|Home | About | Journals | Submit | Contact Us | Français|
FARP2 is a Dbl-family guanine nucleotide exchange factor (GEF) that contains a 4.1, ezrin, radixin and moesin (FERM) domain, a Dbl-homology (DH) domain and two pleckstrin homology (PH) domains. FARP2 activates Rac1 or Cdc42 in response to upstream signals, thereby regulating processes such as neuronal axon guidance and bone homeostasis. How the GEF activity of FARP2 is regulated remained poorly understood. We have determined the crystal structures of the catalytic DH domain and the DH-PH-PH domains of FARP2. The structures reveal an auto-inhibited conformation in which the GEF substrate-binding site is blocked collectively by the last helix in the DH domain and the two PH domains. This conformation is stabilized by multiple interactions among the domains and two well-structured inter-domain linkers. Our cell-based activity assays confirm the suppression of the FARP2 GEF activity by these auto-inhibitory elements.
FARP2 and its close homolog FARP1 are large multi-domain proteins sharing the same domain structures, after which they are named (FERM, RhoGEF, and pleckstrin homology domain proteins). FARP2 is also known as FERM domain including RhoGEF (FIR) and FGD1-related Cdc42-GEF (FRG) (Kubo et al., 2002; Miyamoto et al., 2003). FARP1 is also known as chondrocyte-derived ezrin-like protein (CDEP) and pleckstrin homology domain-containing family C member 2 (PLEKHC2) (Koyano et al., 1997). Functional studies of FARP1 and FARP2 have been focused primarily on their roles in regulation of neuronal development and morphology, motivated by their abundant expression both in neurons at the developmental stage and in the adult brain (Kawakita et al., 2003; Koyano et al., 1997; Kubo et al., 2002; Murata et al., 2006). FARP1 and FARP2 have been shown to interact directly with the neuronal axon guidance receptors plexins, thereby participating in the plexin signaling pathways that regulate dendrite outgrowth and axon guidance respectively (Toyofuku et al., 2005; Zhuang et al., 2009). Recently, FARP2 has been found to be involved in the plexin-mediated regulation of bone homeostasis (Hayashi et al., 2012; Takegahara et al., 2010).
FARPs contain the tandem Dbl-homology (DH) and pleckstrin homology (PH) domains characteristic of the Dbl-family guanine nucleotide exchange factors (GEFs) for RhoGTPases (Rossman et al., 2005). Similar to other Dbl-family GEFs, FARPs contain multiple extra domains in addition to the DH-PH module. The N-terminal 4.1, ezrin, radixin and moesin (FERM) domain is linked to the DH domain by a ~200-residue linker of low sequence complexity. They also have a second PH domain (referred to as PH2), which is connected to the first PH domain (PH1) by a ~70-residue linker. While FARP1 and FARP2 share high sequence identity (~60% excluding the FERM/DH linker), previous studies have suggested that they act on different RhoGTPases for signal transduction. FARP1 has been identified as a RhoA GEF (Koyano et al., 2001), whereas the substrate specificity of FARP2 remains ambiguous. Two studies have reported that FARP2 is a GEF for Rac1 but not for RhoA or Cdc42 (Kubo et al., 2002; Toyofuku et al., 2005), whereas several others have shown that FARP2 is exclusively active towards Cdc42 (Fukuhara et al., 2004; Fukuyama et al., 2005; Miyamoto et al., 2003; Murata et al., 2006).
The DH domain in all the Dbl-family RhoGEFs is responsible for catalyzing GTP/GDP exchange, through interacting with the guanine nucleotide-binding switch I and II regions in the RhoGTPase and thereby ejecting the bound guanine nucleotide (Rossman et al., 2005). The PH domain in the conserved DH-PH module of many RhoGEFs regulates GEF activity through interactions with the DH domain and/or the RhoGTPase substrate (Chhatriwala et al., 2007; Derewenda et al., 2004; Kristelly et al., 2004; Lutz et al., 2007; Rossman et al., 2003; Rossman et al., 2002). The composition and organization of domains outside of the DH-PH module diverge greatly among the ~70 Dbl-family RhoGEFs (Rossman et al., 2005). These extra domains provide additional diverse mechanisms for regulating the catalytic activity of the DH domain (Ahmad and Lim, 2010; Bielnicki et al., 2011; Chen et al., 2011; Chrencik et al., 2008; Mitin et al., 2007; Murayama et al., 2007; Rapley et al., 2008; Solski et al., 2004; Yohe et al., 2008; Yohe et al., 2007; Yu et al., 2010; Zheng et al., 2009).
