersham Biosciences) per mg of protein at room temperature overnight. DCB-HUS-Sec7 BIG1 and its proteolysis products were analyzed by SDS-PAGE and Western blot.
Biophysical Assays-Sedimentation velocity was measured at 40,000 ϫ g for 24 h and analyzed with the SVEDBERG software (available on the World Wide Web). Sedimentation equilibrium experiments were carried out at 10,000, 15,000, and 20,000 ϫ g for 46 h and analyzed with the Origin software (Beckman Coulter). CD scans were recorded between 185 and 260 nm. Thermal denaturations were carried out in the temperature range of 5-95°C at a rate of 2°C/min. Secondary structure composition was estimated with the CDDSTR software (19). The effect of protein concentration on its thermal denaturation was measured with 0.3, 0.75, and 3 M DCB BIG1 and analyzed at 222 nm, a wavelength minimum that is characteristic of ␣-helices.
Biochemical Assays-Liposome binding experiments were performed with liposomes prepared as described in Ref. 22 (Table 1). DCB BIG1 (1 M) was incubated at room temperature in 50 mM Hepes, pH 7.2, and 120 mM potassium acetate with sucrose-loaded vesicles (final lipid concentration, 1 mM) in small polycarbonate tubes. The samples were centrifuged at 360,000 ϫ g for 20 min, and the supernatants and the pellets were analyzed by SDS-PAGE with Sypro-orange staining. The Sec7 domain of BIG1 was used as a negative control. Exchange reaction assays were performed by tryptophan fluorescence kinetics using ⌬17Arf1 as described in Ref. 23. The effect of DCB BIG1 on the exchange rate of Sec7 BIG1 was analyzed by comparing the results of experiments done in the absence or presence of DCB BIG1 (10 M).
Co-immunoprecipitation Assays-COS7 cells in 10-cm culture dishes were cotransfected with the plasmids pHA-GBF1 expressing human HA-GBF1 and either Venus-GBF1, Venus-GBF1⌬889, or YFP-GBF1-C expressing, respectively, human GBF1 and GBF1 deleted of the first 297 (⌬DCB) and 710 amino acids (⌬DCB-HUS). 5 After 20 h of expression, cells were washed two times with 5 ml of cold phosphate-buffered saline (137 mM NaCl, 2.7 mM KCl, 19 mM Na 2 HPO 4 , 1.8 mM KH 2 PO 4 ) 5 T. K. Niu and C. L. Jackson, unpublished data. and then disrupted in 0.5 ml of cold lysis buffer (50 mM Tris-HCl, pH 7.5, 100 mM NaCl, 1 mM EDTA, 1% Nonidet P-40). After centrifugation at 4°C, soluble cellular extracts were precleared with 20 l of protein G-Sepharose 4 Fast Flow (GE Healthcare) at 4°C for 30 min. Supernatants were incubated with 3 g of anti-GFP antibodies (Roche Applied Science) for 1.5 h at 4°C. Then 30 l of protein G-Sepharose 4 Fast Flow was added, and the mixtures were incubated at 4°C for 1.5 h. The resin was washed two times with 1 ml of W100 buffer (50 mM Tris-HCl, pH 7.5, 100 mM NaCl, 1 mM EDTA), followed by two washes with 1 ml of phosphate-buffered saline. Proteins were then eluted by incubation with 60 l of SDS-PAGE sample buffer for 5 min at 95°C. Eluted proteins were separated on a 6% SDS-polyacrylamide gel and analyzed by Western immunoblotting using anti-HA antibodies as primary antibody (Sigma). Plasmid Shuffle Assays in Yeast-The gea1-⌬DCB construct (coding for residues 233-1408 of Gea1p) was obtained by introducing a gap in the pAP22 (CEN, TRP1, GEA1) (gift of A. Peyroche, Commissariat à l'Energie Atomique, Saclay, France). The PCR product extended to 1825 and 154 base pairs past the 5Ј and 3Ј ends of the gap, respectively. The mutated fragment and gapped plasmid were used to transform CJY52-10-2 yeast cells that contained the pAP23 plasmid (CEN, URA3, GEA1) (24). Transformants were plated onto synthetic medium plates lacking tryptophan. Trp ϩ clones were grown at 30°C in minimal medium (YNB) containing 0.67% yeast nitrogen base without amino acids (BD), supplemented with appropriate nutrients and with 2% glucose. 5-Fluoroorotic acid monohydrate (Toronto Research Chemicals) was added to a final concentration of 0.1% to counterselect the URA3-containing cells. The presence of protein product of the gea1-⌬DCB allele was controlled by Western immunoblotting as follows. Total yeast cell protein extracts were prepared using the NaOH/trichloroacetic acid lysis technique (25). Proteins were separated by SDS-PAGE in 10% Tricine gels and were analyzed by immunoblotting. The primary antibody was either a polyclonal anti-Gea1p antibody or a monoclonal anti-Vat2p antibody directed against a subunit of the vacuolar ATPase (Invitrogen).

