Protein Science (2003), 12:1934-1953.
Copyright © 2003 The Protein Society
Molecular modeling of the membrane targeting of phospholipase C pleckstrin homology domains
Shaneen M. Singh and
Diana Murray
Department of Microbiology and Immunology, Weill Medical College of Cornell University, New York, New York 10021, USA
Reprint requests to: Diana Murray, Department of Microbiology and Immunology, Weill Medical College of Cornell University, 1300 York Avenue, Box 62, New York, NY 10021, USA; e-mail: dim2007{at}med.cornell.edu; fax: (212) 746-8587.
(RECEIVED March 3, 2003;
FINAL REVISION June 3, 2003;
ACCEPTED June 5, 2003)
Article and publication are at http://www.proteinscience.org/cgi/doi/10.1110/ps.0358803.
 |
Abstract
|
|---|
Phospholipases C (PLCs) reversibly associate with membranes to hydrolyze phosphatidylinositol-4, 5-bisphosphate (PI[4,5]P2) and comprise four main classes: ß, 
, 
, and 
. Most eukaryotic PLCs contain a single, N-terminal pleckstrin homology (PH) domain, which is thought to play an important role in membrane targeting. The structure of a single PLC PH domain, that from PLC1, has been determined; this PH domain binds PI(4,5)P2 with high affinity and stereospecificity and has served as a paradigm for PH domain functionality. However, experimental studies demonstrate that PH domains from different PLC classes exhibit diverse modes of membrane interaction, reflecting the dissimilarity in their amino acid sequences. To elucidate the structural basis for their differential membrane-binding specificities, we modeled the three-dimensional structures of all mammalian PLC PH domains by using bioinformatic tools and calculated their biophysical properties by using continuum electrostatic approaches. Our computational analysis accounts for a large body of experimental data, provides predictions for those PH domains with unknown functions, and indicates functional roles for regions other than the canonical lipid-binding site identified in the PLC1-PH structure. In particular, our calculations predict that (1) members from each of the four PLC classes exhibit strikingly different electrostatic profiles than those ordinarily observed for PH domains in general, (2) nonspecific electrostatic interactions contribute to the membrane localization of PLC-, PLC-, and PLCß-PH domains, and (3) phosphorylation regulates the interaction of PLCß-PH with its effectors through electrostatic repulsion. Our molecular models for PH domains from all of the PLC classes clearly demonstrate how a common structural fold can serve as a scaffold for a wide range of surface features and biophysical properties that support distinctive functional roles.
Keywords: Phospholipase C; pleckstrin homology domain; bioinformatics; continuum electrostatics; membrane association; computational biology
|
Introduction
|
|---|
Pleckstrin homology (PH) domains are small protein modules (
130 residues) that are present in many proteins involved in cellular processes in which membrane association plays an integral role, for example, signal transduction, vesicular trafficking, and cytoskeletal rearrangements (Lemmon and Ferguson 2000; Hurley and Meyer 2001). A number of PH domains have been implicated in membrane targeting through specific interactions with particular phosphoinositides, but in most cases, the structural basis for these interactions has not been elucidated. Indeed, function for the majority of PH domains remains to be established. To date, structures for 14 PH domains have been solved. In each case, the structural core consists of a ß-sandwich of two nearly orthogonal ß-sheets. One end of the ß-sandwich is capped by a long C-terminal 
-helix, and the other end contains three "variable loops," which are highly diverse in both length and amino acid composition (Lemmon and Ferguson 2000). Despite having the same overall structural fold, PH domains have sequences that share quite low similarity, reflecting the diversity in membrane-binding behavior observed throughout the family (Rebecchi and Scarlata 1998).
Most of the PH domains of known structure are electrostatically polarized, with a highly basic surface corresponding to the position of the three variable loops; structural and biochemical studies implicate this region as the lipid-binding site. Comprehensive model building of the PH domain family indicates that this electrostatic polarization is a feature common to many PH domains (Blomberg and Nilges 1997; Blomberg et al. 1999; Lemmon and Ferguson 2000). Recently, however, a class of PH domains from Caenorhabditis elegans has been characterized that exhibits a highly negative electrostatic profile (Blomberg et al. 2000), consistent with the notion that PH domains may perform functions other than directly mediating membrane association (Blomberg et al. 1999; Rhee 2001; Philip et al. 2002). Given the large size and extreme diversity in sequence and function of the PH domain family, it is not practical to characterize in detail the family as a whole. Therefore, we further subclassify PH domains according to the protein families in which they occur and, in the present study, focus on the set of PH domains found in one such family, the phosphoinositide-specific phospholipases C (PLCs). However, as shown in the literature and further indicated by the work presented here, even the PH domains from this single protein family exhibit a rich array of biophysical properties and functions.
