The Solution Structure and Dynamics of the Pleckstrin Homology Domain of G Protein-coupled Receptor Kinase 2 ( b -Adrenergic Receptor Kinase 1) A BINDING PARTNER OF G bg SUBUNITS*

The solution structure of an extended pleckstrin homology (PH) domain from the b -adrenergic receptor kinase is obtained by high resolution NMR. The structure establishes that the b -adrenergic receptor kinase extended PH domain has the same fold and topology as other PH domains, and there are several unique features, most notably an extended C-terminal a -helix that behaves as a molten helix, and a surface charge polarity that is extensively modified by positive residues in the extended a -helix and the C terminus. These observa-tions complement biochemical evidence that the C-ter-minal portion of this PH domain participates in protein-protein interactions with G bg subunits. This suggests that the C-terminal segment of the PH domain may function to mediate protein-protein interactions with the targets of PH domains.

The solution structure of an extended pleckstrin homology (PH) domain from the ␤-adrenergic receptor kinase is obtained by high resolution NMR. The structure establishes that the ␤-adrenergic receptor kinase extended PH domain has the same fold and topology as other PH domains, and there are several unique features, most notably an extended C-terminal ␣-helix that behaves as a molten helix, and a surface charge polarity that is extensively modified by positive residues in the extended ␣-helix and the C terminus. These observations complement biochemical evidence that the C-terminal portion of this PH domain participates in proteinprotein interactions with G ␤␥ subunits. This suggests that the C-terminal segment of the PH domain may function to mediate protein-protein interactions with the targets of PH domains.
G protein-coupled receptor kinases (GRKs) 1 are a unique family of serine-threonine kinases, which are responsible for activator-dependent phosphorylation of G protein receptors and provide rapid desensitization of the agonist occupied receptors (1). The GRKs are recognized to have three functional components: an N-terminal section believed to interact directly with the seven-trans-membrane helical receptor protein and/or other membrane targets, a central section, which is the catalytic domain, and a C-terminal section containing a generally conserved autophosphorylation region and a variable region that mediates membrane association by various means. In GRK2 (also known as ␤-adrenergic receptor kinase-1) or GRK3 (␤-adrenergic receptor kinase-2), the C-terminal variable region contains a pleckstrin homology (PH) domain (2,3), conferring binding specificity to G ␤␥ proteins (reviewed in Ref. 1).
The PH domain family (reviewed in Refs. 4 -6) appears to be a very large family of structurally homologous protein domains of moderate to low sequence similarity. The PH domain is believed to play a role in intracellular signal transduction, and the functional role of the PH domain has been characterized for several systems. In phospholipase C␦, the PH domain has a high affinity (K d Ͻ 1 M) site for phosphatidylinositol 4,5bisphosphate (PI(4,5)P 2 ) and inositol 1,4,5-trisphosphate (Ins(1,4,5)P 3 ) (7), which forms a crystallographically observed, well defined structural interaction (8). Other PH domains have different lipid specificities, and a well defined set of binding motifs does not readily emerge (9 -15). One hypothesis is that PH domains present a framework with a polymorphic surface used for specific recognition, analogous to immunoglobulins (5,9). In addition, the overall fold of the PH domain was observed to be common with that of the PTB (phosphotyrosine binding) domain (16,17), a protein domain that shows little sequence homology to PH domains. In light of these developments, it is of significance to establish whether the nominal PH domain of GRK-2 (␤-adrenergic receptor kinase (␤ARK-1)), which clearly binds (both in vivo and in vitro) to a protein partner, G ␤␥ subunits of the heterotrimeric G protein family (18), truly has the common structural motif of the PH/PTB domains, and what the relationship of putative lipid and protein binding sites might be in such a structure.
