Allosteric properties of PH domains in Arf regulatory proteins

ABSTRACT

Pleckstrin Homology (PH) domains bind phospholipids and proteins. They are critical regulatory elements of a number enzymes including guanine nucleotide exchange factors (GEFs) and GTPase-activating proteins (GAPs) for Ras-superfamily guanine nucleotide binding proteins such as ADP-ribosylation factors (Arfs). Recent studies have indicated that many PH domains may bind more than one ligand cooperatively. Here we discuss the molecular basis of PH domain-dependent allosteric behavior of 2 ADP-ribosylation factor exchange factors, Grp1 and Brag2, cooperative binding of ligands to the PH domains of Grp1 and the Arf GTPase-activating protein, ASAP1, and the consequences for activity of the associated catalytic domains.

Introduction

Pleckstrin homology (PH) domains are components of hundreds of human proteins that control signaling, membrane trafficking and actin cytoskeleton remodeling.^1-3 The domain is defined by its structure of a 7-β strand sandwich capped at one end by an α helix. One function of PH domains is to target proteins to specific cellular membranes by binding to specific phosphoinositides.^3-5 For example, some proteins containing PH domains will be recruited to membranes in which phosphatidylinositol 3,4,5-triphosphate (PI(3,4,5)P₃) is produced. PH domains can also be recruited to specific sites by binding to proteins, such as guanine nucleotide binding proteins and SH3 domains.^6-9 However, the model of the PH domain binding a single ligand to mediate protein targeting to specific membranes may not apply to most PH domains.^4,5,10

Membrane-associated proteins commonly bind more than one membrane-constrained ligand simultaneously.⁵ Two domains, such as PH, PX, C2 and BAR, that bind membrane components may be constituents of a single protein. Examples include proteins containing a PH and C2 domain such as RASAL and Ras-GAP, PX and BAR domain such as SNX1 or 2 PH domains such as TAPP1. A single PH domain may also bind to 2 ligands simultaneously. In a recent large-scale analysis of the binding of PH domains to liposomes,¹⁰ 56 of the 60 PH domains found to bind liposomes did so cooperatively, leading the authors to conclude that cooperative binding of 2 lipids may be common among PH domains.

A protein is said to have allosteric behavior when binding of a molecule (a ligand or substrate) to one binding or catalytic site affects a second site in the same protein.^11,12 Cooperativity is a property of allosteric proteins and refers to the binding to the sites being codependent.^11,12 The allosteric sites may bind different molecules or identical molecules. In the latter case, the dependence of ligand-protein complex formation or enzymatic activity on ligand or substrate concentration is sigmoidal. In the case of catalytic sites, affinity and/or turnover number may be affected. Allosteric behavior can extend to proteins with more than 2 binding sites. Often, allosteric proteins are composed of multiple identical subunits. However, allosteric behavior is also well described for single polypeptides or protein domains.^11-15

Several mechanisms could account for allosteric behavior of membrane-associated proteins (Fig. 1). Cooperative binding can be the consequence of binding 2 ligands constrained to a surface. When two membrane-constrained ligands bind to a single polypeptide, the entropic cost for binding the second ligand is reduced. The difference in binding energy may result in a sigmoidal dependence of ligand binding.¹¹ Thus, the allosteric behavior in this case is binding of ligand to one site increasing the likelihood of a second binding site encountering ligand (Fig. 1A). The phenomenon of binding 2 distinct ligands enriched on a surface is sometimes called coincidence detection.⁵ Allosterism may also be the consequence of conformational changes or domain rearrangement in the protein on binding ligands.^12-14 The conformational change or domain rearrangement may be within the binding or catalytic site (Fig. 1B), or may expose a binding or catalytic site that would be otherwise occluded by another part of the protein (Fig. 1C). The Arf regulatory proteins described in this review illustrate these mechanisms that underlie allosteric behavior.

Figure 1.

Examples of mechanisms that contribute to allostery. (A) Two binding sites in a protein binding to ligands constrained to a membrane. Binding of ligand 1 (L1) to site 1 (S1) concentrates the protein on the surface, increasing the likelihood of site 2 (S2) encountering ligand 2 (L2). (B) Coordinated conformational changes in binding sites. Binding of ligand 1 to site 1 causes conformational changes in the protein resulting in optimization of binding site 2 for ligand 2. (C) Release of autoinhibition Binding of ligand 1 to site 1 induces a rearrangement of the protein, which exposes site 2 with subsequent binding of ligand 2.

