doi:10.1016/j.jmb.2009.07.076 J. Mol. Biol. (2009) 392, 1044–1054

Available online at www.sciencedirect.com

Prediction of Protein–Protein Interfaces on G-Proteinβ Subunits Reveals a Novel Phospholipase C β2Binding Domain

Erin J. Friedman1, Brenda R. S. Temple2,3⁎, Stephanie N. Hicks4,John Sondek3,4,5, Corbin D. Jones1,6 and Alan M. Jones1,4

1Department of Biology,University of North Carolinaat Chapel Hill, Chapel Hill,NC 27599, USA2The R. L. Juliano StructuralBioinformatics Core Facility,University of North Carolinaat Chapel Hill, Chapel Hill,NC 27599, USA3Department of Biochemistryand Biophysics, Universityof North Carolina School ofMedicine, Chapel Hill,NC 27599, USA4Department of Pharmacology,University of North CarolinaSchool of Medicine, Chapel Hill,NC 27599, USA5Lineberger ComprehensiveCancer Center, Universityof North Carolina Schoolof Medicine, Chapel Hill,NC 27599, USA6Carolina Center for GenomeSciences, University of NorthCarolina at Chapel Hill,Chapel Hill, NC 27599, USA

Received 10 February 2009;received in revised form8 July 2009;accepted 27 July 2009Available online30 July 2009

*Corresponding author. The R. L. JuHill, Genetic Medicine Building, 120E-mail address: btemple@med.unc.eAbbreviations used: G-protein, gu

multiple sequence alignment; PLC-βinterest; RSA, relative solvent access

0022-2836/$ - see front matter © 2009 E

Gβ subunits from heterotrimeric G-proteins (guanine nucleotide-bindingproteins) directly bind diverse proteins, including effectors and regulators,to modulate a wide array of signaling cascades. These numerousinteractions constrained the evolution of the molecular surface of Gβ.Althoughmammals contain five Gβ genes comprising two classes (Gβ1-likeand Gβ5-like), plants and fungi have a single ortholog, and organisms suchas Caenorhabditis elegans and Drosophila melanogaster contain one copy fromeach class. A limited number of crystal structures of complexes containingGβ subunits and complementary biochemical data highlight specific siteswithin Gβs needed for protein interactions. It is difficult to determine fromthese interaction sites what, if any, additional regions of the Gβ molecularsurface comprise interaction interfaces essential to Gβ's role as a nexus innumerous signaling cascades. We used a comparative evolutionaryapproach to identify five known and eight previously unknown putativeinterfaces on the surface of Gβ. We show that one such novel interfaceoccurs between Gβ and phospholipase C β2 (PLC-β2), a mammalian Gβinteracting protein. Substitutions of residues within this Gβ–PLC-β2interface reduce the activation of PLC-β2 by Gβ1, confirming that our denovo comparative evolutionary approach predicts previously unknownGβ–protein interfaces. Similarly, we hypothesize that the seven remaininguntested novel regions contribute to putative interfaces for other Gβinteracting proteins. Finally, this comparative evolutionary approach issuitable for application to any protein involved in a significant number ofprotein–protein interactions.

© 2009 Elsevier Ltd. All rights reserved.

Keywords: heterotrimeric G-proteins; phospholipase C-β2; protein surfaceevolution; interface prediction; PLC-β2–Gβγ interaction interface

Edited by M. Sternberg

liano Structural Bioinformatics Core Facility, University of North Carolina at ChapelMason Farm Road, Suite 3010, Campus Box 7260, Chapel Hill, NC 27599-7260, USA.du.anine nucleotide-binding protein; GRK, G-protein-coupled receptor kinase; MSA,2, phospholipase C β2; RGS9, regulator of G-protein signaling 9; ROI, region ofibility.

lsevier Ltd. All rights reserved.