In this paper, we present the solution structure of an extended PH domain from human ␤ARK1, determined at 0.4 Å r.m.s.d. by high resolution NMR using heteronuclear triple resonance methods. Although the overall fold and topology clearly establishes that the ␤ARK1 extended PH domain is of the same family as other PH domains, there are several significant alterations (most notably an extension of the C-terminal ␣-helix, which in solution presents as a "molten helix" having a clear gradient of mobility) on the subnanosecond, as well as millisecond to microsecond, time scales, increasing toward the free C terminus. The polarity of the surface charge observed in other PH domains is altered by positively charged residues in the extended ␣-helix. This unusual clustering may be complemented by a highly negatively charged area on G ␤␥ subunits. Although a direct study of the G ␤␥ /PH domain complex is beyond the range of current NMR technology, the structure presented here supports a model in which the C-terminal portion of ␤ARK PH domain in particular, and PH domains in general, participate in protein-protein interactions.

MATERIALS AND METHODS
Sample Preparation-Recombinant human ␤ARK PH domain (h␤ARK1-(556 -670)) was obtained by GST fusion expression from pGEX-2T (Pharmacia Biotech Inc.) in BL21 (DE3) Escherichia coli cells (Novagen, Madison, WI) and subsequent bacterial expression and protein purification as described previously (19) on a larger scale. The full-length h␤ARK1 cDNA clone was provided by Dr. Antonio DeBlasi (Mario Negri Sud, Santa Maria Imbaro, Italy). The sequence of the 119-residue construct used in the present study is shown in Fig. 1. It contains both the PH domain and the G ␤␥ -binding motif (20). The first four residues are not from the natural sequence. Uniform 15 N and 15 N, 13 C labeling was achieved by growing the cells in M9 minimal medium using standard procedures.
Solutions used for NMR studies contained 1-2 mM protein in 10 mM acetate buffer at pH 4.5 (uncorrected for isotope effects), 0.02% sodium azide, 1 mM [U-2 H]EDTA, 5 mM [U-2 H]dithiothreitol, and 10% 2 H 2 O in the H 2 O samples. These low salt and low pH conditions were necessary to prevent protein aggregation. CD data indicated no changes in the protein secondary structure, between the buffer used for NMR studies and phosphate-buffered saline, pH 7.2. The external 1 H chemical shift reference used was sodium 2,2-dimethyl-2-silapentane-5-sulfonate, and indirect referencing was used for 15 N (21) and 13 C. Spectra were essentially identical among several preparations of the PH domain.
NMR Spectroscopy-NMR experiments were run on Bruker DMX-500 and DMX-600 spectrometers. Quadrature detection was achieved by the States or States-time proportional phase incrementation methods. Some of the pulse schemes implemented pulse field gradients for coherence selection (HCCH-TOCSY, 13 C-separated NOESY-HMQC), and some used the sensitivity enhancement method (HSQC, heteronuclear NOE) (22). The water signal was suppressed either by the WATERGATE method (23) or by using selective on-resonance irradiation during a relaxation delay of ϳ1.3 s. Experiments were run at 35°C with sweep widths of 8000 and 2000 Hz for 1 H and 15 N (at 600 MHz), respectively, unless indicated otherwise.
The homonuclear experiments, HOHAHA and NOESY, were run in both H 2 O and 2 H 2 O using standard pulse sequences and phase cycling. A range of t 1 increments from 200 to 512, each consisting of 2048 complex points, was typically acquired with 32-128 scans/increment.
The heteronuclear experiments consisted of two-dimensional HSQC ( 1 H-15 N and 1 H-13 C), HSQC-J, and HTQC, three-dimensional CBCA-(CO)NH, CBCANH, HNCA, HN(CO)CA, HNCO, HCCH-TOCSY, and 13 C-separated NOESY-HMQC, and two-and three-dimensional 15 Nseparated NOESY-HMQC (in H 2 O and in 2 H 2 O) and HOHAHA-HMQC. The mixing time in the NOESY-HMQC experiments was 100 and 150 ms, and the spin lock duration was 30 ms in the HOHAHA-HMQC and 19 ms and 6 ms in the HCCH-TOCSY. The three-dimensional spectra were recorded with a 32 ϫ 100 ϫ 1000 hypercomplex matrix, with 32 scans/increment. The degree of amide hydrogen protection was assessed (a) by measuring hydrogen-deuterium exchange rates by following the intensity of cross-peaks in HMQC experiments after exchanging a fully protonated, lyophilized sample with 99.996% 2 H 2 O, and (b) by comparison of cross-peak intensities in two HSQC experiments, with and without water presaturation (24). The three-bond HЈ-H␣ coupling was assessed by the method of (25). Heteronuclear 15 N{ 1 H} NOEs were measured using standard methods as described elsewhere (26). Twodimensional H 2 O-selective heteronuclear 15 N-edited ROESY experiments (27) were performed to map those amide hydrogens in the ␤ARK PH domain that are exposed to and interacting with water molecules. Signal processing and assignment were done as discussed previously (26,28).
