Structure of the tandem PX-PH domains of Bem3 from Saccharomyces cerevisiae

The structure of the putative membrane-binding tandem PX-PH domain module of the yeast protein Bem3 is reported.


Introduction
The Bem3 protein from Saccharomyces cerevisiae (UniProt ID P32873) is a large multi-domain protein consisting of 1128 amino acids (Zheng et al., 1994). Three domains within the protein can be deduced from the amino-acid sequence: a PX domain (amino acids 504-630), a PH domain (amino acids 630-741) and a RhoGAP domain at the C-terminus of Bem3 (amino acids 913-1128) (Fig. 1). It has also been suggested that the N-terminal part of the protein (amino acids 1-114) is an auto-inhibitory module which results in a constitutively active variant of Bem3 upon deletion ( Fig. 1; Kadota et al., 2004;Mukherjee et al., 2013).
The PX domain, which was discovered as a conserved sequence of 100-140 amino acids (Ponting, 1996), has been identified in more than 100 yeast and mammalian proteins belonging to diverse families (Ellson et al., 2002;Sato et al., 2001;Seet & Hong, 2006;Xu et al., 2001). The conserved three-dimensional structure contains an N-terminal -sheet followed by a helical subdomain with three -helices (Lu et al., 2002). Conserved basic residues present in the 3-1 region of the PX domain are believed to mediate interaction with the phosphate groups of phosphatidylinositol phosphates (PIPs) within membranes (Cheever et al., 2001;Ellson et al., 2001;Song et al., 2001;Xu et al., 2001). These PIPs are phosphorylated derivatives of phosphatidylinositol that constitute a small fraction of the total cellular lipids but are important components of cellular organization (Lenoir & Overduin, 2013). Phosphorylation at the 3 0 -, 4 0 -and 5 0hydroxyl groups of the inositol ring leads to seven different phosphatidylinositol phosphates. Each of these singly, doubly or triply phosphorylated PIPs displays a distinct localization in cellular organelles, where they recruit specific PIP-binding effectors to exert their regulatory functions, such as lipid signalling, vesicle transport and cell growth (Lenoir & Overduin, 2013;Mayinger, 2012). The specificity and affinity towards certain PIPs can vary between different PX domains ISSN 2053-230X (Song et al., 2001). Whereas many PX domains exhibit a common preference for PI3P as the primary binding partner on membranes, others bind less specifically and recognize different PIPs (Mayinger, 2012). The low binding affinity for PI3P observed in some PX-domain-containing proteins is associated with the presence of a secondary lipid-binding site within this domain which specifically binds to phosphatidic acids (PAs) or phosphatidylserine (PS), as well as other PIPs, and can synergistically cooperate with the primary PI3P binding of the PX domain (Karathanassis et al., 2002;Stahelin et al., 2004). Apart from the secondary lipid-binding site, the possible oligomerization of some proteins owing to the PX domain can further enhance the interaction between these proteins and membranes (Xing et al., 2004).
The second important PIP-binding domain found in Bem3 is the pleckstrin-homology (PH) domain, which directly follows the PX domain within the primary sequence of Bem3 and consists of approximately 100 amino acids (Haslam et al., 1993). The structure of PH domains consists of seven strongly bent antiparallel -strands, which are terminated by a C-terminal -helix (Blomberg et al., 1999;Macias et al., 1994;Saraste & Hyvö nen, 1995;Yoon et al., 1994). Sequencesimilarity searches in the human protein database led to the identification of this domain in more than 250 proteins,   making it the 11th most abundant domain in the human proteome (Letunic et al., 2006). PH domains preferably bind to PIPs that have adjacent phosphates, such as PI(4,5)P 2 , PI(3,4)P 2 and PI(3,4,5)P 3 . However, only a minor fraction of these domains show high affinity and specificity for one PIP, whereas other PH domains either do not recognize PIPs or interact very weakly and the functional significance of their interactions is not clear Kavran et al., 1998;Takeuchi et al., 1997;Yu et al., 2004). In addition to PIP binding, the -sandwich structure of PH domains is involved in various other functions, including protein-protein interactions as well as oligomerization (Baumeister et al., 2003;Klein et al., 1998;Worthylake et al., 2004, Lemmon, 2007.
In this study, we have determined the X-ray crystallographic structure of the interesting tandem PX-PH domain module of Bem3 to a resolution of 2.2 Å , showing that the putative membrane-binding regions of both domains are oriented in the same direction and thus might act synergistically in membrane recognition and binding.