Mammalian BIG and GBF ArfGEFs Have an N-terminal
Dimerization DCB Domain-Using a bioinformatics approach, we predicted previously that the DCB domain, originally identified as a dimerization domain in the GBF group member GNOM (16), is also present at the N terminus of large ArfGEFs from the BIG group, where it should form an all-helical domain (15). In order to address this question experimentally, we expressed in E. coli the N terminus of human BIG1, encompassing the helical subdomain of highest sequence homology and the more variable N-terminal subdomain (Fig. 1A), and purified it to homogeneity. The recombinant protein behaved as a 2 ϫ 26-kDa dimer ( Fig. 2A), a molecular mass that was confirmed by analytical ultracentrifugation equilibrium sedimentation (54.3 kDa). A similar construct from human BIG2 was also expressed and purified to homogeneity and, as for BIG1, eluted as a dimer on a gel filtration column ( Fig. 2A). Deconvolution of CD spectra for both proteins was consistent with a mostly helical secondary structure. We thus conclude that the N-terminal domain of human BIGs qualifies as the bona fide homolog of the DCB domain of the GBF group member GNOM and refer to it as DCB BIG hereafter.
In order to assess a possible dimer/monomer equilibrium, we analyzed the CD thermal denaturation spectra of DCB BIG1 (Fig.  2B). The model providing the best fit for the data was a twostate transition between an all-helical structure and a random coil denaturated state, without formation of a monomeric intermediate. This and a denaturation temperature (69°C) independent of the protein concentration and higher than the average for proteins (around 55°C) suggest that the dimer is stable with a dissociation constant below the concentration used in the experiment (100 nM). Analytical ultracentrifugation confirmed the predominance of a dimeric species and the absence of monomer.
To determine whether dimerization is a general feature of DCB-like domains of the large ArfGEFs, we tested mammalian GBF1 and BIG2 DCB/DCB interactions using the yeast twohybrid system. We observed a strong interaction between DCB domains of hamster GBF1 (98% identity with human GBF1) and between human BIG2 DCB domains, indicating that the DCB domain mediates dimerization in all mammalian large ArfGEFs (Fig. 3, sectors 1 and 12). We next looked for residues that contribute to the DCB/DCB interaction. Two highly conserved residues, Lys 91 and Glu 130 in DCB GBF1 , are found in both large ArfGEF groups (Fig. 1A). Mutation of either residue to alanine in DCB GBF1 abolished the interaction between mutant and wild-type DCB GBF1 in the two-hybrid system (Fig. 3, sectors 2 and 11). Thus, these residues are either part of the dimer interface or induce an abnormal structure in this interface. We then analyzed a mutation found in the C terminus of the DCB domain of human BIG2 (Fig. 1A), which has been associated with a congenital disease, autosomal recessive periventricular heterotopia with microencephaly (26). DCB BIG1 carrying the equivalent mutation, E221K, was expressed in E. coli with a solubility similar to that of wild-type DCB BIG1 . Analytical ultracentrifugation experiments showed that DCB BIG1 E221K forms a dimer (data not shown). Thus, functions of DCB BIG2 other than dimerization are affected by this mutation in the autosomal recessive periventricular heterotopia with microencephaly disorder.