The domain structure of FARP1 and FARP2 is unique in the Dbl family, and their regulation mechanisms are poorly understood due to lack of structural studies. Previous studies have shown that the interaction with plexins and phosphorylation by the Src kinase regulate the GEF activity of FARPs, but the detailed mechanism for these regulation processes are not known (Fukuhara et al., 2004; Fukuyama et al., 2005; Miyamoto et al., 2003; Murata et al., 2006; Toyofuku et al., 2005; Zhuang et al., 2009). We have determined crystal structures of the DH domain and the DH-PH-PH domains of FARP2 at 2.9 and 3.2 Å respectively, and a 4.1 Å structure of the DH-PH-PH domains of FARP1. The structures reveal a multi-layered auto-inhibition mechanism that involves the C-terminal portion of the last helix in the DH domain, the two PH domains and two inter-domain linkers. This auto-inhibited conformation is distinct from those of other RhoGEFs characterized by previous structural studies. The results of our cell-based GEF assays support the regulation of the FARP2 activity by this auto-inhibition mechanism.
The structure of the DH domain of mouse FARP2 displays the typical DH domain fold of an elongated 6-helix bundle (Figure 1A). A protein structure database search using the Dali server suggested that it is most similar to the DH domain of intersectin, a Cdc42-specific GEF (Holm et al., 2008). However, FARP2 shows an unexpected conformational difference in the last helix in the DH domain (Helix α6). While Helix α6 in many other DH domain structures is largely straight, Helix α6 in FARP2 undergoes a sharp kink of ~60° between Leu726 and Gln727. The consequent direction change positions the C-terminal portion of Helix α6 (referred to as Helix α6C) in the groove between Helices α3 and α5. The asymmetric unit of our crystal contains nine protomers of the FARP2 DH domain, which all exhibit the same kinked conformation of Helix α6, supporting that this conformation is not induced by a particular set of crystal contacts (Figure S1).
A superimposition of the FARP2 DH domain with the structure of intersectin bound to its substrate Cdc42 (PDB ID: 1KI1) revealed that part of Helix α6C closely resembles the helical segment in switch II of Cdc42 (Figure 1B) (Snyder et al., 2002). Particularly, Leu730 and Leu733 in Helix α6C superimpose well with Leu67 and Leu70 in Cdc42 and make similar interactions with the hydrophobic groove between Helices α3 and α5 in the DH domain. The interface between the switch II helix and the DH domain is a conserved feature that is essential for the exchange activity (Rossman et al., 2005). Helix α6C in FARP2 therefore acts as a pseudosubstrate-like inhibitor of the GEF activity. Helix α6C in the DH domain of son of seven-less (SOS) (PDB ID: 1DBH) and faciogenital dysplasia gene product 5 (FGD5) (PDB ID: 3MPX) also adopts kinked conformations (Soisson et al., 1998) (Figure 1C). However, in those cases the kink is more acute, and Helix α6C interacts with the middle portion of Helix α5 and does not resemble the RhoGTPase substrate. To the best of our knowledge, no detailed mutational analyses of the inhibitory role of Helix6αC in SOS and FGD5 have been reported.
The structure comparison with intersectin also revealed that the FARP2 DH domain contains substitutions of several conserved residues that are known to be important for RhoGTPase binding and the GEF activity. Most DH domains have a lysine or arginine residue in Helix α5 (Arg1384 in intersectin), which makes electrostatic interactions with both switches I and II in the RhoGTPase (Figures 1B and 1D). The corresponding position in FARP2 and its close homolog FARP1 is a histidine (His690 and His691, respectively) (Figures 1B and 1D). Gln727 in Helix α6 of the FARP2 DH domain replaces a highly conserved asparagine in other DH domains (Asn1421 in intersectin). The asparagine residue makes critical contributions to the GEF activity by forming hydrogen bonds with the N-terminal end of the switch II helix in the RhoGTPase substrate. The longer side-chain of Gln727 in FARP2 may clash with the switch II helix unless the interface undergoes an adjustment. The equivalent residue in FARP1 is a histidine (His728), representing a further deviation. These substitutions of key residues suggest that FARPs diverge from typical DH domains and have compromised active sites, which may partially explain the lack of activity in our in vitro assays (see below).