Structure Calculation-Structure calculations used DIANA with RE-DAC strategy (29) or DYANA (30) with ECEPP stereochemistry, with structurally significant constraints of 1956 upper and 76 lower distance bounds (from ϳ3000 NOEs), 38 hydrogen bonds chosen within strands or helix with slowed exchange, 99 -angle constraints derived from 3 J HNH␣ coupling constants, and 99 -angle constraints (derived from C␣ chemical shift data) corresponding to conservative ranges of allowed torsion angles, in those regions of strand or helix that were well defined (Fig. 2). All peptide bonds were assumed to be trans. A final selection of 20 structures from 400 starting structures was done by using the lowest target functions (the ensemble statistics are shown in Table I). DIANA and DYANA use no assumptions about protein energetics, other than van der Waals repulsion; structures are unrefined and only adjusted by rotation/translation for comparison purposes. Structures were aligned using XPLOR or in-house software written in MATLAB (MathWorks) and displayed and analyzed with the INSIGHTII package (Biosym) or with MOLMOL (31).  (50) to be sufficient and optimal for G ␤␥ binding, below which is the construct used here, which has the same G ␤␥ binding. The lowercase "gshm" residues are from the GST construct, and are not further referred to. At the bottom, the complete sequence of h␤ARK1 PH domain, and the similar h␤ARK2 are compared, with the secondary structural elements of h␤ARK1 superimposed in color. The more flexible region of the C-terminal ␣-helix is shown in light blue.   Table II. The ␣-helical insertions in the loop regions of the PH domains were not taken into account. The alignment was done by direct calculation of r.m.s.d. values and optimized by relative shift of the protein sequences within each secondary structure element (␤1-␤7 strands, ␣-helix), as well as by adding or removing individual residues. The resulting alignment and r.m.s.d. values are presented in Tables II and III, respectively. Protein Backbone Dynamics-The backbone dynamics were assessed via 15 N spin relaxation studies comprising T 1 , T 2 , and heteronuclear steady state NOE measurements using previously described protocols (26). Fifteen two-dimensional spectra with the relaxation delays of 4 (ϫ2), 200, 400 (ϫ2), 600, 900 (ϫ2), and 1200 ms (positive initial 15 N magnetization), and 4, 200 (ϫ2), 400, 600 (ϫ2), and 900 ms (negative initial 15 N magnetization) were acquired in the alternate-sign T 1 experiment (duplicate experiments are indicated by ϫ2) (26). Eleven two-dimensional spectra were collected for the T 2 measurements, with the relaxation delays of 8 (ϫ2), 16, 32 (ϫ2), 48, 64 (ϫ2), 80, 96 (ϫ2), 112, 128 (ϫ2), and 160 ms. The heteronuclear { 1 H} 15 N steady state NOEs were assessed as a ratio of cross-peak intensities in the experiments with and without proton presaturation. The relaxation data analysis was performed using programs RELAXFIT and DYNAMICS (26), ex-tended to include anisotropic character of the overall motion of the protein (32).

RESULTS
h␤ARK-PH domain corresponding to residues 556 -670 of human ␤ARK1 (Fig. 1) was produced in E. coli and isolated as a GST fusion protein, cleaved, and purified. Solubility and stability limited observation to a narrow range of conditions, and the majority of studies were conducted in 20 mM acetate buffer at pH 4.5, 35°C. Under these conditions, binding of the construct to G ␤␥ is maintained (data not shown). The 546 -670 construct was also produced, and NMR spectra indicated that the additional N-terminal residues did not belong to the domain fold, and were apparently unstructured. It was concluded that the first construct contained the essential domain. Assignment used standard triple resonance methods, complemented by study of the [U 13 C, 15 N; 12 C, 14 N-Met]PH domain to help identify methionine residues that underwent partial oxidation during sample preparation. Assignment and NOE data are summarized in Fig. 2.