Experimental procedures 2.1. Plasmid construction
Bem3 constructs encompassing the regions containing amino acids 500-765 or 500-750 (comprising the PX and PH domains of Bem3) were amplified from yeast genomic DNA (strain S288c) using Pfu polymerase (NEB, England) and the primers listed in Table 1. The amplicons were digested with the respective restriction enzymes (NdeI, BamHI and XhoI; the recognition sequences are underlined in the primer sequences) and incorporated into a modified bacterial expression plasmid . The resulting constructs were used for the expression of Bem3 including an N-terminal hexahistidine tag followed by a Tobacco etch virus (TEV) protease cleavage site. For phasing and unambiguous assignment of the amino-acid sequence in the structure, point mutations were introduced by site-directed mutagenesis at positions 680 (K680M) and 688 (L688M) to incorporate l-selenomethionine (SeMet).

Protein purification
To express the recombinant proteins, Escherichia coli BL21-CodonPlus (DE3)-RIL cells were transformed with the corresponding plasmid and allowed to grow at 37 C until the OD 600 reached approximately 0.60-0.65. Expression was induced by adding isopropyl -d-1-thiogalactopyranoside (IPTG) to a final concentration of 0.2 mM and growth was continued overnight at 20 C. The SeMet-labelled mutants were expressed using conditions described elsewhere (Van Duyne et al., 1993).

Crystallization
To determine the initial crystallization conditions, Bem3 was subjected to crystallization trials at 293 K by the sittingdrop vapour-diffusion method using the following commercially available protein crystallization screens: The PACT Suite, The PEGs Suite and The JCSG Core I-IV Suites  Table 1 Macromolecule-production information.
Recognition sequences are underlined in the primer sequences.

Data collection and processing
Individual hexagonal rods were harvested, cryoprotected in mother liquor supplemented with 10%(v/v) glycerol and flashcooled in liquid nitrogen prior to data collection. The Bem3 crystals were maintained at 100 K and the collection of X-ray diffraction data was performed on beamline X10SA at the Swiss Light Source (SLS), Paul Scherrer Institute, Villigen, Switzerland. The data were indexed and integrated with XDS and scaled with XSCALE (Kabsch, 2010a,b). Data statistics are described in Table 3.

Structure solution and refinement
Initial phases and an initial model were obtained using the data set from the SeMet derivative of L688M Bem3 500-750 (containing a total of six SeMet residues) with phenix.autosol (Terwilliger et al., 2009). This resulted in an initial model with 82 amino-acid residues built (BAYES-CC = 44.3 AE 9.6, R work = 34.3%, R free = 37.9%) and was followed by automated model building in phenix.autobuild (R work = 27.2%, R free = 30.0%) (Terwilliger et al., 2008). The model was iteratively improved by manual model building in Coot (Emsley et al., 2010) and refinement with phenix.refine (Afonine et al., 2005) against the native data set for Bem3 500-765 . Validation of the structure was performed with MolProbity (Chen et al., 2010). Refinement statistics can be found in Table 4.