To uncover the organization of the tandem PH domains in relation to the DH domain and their potential regulation of the GEF activity, we determined the structure of the DH-PH-PH domains of mouse FARP2 at 3.2 Å resolution (Figure 2). We also determined the structure of the DH-PH-PH domains of human FARP1 at 4.1 Å resolution, which is very similar to that of FARP2 (Figure S2). The following discussion will be focused on the higher resolution FARP2 structure unless otherwise stated.
Helix α6C in the DH-PH-PH structure of FARP2 adopts the kinked conformation similar to that in the isolated DH domain, although it tilts more towards and interacts more intimately with the main body of the DH domain (Figures 2B and 2C). This may be a result of the interactions between PH2 and the DH domain, which pull Helix α6C indirectly through PH1 toward the DH domain. The FARP1 structure exhibits the same kinked Helix α6C, which makes similar hydrophobic interactions with the main body of the DH domain (Figure S2).
Similar to many other DH-PH modules, the first PH domain (PH1) in FARPs interacts extensively with Helix α6C (Figure 2B). Unique to FARPs is the second PH domain (PH2), which docks onto the RhoGTPase binding surface on the DH domain. The C-terminal portion of the ~70-residue linker between PH1 and PH2 (the PH1-PH2 linker) forms a well-defined structure motif and interacts with both the PH domains. These inter-domain interactions hold the DH-PH-PH domains together to adopt a highly compact structure. A superimposition with the intersectin/Cdc42 complex structure shows that the entire substrate-binding site in the DH domain of FARP2 is occupied collectively by Helix α6C, the two PH2 domains and the PH1-PH2 linker (Figure 2D).
Our three structures together demonstrate that FARP1 and FARP2 use a conserved auto-inhibition mechanism that involves both Helix α6C in the DH domain and the tandem PH domains. The auto-inhibition is stabilized by the multiple interactions among the domains and the inter-domain linkers. Activation of the GEF activity would require a large-scale conformational change that removes these auto-inhibitory elements from the active site of the DH domain.
PH1 adopts the typical PH domain fold characterized by a 7-strand β-sandwich that is covered at one side by a C-terminal helix (Figure 3A). It is connected to Helix α6 in the DH domain by the 14-residue linker (the DH-PH1 linker) that is highly conserved between FARP1 and FARP2 (Figures 3A and 3C). The surface of the β-sheet formed by Strands β1–4 contains a large number of hydrophobic residues, including Phe779, Phe781, Met784, Leu786, Ile800. These residues together with the conserved non-polar residues in the DH-PH1 linker make hydrophobic interactions with the outward-facing side of Helix α6C in the DH domain (Figure 3B). The interface also includes several hydrogen bonds and charge/charge interactions. These interactions fix the position of PH1 in relation to the DH domain such that it occupies part of the space for binding of the RhoGTPase substrate (Figure 2D). Helix α6C, PH1 and the DH-PH1 linker held together by these interactions likely swing away from the DH domain as one rigid body when FARP2 undergoes activation.
PH domains in many RhoGEFs bind phosphatidylinositol mono- or poly-phosphates (PtdInsP(s)) and thereby recruit the proteins to specific membrane compartments where their RhoGTPase substrates reside (Viaud et al., 2012). The typical binding site is located near the loop between Strands β1 and β2, which contains several positively charged residues that make electrostatic interactions with the negatively charged phosphoinositide head group (Ferguson et al., 2000). PH1 in FARP2 has 7 lysine or arginine residues at the potential binding site, which are mostly conserved in FARP1 (Figures 3A, 3C and S3A). Interestingly, strong difference electron density is present at this site in the FARP2 DH-PH-PH structure, demonstrating binding of a certain compound (Figure S3B). The compound was probably from the expression host E. Coli BL21 (DE3) strain, and was bound to and co-crystallized with the FARP2 protein. The identity of the compound is not known at present, although it is likely negatively charged given the environment of the site. Regardless, these observations together support that PH1 is capable of binding PtdInsP(s) and facilitating membrane localization of FARPs.