PLCs constitute a large family of closely related enzymes that reversibly associate with membranes to carry out hydrolysis of its membrane resident phosphoinositide substrate, phosphatidylinositol 4,5-bisphosphate (PI[4,5]P2; Rhee 2001). PLCs can be organized into four classes—ß, , , and —each containing various isoforms (Rebecchi and Pentyala 2000; Rhee 2001). In addition, another protein similar to PLCs, PLC-L or p130, has recently been identified (Kanematsu et al. 1996; Takeuchi et al. 2000). All of the PLC isozymes are soluble multidomain proteins comprised of a core catalytic X/Y domain and various combinations of regulatory domains, for example, PH, C2, SH2, and SH3 domains. The regulatory domains serve to target PLCs to the vicinity of their substrates, activators, and/or effectors through protein–lipid or protein–protein interactions (Rhee 2001). Most eukaryotic PLCs contain a single PH domain located in their N-terminal region. The structure of the PLC1 PH domain was determined experimentally and has been instrumental in explaining how its interaction with PI(4,5)P2 targets PLC1 to membrane surfaces (Ferguson et al. 1995). Other structural studies indicate that the PLC1 PH domain is attached to the rest of the enzyme by a flexible linker region and may, thus, serve a membrane tethering role (Essen et al. 1996). Furthermore, PH domains from a number of PLC classes have been shown to facilitate the interaction of the enzyme with membrane surfaces (Falasca et al. 1998; Pawelczyk and Matecki 1999; Wang et al. 1999; Razzini et al. 2000; Matsuda et al. 2001; Varnai et al. 2002). Moreover, experimental studies provide evidence that the membrane-binding properties of PH domains from different PLC classes are quite distinct: The PLC- PH domains bind specifically and with high affinity to membranes containing PI(4,5)P2 (Rebecchi et al. 1992; Pawelczyk and Lowenstein 1993; Ferguson et al. 1995; Garcia et al. 1995), and the PLC- PH domains target membranes containing phosphatidylinositol 3,4,5-triphosphate (PI[3,4,5]P3; Falasca et al. 1998; Pawelczyk and Matecki 1999). In contrast, the PLC-ß PH domains bind nonspecifically to membranes of a wide range of lipid compositions (Runnels et al. 1996; Wang et al. 1999). However, in some cases, PLC PH domains have also been shown to mediate protein–protein interactions, which indicates that they perform functions beyond that of a simple membrane tether (Wang et al. 1999; Thodeti et al. 2000; Chang et al. 2002; Philip et al. 2002). The goal of the present study is to use computational models to rationalize the observed functional differences for the PLC PH domains and to provide predictions when function is unknown.
Recent studies using homology modeling and calculations of electrostatic properties of both experimentally determined structures and computational models have been successful in describing the membrane-binding behaviors of a number of other lipid-interacting signal transduction domains, namely, C2, FYVE, and PX domains. For example, finite difference Poisson-Boltzmann (FDPB) calculations explained how calcium binding can drive the association of some C2 domains to negatively charged membranes and others to neutral zwitterionic membranes (Murray and Honig 2002). In addition, the biophysical properties of homology models for the C2 domains from PLC isoforms and 5-lipoxygenase were shown to correlate with the calcium-dependent lipid-binding preferences of the C2 domains in both biochemical and cellular assays (Ananthanarayanan et al. 2002; Kulkarni et al. 2002). Electrostatic calculations also provide a model for how the binding of both FYVE and PX domains to poly-phosphoinositides facilitates the membrane penetration of these domains: The multivalent, negatively charged poly-phosphoinositides neutralize strong regions of positive potential surrounding hydrophobic residues that are adjacent to the lipid-binding sites, thus decreasing the unfavorable desolvation that occurs upon inserting these residues into the membrane interface (Stahelin et al. 2002; Diraviyam et al. 2003; Stahelin et al. 2003). These studies illustrate the utility of computational approaches in describing, at the molecular level, the regulation of lipid-interacting protein domains.
As illustrated by the multiple sequence alignment in Figure 1
, the sequences of the PLC PH domains share a limited number of conserved residues, many of which correspond to hydrophobic residues that contribute to the structural core of the domains. This lack of sequence conservation is illustrated more dramatically by the all-on-all pairwise sequence comparisons for these domains (see Electronic Supplemental Material). It is clear that sequence similarity is relatively high within classes, but is much less statistically significant for sequences from different classes. In agreement with experiments, this diversity in sequence indicates the existence of class-specific properties and regulatory roles. By considering sequence information alone, it is possible to explain lipid-binding specificity for only the 2, 3, and 4 PH domains (see Results). However, this characterization relies on the knowledge obtained from the structure of the 1 PH domain complexed with Ins(1,4,5)P3. Furthermore, although the lipid-binding specificity of the PH domains is known, it has previously been noted that it is difficult to extract a sequence motif indicative of PI(3,4,5)P3 binding due to a number of alternative ways in which PH domains have been shown to bind this lipid (Baraldi et al. 1999; Ferguson et al. 2000; Lietzke et al. 2000). Moreover, sequence information alone cannot account for nonspecific contributions to membrane-binding, nor can it distinguish those PH domains that do not target membranes. Thus, it would be of great utility to examine how biophysical properties are arrayed on the surfaces of the PH domains. To this end, we modeled the three-dimensional structures of the PLC PH domains by using a number of different computational tools, including fold recognition, sequence-to-profile alignments, and homology modeling. We then calculated the biophysical properties of our models by using continuum electrostatic approaches (Honig and Nicholls 1995). The results of this analysis are robust with respect to alternative models for a given sequence. This is significant because although the core of the PH domains is expected to be well conserved structurally, there is uncertainty in modeling the surface loops, which are quite variable across the PLC classes. Importantly, we found that many members of the PLC-PH family do not exhibit the electrostatic polarity characteristic of the majority of PH domains (Blomberg and Nilges 1997; Blomberg et al. 1999; Lemmon and Ferguson 2000), something that could not be predicted on the basis of sequence alone. In addition, the analysis of our models shows that their biophysical and structural features correlate well with observed functional behaviors and provides the basis for experimentally testable hypotheses for those domains with unknown functions.