A Novel Interaction between the DCB and HUS Domains in Large ArfGEFs-To determine whether interactions exist between the different domains of the large ArfGEFs, we carried out an extensive yeast two-hybrid analysis of domain-domain interactions for mammalian GBF1 and BIG1. All five noncatalytic domains in addition to the catalytic Sec7 domain were considered. For GBF1 (Table 2A) and BIG1 (Table 2B), only one interaction was detected in addition to the DCB/DCB interaction described above. In both cases, a strong interaction between the DCB and HUS domains of each ArfGEF was observed (Fig. 3, sectors 3 and 4). This interaction was independent of the variable DCB-HUS linker added on either the DCB or HUS side (Table 2A). A strong interaction was also found between the DCB and HUS domains of human BIG2 (Fig.  3, sector 5).
The HUS box is an almost invariant N(Y/F)DC(D/N) motif, which is predicted to lie between two ␣-helices (Fig. 1B). Muta-tion of the central aspartate to alanine in this motif abolished the DCB/HUS interaction in both GBF1 and BIG1 (Fig. 3, sectors 18 and 19). Thus, the HUS box supports the DCB/HUS interaction, which is the first molecular function to be associated with this motif. We then analyzed whether mutations that impair the DCB/DCB interaction (see above) also affect the DCB/HUS interaction. The E130A mutation in GBF1, but not the K91A mutation, abolished the DCB/HUS interaction (Fig.  3, sectors 16 and 17). These results indicate that residue Lys 91 is involved only in the DCB/DCB interaction, whereas the Glu 130 residue is involved in both interactions.
Next, we used the yeast two-hybrid system to analyze DCB/ DCB and DCB/HUS interactions within GBF1 constructs spanning more than one domain. We first analyzed the ability of the DCB domain alone to interact with larger ArfGEF fragments. DCB GBF1 interacted with DCB-HUS GBF1 , DCB-HUS-Sec7 GBF1 , and full-length GBF1 (Fig. 3, sectors 7, 8, and 10). Although these yeast two-hybrid interactions appeared to be weaker than the interaction between two DCB domains alone, they strongly suggest that the DCB/DCB interaction occurs also in the context of the full-length GBF1 protein. We then analyzed the DCB/HUS interaction in the context of multiple domains. A strong interaction was found between the DCB domain and a HUS-Sec7 construct (Fig. 3, sector 6). HUS GBF1 also interacted with DCB-HUS-Sec7 GBF1 and full-length GBF1 although more weakly than with DCB GBF1 alone (Fig. 3, sectors 14 and 15). In support for the dimerization of the N-terminal region taking place in larger constructs, recombinant DCB-HUS-Sec7 BIG1 eluted on a gel filtration column with a molecular weight consistent with a dimer (Fig. 2A).
No interactions between the noncatalytic domains other than the DCB/DCB and DCB/HUS interactions were identified with the yeast two-hybrid assay. We also failed to detect an interaction of the noncatalytic domains with the Arf1 substrate (data not shown). In addition, we did not observe any DCB/ DCB or DCB/HUS cross-interactions between BIG1, BIG2, and GBF1 in the yeast two-hybrid system (data not shown).