The ~70-residue PH1-PH2 linker is disordered except for the 864–868 and the 907–931 segments. The 864–868 segment adopts an extended conformation and docks onto the groove between the DH-PH1 linker and Strand β4 in PH1 (Figure 3A). These interactions may not form in FARP1, since no density is present in the same area in the FARP1 structure and the equivalent segment (the 867–870 segment) is not conserved (Figure 3C). In contrast, the 907–931 segment in FARP2 shows high sequence identity with the 909–933 segment in FARP1 (Figure 3C). This segment forms a well-ordered helix-strand-helix motif and is sandwiched between PH1 and PH2 in both the FARP1 and the FARP2 structures (Figures 2B, ,44 and S2). In the first helical segment, Met910, Cys913 and Trp914 interact with the part of the β1–4 surface in PH1 adjacent to the area contacting Helix α6C (Figures 3A and and4A).4A). The interaction with PH1 is further strengthened by two hydrogen bonds (Figure 4A). The other side of this helix interacts with the β2–3, β4–5 and β6–7 loops in PH2. The interface includes Van der Waals interactions formed by Val912 with Pro975 and Leu977, and two hydrogen bonds between the side-chain of Arg916 and two mainchain carbonyls in PH2. The middle strand segment (residues 918–921) of the linker joins the β1–4 sheet in PH1 and turns it into a 5-strand sheet (Figure 4B). The second helical segment in the linker connects directly to Strand β1 in PH2 and interacts with all the three domains (Figure 4B). At the N-terminal end, Arg922 forms hydrogen bonds with the main-chain carbonyl of Phe759 and the side-chain of Glu758 in PH2. One side of the helix packs against the middle portion of Helices α5 and α6 in the DH domain (Figure 5A), while the opposite side packs against the β2–3 loop in PH2 (Figure 4B). These interactions made by the PH1-PH2 linker effectively glue the two PH domains together, coupling these two inhibitory elements for stronger inhibition of the DH domain.
PH2 docks its surface composed of the β1–4 sheet and the C-terminal helix onto the substrate-binding surface of the DH domain (Figure 5). Gln932 and Arg1017 in PH2 make polar interactions with His690 and Glu545 in the DH domain, respectively. Tyr1013 in PH2 packs its aromatic ring with Phe541 in the DH domain (Figure 5A). At the other end of the interface, Strands β2–4 and the connecting loops in PH2 contribute numerous hydrophobic interactions to the PH2/DH interface (Figure 5B). Particularly, Trp951, Val953, Phe960, Tyr962, Tyr969 and Pro970 in PH2 form a large hydrophobic patch, which packs against the surface around Tyr674 in the loop between Helices α4 and α5 in the DH domain. The PH2/DH interface buries 2170 Å2 surface area and covers virtually the entire substrate-binding face of the DH domain, rendering it completely inaccessible to the substrate. Previous studies have shown that Src-mediated tyrosine phosphorylation of FARP2 contributes to its activation (Fukuhara et al., 2004; Fukuyama et al., 2005; Miyamoto et al., 2003; Murata et al., 2006). Intrinsic dynamics of the FARP2 protein may transiently expose some of the tyrosine residues in the PH2/DH interface, allowing them to be phosphorylated. This would lead to disruption of the interface, providing a plausible mechanism for phosphorylation-triggered activation of the GEF activity.
A search by Dali suggested that the structure of PH2 in FARP2 is most similar to the PH domain in DAPP1/PHISH (PDB ID: 1FAO) (r.m.s.d of 1.5 Å for 85 aligned Cα atoms) (Ferguson et al., 2000) (Figure S4A). It deviates from PH1 substantially, mainly at Strands β6–7 and the β3–4 loop (Figure S4B). In contrast to the relatively straight conformation in PH1, the β3–4 loop plane in PH2 is nearly orthogonal to the axes of the strands. As described above, the β3–4 loop in PH2 uses this conformation to make a major contribution to the PH2/DH interface (Figure 5B). Another notable difference is at the potential PtdInsP(s) binding site. PH2 only has three positively charged residues at this site (Figure S4A), suggesting that it may have lower binding affinity and/or different specificities for PtdInsP(s).