In Figs. 3 and 4, the overall fold and the electrostatic poten- , and carbonyl carbons (⌬CЈ) (54); 3 J HNH␣ , the magnitude (in Hz) of the three-bond scalar (intraresidue) spin-spin coupling between the ␣and the amide hydrogens; intensities of the NOE cross-peaks (on an arbitrary log scale) between the ␣and amide hydrogens, d ␣,N (i,iϩ1), and between amide hydrogens, d N,N (i,iϩ1), of the adjacent residues; horizontal bars indicate NOEs observed between the ␣and amide hydrogens three (d ␣,N (i,iϩ3)) or four (d ␣,N (i,iϩ4)) residues apart, and between the ␣and ␤-hydrogens three residues apart (d ␣,␤ (i,iϩ3)) characteristic for the ␣-helix; heteronuclear 15 N{ 1 H} steady state NOE; circles indicate amides protected from exchange with (solid circles) or exposed to (open circles) solvent. ␣-Helices are typically characterized by ⌬H␣ Ͻ 0, ⌬C␣ Ͼ 0, ⌬C␤ Ͻ 0, and ⌬CЈ Ͼ 0, whereas chemical shift deviations of the opposite sign are expected for the ␤-strands (54). 3 J HNH␣ is directly related to the intervening torsion angle , so that 3 J HNH␣ Ͻ 6 Hz is characteristic of the ␣-helix. Based on all the data, seven ␤-strands (␤1, 561-568, ␤2, 578 -584; ␤3, 587-591; ␤4, 600 -602; ␤5, 606 -613; ␤6, 618 -624; ␤7, 628 -634) and a C-terminal ␣-helix (residues 638 -655) were identified. A representation of the secondary structure of the ␤ARK PH construct is shown on the bottom. Shaded in gray is the C-terminal extension of the ␣-helix (656 -658) where some of the helical features are still preserved (see text), characteristic of a "molten" helix. tial of the h␤ARK PH domain are shown, along with the PH domain of PLC␦ (8) for comparison. The topology of the fold is typical for PH domains, and consists of seven ␤-strands forming a ␤-sandwich flanked on one end by a C-terminal ␣-helix. The termini of the construct are disordered and highly flexible, with large amplitudes of backbone motion on a nanosecond time scale (Fig. 3). The ␤1/␤2 and ␤3/␤4 loops are disordered and display increased amplitudes of backbone dynamics on a subnanosecond to nanosecond time scale, as well as motions on a millisecond to microsecond time range (data not shown).
Of specific note, the C-terminal ␣-helix is clearly extended by more than one turn compared with C-terminal ␣-helices of most previously determined PH domain structures. The position/orientation of the helix appears to be fixed by interactions with the protein core, namely by a hydrophobic strip formed by Leu-640, Trp-643, Leu-647, Ala-650, Tyr-651, and Ala-654, which are located on the side of the helix facing the ␤-sandwich and are involved in contacts with several residues in the first two ␤-strands. The aromatic ring of Trp-643, the only conserved residue among PH domains, is buried in the protein core and exhibits numerous NOE contacts to residues in ␤1 and ␤2. Another aromatic residue in the helix, Tyr-651, is also oriented toward the interior of the ␤-sandwich. The NOESY data indicate several close contacts between the aromatic ring of Tyr-651 and the residues in the ␤4/␤5 loop and in the strand ␤5. Both the structure of this loop and the orientation of the Tyr-651 ring are well defined, as indicated by low r.m.s.d. values in these parts of the structure, and by chemical shift non-equivalence of all four ring hydrogens of Tyr-651.

Relation to Other PH/PTB Domain Structures-
The h␤ARK1 PH domain has very low sequence similarity to other PH domains of known three-dimensional structure, and therefore cannot be satisfactorily homology-modeled from known structures. The structure-based alignment of the h␤ARK1 PH with these PH domains (Tables II and III) demonstrates the same overall topology of the protein fold. The expected range of r.m.s.d. values between sequences of the same structural class, but with varying degrees of homology, has been derived previously (33). The r.m.s.d. values between different members of the PH/PTB domain family (Fig. 7) are within the range expected for such homologous sequences of low identity. The charge distribution is, however, different among the PH domains, and the large positive charge associated with the Cterminal helix of ␤ARK is unusual.