View larger version (89K):
[in this window]
[in a new window]
|
Figure 1. Multiple sequence alignment of the N-terminal PLC PH domains. The sequences were aligned with the program Pattern-Induced (local) Multiple Alignment (PIMA) using the default parameters (Smith and Smith 1992). Columns in the alignment are shaded from light gray to dark gray to black to denote increasing numbers of conserved residues at those positions. White lettering indicates residues that are conserved across all or nearly all sequences. Residues of the PLC1 PH domain that directly contact Ins(1,4,5)P3 are highlighted in orange. Note that the alignments with respect to PLC1-PH are not necessarily the alignments used in homology modeling (see Materials and Methods). The Swiss-Prot, PIR, and NCBI accession numbers and residues delimiters for the sequences in the alignment are as follows: beta1_bov, P10894
[GenBank]
(16–139); beta2_hum, Q00722
[GenBank]
(11–135); beta3_mus, P51432
[GenBank]
(16–140); beta4_rat, Q9QW07(11–134); epsilon_hum, AAG28341
[GenBank]
(544–599); gamma1a_bov, P08487
[GenBank]
(18–144); gamma2a_hum, P16885
[GenBank]
(11–133); p130_rat, NP_445908
[GenBank]
(105–224); delta2_bov, S14113
[GenBank]
(10–127); delta3_hum, AC002117
[GenBank]
; delta4_rat, Q62711(10–126); and delta1_rat, P10688
[GenBank]
(12–131). For more detail, see Electronic Supplemental Material.
|
|
The focus of our computational analysis on a subset of PH domains, those from the PLC isoforms, constitutes a family-specific strategy for functional annotation in the absence of experimentally determined structures. Consistent with this notion, out of 14 possible PH domain structures, the structure of PLC1-PH was identified as the best structural representative for all of the PLC PH domain sequences we examined, and homology models constructed based on the alignment of the PLC PH domain sequences to this structural template all scored well according to structure evaluation analysis. This indicates that our subclassification is reasonable. Our approach provides detailed models for the molecular basis of the interactions of the PLC PH domains with membranes, as well as with other proteins, through the detection of biophysical similarities and differences both within and across the PLC classes. The examination of the PH domains from each PLC class, in turn, results in a more complete picture of their role in the regulation of PLC isozymes than is possible through the analysis of the domains taken either individually or more broadly in the context of the entire PH domain family.
|
Results
|
|---|
The analysis of the structure and function of PH domains is a challenging task due to the diversity in their sequences and membrane-binding modes. However, by integrating information available on sequence, structure, and function with homology model building and the calculation of electrostatic potentials, we derive computational models that can account for experimental observations and lead to useful biological predictions that can be tested experimentally. All of our models score well according to the structure evaluation programs (see Materials and Methods), indicating that the conclusions based on them are reliable. The sequences and coordinate files representing our models for all mammalian PLC PH domains, as well as other supplementary information, for example, GRASP images and Verify3D profiles, are available at our Web site (http://maat.med.cornell.edu/ph_PVsuppl.html).
PLC-
Although four distinct PLC isoforms (1, 2, 3, and 4) have been identified, the mechanisms that activate and regulate these enzymes remain unclear. The sensitivity of PLC isozymes to Ca2+ is greater than that of the other classes. Hence, increases in the intracellular Ca2+ concentration alone may be sufficient to trigger activation of PLC (Rhee 2001). Studies of the role of the PH domain in regulating PLC activity have focused on PLC1 due to the availability of experimentally determined structures for both its PH domain and catalytic core (Ferguson et al. 1995; Essen et al. 1996). Experiments show that both PLC1 and PLC1-PH bind with high affinity and stereospecificity to membranes containing PI(4,5)P2 (Rebecchi et al. 1992; Pawelczyk and Lowenstein 1993; Garcia et al. 1995). It is, therefore, thought that the PH domain localizes or tethers the enzyme to membranes containing its substrate (Essen et al. 1996). Ligand-binding studies demonstrate that Ins(1,4,5)P3, a product of the hydrolysis of PI(4,5)P2 by PLC1, competes effectively with PI(4,5)P2 for binding to PLC1-PH and may, thus, interfere with membrane attachment and serve as a negative feedback regulator of catalysis (Cifuentes et al. 1994; Lemmon et al. 1995; Hirose et al. 1999).
Figure 2A shows the backbone structure of PLC1-PH along with electrostatic equipotential contours, which illustrate the electrostatic polarity characteristic of many PH domains (Lemmon and Ferguson 2000). The PLC1-PH structure revealed that the 4- and 5-phosphoryl groups of Ins(1,4,5)P3 interact with the side-chains of specific amino acids present in the variable loops of the domain (Fig. 1
, orange highlights), which accounts for the preference for PI(4,5)P2 over other poly-phosphoinositides. Although the binding of 1-PH to PI(4,5)P2 has been well characterized, not much is known about the phosphoinositide or membrane-binding properties of the other isoforms. As illustrated in Figure 2, A–D, the structure and our models for the PLC PH domains share similar biophysical features: All have a deep PI(4,5)P2-binding pocket that is highly positively charged. Consistent with this, the sequences of the PH domains from the 2, 3, and 4 isozymes share a relatively high level of identity with the 1-PH sequence (33%–38% identity). Based on the multiple sequence alignment of the PLC PH domains (Fig. 1), many of the residues identified in the PLC1-PH structure as making direct contacts with Ins(1,4,5)P3 are conserved in the other isoforms. Therefore, it is expected that all PH domains will bind PI(4,5)P2, although the affinity may vary among the different isoforms because not all Ins(1,4,5)P3-binding residues are strictly conserved across the multiple alignment. For example, nine of the 10 key ligand-binding residues observed in the PLC1-PH/Ins(1,4,5)P3 complex (highlighted in Fig. 1: K30, K32, W36, R38, R40, E54, S55, R56, K57, and T107) are either identical or similar at the corresponding positions in the 3-PH sequence. Based on the conservation of amino acids at the 1-PH ligand-binding positions, we predict that the PH domains have decreasing affinity for PI(4,5)P2 in the order 1 3 > 4 > 2.