We performed sequence and structural comparisons of FARPs with other well-studied Dbl-family GEFs to identify residues that are potentially involved in determining substrate specificity. One major specificity determining residue in the DH domain is located in Helix α5 (Snyder et al., 2002). While most Cdc42-specific GEFs contain a leucine residue at this position (Leu1376 in intersectin), which makes favorable interactions with Phe56 in Cdc42, Rac1-specific GEFs contain an isoleucine residue to accommodate the larger side-chain of Trp56 in Rac1. RhoA-specific GEFs often possess one or more positively charged residues in the loop between Helices α4 and α5, which interact with the unique negatively charged patch consisting of Asp45 and Glu54 in RhoA (Kristelly et al., 2004; Oleksy et al., 2006; Snyder et al., 2002). Both FARP2 and FARP1 have a leucine (Leu682 and Leu683 respectively) at the Phe/Trp interacting position and lack the RhoA interacting positively charged residues in the α4-α5 loop (Figure 1D), indicating that they may prefer Cdc42 as their substrates. However, as mentioned above, various previous experimental investigations have led to conflicting results regarding the substrate specificities of FARPs.
We used an in vitro GEF assay to measure the activity of FARP2 to Rac1 and Cdc42, each of which has been identified as the exclusive substrate of FARP2 by different studies (Fukuhara et al., 2004; Fukuyama et al., 2005; Kubo et al., 2002; Miyamoto et al., 2003; Murata et al., 2006; Toyofuku et al., 2005). We designed a series of constructs to test the auto-inhibition mechanisms revealed by the structures. These include both the wild type and the Leu730Arg/Leu733Gln (L730R/L733Q) double mutants of the isolated DH domain and the DH-PH domains. The L730R/L733Q mutations are expected to disrupt the pseudosubstrate-like interactions made by Helix α6C, whereas truncation of PH1 and PH2 releases the substrate-binding surface occupied by these domains. The results showed that none of these FARP2 proteins has detectable GEF activity to either Rac1 or Cdc42 (Figure S5A and S5B). Additional assays using nine other RhoGTPases also failed to detect significant activity for FARP2 (Figure S5C-K).
The lack of activity may be partially due to the low intrinsic activity of FARP2 as a result of replacement of several key catalytic residues in the DH domain mentioned above (Figure 1D). It is also possible that the tested mutations are not sufficient for releasing the auto-inhibition and meanwhile stabilizing the active conformation. Particularly, the proper orientation of Helix α6, which is known to be critical for the activity of many RhoGEFs (Rossman et al., 2005), may not be achieved by the L730R/L733Q mutations. In addition, it has been shown that FARP2 activation involves membrane localization, binding of activators and tyrosine phosphorylation (Fukuhara et al., 2004; Fukuyama et al., 2005; Miyamoto et al., 2003; Murata et al., 2006), which were missing in the in vitro assays.
We therefore turned to a GTPase pull-down assay to measure the GEF activity of FARP2 in cells. We introduced the same L730R/L733Q double mutations to both the full-length (FL) and a PH2-truncated version (ΔPH2) of FARP2. The results show that none of these FARP2 constructs significantly increased the levels of GTP-bound Cdc42 in cells (Figure 6A). FARP2 may be indeed not active to Cdc42, or the assay condition used here does not allow detection of this activity, as it has been reported that Cdc42 activation is difficult to detect through the pull-down assay (Gotthardt and Ahmadian, 2007; Mitin et al., 2007; Nalbant et al., 2004). In contrast, expression of the FL construct increased the level of GTP-bound Rac1 when compared with the control (Figure 6B). Similar activation levels were also observed for both ΔPH2 and FLL730R/L733Q. Notably, GTP-bound Rac1 increased much more substantially when ΔPH2L730R/L733Q was expressed. These results demonstrate that Helix α6C and PH2 domain each can independently inhibit GEF activity of FARP2, and full activation requires releasing of both these two inhibitory elements.
Our crystal structures of FARPs reveal a complex auto-inhibitory mechanism that involves multiple domains. The results of the cell-based activity assays suggest that Helix α6C and PH2 are able to exert their inhibitory effects independently of each other. The auto-inhibition is likely strengthened by the inter-domain interactions among the auto-inhibitory domains and the two well-structured linkers. This mechanism differs from the multi-layered auto-inhibition of Vav, in which the secondary interactions all exert their effects indirectly through the core inhibitory element, the so-called acidic domain adjacent to the DH domain (Yu et al., 2010). These complex auto-inhibition mechanisms raise the question of how the GEFs are activated efficiently in response to upstream signals. In the case of Vav, activation is achieved by stepwise phosphorylation events. Initial phosphorylations disrupt the secondary auto-inhibition, facilitating the subsequent phosphorylation and eventual release of the acidic domain. FARP2 has been shown to be activated by Src-mediated phosphorylation, but the sites of phosphorylation and how it triggers activation are not known (Fukuhara et al., 2004; Fukuyama et al., 2005; Miyamoto et al., 2003). The structure of the DH-PH-PH domains of FARP2 reveals a number of tyrosine residues at the PH2/DH interface. Transient exposure of these residues due to intrinsic dynamics of the protein may allow them to be phosphorylated by Src or other kinases, leading to disruption of the PH2/DH interface and initiation of the activation process.