Validity of the Derived Structure of the ␤ARK PH Domain-The low pH (4.5) required for this study is close to the pK a values for both glutamate and aspartate, so variations in the side chain charges of these residues compared with physiological conditions are expected. Since this might result in a perturbed structure in those regions containing negatively charged residues, a question arises of whether the NMR structure derived under these conditions represents the protein structure under physiological conditions. As mentioned above, the circular dichroism data indicate no changes in the protein structure as compared with the more physiological conditions in a phosphate-buffered saline pH 7.2. To address this issue in greater detail, the 1 H-15 N correlation maps (HSQC) were also recorded for the PH domain dissolved in the phosphate-buffered saline (pH 6.0, temperature 25°C) or in 0.1 M Tris buffer (pH 7.9, 35°C). The minor chemical shift changes (up to 0.06 ppm in 1 H and 0.6 ppm in 15 N) are consistent with expectations of variations in pH, temperature, and buffer content. The absence of significant chemical shift perturbations in this fingerprint region suggests no significant changes in the protein structure. The ␤ARK PH domain tertiary structure here is also generally similar to other PH domain structures measured in the range pH 6.0 -9.0 (8, 11, 34 -39). The high flexibility of the C terminus in the extended ␤ARK PH domain construct reported here is also preserved at the more physiological conditions, as indicated by negative steady-state heteronuclear NOEs observed for the C-terminal residues (663-670) in phosphate-buffered saline (pH 6.0, 25°C). The binding to G ␤␥ subunits is also retained at pH 4.5, with c. 100 nM affinity of the GST fusion protein at pH 4.5 and 7.5, from an immunoblotted Western assay (19).
Structure of the C-terminal Extension: Molten Helix-The C-terminal segment shows an unusual structural feature that, to our knowledge, has not been reported previously in proteins. NOEs characteristic for an ␣-helix are preserved for residues toward the C terminus despite the gradual loss of other NMR characteristics of helical structure (deviations from standard chemical shifts, heteronuclear 15 N{ 1 H} NOEs, 3 J HNH␣ coupling) (Fig. 2). Increased mobility is indicated by both relaxation and solvent exchange/accessibility data, suggesting that the C-terminal part of the ␣-helix is present as a "molten helix" in solution. Molecular dynamics calculations (40) of ␣-helical melting appear to be qualitatively consistent with our NMR observations.
The hydrophobic residues C-terminal to Gln-656 (Leu-657, Val-658, and Val-661) are located at proper sequence positions to extend the already existing hydrophobic strip on the helix surface. However, being extended beyond possible interaction with the protein core's ␤-sandwich and therefore exposed to solvent, these residues lack proper hydrophobic contacts to other residues that might stabilize the helix formation. It is possible, however, that the C-terminal extension of the helix becomes structured under certain conditions, e.g. in the presence of a binding partner such as G ␤␥ subunits. It is possible also, that the apparent "molten helix" is a consequence of the expression of the C-terminal region of ␤ARK in the absence of its native protein context, and therefore might be stabilized through tertiary contacts with residues N-terminal to the PH domain. It is unlikely, however, that the thermodynamic stability of the protein is a result of special conditions of this study (low salt, pH4.5), since both the CD and NMR data (see above) indicate no significant perturbations in the ␤ARK PH domain structure upon exchange into phosphate buffer.
In the human SOS PH domain, residues N-terminal to the normal PH domain (in the Dbl homology domain) make well defined structural contacts with the PH domain, (41) and in the Btk PH domain, the "Btk motif" C-terminal to its nominal PH domain packs against the C-terminal ␤-sheet (39).