Functions of DCB/DCB and DCB/HUS Interactions in Arf-GEF Dimers-
The DCB domain has features of a dimeric helical bundle, which is a frequent arrangement in cytosolic proteins involved in membrane recruitment, such as the mem-

TABLE 2 Summary of domain/domain interactions in mammalian ArfGEFs analyzed with the yeast two-hybrid assay
Strong interactions are indicated in dark gray, and weaker interactions are shown in gray. Baits (BD) and preys (AD) are indicated in rows and columns, respectively. brane curvature-sensing BAR domain found in amphiphysins and Arfaptin/POR, an Arf effector (27). We thus investigated the binding of DCB BIG1 to liposomes of various compositions using a sedimentation assay (Table 1 and Fig. 4A). However, no such interaction could be observed regardless of the liposome composition, suggesting that the DCB homodimer does not have membrane-binding properties on its own. The Sec7 domain did not interact with the constructs tested in our yeast two-hybrid analysis ( Table 2, A and B, and Fig. 3,  sectors 9 and 20). To further analyze whether the N terminus could regulate the catalytic exchange activity of the Sec7 domain, we used recombinant proteins and a fluorescence kinetics assay. We first analyzed the effect of excess DCB BIG1 on the exchange rate of the Sec7 domain of human BIG1 (Sec7 BIG1 ) using ⌬17Arf1 as substrate. No inhibition or stimulation of the exchange rate by DCB BIG1 could be observed, regardless of whether ⌬17Arf1, Sec7 BIG1 , or both had been preincubated with DCB BIG1 (Fig. 4B). We then analyzed the catalytic activity of a BIG1 construct spanning the DCB-HUS-Sec7 domains using the same fluorescence assay. This construct was active at stimulating GDP/GTP exchange on ⌬17Arf1 (Fig. 4C), and it was inhibited by brefeldin A with a K i of 23.9 Ϯ 7.2 M, which is similar to that measured for the Sec7 of BIG1 alone (23). To confirm that the DCB-HUS tandem had no effect on the catalytic activity, we took advantage of a unique thrombin cleavage site located at residue 622 between the HUS and Sec7 domains, which allowed us to generate free DCB-HUS BIG1 and Sec7 BIG1 by limited proteolysis. Exchange rates measured with a BIG1 peptide concentration of 0.5 M were in the same range for the uncleaved and cleaved fragments (0.073 Ϯ 0.005 and 0.098 Ϯ 0.012 s Ϫ1 , respectively), suggesting that the DCB-HUS tandem does not have a simple one-to-one regulatory activity toward the Sec7 domain.
The N terminus of large ArfGEFs interacts with several large ArfGEF protein partners (reviewed in Ref. 1). We thus investigated whether the DCB/HUS structure may be required for protein-protein interactions. To this end, we took advantage of the fact that the N terminus of GBF1 binds to 3A, a protein from enteroviruses that blocks host cell secretion by inhibiting GBF1 function (28). The cytosolic portion of 3A (residues 1-60) interacts with DCB-HUS GBF1 and DCB-HUS-SEC7 GBF1 in the yeast two-hybrid assay (Fig. 3, sector 21). In contrast, no interaction was observed with individual DCB or HUS domains (Fig.  3, sectors 22 and 23). This is consistent with data showing that deletion of either the first 50 amino acids of GBF1 or deletion of the HUS domain and downstream sequences abolishes interaction with the 3A protein in the mammalian two-hybrid system (29). These results show that portions of both the DCB and HUS domains of GBF1 are required for binding to the viral 3A protein and suggest the possibility that an integral DCB-HUS structure is necessary for binding of the 3A protein.
Dimerization of Large ArfGEFs in Vivo-The above analysis suggests that the DCB domain supports the dimerization of large ArfGEFs and organizes a structure that can bind protein partners. We thus assessed the dimerization and function of this domain in cells for two large ArfGEFs of the GBF group.
We first analyzed the formation of human GBF1 dimers in mammalian cells by pull-down assays (Fig. 5A). We found out that full-length GBF1 can easily be isolated as a dimer from  Table 1). The last lane represents the total amount of protein in the experiment (T). B, effect of DCB BIG1 on the kinetics of Sec7 BIG1 -stimulated GDP/GTP exchange on ⌬17Arf1 measured by tryptophan fluorescence. 1, no DCB; 2, DCB BIG1 was incubated with Sec7 domain for 5 min before ⌬17Arf1-GDP was added; 3, DCB BIG1 was incubated with Arf1 for 5 min before the Sec7 domain was added. In all cases, GTP (100 M) was added 2 min after all proteins were mixed together. C, GDP/GTP exchange activity of DCB-HUS-Sec7 BIG1 on ⌬17Arf1 measured at different GEF concentrations. mammalian cells. Next, we examined dimer formation between the full-length GBF1 and forms of GBF1 deleted of the DCB domain alone or of both the DCB and HUS domains. Clearly, whereas deletion of the DCB domain alone reduced somewhat the formation of a dimer with full-length GBF1, both the DCB and HUS domains had to be deleted to nearly abolish dimer formation. Thus, the DCB and the HUS domains are both involved in the dimerization of GBF1.