Tyrosine phosphorylation alone is unlikely capable of triggering full activation of FARP2, due to the pseudosubstrate-like auto-inhibition by Helix α6C, which does not contain any tyrosine residue. The weakened auto-inhibition upon phosphorylation may facilitate binding of an allosteric activator, which converts FARP2 to the active conformation in a manner similar to activation of p63RhoGEF by Gαq (Lutz et al., 2007). Mapping the sequence conservation between FARP1 and FARP2 to the FARP2 DH-PH-PH structure reveals a prominent conserved surface patch on the side of the two PH domains distal to the DH domain (Figure S2C), which may serve as the binding site for the allosteric activator. The Ras homolog Rap has been shown to stimulate FARP2 activity in the nectin signaling pathway, although it is not clear whether it does so through direct interaction (Fukuyama et al., 2005). Consistent with our analyses, FARP2 activation in this pathway appears to require both Src-mediated phosphorylation and the presence of GTP-bound active Rap, whereas either one alone is insufficient (Fukuyama et al., 2005). While it has been suggested that semaphorin-triggered dissociation of FARP2 from plexin leads to activation of FARP2 (Toyofuku et al., 2005), the tight auto-inhibition of FARPs shown here argues that dissociation from plexin per se is unlikely to induce spontaneous activation. We have recently shown that plexins are non-canonical GTPase activating proteins for Rap, and can cause localized inactivation of Rap (Wang et al., 2012). Dissociation from plexin may allow FARP2 to diffuse to areas with higher levels of active Rap, providing a putative mechanism for semaphorin/plexin-controlled FARP2 activation.
Lastly, the lack of GEF activity in the in vitro assays and the replacement of key catalytic residues in the DH domain indicate the possibility that FARP2 has impaired GEF activity and functions in signal transduction primarily by acting as a protein-protein interaction module. For example, FARP2 may bind and activate other RhoGEFs in the cell, and thereby indirectly activate RhoGTPases. The conserved surface patch on the DH-PH-PH domains (Figure S2C) may mediate such bindings, or make intra-molecular interactions with the N-terminal FERM domain and control its interactions with other proteins. Regardless whether FARPs are active GEFs or scaffold-like proteins, the structures presented here provide a framework for future analyses of the interactions of FARPs with their binding partners and the mechanisms by which they regulate various signaling pathways.
The coding regions of the DH domain (residues 536–749), the DH-PH domains (residues 536–860) and the DH-PH-PH domains (residues 536–1032) of mouse FARP2 and the DH-PH-PH domains (residues 538–1034) of human FARP1 were all cloned into a modified pET-28(a) vector (Novagen) that encodes an N-terminal His6-tag followed by a recognition site for human rhinovirus 3C protease. Point mutations were introduced by QuikChange reactions (Stratagene). The plasmids were transformed into the E. coli. strain BL21(DE3) and protein expression was induced by 0.1 mM IPTG at 16 °C overnight. Seleno-methoinine replaced proteins were expressed in the same bacterial strain using the protocol as described by Van Duyne et al (Van Duyne et al., 1993). The proteins were purified by using a 1 ml HisTrap column (GE healthcare), and treated with recombinant human rhinovirus 3C protease at 4 °C overnight to remove the N-terminal tag. Further purification was performed by using a Superdex 200 HR 10/30 (GE Healthcare) equilibrated with the buffer containing 20 mM Tris, (pH 8.0), 150 mM NaCl, 10% glycerol and 2 mM DTT. The purified proteins were concentrated and stored at −80 °C.