Binding of Phosphatidylinositides and Inositol Phosphates-A general hypothesis that has been advanced in the literature is that the PH domain recognizes specifically and with high affinity highly anionic phospholipids, especially phosphatidylinositol (4, 5) bisphosphate (42). This has been illustrated for the PLC␦ PH domain, in which the binding of PI(4,5)P 2 and Ins(1,4,5)P 3 is submicromolar, and a well defined structural interaction with the PH domain has been characterized (8). However, other PH domains appear to have significantly weaker affinity (11,12,42). There is also sensitivity to the isomer identity of the inositide, since the PH domain of Akt (13,14) binds to phosphatidylinositol 3,4-bisphosphate, and a newly identified GRP1 binds to phosphatidylinositol 3,4,5triphosphate (43). It is evident that the basic residues in the N-terminal section involved in the PLC␦ PH domain/ligand complex (8) are not generally present in the structure-based sequential alignment of  (b and c (50) and d (19)). The most Cterminal residues, upon truncation, are indicated. For comparison, the electrostatic potential of the PH domain from PLC␦ (Protein Data Bank entry 1MAI) is shown in f; the arrow indicates a positively charged area at the opening of the ␤-barrel, which is involved in the phospholipid binding (8). The molecular orientations are similar, as indicated by the backbone tube diagrams.  (35); PH domain of human SOS (41); PH domain of mouse ␤-spectrin (36); PH domain of Drosophila ␤-spectrin (37); N-terminal PH domain from pleckstrin (38); PLC-␦ PH domain (8); PTB domain of Shc (16); PTB domain of IRS (17). The alignment was done by pair-wise superposition of the structures and direct calculation of aligned RMSD (Table  III), based on the elements of regular secondary structure, as described in the text. Residues belonging to the elements of secondary structure used for the alignment are enclosed in boxes, labeled on the top. Numbers indicate residue positions within the corresponding secondary structure element. Conserved hydrophobic residues are colored green. The only conserved residue in PH domains, Trp ␣1-7, is colored blue and underlined. In the Shc PTB domain, residues 58 -114, belonging to an insertion in the ␤1/␤2 loop are omitted.

TABLE III Pairwise root-mean-square deviations (in Å) between the PH/PTB domain structures
The elements of regular secondary structure of the proteins used for the comparison and their alignment are indicated in Table II. Numbers shown above the diagonal were obtained with only C␣ atoms selected for the alignment and r.m.s.d. evaluation, whereas numbers below the diagonal correspond to all heavy backbone atoms (N, C␣, CЈ, and O) in the selected core elements taken into account. The percent of sequence identity between core elements of the compared proteins is indicated in the parentheses. PH/PTB domain notation is the same as in Table 2 ing to PH domains remains unresolved for the PH domains of ␤-spectrin, N-pleckstrin, dynamin, and ␤ARK.
Possible PI(4,5)P 2 Binding Site on ␤ARK-1 PH Domain and Its Significance-Using Ins(1,4,5)P 3 as a model compound for PI(4,5)P 2 , the 15 N and 1 H spectral perturbations upon titration were mapped for h␤ARK-1 PH domain using a previously published procedure (12). Under the experimental conditions, the ␤ARK PH domain binds Ins(1,4,5)P 3 with K d of 207 M, according to our protein fluorescence titration measurements using the protocol described in Ref. 12. The data (not shown) indicated maximal shift perturbations at residues Gly-569, Trp-576, Arg-578⑀, Tyr-580, and Ala-596, located in the N-terminal segment of the domain, a pattern seen similar to, but different in detail from, other PH domains. The amide 1 H chemical shift of Asp-635 is also perturbed (0.02 ppm), that is probably caused by variation in the distance between this site and the closely located (in the unbound state) positively charged side chain of Arg-578.
The association of ␤ARK with membranes is complex. It has been suggested (44) that the high affinity binding of ␤ARK to microsomal membranes depends on a segment of the N termi-nus of ␤ARK, distinct from the putative PH domain. However, other investigators have shown that the ␤ARK PH domain binds with moderate specificity to PI(4,5)P 2 and suggest that synergistic interactions of binding to both PI(4,5)P 2 and to G ␤␥ proteins via the PH domain are required for activation of ␤ARK. (45). This synergism was not observed in a model system of higher turnover, where PI(4,5)P 2 was inhibitory (46).