View larger version (59K):
[in this window]
[in a new window]
|
Figure 2. Electrostatic properties of the PLC1 PH structure and homology models for the 2, 3, and 4 PH domains. In all panels, the electrostatic potentials were calculated in 0.1 M KCl and contoured at +1 kT/e (blue) and –1 kT/e (red) by using the program GRASP (Nicholls et al. 1991). The structure and models are represented as C backbone traces. (A) PLC1-PH structure (PDB identifier: 1mai
[PDB]
). (B–D) Homology models for the 2, 3, and 4 PH domains. (E–H) Same as in A through D, except that Ins(1,4,5)P3 (yellow spheres) is docked in the lipid-binding loops. The net charge assigned to Ins(1,4,5)P3 for the electrostatic potential calculations is -5, the net charge it would have as the headgroup of intact PI(4,5)P2.
|
|
Experiments indicate that interactions in addition to those directly mediated by PI(4,5)P2 may be important for membrane association. For example, mutating basic residues on the variable loops that are not involved in direct interactions with PI(4,5)P2 moderately weakens the binding to membranes containing PI(4,5)P2; this implies the existence of secondary, lower-affinity membrane association sites within the PH domain (Yagisawa et al. 1998). These secondary sites may mediate interactions with anionic phospholipids in the membrane and contribute to membrane association through nonspecific electrostatic interactions. A role for electrostatic interactions in the membrane association of PLC1-PH is also indicated by observations that the presence of monovalent acidic phospholipids in PI(4,5)P2-containing vesicles increases the binding of the isolated PH domain by 10-fold and that PLC-1 activity is stimulated by interaction with the monovalent acidic lipid phosphatidic acid (Rebecchi et al. 1992; Garcia et al. 1995; Henry et al. 1995). Interestingly, the 3 PH domain is predicted to lack the electrostatic polarity of 1-PH and is, instead, engulfed in a strong, contiguous positive potential profile (Fig. 2C), which indicates that 3-PH should interact strongly and nonspecifically with negatively charged membrane surfaces (see below). This is consistent with the observation that PLC3 interacts with membranes containing PI(4,5)P2 or phosphatidic acid in a PH domain–dependent manner (Pawelczyk and Matecki 1999).
Our calculations based on the FDPB method (see Materials and Methods) predict that nonspecific electrostatic interactions between the positively charged regions of the -PH domains and acidic phospholipids in the membrane contribute significantly to the membrane localization of these PH domains. Figure 3 illustrates the electrostatic component of the membrane-binding free energy for the PLC1-PH structure and the PLC3-PH homology model as a function of distance between the surfaces of the domains and membrane. As depicted by the free energy curves, the electrostatic attraction to a membrane containing 33 mole % acidic lipid; that is, 2 : 1 phosphatidylcholine (PC)/ phosphatidylserine (PS; PC, z = 0; PS, z = -1), in 0.1 M KCl is long-range and, thus, could facilitate the diffusion of the PH domains to membranes containing acidic phospholipids. As seen in Figure 3A, the predicted minimum electrostatic free energy of interaction between PLC1-PH and a 2 : 1 PC/PS membrane is quite strong (–4 kcal/mole), and that for PLC3-PH even stronger (-9 kcal/mole). The minimum electrostatic free energy of interaction with a membrane of lower negative surface charge density (5 : 1 PC/PS; Fig. 3B) is correspondingly weaker. These calculations indicate that nonspecific electrostatic interactions alone may be sufficient to localize these PH domains, especially PLC3-PH, to membranes containing acidic phospholipids, and that the interaction with membrane regions containing higher negative surface charge density is significantly favored. The nonspecific accumulation of a PH domain at the membrane surface may then facilitate the specific 1 : 1 interaction with membrane-embedded PI(4,5)P2.
Our FDPB calculations also provide an explanation for why 1-PH prefers soluble Ins(1,4,5)P3 to membrane embedded PI(4,5)P2 (Lemmon et al. 1995). As described above, the nonspecific association of the PH domains with membrane surfaces involves two main electrostatic components: (1) a "coulombic" attraction between the positively charged PH domain and acidic lipids in the membrane and (2) a short-range desolvation penalty due to the loss of favorable interactions between the aqueous solvent and the charged and polar groups on both the PH domain and membrane when they are in close apposition. These two components are illustrated in Figure 3 by the electrostatic free energy curves for the 1 and 3 PH domains. Far from the membrane, the PH domains experience an electrostatic attraction toward the negatively charged membrane surface, but at short distances, the desolvation penalty dominates and the electrostatic interaction becomes much less favorable and, in some cases, repulsive (Fig. 3B). Thus, the minimum electrostatic free energy of interaction with the membrane is attained when the surfaces of the PH domains and membrane are separated by a distance of about the thickness of a layer of water. This orientation maximizes the coulombic attraction and minimizes the desolvation repulsion (Ben-Tal et al. 1996). When the PH domains bind PI(4,5)P2, they are anchored at the membrane surface and are no longer necessarily free to assume an orientation that minimizes the nonspecific electrostatic contribution to the membrane association. Importantly, as illustrated in Figure 2, E through H, the PI(4,5)P2 headgroup dramatically alters the electrostatic profile of the PH domains. In particular, the electrostatic potential in the membrane-binding region of 1-PH becomes predominantly negative. The PI(4,5)P2-bound PH domain would, thus, experience highly repulsive electrostatic interactions with the surrounding membrane environment due both to unfavorable coulombic interactions with negatively charged lipids and the desolvation penalty that occurs close to the membrane surface. These nonspecific interactions would oppose the energetically favorable specific interactions with PI(4,5)P2.