The regulation of Arf proteins requires guanine nucleotide exchange factors (GEFs) and GTPase-activating proteins (GAPs)

Arf proteins are members of the Arf-family within the Ras-superfamily of guanine nucleotide binding proteins.^16-19 Five genes in humans encode Arf proteins, which are divided into 3 classes based on primary sequence. Arf1 and 3 are Class 1, Arf4 and 5 are Class 2 and Arf6 is Class 3. The Arf proteins are most extensively characterized as regulators of membrane trafficking but also affect the actin cytoskeleton and cellular signaling. Arf proteins function as molecular switches. The active GTP bound form associates with membranes where it binds to effectors, while the GDP bound form is inactive. Arf proteins cycle between states by exchanging GDP for GTP and hydrolyzing GTP (Fig. 2). Arf proteins are unique among the Ras-superfamily members in that their intrinsic exchange rates are slow and intrinsic GTPase activity of Arf is undetectable. The cycle of GTP binding and hydrolysis for Arf proteins is therefore exclusively controlled by guanine nucleotide exchange factors (GEFs) and GTPase-activating proteins (GAPs).

Figure 2.

The cycle of GDP for GTP exchange and GTP hydrolysis. Nucleotide exchange is catalyzed by a guanine nucleotide exchange factor (GEF) and GTP hydrolysis is catalyzed by a GTPase-activating protein (GAP).

The GEFs and GAPs for Arf that control the GTP binding/GTP hydrolysis cycle are regulated through multiple mechanisms. Subsets of the GEFs and GAPs have PH domains that are both remarkably diverse between groups and critical to function and regulation.^17,19-22 Figure 3 is a schematic of the domain structure of the GEFs and GAPs that contain PH domains. The diversity of this group of PH domains is evident when the primary sequences and structural models are compared. For example, the PH domains of the AGAPs have an approximately 100 amino acid insert of undetermined structure and function between β strands 5 and 6. The PH domains of cytohesins, namely Cytohesin 1, Grp1, Arno and PSCD4 (Cytohesins 1-4), have an insert in the loop between β strands 6 and 7, which forms a β hairpin (βi1 and βi2) and provides greater specificity for binding the phosphoinositide ligand. Despite the structural diversity, a common theme of cooperative ligand interaction and allosteric behavior has been observed for those PH domains in Arf regulatory proteins that have been studied, including AGAP1 (Luo, Roy and Randazzo, unpublished), ASAP1,²³ Grp1^24,25 and Brag2^26,27; however, consistent with the structural diversity, there is diversity in allosteric mechanism.

Figure 3.

Schematic of Arf GEFs and Arf GAPs containing PH domains. Arf GAPs and GEFs that do not have PH domains are excluded from the figure. Accession numbers: Cytohesin 1, NM_004762; Cytohesin 2/ ARNO, NM_017457; Cytohesin3/Grp1, NM_004227; Cytohesin 4/ PSCD4, NM_013385; EFA6A/ PSD1, NM_002779; EFA6B/ PSD4, NM_012455; EFA6C/ PSD2, NM_032289; EFA6D/ PSD3, NM_015310; BRAG1/IQSEC2, NM_001111125; BRAG2/IQSEC1, NM_001134382; BRAG3/IQSEC3, NM_001170738; ADAP1, NM_006869; ADAP2, NM_018404; ASAP1, NM_018482; ASAP2, NM_003887; ASAP3, NM_017707; ACAP1, NM_014716; ACAP2, NM_012287; ACAP3, NM_030649; ARAP1, NM_001040118; ARAP2, NM_015230; ARAP3, NM_022481; AGAP1, NM_014914; AGAP2, NM_014770; AGAP3, NM_031946; AGAP4, NM_133446; AGAP5, NM_001144000; AGAP6, NM_001077665; AGAP7P, NR_126580 (pseudogene); AGAP9, NM_001190810; AGAP10P, NG_005805 (pseudogene); AGAP11, NM_133447.