We then analyzed the effect of deleting the DCB domain of Gea1p, a member of the GBF group of large ArfGEFs in yeast, using a plasmid shuffle strategy. The strain used contains the wild-type GEA1 gene on a URA3 plasmid with both gea1⌬ and gea2⌬ deletions of the chromosomal copies of the genes (24). The gea1-⌬DCB allele was introduced into a low copy TRP plasmid. The Gea1p-⌬DCB protein was expressed and was not degraded (Fig. 5B, left). Clones failed to grow at 30°C upon the loss of the wild-type GEA1 plasmid when the gea1-⌬DCB plasmid became the sole copy of the redundant GEA1 and GEA2 genes (Fig. 5B, right). This result indicates that the DCB domain of Gea1p is essential for yeast viability.

A Conserved DCB/DCB and DCB/HUS Structure in Eukaryotic Large ArfGEFs and the Related Mon2p
Family-In this study, we investigated the domain/domain interactions within the BIG and GBF groups of large ArfGEFs, which we predicted previously to share a common architecture (15). Based on biochemical and yeast two-hybrid analyses of mammalian BIG1, BIG2, and GBF1, we establish that all three members share a similar DCB-HUS organization upstream of their Sec7 domains, in which the DCB domain interacts with itself and with the HUS domain. The DCB/HUS interaction requires the highly conserved HUS box, a five-amino acid motif found in all members of the BIG and GBF groups of ArfGEFs.
Because of its bipartite organization, the DCB-HUS tandem provides different ways for large ArfGEFs to form multimers. One is through the DCB/DCB interaction, which is an obligate intermolecular interaction. Since the DCB domain forms a strong homodimer in vitro, we propose that it supports constitutive homodimerization of large ArfGEFs. The existence of this interaction in native BIG and GBF ArfGEFs is supported by its formation in a range of yeast two-hybrid GBF1 constructs, the dimerization of the recombinant BIG1 and BIG2 constructs, and our in vivo data on GBF1. It is also consistent with the molecular weight of several large ArfGEFs of both the BIG and GBF groups as measured by size exclusion chromatography, including yeast Gea1p (30), human BIG1 and BIG2 (31,32), and plant GNOM (16). All elute as large molecular weight complexes, which, given the uncertainty of this technique for nonglobular proteins, is consistent with their association as homodimers.
In contrast, the DCB/HUS interaction can occur either between two monomers (intermolecular) (Fig. 6A) or within a single ArfGEF polypeptide (intramolecular) (Fig. 6B). An intermolecular DCB/HUS interaction would provide a second contribution to dimerization in addition to the DCB/ DCB interaction. This possibility is supported by our coimmunoprecipitation results, which show that the ⌬DCB form of human GBF1 formed a dimer with full-length GBF1 almost as efficiently as full-length GBF1 in mammalian cells, whereas deletion of both DCB and HUS domains practically eliminated dimerization with full-length GBF1. Interestingly, the HUS box has an unusual level of sequence conservation and content of polar residues within a protein interface, pointing to a potential for the DCB/HUS interaction to open up and expose the HUS box (Fig. 6C). The HUS box could then carry out other functions, allowing in particular the formation of ArfGEF tetramers through three-dimen- FIGURE 5. In vivo assays. A, the DCB and HUS domains of human GBF1 are both involved in its dimerization in mammalian cells. HA-tagged GBF1 was coexpressed with either GFP, GFP-tagged GBF1, GFP-tagged GBF1⌬DCB, or GFP-tagged GBF1⌬(DCB-HUS). After incubation with anti-GFP antibodies followed by an incubation with protein G-Sepharose, the resin was washed several times, and proteins were eluted by incubation with SDS-PAGE sample buffer. Eluted proteins were separated on a 6% SDS-polyacrylamide gel and analyzed by Western blotting using anti-HA antibodies. B, the DCB domain of Gea1p is