To get an idea of the magnitude of the nonspecific repulsion, we performed FDPB calculations of the membrane interaction of 1-PH when it is bound to Ins(1,4,5)P3. Table 1 compares the electrostatic free energy of interaction of both the Ins(1,4,5)P3-free and Ins(1,4,5)P3-bound forms with a 2 : 1 PC/PS membrane in 0.1 M KCl for small separations between the domain and membrane. In all cases, the interaction of the Ins(1,4,5)P3-bound form is highly repulsive. These calculations indicate that when the bulk membrane contains acidic phospholipid, the interaction of 1-PH with soluble Ins(1,4,5)P3 is energetically more favorable than the interaction with membrane-embedded PI(4,5)P2. In the former case, the PH domain retains the favorable specific interactions with the lipid headgroup without the repulsive electrostatic interactions due to the surrounding membrane environment that are experienced in the latter case.
View this table:
[in this window]
[in a new window]
|
Table 1. Comparison of the electrostatic free energy of interaction of the PLC1 PH domain in the absence and presence of bound Ins(1,4,5)P3 with a 2 : 1 PC/PS membrane in 0.1 M KCl
|
|
PLC-L/p130
p130 was originally identified as a catalytically incompetent PLC with lipid-binding properties similar to those of PLC1 (Kanematsu et al. 1992; Yoshida et al. 1994). Although it has the same domain organization as PLC1, that is, PH, EF-hand, catalytic X/Y, and C2 domains, it exhibits distinct functional characteristics. Even though p130-PH is capable of binding PI(4,5)P2 in membranes, both the PH domain and intact protein are found principally in the cytoplasm of cells (Takeuchi et al. 2000; Várnai et al. 2002). This is in sharp contrast to PLC1 and the 1-PH domain, which are both predominantly localized to the plasma membrane. p130 has been implicated in the down-regulation of Ins(1,4,5)P3-mediated Ca2+ signaling (Kanematsu et al. 1996; Takeuchi et al. 2000). Studies show that the PH domain, which binds with high affinity to Ins(1,4,5)P3, is responsible for this functionality (Takeuchi et al. 2000): The sequestration of cellular Ins(1,4,5)P3 by the p130 PH domain prevents the activation of Ins(1,4,5)P3 receptors and the subsequent increase in intracellular calcium.
We compared the biophysical properties of our homology model for p130-PH with those of the PLC1-PH structure. As described above, the membrane targeting of PLC1-PH is mediated by specific interactions with PI(4,5)P2 as well as nonspecific electrostatic interactions between basic residues surrounding the lipid-binding pocket and acidic lipids in the bulk membrane environment. In addition, one of the membrane-binding loops of PLC1-PH contains a pair of hydrophobic residues (Val and Met) that are well positioned to penetrate the membrane interface when the PH domain binds PI(4,5)P2 (Fig. 4, green). The corresponding elements in p130-PH combine to produce a PH domain that we predict should interact significantly more weakly with PI(4,5)P2-containing membranes than does PLC1-PH. First, as shown in Figure 5, the potential profile surrounding the membrane-binding loops of p130-PH is significantly less positive than that of 1-PH and contains a region of negative potential. Therefore, the favorable nonspecific electrostatic free energy between p130-PH and the membrane is expected to be diminished with respect to 1-PH. Second, although the sequences are quite similar, p130-PH is missing some of the residues implicated in mediating interactions with the 1- and 5-phosphates of PI(4,5)P2 (Fig. 4). Third, the residue corresponding to the valine in PLC1-PH is aspartic acid in p130-PH (Fig. 4). In contrast, all of the PH domains contain a nonpolar residue in this position (Fig. 1). The aspartate contributes to the negative potential profile of p130-PH (Fig. 5) and is adjacent to basic residues that are expected to directly contact the PI(4,5)P2 headgroup (Fig. 4). Therefore, if the domain was bound to PI(4,5)P2, the aspartate would be positioned close to the membrane surface and would, thus, result in electrostatic repulsion from the negatively charged membrane surface due both to unfavorable charge-charge interactions and desolvation. Given its proximity to potential PI(4,5)P2-binding residues, it may also disrupt specific interactions with the lipid headgroup.