Arf GEFs are defined by the presence of a Sec7 domain, which is responsible for catalyzing the exchange of GDP for GTP on Arf family members.^17,22,28,29 Sixteen genes encode Arf GEFs in humans. Three subtypes, namely the Cytohesin, Brag and EFA6 proteins, contain PH domains C-terminal to the Sec7 domain (Fig. 3). The Arno/cytohesin/Grp1 subgroup regulates cell adhesion, migration, cytoskeleton remodeling, positioning of Glut-4 transporters, endocytosis and lysosomal maturation.^30-36 The IQsec/Brag proteins regulate cellular adhesions, at least in part, by regulating the endocytosis of integrins and E-cadherin,^37-41 myoblast fusion^42,43 and endocytosis of AMPA receptors in excitatory synapses.⁴⁴ Brag proteins contribute to cancer cell invasion^45,46 and antiangiogenic signaling,^39,40 and mutations are associated with intellectual disability.⁴⁷ EFA6 proteins control tight junctions, cell adhesion and cytokinesis by activating Arf6, which affects endocytosis and actin remodeling.^48-52 The molecular basis of ligand binding and substrate recognition has been examined for Grp1 and Brag2 as discussed below.

Thirty-one genes in humans encode proteins with Arf GAP domains (Fig. 3).²¹ Of these, 21 have a PH domain immediately N-terminal to the Arf GAP domain. ADAP1 and 2 have 2 PH domains C-terminal to the Arf GAP domain, but GAP activity has yet to be detected in these proteins. In contrast, GAPs with an N-terminal PH domain are active with k_cat/K_m values approximately 10⁸M⁻¹sec⁻¹.⁵³ For the Arf GAPs examined, the PH domains do not have a role in membrane recruitment, but are critical for activity and for regulation of activity.^54,55 Cooperative binding of ligands controls activity of the Arf GAP ASAP1.²³

ASAP1 is composed of BAR, PH, Arf GAP, Ank repeat, proline rich and SH3 domains (Fig. 3). It is targeted to sites of active actin remodeling through interactions with Src, FAK, CrkL and cortactin.^56-59 The ASAP1 gene is amplified in a number of human tumors and the protein is expressed at high levels in cancers, either as a consequence of gene amplification or transcriptional control.^60-63 ARAPs, which bind phosphatidylinositol 3,4,5-trisphosphate (PtdIns(3,4,5)P₃), are composed of a SAM domain, 5 PH domains, an Arf GAP domain, Ank repeat, a Rho GAP domain and a Ras-association domain. The ARAPs control trafficking of a number of transmembrane receptors including α₅β₁ integrin and EGFR and affect cell adhesion, migration, insulin secretion and lymphatic vascular development.^64-69 ARAP3 has been reported to inhibit dissemination of scirrhous gastric cancer.⁷⁰ AGAPs are composed of a G-protein like, PH, Arf GAP and Ank repeat domains. AGAPs bind directly to clathrin adaptor proteins^71,72 and to muscarinic receptor, affecting its trafficking.⁷³ AGAP2 binds to β arrestin, which affects Erk signaling,⁷⁴ and to focal adhesion kinase, which controls focal adhesions.⁷⁵ ACAPs are composed of BAR, PH, Arf GAP and Ank repeat domains and regulate cell adhesion and migration at least in part by affecting trafficking of integrins.^76,77 All 4 subtypes of Arf GAPs may be allosterically regulated.^23,54,76,78 ASAP1 has been the most extensively examined Arf GAP and will be discussed here.

The PH domain of the Arf GEF Grp1 binds PI(3,4,5)P₃ and Arf6•GTP leading to domain rearrangement that relieves inhibition of the associated Sec7 domain

The structures of cytohesin family PH domains with the associated Sec7 domain and in complex with Arf have revealed PH domain rearrangements important to the control of activity of the Sec7 domain. Initial structures of the isolated PH domains of ARNO (cytohesin 2) and Grp1 (cytohesin 3) identified a unique structural element.^79,80 The PH domains have a β hairpin inserted into the loop between β6 and β7, which forms an extension of the β1-β4 sheet and contributes to phosphoinositide binding. In one splice variant of Grp1, the additional coordination of phosphates on the inositol ring provides specificity for PtdIns(3,4,5)P₃ not observed in other PH domains. The insert is also critical for binding a second ligand, Arf6•GTP, important for PH domain rearrangement in the control of exchange factor activity as discussed below.