Residues perturbed by the Ins(1,4,5)P 3 binding are located mostly in the ␤1/␤2 loop and in the N-terminal part of the ␤2-strand. This rather flexible loop is relatively distant from both the putative G ␤␥ binding site and the area of hydrophobic contacts between the ␣-helix and the protein core. The present data provide no direct structural evidence for a possible relationship between the phospholipid and G ␤␥ binding regions.
Protein Interaction-The clustering of positive residues in the extended C-terminal helix creates a positively charged site on the ␤ARK PH domain surface (Fig. 4, a-e), in addition to a cluster of positive charges at the opening of the ␤-barrel, a site implicated for phospholipid binding (11) in other PH domains (Fig. 4f). This causes a different polarity of the surface charge in the case of ␤ARK PH domain, so that in the fully extended construct, the dipole moment of the molecule is aligned at ϳ30°a ngle relative to the helical axis (Fig. 5). Truncation of a few C-terminal residues causes a 2.5-fold reduction in the protein dipole moment and alters the orientation of the dipole vector to close to perpendicular to the helix axis. Further truncation causes only minor variations of the dipole vector. This reduction could explain the observed differences of G ␤␥ binding to truncated ␤ARK1 PH domains (19).
A depression on the ␤ARK PH domain surface between the ␣-helix and the ␤5-strand, flanked by positively charged side chains of Lys-644, Lys-645, Arg-648, and Arg-652 (helix); Lys-623 (␤6); and Arg-625 (␤6/␤7) (Fig. 4) resembles the site in the PTB domains involved in the phosphopeptide binding of that protein. This topological similarity suggests that these residues might be involved in electrostatic interaction with a negatively charged cluster on the G ␤ surface formed by Glu-226, Asp-228, Asp-246, and Asp-247 of the WD5 and Asp-267, Asp-290, and Asp-291 of the WD6 subunits (Fig. 6). With the exception of the Asp-247 and Asp-291, which are highly conserved among the WD40 subunits and are involved in the formation of an interand intra-blade hydrogen bonds (47,48), other negatively charged residues in this cluster are unique for WD5 and WD6, and highly conserved in eukaryotes. Asp-228 and Asp-246 are directly involved in ionic interactions with the switch-II region (␣-Lys-210, Fig. 6) of G ␣ , which plays a critical role in G protein heterotrimer formation (47). It is possible that in the absence of G ␣ the PH domain of ␤ARK interacts (via its positively charged C terminus) with the same residues on the top of G ␤ surface and therefore either directly or indirectly interferes with G ␣ binding to G ␤␥ .
The ␤ARK2 (GRK-3) PH domain has a different receptor and G ␤␥ selectivity, but possesses a similar pattern of basic residues (Lys-644, Lys-645, Lys instead of Arg at 625 and 652, and Arg instead of Lys at 623) (Fig. 1). The difference of sequence in the C-terminal segment of the ␤ARK2 PH domain as compared with ␤ARK1 (Arg/Asn-648, Gln/Arg-656, and Gln/Arg-659) results in an increased total positive charge and thus a larger polarization of the PH domain. Consistent with the above electrostatic model of the ␤ARK -G ␤␥ interaction, a recent G ␤␥ binding assay (49) involving the C-terminal-peptide sequences (corresponding to the segments 643-670 and 648 -665) indicates a higher potency of the PH domain of ␤ARK2 compared with that of ␤ARK1.
The structure of the ␤ARK PH domain allows rationalization of the results of ␤ARK truncation studies (50). The modifica-FIG. 5. The dipole moment as a function of C-terminal truncation. The magnitude of the dipole moment (A) and its orientation (B) relative to the ␣-helix (the polar angle between the dipole vector and the ␣-helix axis), for various lengths of the C-terminal extension in ␤ARK1 PH domain. Upon truncation, the C terminus was capped with the COOgroup. The following partial charges were assigned to the side chain atoms: ϩ1 (NZ in Lys), ϩ0.5 (NH2, NH3 in Arg), Ϫ0.5 (OG1, OG2 in Asp and OD1, OD2 in Glu) (56), and the protein mass centroid was used as the origin (57). Each peptide bond was assigned a dipole moment of 3.5 Debye aligned along the CO bond (57,58). To account for the observed flexibility in the loops and in the termini, the results are the average of the values from the ensemble of 20 ␤ARK PH domain structures (Fig. 3A). Similar results were obtained using a much larger ensemble of 100 NMR-derived lowest-target-function structures from the same distance-geometry calculation. Positions corresponding to the charged residues are labeled in A. Insets show the orientation and relative strength (in arbitrary units) of the effective electric dipole vector, colored green, in the full-length (B) and in the residue 556 -656 construct (A). tions in ␤ARK that affect the G ␤␥ binding in that assay can be explained either by deletion of one or more core elements, hence causing a disruption of the overall fold of the PH domain, or by deletion of positively charged residues on the C terminus (see below).