All crystallization experiments were conducted through hanging drop vapor diffusion at 20 °C. The seleno-methionine replaced protein of the DH domain of FARP2 at 9 mg/ml was crystallized in 100 mM HEPES (pH 7.0), 1.5 M LiSO4. The DH-PH-PH domains of FARP2 at 7 mg/ml was crystallized in 100 mM MMT (containing DL-malic acid, MES and Tris base at the 1:2:2 molar ratio, and adjusted to pH 7.0 by NaOH titration), 20% PEG3350, 0.2M Li2SO4. The seleno-methionine DH-PH-PH domains of FARP1 at 15 mg/ml was crystallized in 100 mM MES (pH 6.4), 0.2 M magnesium formate, 14% PEG3350. Crystals were flash frozen in the crystallization buffers supplemented with ~25% glycerol. All diffraction data were collected at −173 °C at beamline 19ID at advanced photon source (APS, Argonne national laboratory). Data were indexed, reduced and scaled using the HKL2000 package (Otwinowski and Minor, 1997). Detailed information of the crystals and data collection statistics are summarized in Table 1.
The structure of the FARP2 DH domain was solved by single wavelength anomalous dispersion (SAD) by using the Autosol module of the Phenix program package (Adams et al., 2002). The initial experimental map showed clear density for all 9 protomers of the DH domain in the asymmetric unit. Attempts of solving the structure of the DH-PH-PH domains of FARP2 through molecular replacement by using various individual DH and PH domain structures as searching models failed. We then used a model of the entire DH-PH-PH module assembled based on the experimental electron density map of the FARP1 structure (see below). Two copies of the ensemble model were located in the asymmetric unit by Phaser in the Phenix package (Mccoy et al., 2007). Strong difference density appeared at the potential PtdInsP(s) binding site in PH1 after a few rounds of refinement, suggesting binding of a certain small molecule compound. Attempts at determining the identity of the compound by mass spectrometry failed. Several water molecules were placed in the density for refinement and are included in the final model.
A SAD data set to 4.1 Å resolution was collected on a crystal of the seleno-methionine replaced DH-PH-PH domains of FARP1. The Phenix Autosol module was used to locate the selenium sites and generate the initial experimental electron density map. Despite the low resolution, the map clearly defined two protomers of the DH-PH-PH domains in the asymmetric unit. Two copies of the structure of the FARP2 DH domain were docked manually into the density. Structures of the PH domains from PEPP-3 (PDB ID: 2D9Y) and Grp1 (PDB ID: 1FHW) were then docked as PH1 and PH2, respectively. The three docked domains served as the ensemble model for the entire DH-PH-PH module, which was used for solving the structure of the FARP2 DH-PH-PH domains by molecular replacement. After the higher resolution structure of FARP2 was refined, the final model was docked into the SAD electron density of the FARP1 crystal. Due to the low resolution, many side-chains and less well-ordered loops were removed from the model before refinement. Rigid body refinement with each domain treated as one rigid body led to excellent fit without significant changes of the overall organization. Subsequent refinement steps were performed by using the mlhl target function in Phenix, with tight restraints to prevent overfitting.
Model building and structure refinement were performed by using the programs Coot and Phenix, respectively (Adams et al., 2002; Emsley and Cowtan, 2004). Refinement statistics are shown in Table 1. Buried surface area was calculated in the CNS package (Brunger et al., 1998). Molecular structure figures were rendered by the program Pymol (the PyMOL Molecular Graphics System, Schrödinger, LLC.). Sequence alignments were rendered by using ESPript (Gouet et al., 1999).
The in vitro GEF assays were based on (Eberth and Ahmadian, 2009). Cdc42, Rac1, RhoA or RhoC preloaded with N-methylanthraniloyl-GDP (mant-GDP) at 200 nM was incubated with the reaction buffer containing 400 nM unlabeled GDP, 50 mM HEPES (pH7.5), 100 mM NaCl, 5 mM MgCl2. Purified proteins from various constructs of FARP2 were added to the final concentration of 10 µM. The Salmonella virulence effector SopE, a highly active GEF, was used at 0.5 µM as a positive control. Decrease of the fluorescence signal at 440 nm with 360 nm excitation was monitored by a Fluorolog-3 spectrofluorometer at 25 °C. For assays with RhoB, RhoE, RhoG, Rho6, Rho7, Rif and TCL, mant-GDP at 200 nM was in the reaction buffer instead and increase of the fluorescence signal upon its binding to the GTPases was monitored.