Cytohesin family proteins have at least 2 allosteric behaviors. The first is related to ligand binding, in which PI(3,4,5)P₃ binding to one site on the PH domain is necessary for Arf6•GTP binding to a second site on the PH domain.^6,25 The allosteric effect of PtdIns(3,4,5)P₃ is independent of concentration on the surface of the lipid bilayer. Native Arf has a myristoylated N-terminal extension that associates with hydrophobic surfaces. When GTP is bound, the N-terminus tightly associates with hydrophobic surfaces, an interaction that is required for stability of Arf•GTP. Truncating the N-terminus of Arf renders the GTP bound form water-soluble. Lambright and colleagues²⁵ used a truncated form of Arf6 ([Δ13]Arf6) and the soluble headgroup of PI(3,4,5)P₃, I(1,3,4,5)P₄, to study Arf6•PH association. They discovered that I(1,3,4,5)P₄ was necessary to detect binding of [Δ13]Arf6•GTP to the Grp1 PH domain.

The mechanistic basis for cooperativity of PtdIns(3,4,5)P₃ and Arf6•GTP binding to the PH domain has yet to be determined. Malaby et al.²⁵ consider 2 mechanisms. One is conformational change in the βi1 –βi2 loop and/or stabilization of the β3/4 loop by PI(3,4,5)P₃ binding to the PH domain could favor Arf6•GTP binding. Neutralization of the charge on the polybasic region by the phosphates in the inositol ring of PtdIns(3,4,5)P₃ may also contribute to binding Arf6•GTP, which has a pI of 8.8 and, therefore, is positively charged at neutral pH. Comparison of splice variants of Grp1 that differentially bind PI(4,5)P₂ and PtdIns(3,4,5)P₃ with different bound inositide geometries⁷⁹ may provide insights to distinguish between the 2 mechanisms.

Cytohesin family proteins have a second allosteric behavior directly involving the PH domain. Arf6•GTP and PI(3,4,5)P₃ bind to the PH domain, inducing a domain rearrangement, resulting in release of autoinhibition of the catalytic Sec7 domain.^24,25 Sec7 domains are composed of 10 α helices^a .^82-86 They bind to Arf through a hydrophobic groove formed by α helices F, G and H and a hydrophilic loop between α helices F and G (Fig. 4B). Switch 2 of Arf forms a 3₁₀ helix that packs against α H of the Sec7 domain and a lysine at the N-terminus of switch 2 ion pairs with a glutamate from α helix G and an aspartate from α helix H of the Sec7 domain. Switch 1 of Arf inserts into the hydrophobic pocket bounded by α helices H and G with the loop between α helices F and G forming a clamp. The PH domains of Grp1 and ARNO position 2 pseudosubstrate motifs to block Arf•GDP from accessing the catalytic Sec7 domain.²⁴ The helix and polybasic region that are C-terminal to the PH domain block the switch 2 binding site of the Sec7 domain by forming hydrophobic interactions and salt bridges with α helices G and H (Fig. 4A). The linker between the Sec7 and PH domains blocks the Switch 1 binding site, binding to the hydrophobic pocket formed by α helices H and G, with acidic residues contacting the α helix F/G loop (Fig. 4A).

Figure 4.

Speculative model complex with the Sec7-PH domain tandem of Grp1 (A) Grp1 PH (orange) and Sec7 (dark blue) domains shown prior to binding Arf6 (pink), the lipid PIP₃ (orange) and Arf1 (green). Two regions important for allostery are indicated by the cyan and yellow ovals, the linker between the Grp1 PH and Sec7 domains (cyan), and the C-terminal helix following the PH domain (yellow, with the polybasic region (blue) comprising the last turn of the helix). These two regions interact with the Sec7 helices, αG and αH, as well as helix αF, which lies beneath αG in this view. Arf6 is in its GTP (red) bound state, with its N-terminal helix and myristoyl group (yellow-green) inserted into the membrane. Arf1 is in its GDP (magenta) bound state with the non-membrane bound N-terminal in coil conformation, its switch 1 and 2 regions labeled s1 and s2 (pale green), and its myristoyl moiety bound in a groove of the protein. (B) Grp1 after its PH domain binds PI(3,4,5)P₃ and the PH domain and C-terminal helix bind Arf6. The displacement of the linker and C-terminal helix from the Sec7 domain allow it to bind Arf1 and remove GDP. The displaced linker and helix locations are indicated by the ovals with solid lines. The former locations of the linker and helix are indicated by the ovals with dashed lines, which are now occupied by the switch 1 and 2 regions of Arf1. (C) Upon binding GTP, Arf1 dissociates, and its N-terminal myristoylated helix forms with insertion into the membrane. The structures incorporated in the model are the crystal structure of the Grp1 Sec7-PH tandem (2R09), crystal structure of the Grp1 PH/Arf6•GTP complex (4KAX), NMR structure of bicelle-bound myristoylated Arf1•GTP (2KSQ), and NMR solution structure of myristoylated Arf1•GDP (2K5U).