PLC-ß
The PLCß class is comprised of four isoforms: ß1, ß2, ß3, and ß4, and is distinguished from the other PLC classes by the presence of a long C-terminal coiled-coil sequence (Singer et al. 2002) that is implicated in membrane association (Kim et al. 1996) and in activation by heterotrimeric G-protein subunits (Rhee 2001). Experiments indicate that the ß class exhibits a broad subcellular distribution. PLCß has been shown to be present in the nucleus (Martelli et al. 1992) and associated with both the soluble and particulate fractions from cells (Smrcka and Sternweis 1993). Biochemical measurements indicate that PLCß1 and PLCß2 bind strongly and nonspecifically to phospholipid (PC, PS) membranes and that the interaction is independent of the presence of poly-phosphoinositides (Romoser et al. 1996; Wang et al. 1999). In addition, complementary experiments with the isolated PH domains indicate that they have similar membrane-binding properties as their intact enzymes (Wang et al. 1999). However, a recent study indicates that PLCß1-PH specifically recognizes PI(3)P (Razzini et al. 2000).
Figure 6 depicts our alignment of the sequences and secondary structure elements of the ß3 and 1 PH domains. The correspondence in the locations of the predicted secondary structure elements of ß3-PH with those from the 1-PH structure indicates that the alignment is suitable for homology modeling despite the low sequence similarity. Similar results were obtained for the PH domains from other ß isoforms (see Electronic Supplemental Material).
As illustrated in Figure 7, the electrostatic profiles of our ß PH domain models are strikingly different from those of other PH domains. In contrast to the observations of Razzini et al. (2000), our models predict that these domains are unable to interact specifically with phosphoinositides. In all cases, the canonical lipid-binding site is predominantly negatively charged (Fig. 7). Moreover, our models predict that regions other than this site may mediate membrane partitioning. As depicted in Figure 7, all of the models have a significant patch of surface-exposed hydrophobic residues (Leu, Phe, Tyr, Trp) located mainly on ß-strands 3 and 4. Leu, Phe, Tyr, and Trp residues have been shown, in the context of model peptides, to partition favorably into the interface of phospholipid membranes (Wimley and White 1996). In addition, the models for ß2- and ß3-PH have significant patches of surface-exposed basic residues (Fig. 7, blue meshes). Figure 8 depicts the electrostatic free energy of interaction of our model for the PLCß2 PH domain with a 2 : 1 PC/PS membrane in 0.1 M KCl as a function of the distance between the van der Waals surfaces of the domain and membrane. Nonspecific electrostatic interactions are predicted to contribute significantly to the membrane-binding free energy, that is, -5 kcal/mole; this is similar to what is predicted for the PLC1-PH domain (Fig. 3). The nonpolar residues may partition hydrophobically into the interface of both electrically neutral and negatively charged membranes, whereas the basic residues may mediate electrostatic interactions with acidic phospholipids. Thus, our models support the observation that the ß PH domains bind nonspecifically to phospholipid membranes (Wang et al. 1999).
View larger version (115K):
[in this window]
[in a new window]
|
Figure 7. Electrostatic properties of the homology models for the PLCß PH domains. Electrostatic equipotential profiles for the PLCß PH domain models were calculated and visualized in GRASP (Nicholls et al. 1991): ß1-PH (A), ß2-PH (B), ß3-PH (C), and ß4-PH (D). The loops that correspond to the PI(4,5)P2-binding loops in the PLC1-PH structure are located at the bottom of the images. All ß PH domains are in an orientation similar to that as the PLC PH domains in Figure 2, except for ß2-PH (B), which is rotated 180 degrees about the vertical axis to highlight its large basic surface patch. Clusters of surface-exposed hydrophobic residues (Leu, Phe, Tyr, Trp) are colored green. The blue and red meshes represent, respectively, the +1 kT/e and -1 kT/e equipotential contours for 0.1 M KCl.
|
|
View larger version (14K):
[in this window]
[in a new window]
|
Figure 8. Electrostatic free energy curve for the membrane interaction of the ß2 PH domain. The electrostatic free energy of interaction (in kcal/mole) with a 2 : 1 PC/PS membrane in 0.1 M KCl as a function of the distance between the van der Waals surfaces of the PH domain and membrane. The free energies were calculated by using the finite difference Poisson-Boltzmann (FDPB) method as described in Materials and Methods (Gallagher and Sharp 1998).
|
|
The basic surface patches found on our PLCß-PH models may also mediate the observed interaction with Gß-subunits of heterotrimeric G proteins (Wang et al. 1999, 2000; Barr et al. 2000). Our previous computational studies indicate that Gß is oriented at the membrane surface such that the basic surface patch surrounding the site of prenylation on the -subunit interacts with acidic lipid headgroups (Murray et al. 2001). Gß has a net charge of -12 and is dramatically electrostatically polarized. The basic surface patch on Gß that is implicated in membrane binding is localized to the region surrounding the prenyl group. The rest of the protein is highly negatively charged, indicating that Gß and PLCß PH domains interact through the complementarity of their electrostatic surfaces (positive on the PH domain, negative on Gß). Indeed, there is a serine residue in PLCß3-PH whose phosphorylation is implicated in disrupting the interaction with Gß (Xia et al. 2001). Our model for PLCß3-PH predicts that this serine is located in the center of a basic surface patch that could serve as a potential Gß-binding site. As shown in Figure 9, phosphorylation of this serine significantly reduces the positive potential in this region of the PH domain and may, thus, disrupt the interaction with Gß through an electrostatic repulsion mechanism.
PLC-
The class of PLCs consists of two isoforms, PLC1 and PLC2, both of which are activated by polypeptide growth factor stimulation of receptor and nonreceptor protein tyrosine kinases (Cockcroft and Thomas 1992; Rhee and Bae 1997). Both isoforms have two PH domains, one in their N terminus and one occurring in the linker sequence between the two halves of the catalytic domain. Experimental studies indicate that the N-terminal PLC PH domains are targeted to the plasma membrane and bind specifically to PI(3,4,5)P3 in a PI3 kinase–dependent manner (Falasca et al. 1998; Matsuda et al. 2001).