Results and discussion
The crystal structure of the tandem PX-PH domains shows the well characterized folds of both protein domains as reported previously (Riddihough, 1994;Bravo et al., 2001). The model contains amino acids 502-629 belonging to the PX domain and 636-737 of the PH domain, but the linker region (amino acids 630-635) between the domains as well as the loops between 1 PH and 2 PH (amino acids 647-651) and 6 PH and 7 PH (amino acids 709-715) within the PH domain were flexible and could not be resolved. Even though six amino acids are missing between the two domains, the presented structural assembly within the asymmetric unit represents the only possible arrangement, and a different connectivity between the PX and PH domains is excluded because the positions of the respective C-and N-termini are too remote. Using the native data set only, the sequence could not be unambiguously assigned within parts of the PH domain because of a lack of    electron density for the side chains (especially within 4 PH ; see Fig. 1). We therefore used the anomalous signal from crystals of SeMet-labelled K680M Bem3 500-750 and L688M Bem3 500-750 in order to correctly assign the positions of amino acids in this region (the electron density is shown in Fig. 1a). As outlined in x1, both the PX and PH domains have been implicated in binding to PIPs and, besides other possible functions, mediate membrane binding and localization in cells. Strikingly, the PX domain of Bem3 contains basic residues (for example Lys551 and Arg592/Lys593) at similar positions compared with other PX domains, which have previously been termed basic motifs I and II, respectively, and have been shown to be involved in PIP binding . A similar observation also holds true for the PH domain (basic residues Arg644, Arg645 and Arg658), with similar relative positions compared with the PH domain of, for example, PLC (see the alignment in Fig. 1c; Lemmon, 2007). Positively charged residues within this part of the protein consisting of strands 1 PH and 2 PH and the connecting loop have also previously been identified as an important lipid-binding motif . This motif consists of the first lysine at the end of 1 PH and an arginine in the middle of 2 PH , which both interact with adjacent phosphate head groups of PIP2 (Ferguson et al., 1995. The canonical lysine is, however, replaced by an arginine in the PH domain of Bem3. This arginine (Arg644) is well conserved among PH domains lacking the canonical lipid-binding sequence (Anand et al., 2012;Macia et al., 2008). The crystal structure of the PH domain of Slm1 (PDB entry 4a6k) revealed that the corresponding Arg478 interacts with PIPs (Anand et al., 2012).
Specific point mutations introduced at this position in Slm1 (Gallego et al., 2010) and the PH domain of EFA6 (Macia et al., 2008) resulted in mislocalization of the respective proteins, further signifying the role of this arginine in ligand binding. In line with these findings, an analysis involving the mutation of basic residues within the 1 PH -2 PH loop of yeast Bem3 (R644S, R645S and K647D) showed that this abolished binding of the PH domain to PIP2, rendering the protein incapable of correctly localizing to the polar site at the cell membrane (Mukherjee et al., 2013). In contrast to these mutations within the PH domain, a quadruple mutant (Y524W, R578S, L580S and F581M; Mukherjee et al., 2013) involving amino acids within the so-called 'membrane-interaction loop' (Cheever et al., 2001) of the PX domain exhibits wild-type localization of Bem3. Although the critical residues Lys551 and Arg592 of basic motifs I and II of the Bem3 PX domain are conserved in the majority of PX domains (Cheever et al., 2001), their role in the direct binding of PIPs has yet to be elucidated.
Interestingly, the putative membrane-binding sites of both domains (basic motifs I and II within the PX domain and the PIP-binding site within the PH domain) are oriented in a similar direction (Fig. 1a) and this part of the protein displays a large positively charged basic surface (Fig. 2b). The similar orientation thus suggests a possible synergy between the domains with regard to the recognition of PIPs within a membrane, both with respect to the affinity of binding and the possibility of the recognition of a distinct membrane or compartment within the cell by the relative abundance of PIPs.  The interface of $620 Å 2 between the PX and PH domains (calculated using PDBePISA; Krissinel & Henrick, 2007) contains several hydrophobic interactions between Leu523, Leu531, Leu584, Pro586 and Val589 in the PX domain, and Leu642, Trp655 and Val657 in the PH domain, as well as hydrogen bonds (Thr587-Gly698, Ser521-Glu724, Thr522-Glu724 and Asp590-Asn701) and ionic interactions (Arg525-Asp671, Asp529-Lys672 and Asp590-Arg700) (Fig. 2), and many of these amino acids are conserved among Bem3 proteins from different organisms. Interestingly, a flexible relative mobility of the domains is indicated by the elevated B factors, especially at the periphery of the PH domain facing away from the PX domain (Fig. 2a), resulting in average B factors of 71 and 96 Å 2 for the PX and PH domains, respectively. The elevated B factors of the PH domain can be understood by considering the different crystal contacts. Whereas the PX domain has large contact areas ($1210 Å 2 ) with symmetry-related molecules, the PH domain has only small contact areas ($310 Å 2 ; contact areas were calculated with PDBePISA; Krissinel & Henrick, 2007;Fig. 2b), thus allowing the increased mobility. While we can only speculate on the possible role of this flexibility, it is conceivable that a less-than-rigid relative orientation of the two lipid-binding domains might allow different membrane curvatures in different specific situations.

Conclusions
In this publication, we present the three-dimensional structure of the tandem PX-PH domains of the yeast protein Bem3. The structure shows that the putative lipid-binding domains adopt a relative orientation that enables them to simultaneously bind to a membrane. Therefore, future studies should address the possibility of cooperativity between the domains with regard to membrane interaction as well as the resulting specificity with respect to the relative abundances of PIPs in a target membrane.