Autoinhibition is relieved by PtdIns(3,4,5)P₃ and Arf6 binding to the PH domain,²⁵ which was revealed in the structure of Arf6•GTP bound to Grp1 (Fig. 4). The Arf6•GTP interface is centered on the hydrophobic triad at the switch 1-interswitch-switch 2 junction, which binds to the β sheet β1-β4-βi1-βi2 of the PH domain. The linker, C-terminal helix and polybasic region of Grp1 lie in grooves formed at the periphery of the Arf6-PH domain interface, removing the barrier for Arf•GDP binding to the Sec7 domain (Fig. 4B and C).

Autoinhibition of Grp1 and ARNO is also relieved by phosphorylation. PKC phosphorylates sites within the polybasic region (pbr), which increases activity and increases Arf6•GTP-dependent activation.²⁴ Thus, phosphorylation is an allosteric effect in 2 respects, with phosphorylation at one site affecting the catalytic Sec7 domain and the Arf6•GTP binding site.²⁴ Akt phosphorylates ARNO within the sec7 domain to increase exchange factor activity,⁸⁷ although the molecular basis of the change in activity is yet to be determined. Akt also phosphorylates ARNO within the PH domain, resulting in dissociation of the N-terminal coiled-coil domain of ARNO from the C-terminal part of ARNO with a consequent increase in membrane association.⁸⁸ However, phosphomimetic mutation at the phosphorylation site did not affect in vitro activity ⁸⁷ and phosphorylation by PKC of the pbr, which increases activity, stabilized association of the N-terminal coiled-coil domain with the C-terminus of ARNO,⁸⁸ which is predicted to decrease association with membranes containing substrate Arf•GDP. Further studies directly examining the consequences of phosphorylation of the PH domain on PI(3,4,5)P₃ and Arf6•GTP stimulated GEF activity will be useful for resolving this apparent paradox.

Autoinhibition is a regulatory mechanism that extends to a number of GEFs including VAV, Tim, ASEF, FARP2 and SOS.^89-94 For FARP2, PH domains are central to the mechanism.⁹³ A Rho family GEF, FARP2 contains a catalytic Dbl homology (DH) domain and 2 PH domains. The PH domains stabilize an autoinhibited conformation and directly occlude the exchange factor active site.⁹³ Another Rho GEF, DBS, is similar to Grp1 in binding a phosphoinositide and a GTP-binding protein in the GTP-bound form (in this case Rac1•GTP), simultaneously. ^95,96 The molecular consequences have not been explored in detail. The PH domains of other RhoGEFs, from the LBC family, directly bind to RhoA•GTP, which may increase activity by targeting the GEF to a membrane containing RhoA•GDP, thereby acting in a positive feedback mechanism, which can be considered allosteric (binding to the PH domain affects activity of the associated catalytic domain).⁹⁷

Binding of both phosphoinositide and Arf6•GTP has been reported to recruit Grp1 and ARNO to the plasma membrane, ⁶ where it may encounter the substrates Arf1•GDP and Arf6•GDP, which could be considered another allosteric behavior as defined as binding of ligand to one site, the PH domain, affecting activity of a second site, the Sec7 domain. The PH domain of FAPP1 functions similarly to specifically target FAPP1 to the Golgi apparatus by binding a lipid (PI4P) and protein (Arf1•GTP) simultaneously. ⁸¹

The PH domain of Brag2 forms part of the catalytic interface with Arf•GDP under control of acidic phospholipids

Brag2 is another example of an Arf GEF with a PH domain that contributes to allosteric behavior. Phospholipid binding to the PH domain increases association of substrate, Arf•GDP, with the catalytic Sec7 domain. The specific mechanism, however, is different than described for Grp1. Removing the PH domain of Brag2 results in a loss of Arf GEF activity. The loss of activity is greater if the linker between the GEF and PH domains is also removed.^26,27 These results indicate that the PH domain and linker do not contribute to autoinhibition but, rather, have a positive role in substrate interaction. The crystal structure of Brag2 in complex with [Δ17]Arf1 in part explained the structural basis for contribution of the PH domain and interdomain linker to GEF activity.²⁷