Cell-based GEF activity assays of FARP2 toward Rac1 and Cdc42 were based on the selective binding of GTP-bound Rac1 and Cdc42 by Cdc42/Rac1 interactive binding motif (CRIB) from p21 activated kinase (Malliri et al., 2002). All the FARP2 constructs with a C-terminal FLAG-tag were cloned into the pcDNA3.1(+) vector (Invitrogen). Full-length Rac1 and Cdc42 with a N-terminal Myc-tag were also cloned into pcDNA3.1(+). HEK293T cells were cultured in Dulbecco's modified Eagle's medium (DMEM, Invitrogen) supplemented with 10% fetal bovine serum (Invitrogen) and transfected at 80% confluency in 6-well plates using Fugene HD transfection reagent (Promega) according to the manufacturer's instruction. The amounts of plasmids for transfection were optimized to ensure equal protein expressions among different transfection sets. For Rac1 transfections, 0.5 µg of the Rac1 plasmid was used. The amounts of FARP2 plasmids used were: full-length wild type (FL), 3.75 µg; FL with the L730R/L733Q mutations (FLL730R/L733Q), 3.75 µg; FARP2 with PH2 truncated (ΔPH2, residues 1–857), 5.5 µg; ΔPH2L730R/L733Q, 5.5 µg. For Cdc42, 0.25 µg and 0.50 µg of the Cdc42 plasmid were used when transfected with and without the FARP2 plasmids, respectively. The amounts of the FARP2 plasmids were: FL, 4 µg; FLL730R/L733Q, 4 µg; ΔPH2, 5.75 µg; ΔPH2L730R/L733Q, 5.75 µg. The empty pcDNA3.1(+) plasmid was added to adjust the final total amount of DNA to 6 µg per transfection.
Twenty-four hours after transfection, cells were serum starved by replacing the original medium with DMEM containing 0.1% serum for additional 24 hours. Cells were then lysed in the pull-down buffer (50 mM HEPES, pH 7.2, 150 mM NaCl, 2 mM MgCl2, 5% glycerol, 1% (w/v) Triton X-100, 2 mM DTT) containing a protease inhibitor tablet (Pierce) at 1 tablet/10 ml. Lysates were cleared by centrifugation at 14,000 rpm for 10 minutes at 4 °C and incubated with glutathione sepharose beads loaded with purified GST-CRIB fusion (10 µg) for 1 hour at 4 °C. Beads were washed three times with the pull-down buffer and resuspended in a SDS-PAGE sample buffer. The samples were run on SDS-PAGE and analyzed by western blots. Expression levels of FARP2 were detected with the mouse monoclonal anti-FLAG antibody M2 (Sigma). RhoGTPases were detected with a mouse monoclonal anti-Myc primary antibody (Cell Signaling). A rabbit anti-mouse IgG HRP-conjugated antibody (Cell Signaling) was used as the secondary antibody. Levels of GTP-bound Rac1 or Cdc42 were quantified and normalized by the total expression levels of the proteins.
We thank Paul Sternweis and members in the Zhang laboratory for discussions and technical assistance, Michael Rosen for the SopE protein, John Sondek for some of the RhoGTPase plasmids and David King for mass spectrometry. We also thank the staff at the structural biology laboratory at UTSW and at beamline 19ID of APS for assistance with X-ray data collection. X. Z. is a Virginia Murchison Linthicum Scholar in Medical Research at UTSW. The work is supported in part by grants to X.Z. from the American Heart Association (10GRNT3430013), NIGMS (5R01GM088197) and the Welch foundation (I-1702). Use of the beamline at the structural biology center at APS (Argonne National Laboratory) was supported by the United States DOE under contract DE-AC02-06CH11357. X.Z, and X.H. conceived the project. X.H. and X.Z. determined the structures. Y.-C.K. performed the cell-based assays. X.H. and Y.-C.K. performed the in vitro GEF assays. T.R. optimized the protein purification procedure. X.Z., X.H. and Y.-C. K. wrote the paper.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Atomic coordinates and structure factors for the DH domain of FARP2, the DH-PH-PH domains of FARP2 and the DH-PH-PH domains of FARP1 have been deposited in the Protein Data Bank (http://www.pdb.org) with the accession codes 4GYV, 4GZU and 4H6Y respectively.
Supplemental information includes 5 figures.