Several fusion proteins, containing sequences encompassing a PH domain (from Ras-GRF, Ras-GAP, OSBP, ␤-spectrin, IRS-1, and others), were shown to bind G ␤␥ in vitro with varying affinities, and some of these PH domains were able to compete with the ␤ARK PH domain and G ␣ for binding to G ␤␥ (18). The C-terminal extended PH domains of Btk (51), IRS-1, and Dbl (19) were also demonstrated to interact with the G ␤␥ heterodimer in vitro. The results of indirect assays (51,52) suggest that some of these PH domain constructs may also interact with G ␤␥ in vivo. The C-terminal parts of all these proteins, in particular region corresponding to residues 644 -670 in ␤ARK (Figs. 1 and 4), are rich in basic residues, and this supports the proposed ␤ARK-G ␤␥ interaction model. Other literature data are also consistent with this model. Mutation of the last four of the five basic residues in the highly charged C terminus (Arg-660, Lys-663, Lys-665, Lys-667, and Arg-669) to acidic ones leads to almost complete loss of the G ␤␥ binding (20), as does a deletion of the last nine residues (662-670) (50). A truncated fragment, ending with Asn-666, does retain an FIG. 6. Electrostatic polarization of the G ␤␥ heterodimer. A, molecular surface of the G ␤␥ complex colored by the electrostatic potential, from red (Ϫ10 kT/e) to blue (ϩ10 kT/e). A fragment of G ␣ (green) contains the switch II region (␣2 helix), which is intimately involved in a direct contact with the top of the G ␤ propeller in the heterotrimer (47,48). Those positively charged residues of G ␣ (Lys-209, Lys-210) in the G ␣␤␥ complex less than ϳ3 Å between side chain heavy atoms from Asp-228 and Asp-246 of G ␤ are labeled. The highly negatively charged area around this position may serve as a complementary charged surface to the ␤ARK1 PH domain. B, a diagrammatic representation of the location of negative charges on the G ␤ surface. The surface corresponding to the highly negatively charged area in A is indicated by a broken line. The general sequence consensus of a WD motif and location of negatively charged residues in the sequences of WD5 and WD6 subunits are also shown (B, top). The residues in the WD repeat sequence are marked: x, a non-conserved position; h, a conserved hydrophobic position; r, a conserved aromatic; p, a conserved polar position; t, a tight turn containing Gly, Pro, Asp, or Asn. The superscripts indicate the range of residues observed in the various known G ␤ subunits. Those acidic residues that are unique for the WD5 and WD6 subunits are shaded in gray. A was generated using GRASP (56); atom coordinates were extracted from PDB entry 1GG2; the schematic representation of G ␤ and the general sequence consensus of WD motifs were adopted from Ref. 59 Table III). intermediate level of G ␤␥ activation, whereas further truncation up to Val-661 leads to a significantly lower amounts of G ␤␥ binding compared with the full-length construct (50). On the other hand, the results of deletion analysis (19) indicate that the C-terminal extension beyond Gln-656 is not absolutely required for binding to G ␤␥ ; the full-length C-terminal extension (553-689 construct) dramatically increases the maximal extent of G ␤␥ binding although does not significantly alter the binding affinity. A 18-residue peptide comprising the C-terminal amino acids 648 -665 was recently shown to bind G ␤␥ , also suggesting that the critical C-terminal extension in ␤ARK1 PH domain required for G ␤␥ binding might be shorter than initially suggested by (50). However, other functions may be associated with the extended domain.