In Brag2, the linker between the PH and Sec7 domains forms a subdomain of the PH domain, packing against the β1, β2 and β3 strands of the PH domain (Fig. 5). This linker makes contact with switch 1 of the substrate Arf•GDP, which contributes to Arf GEF activity. The PH domain, through the C-terminal α helix, contacts the N-terminus of the catalytic Sec7 domain and makes edge contacts with the substrate Arf (Fig. 5B). Kinetic analysis revealed a combination of conformational changes on lipid binding with membrane targeting results in 2000-fold stimulation of GEF activity, leading to the conclusion that lipid binding to the PH domain at least in part controls binding of the substrate Arf•GDP.²⁷ Although not regulated by phospholipids, the PH domains of some Rho exchange factors also contact the substrate Rho family protein.⁹⁸

Figure 5.

Speculative model complex with the Sec7-PH domain tandem of Brag2. (A) Brag2 PH (orange) and Sec7 (dark blue) domains shown prior to binding PIP₂ (light orange) and Arf1 (green). Prior to binding, the Sec7-PH linker (cyan) is presumably unstructured and flexible, though the first portion, indicated by the cyan circle, could have propensity to form helix. Arf1 is in its GDP (magenta) bound, non membrane-bound state, with its myristoyl (yellow-green) lying within a groove at the Arf1 surface. Switch regions 1 and 2 are indicated (pale green). (B) After the Brag2 PH domain binds PIP₂, the linker adopts a folded helical and coil structure, interacting with the PH domain and allowing binding of Arf1 to the Sec7 domain which removes GDP. The linker region contacts the switch 1 region (pale green) and lies near the myristoyl moiety (yellow-green), indicated by the cyan circle. The PH domain C-terminal helix contacts Sec7 and also Arf1, indicated by the white double arrow. (C) Upon binding GTP (red), Arf1 dissociates and its myristoylated N-terminal helix forms with insertion into the membrane. The structures incorporated in the model are the crystal structure of the Brag2 Sec7-PH tandem Arf1 complex (4C0A), NMR structure of bicelle-bound myristoylated Arf1-GTP (2KSQ), and NMR solution structure of myristoylated Arf1•GDP (2K5U).

Arf6 is also a substrate for Brag2. The allosteric mechanism discovered for Arf1 likely extends to Arf6. Arf1 and Arf6 were directly compared as substrates for Brag2.²⁶ The catalytic parameters were similar for both Arf1 and Arf6 as were the effects of phospholipids when using either Arf as a substrate.

The PH domain of ASAP1 cooperatively binds phospholipids to regulate the associated GAP domain by a mechanism that does not involve autoinhibition

The PH domain of ASAP1 does not function to target the protein to cellular membranes. Instead, protein-protein interactions with domains in the C-terminus of ASAP1, including the proline rich domain binding to CrkL, Src and cortactin and the SH3 domain binding to FAK, mediate ASAP1 targeting to specific cellular sites on the plasma membranes.^57,99 The PH domain, however, is critical for GAP activity. Recombinant ASAP1 lacking the PH domain has 1/100,000 the activity of ASAP1 with the PH domain. Lipid binding to the PH domain of ASAP1 stimulates GAP activity approximately 10,000-fold.^{23,55,100,101}

Like Grp1, ASAP1 has at least 2 behaviors that are examples of allosterism. The first is binding to 2 phospholipids cooperatively. PH domains can bind phosphoinositides through 2 sites, a canonical and alternate site (Fig. 6). The canonical site is the pocket between the β1/β2 loop, β3/β4 loop and β6/β7 loop, which corresponds to the phosphoinositide binding site observed in PH domains such as Grp1, Akt, PDK, Btk, FAPP1, TAPP1 and PLCδ1.^4,5 The alternate site is on the opposite side of the β1/β2 loop, and includes interaction with the β5/β6 loop, corresponding to the phosphoinositide binding site in the PH domains of spectrin, TIAM1 and Slm1, for example.^4,5,102,103 The crystal structures of the unliganded ASAP1 PH domain and the PH domain with diC4-PtdIns(4,5)P₂ (a water soluble form of PI(4,5)P₂ with 4 carbon acyl groups) were determined (Fig. 6).²³ DiC4-PtdIns(4,5)P₂ occupied both the canonical and alternate sites. Furthermore, comparison of the unliganded and liganded structures revealed side chain rearrangements with the β1/β2 loop that allows charged residues to contribute to both binding sites. An isoleucine from the β1/β2 loop lies in the canonical binding site in the unliganded form, where it may reduce the affinity of the charged inositol head group to the site. The isoleucine is surface exposed in the liganded PH domain and may contact the lipid bilayer. The authors hypothesized that binding of ligand to one site stabilized the β1/β2 loop in a conformation promoting binding to the second site. Consistent with the idea, binding of the PH domain to phosphoinositides was cooperative, with disruption of one site through mutation resulting in decreased binding to the other site. The alternate site could be occupied by anionic phospholipids, such as phosphatidylserine (PS), in addition to phosphoinositides (Fig. 6A). Control of GAP activity by cooperative binding could confer switch-like properties to the enzyme.