The PLC-PH sequences are slightly more similar to the 1-PH sequence than are the ß-PH sequences. Our PLC PH domain models exhibit the characteristic electrostatic polarization typical of PH domains that bind poly-phosphoinositides (e.g., cf. Fig. 10 and Fig. 2A). As predicted for the PH domains, nonspecific electrostatic interactions between basic residues in the lipid-binding loops and acidic phospholipids in the membrane are expected to contribute to the membrane localization of the PH domains.
Figure 11 depicts the multiple structure-based sequence alignment for the PLC1 and 2 PH domain models with PI(3,4,5)P3-binding PH domains of known structure. Because the sequences in this group share such low similarity, superposition of the structures with our homology models is the most reliable way of deriving a multiple sequence alignment. From the experimentally determined structures, it is known that residues in the ß1/ß2 loop are important for dictating PI(3,4,5)P3 specificity and that residues in the ß3/ß4 and ß6/ß7 loops also contribute, but to varying degrees (Falasca et al. 1998; Baraldi et al. 1999; Cullen and Chardin 2000; Ferguson et al. 2000; Lemmon and Ferguson 2000; Lietzke et al. 2000; Thomas et al. 2001). For PLC1-PH, experiments determined that the ß3/ß4 loop plays a key role in mediating the interaction with PI(3,4,5)P3 (Falasca et al. 1998). As pointed out by others (Baraldi et al. 1999; Ferguson et al. 2000; Lietzke et al. 2000) and as illustrated by the alignment in Figure 11, the residues in the lipid-binding loops are not well conserved across this group of PH domains and do not provide a simple consensus sequence for PI(3,4,5)P3 specificity.
However, based on the multiple structure–based sequence alignment (Fig. 11), we are able to identify several residues in the PLC-PH sequences that together may define a PI(3,4,5)P3-binding function. In our models, there is a conserved basic residue in the ß3/ß4 loop that has been shown to be important for membrane-binding activity in other PI(3,4,5)P3-binding PH domains. For example, a Glu-to-Lys mutation in Btk-PH (Fig. 11, arrow) has been reported to contribute to the binding to PI(3,4,5)P3-containing membranes by increasing nonspecifically the membrane affinity of the PH domain (Li et al. 1995). In addition, a lysine at this position in DAPP1-PH forms direct contacts with the PI(3,4,5)P3 headgroup (Ferguson et al. 2000). Therefore, the basic residue at the same position in the PLC PH domains may also form specific interactions with the PI(3,4,5)P3 headgroup. Moreover, the alignment in Figure 11 indicates that the PLC PH domains have other functionally important residues at conserved positions: (1) ß-Strand 2 contains basic and aromatic residues at its N and C termini, respectively, that are highly conserved across the PI(3,4,5)P3-binding PH domains, but not the PLC PH domains, which bind PI(4,5)P2; and (2) a number of basic residues occur in the ß1/ß2 and ß6/ß7 loops, which have been shown in other PI(3,4,5)P3-binding PH domains to contribute to lipid specificity. Overall, the characteristic electrostatic polarity of our models (Fig. 10) and the conservation of residues that interact with PI(3,4,5)P3 in other PH domains of known structure (Fig. 11) indicate a similar function for the PLC PH domains and provide a molecular basis for the experimentally observed interaction with PI(3,4,5)P3 (Falasca et al. 1998; Matsuda et al. 2001).
PLC1 and 2 also possess a purported second PH domain that is located within a sequence insert between the two halves of the catalytic domain (X and Y). Our analysis predicts that in both cases, the sequence corresponding to the additional PH domain is split into two parts by a span of sequence that corresponds to the one SH3 and two SH2 domains that are characteristic of the PLC class. Experimental studies (Chang et al. 2002) indicate that the N-terminal portion of this "split PH domain" binds specifically to phosphoinositides, including PI(4)P and PI(4,5)P2. Other studies implicate the domain in protein–protein interactions (Thodeti et al. 2000; Chang et al. 2002).
While attempting to identify the segments of sequence that correspond to these split PH domains, we found significant ambiguities in the sequence assignments provided by different databases and domain detection programs (Schultz et al. 1998; Boekmann et al. 2003; Wheeler et al. 2003). We, therefore, used a combination of fold recognition and secondary structure prediction (see Materials and Methods) to identify the sequence segments corresponding to the two halves of each split PH domain. Once assembled into contiguous pieces (excluding the sequences representing the SH domains), these sequences are predicted to have all of the characteristic secondary structure elements of the PH domain fold (Fig. 12). In addition, these "patched" sequences pick up PH domain structures with very high statistical significance when submitted to fold recognition programs (data not shown). Therefore, these sequences have the potential to assemble into PH domains. Indeed, the homology models representing these split PH domains score very high according to structure validation analysis (see Materials and Methods; Electronic Supplemental Material). Furthermore, a recent study demonstrated that a functional PH domain assembles from two separate sequence segments (both attached to coiled-coil forming regions) which correspond to the N- and C-terminal halves of PLC1-PH, establishing, in principle, the possibility of the formation of a stable PH domain structure from noncontiguous sequence segments when they are tethered together (Sugimoto et al. 2003). However, each of the PLC split PH domains has an insert in the ß3/ß4 loop, which corresponds to the SH3 and SH2 domains, which we have not attempted to include in our models. The impact of these large insertions on the structural organization and function of the split PH domains is unknown.