essential for yeast inability. Left, yeast extracts from ⌬gea1⌬gea2 cells bearing a URA3 plasmid containing the GEA1 wild type gene (plasmid pAP23, lane 1), pAP23 and a TRP plasmid containing the GEA1 wild-type gene (pAP22, lane 2), pAP23 and pAP22 containing the gea1-⌬DCB1 gene (lane 3), or yeast extracts from a wild-type strain (lane 4) were separated on SDS-polyacrylamide gels and analyzed by western immunoblotting using an anti-Gea1p antibody. An anti-Vat2p antibody was used as a loading control. Right, yeast cells deleted for the GEA1 and GEA2 genes and bearing a URA3 plasmid containing the GEA1 wild-type gene were transformed with a plasmid containing either the wild type GEA1 gene (⌬gea2 GEA1 cells) or the GEA1 gene deleted for the DCB domain (⌬gea2 gea1-⌬DCB cells). 10-Fold serial dilutions of yeast cultures were plated on media with or without 5-fluoroorotic acid monohydrate (5FOA) to counterselect the URA3-containing cells. sional domain swapping (Fig. 6D). An interesting corollary is that this could allow large ArfGEFs to form heterotetramers, which are more likely to form than heterodimers, given the stability of the homodimeric DCB/DCB interaction. BIG1 and BIG2 have been shown to co-immunoprecipitate in human cells (32), which could thus be mediated by the formation of heterotetramers containing one BIG1 homodimer and one BIG2 homodimer.
A region homologous to the DCB and HUS domains is present in a novel eukaryotic protein family, Mon2p/Ysl1p/SF21, that is related to the large ArfGEFs (33-35) but lacks the Sec7 nucleotide exchange domain (34,35). Yeast Mon2p has been shown to localize to late Golgi/endosomes (33)(34)(35) and to bind Arl1p (33), a close relative of Arf proteins. We propose that members of the Mon2p family feature a DCB-HUS structure, including a DCB homodimerization domain (Fig. 1A) and a HUS domain with a candidate HUS box containing the central (Y/F)D motif (Fig. 1C) and capable of forming a DCB/HUS interaction. DCB/DCB-mediated dimerization is in agreement with the co-immunoprecipitation of yeast Mon2p as a homodimer (34). The presence of the DCB-HUS module without an associated Sec7 domain in these proteins is consistent with a structural function that is independent of the biochemical nucleotide exchange activity.
The Role of the DCB-HUS Structure in Large ArfGEF Interactions-The N terminus of large Golgi ArfGEFs has been reported to interact with various protein partners. Notably, the N terminus of GBF1 interacts with Rab1 (36) and regions of mammalian BIGs encompassing DCB-HUS interact with AMY-1 (12), protein kinase A (10), FKBP13 (11), the Exo70 subunit of the exocyst (37), and the HSC70 chaperone (12). The requirement of 3A protein for both the DCB and HUS domains suggests that the two-domain DCB-HUS structure could mediate these protein-protein interactions in addition to its dimerization function. Furthermore, phenotypic data in vivo have shown that the DCB-HUS tandem is necessary and sufficient to define the subcellular localization of p200/BIG1 to Golgi membranes (17,18). Interestingly, this is also the case for the DCB-HUS homology region of the related protein Mon2p (34). Furthermore, mutation of the HUS box in Gea2p in yeast, which is likely to disrupt the DCB/HUS interface according to our study, resulted in impaired membrane association together with a severe defect of anterograde ER/Golgi traffic (18). Thus, the DCB-HUS structure is likely to contribute to large ArfGEF functions upstream of their exchange activity, including interactions that define their localization.
Further investigations are now needed to establish whether the DCB/HUS interaction is constitutive or supports a regulated switch between a closed and an open conformation capable of alternative interactions. The DCB/HUS structure characterized here should provide a rational framework to address this issue in the BIG and GBF groups of ArfGEFS.