Figure 6.

PH domain canonical and alternate binding sites and speculative model of interface with a phospholipid bilayer. (A) Phospholipid binding sites in ASAP1 PH domain compared to PLCδ1 and β-spectrin PH domains. The membrane-bound ASAP1 PH domain is shown with PtdIns(4,5)P₂ in the canonical lipid binding site and phosphatidylserine (PS) at the alternate phospholipid binding site. For comparison, the PH domain of PLC-δ  (pdb 1MAI) is shown as an example of a PH domain that binds PI(4,5)P₂ in the canonical site and the PH domain of β-spectrin (pdb 1BTN) that binds PtdIns(4,5)P₂ at the alternate site. The stars mark the β1/β2 loops of the PH domains. (B) Superposition of the PH domain of ASAP1 in its lipid-bound (orange, pdb 5C79) and unbound (blue, pdb 5C6R) forms. The lipid PI(4,5)P₂ is bound in the canonical site and PS in the alternate site. The side chain of Ile353 is shown with light blue spheres for the unbound form and white spheres (white star) for the lipid-bound form. In the unbound form Ile353 projects into the canonical binding site, overlapping with where PIP₂ binds. In the bound form Ile353 is displaced away from the canonical site, and interaction is possible with the membrane and bound lipids. Full length acyl chain PtdIns(4,5)P₂ was modeled in to replace the analogs in the crystal structures in panels A and B.

Other PH domains may have cooperative lipid binding sites similar to ASAP1. As described in the introduction, a high-throughput screen indicated that a high proportion cooperatively bind lipids.¹⁰ In earlier work, simultaneous occupancy of the canonical and alternate sites was proposed for the PH domain of Slm1, a target of TORC2 in the PI -3-kinase pathway and of sphingomyelin signaling.^104,105 Akt, a serine/threonine kinase that mediates signaling in the PtdIns 3-kinase pathway, also contains a PH domain that may bind phospholipids cooperatively.^10,106-108 Akt binds PI(3,4)P₂ and PtdIns(3,4,5)P₃ through a structured canonical site. It also has a patch of positive surface charge contributed by amino acids in the β1/β2 and β5/β6 loops, a position similar to the alternate site seen, for example, in ASAP1, Slm1 and TIAM1. A mutant of Akt with one of the basic amino acids in the patch changed to a neutral amino acid has reduced kinase activity. Akt has also been reported to depend on PS, in addition to phosphoinositides, which may be related to the second site. A constitutively active mutant of Btk ([E41K]) also binds 2 anionic lipids simultaneously. ¹⁰⁹

Other lipid binding domains have been reported to have 2 integrated lipid binding sites. For example, the phox homology (PX) domain of p47phox has sites for PI(3,4)P₂ and phosphatidic acid (PA). A hydrophobic ridge separates the sites and is thought to insert into lipid bilayers when both the PtdIns(3,4)P₂ and PA sites are occupied.¹¹⁰ Another example is the C2 domain of PKCα, which binds to PI(4,5)P₂ and PS cooperatively.¹¹¹ Thus, cooperative binding of multiple lipid ligands may be a general property of lipid binding domains.