The electrostatic profiles of our homology models for the split PH domains are overall negative with minor positive regions (data not shown). Therefore, contrary to previous studies (Chang et al. 2002), our models indicate that these domains do not bind poly-phosphoinositides. However, the experimental studies were based on a construct that corresponds to only the N-terminal portion of the domain rather than the complete domain, which we have modeled here. Hence, the experimental characterization may not be an accurate representation of lipid-binding function if the two portions of the split domain, identified here, do indeed come together and fold up into a complete PH domain structure.
PLC-
PLC is a recently discovered member of the PLC family and has been shown to be regulated by the small GTPase H-Ras and -subunits (G12) of heterotrimeric G proteins (Kelley et al. 2001; Rhee 2001; Wing 2001). PLC was originally distinguished by the absence of an N-terminal PH domain, but in agreement with the recent work of Wing et al. (2001), we have detected a potential PH domain by aligning the PLC sequence with the PH domain sequences from other PLC classes by using multiple sequence and secondary structure alignments. Figure 13 depicts the alignment between the sequences of the PLC1 and PH domains. Although the details in the sequence assignments differ, our alignment and that of Wing et al. (2001) both predict a long insert between the ß3- and ß4-strands. In addition, Wing et al. (2001) predict that the C-terminal -helix of -PH extends significantly beyond that of 1-PH and that this "molten helix" may be important for mediating interactions with Gß subunits as seen for other PH domains (Fushman et al. 1998; Carman et al. 2000). The ß3/ß4-loop typically contributes to lipid binding in those PH domains that are known to associate with poly-phosphoinositides. In our model, the structure of the insertion is not well defined due to the absence of a suitable structural template. However, because the insertion is located in a loop region, it most likely does not affect the core structure of the PH domain.
As shown in Figure 13, the predicted secondary structure elements of PLC-PH are well aligned to the core structural elements of PLC1-PH. The large insertion is predicted to have a short -helical segment flanked by regions with random coil conformation, although its structure and orientation with respect to the PH domain core is uncertain based on our analysis. As illustrated in Figure 14, the electrostatic profile of the PLC-PH domain model is predominantly negative with a positive region in the vicinity of the insert. The ß3/ß4 insertion could contribute to membrane association because it contains 11 hydrophobic residues, several of which (five Leu and one Phe) have been shown, in the context of model peptides, to partition favorably into the membrane interface (Wimley and White 1996), as well as three Lys residues that may interact electrostatically with acidic lipids. Hence, like other long unstructured sequences in proteins such as MARCKS, Gap43, and phospholipase D (Wang et al. 2002), the long ß3/ß4-loop may interact with membrane surfaces through nonspecific electrostatic and hydrophobic interactions. Such membrane-adsorbed sequences have been shown to sequester PI(4,5)P2 in phospholipid vesicles (Wang et al. 2002). Hence, this sequence may function to localize the enzyme in the vicinity of its substrate. The insert contains a serine at position 903 of the PLC sequence (Fig. 13, arrow), which Prosite (Falquet et al. 2002) predicts with high probability to be a protein kinase C phosphorylation site. This indicates a potential mechanism whereby membrane association is regulated: Phosphorylation would introduce negative charge into a region containing basic and hydrophobic residues, produce an electrostatic repulsion from the membrane surface, and, thus, weaken the membrane association of the insert sequence. Although the PLC PH domain may contribute to nonspecific membrane association, it is clear from Figure 14 that its electrostatic profile, like those for the ß and split PH domains, deviates significantly from the electrostatic polarization of most PH domains previously studied (Lemmon and Ferguson 2000; Hurley and Meyer 2001).
|
Discussion
|
|---|
Although PH domains share a common structural fold, they are highly diverse in their sequences, biophysical properties and functions, which makes it particularly challenging to derive general "rules" that characterize their targeting and regulatory roles. Computational studies that examine broadly the entire PH domain family highlight conserved characteristics, but provide a coarse-grained, incomplete, functional annotation (Blomberg and Nilges 1997; Blomberg et al. 1999). In contrast, computational studies that focus on a single PH domain provide significantly more detailed information on the molecular basis of function, but lack context with regard to related PH domains (Rong et al. 2001). Here, we have chosen to focus on a subset of PH domains that belongs to the same protein family, the phosphoinositide-specific PLCs. This type of examination allows for detailed molecular-level descriptions of the individual PH domains and provides an additional dimension to functional annotation through comparisons among related PH domains.
Despite low conservation in the sequences of the PH domains from the different PLC classes (Fig. 1), the most suitable structural template for modeling them was consistently identified as the PH domain from PLC-1 (Protein Databank [PDB] identifier: 1mai
[PDB]
; Ferguson et al. 1995). A major finding of our study is that even though we used a single PH domain of well-defined structure and function as the template for homology modeling, the biophysical properties of our derived models are quite distinct but correlate well with the observed characteristics of the respective PH domains. (See Materials and Methods for a discussion of the reliability of our models and conclusions.) This supports the notion that the structural core of the PH domain serves as a stable scaffold that can support a wide range of functions (Lemmon and Ferguson 2000). Moreover, this study brings to light PH domains with electrostatic profiles that differ markedly from the polarized profile commonly associated with PH domains: (1) Unique among the PH doma
TOP
Abstract
Introduction
K mutation (denoted by arrow); 1b55