The second allosteric behavior of ASAP1 is phospholipid binding to the PH domain leading to increased catalytic activity of the associated GAP domain. The molecular mechanism by which cooperative binding to the PH domain of ASAP1 controls GAP activity remains to be determined. Previous studies have indicated dynamic interaction between the PH domain and the Arf GAP domain in solution,¹⁰¹ and other studies have shown that both domains interact with Arf.¹⁰⁰ In solution, the canonical lipid-binding site is occluded by the interaction between the PH and Arf GAP domains, thus significant reorganization must occur for binding lipid at the membrane (Fig. 7). While ASAP1 constructs consisting of the PH and Arf GAP domains have detectable activity with membrane-bound Arf, the activity is enhanced 15-fold in the presence of PS, and enhanced 1,500-fold further with the addition of PtdIns(4,5)P₂.²³ This suggests that lipid binding plays some role orienting the PH and, consequently the catalytic GAP domain, at the membrane such that its interaction with Arf favors GTP hydrolysis. In line with this idea is the structural reorganization seen for the β1/ β2 loop upon binding the 2 lipids, with the exposed isoleucine likely favoring additional membrane interaction affecting the orientation of the PH domain at the membrane. In its GTP bound form, the Arf N-terminal helix and myristoyl are inserted in the membrane, while in the GDP bound form, the protein dissociates from the membrane with the N-terminal and myristoyl interacting with the protein (Fig. 7).¹¹² Perhaps interaction of Arf with the lipid-bound PH domain facilitates retraction of the N-terminus from the membrane, though whether such retraction follows or precedes GTP hydrolysis is not known.

Figure 7.

Speculative model complex with ASAP1. (A) The model includes the PH domain and the Arf GAP domain, designated ZA for Zn-binding and Ankyrin repeats, and shows one of the many orientations consistent with solution NMR data. ASAP1 PH (orange) and Arf GAP (maroon, labeled ZA) domains shown prior to binding PIP₂ (light orange), phosphatidylserine (PS, yellow), and Arf1 (green) in its GTP (red) bound, membrane bound state. In its unbound state, the PH domain interacts with the Arf GAP domain via dynamic interactions with the PH loop regions. Isoleucine 353 in the β1/β2 loop of the PH domain is shown in white, interacting with the back of the Arf GAP domain in this view (white circle). (B) The PH domain binds PIP₂ in its canonical binding site and a second phospholipid at its alternate site, here PS. This causes a rearrangement in the β1/β2 loop structure, exposing the isoleucine, possibly resulting in insertion into the membrane (white circle). With both lipids bound, we speculate that the PH domain forms part of the interaction interface with Arf1 in a conformation that allosterically enhances Arf1 GTP hydrolysis by the Arf GAP (ZA) domain. (C) After hydrolysis to GDP (magenta), Arf1 dissociates from ASAP1 with its N-terminal helix exiting the membrane and the myristoyl moiety inserting into its groove at the Arf1 surface, allowing Arf1 to diffuse into the cytosol. The structures incorporated in the model are the crystal structure of the ASAP1 Arf GAP domain (1DCQ), the crystal structure of the ASAP1 PH domain (5C79), the NMR structure of bicelle-bound myristoylated Arf1•GTP (2KSQ), and NMR solution structure of myristoylated Arf1•GDP (2K5U).

Conclusions

The PH domain is a common protein fold that regulates protein function through diverse mechanisms. Membrane targeting mediated by binding phospholipid or membrane-bound protein is one mechanism for controlling the activity of the enzymes containing PH domains. Binding to 2 ligands simultaneously can increase specificity for a particular membrane. However, cooperative binding, such as that described for the Arf regulatory proteins Grp1, Brag2 and ASAP1, is accompanied by conformational changes that can also influence the enzymatic activity of the protein containing the PH domains. The results of a high-throughput screen examining 91 PH domains indicated a large proportion bind lipids cooperatively, raising the possibility that the mechanisms described for the Arf regulatory proteins may generalize to other families of enzymes and signaling proteins that contain PH domains or other lipid binding domains.

Abbreviations

Arf

ADP-ribosylation factor

DH

Dbl homology

GAP

GTPase-activating protein

GEF

guanine nucleotide exchange factor

PI(4,5)P₂

phosphatidylinositol 4,5-bisphosphate

PtdIns(3,4,5)P₃

phosphatidylinositol 3,4,5-trisphosphate

PH

pleckstrin homology

Disclosure of potential conflicts of interest

No potential conflicts of interest were disclosed.

Acknowledgment

We thank Tamas Balla for critical review of the manuscript.

Funding

The work was supported by the Intramural Program of the National Cancer Institute, project number BC 007365.