1045Novel Gβ Interfaces

Introduction

Gene duplication is a fundamental source ofgenetic and phenotypic novelty.1 After duplication,one of the new paralogs is freed from functionalconstraint, enabling it to evolve new functions. Forgenes encoding proteins that interact with otherproteins, this process often liberates one copy todevelop a new set of interactions. This is particularlytrue for signaling molecules that are used byorganisms to communicate between cells and toperceive their environment. For example, evolutionof increasingly complex organisms correlates withthe enormous diversification of heterotrimeric G-protein (guanine nucleotide-binding protein) signal-ing complexes and cell surface G-protein-coupledreceptors. The G-protein complex consists of aheterotrimer composed of Gα, Gβ, and Gγ subunits.Upon activation by cell surface receptors, thecomplex dissociates into a free Gα subunit and aGβγ dimer, both of which bind to and signalthrough other proteins. Signaling typically termi-nates when the heterotrimer reforms.2,3 Saccharomy-ces cerevisiae has two Gα genes but only one each ofGβ and Gγ genes; mammalian genomes encode 16Gα, 5 Gβ, and 12 Gγ genes.2 This complex array ofsubunit combinations allows for diverse signalingpossibilities. It was previously thought that Gα wasthe primary signaling molecule in mammals, whilethe sole function of Gβ was to inhibit Gα signalingand to provide for its membrane localization.4 It isnow clear that Gβ also modulates downstreamtargets; a subset of signaling pathways is uniquelyregulated by the Gβ subunit.5Several proteins that bind Gβ have been identified

in mammals. In addition to Gα and Gγ, interactingproteins such as the localization chaperone phosdu-cin, G-protein-coupled receptor kinases (GRKs),phospholipase C β2 (PLC-β2), regulator of G-protein signaling 9 (RGS9), calcium channels,potassium channels, and adenylyl cyclase 2 bindGβ proteins. Mammalian Gβ interacting proteinsarose differentially over evolutionary time as notall eukaryotes contain homologs to all knowninteractors.2 For example, Gα and phosducin arepresent in all eukaryotes beginning with plants.Canonical RGS9, PLC-β2, and GRK2 originated andwere maintained in metazoans at least by the time ofthe formation of annelids since Caenorhabditis eleganscontains bona fide RGS9, PLC-β2, and GRK2orthologs.6–8 As the Gβ subunit acquired additionalbinding partners throughout evolution, new Gβ–protein interfaces likely evolved to accommodatethese interactions. These new interfaces may havepartially overlapped with existing interfaces since,for example, the interface from many Gβ interactorsoverlaps with the Gα–Gβ interface.9 However,these interfaces may have also utilized regions onthe Gβ molecular surface that previously had noassociated function.Two major experimental approaches, structural

and biochemical studies, have characterized somebinding interfaces between Gβ and interacting

proteins. Currently, there are four crystal structuresof mammalian Gβ subunits in complex withsignaling proteins: Gβ1γ1–Gα,10 Gβ1γ2–GRK2,11

Gβ1γ1–phosducin,12 and Gβ5–RGS9.13 These struc-tures provide a three-dimensional view of whereproteins interact but provide limited informationregarding the importance of individual contacts.Additionally, structures of Gβ in complex withpeptides14,15 provide partial information on physi-ologically relevant interfaces within Gβ. Biochemi-cal studies include both targeted mutationalstudies9,16,17 and targeted domain-swapping orpeptide-binding experiments.18–20 Mutational stud-ies directed by the structural studies are limited bythe number of available structures. Interpretation ofthese studies is complicated by overlap between theimplicated binding regions. When crystal structuresof a particular Gβ complex are not available,mutations are frequently targeted to an interfaceidentified in a solved Gβ complex structure. Forexample, Ford et al.9 created several mutations in theGα binding interface to show that this interface wasutilized in part by five other interacting proteins.This type of approach however does not elucidatebinding sites, or even those regions of the bindinginterface, that are not shared with the Gα interface.To locate these alternative sites, groups such asPanchenko et al.17 performed mutational analysesoutside of known binding areas. Although thisstudy identified mutant regions of Gβ with reducedability to activate PLC-β2, much refinement remainsnecessary to identify individual critical interactionsites. Moreover, this study left many regions of theGβ surface unexplored.Bioinformatic analyses have been developed to

identify functional residues, including those com-posing conserved patches on the surface of a proteinthat may function as a binding interface. Three suchanalyses include evolutionary trace,21 DIVERGE,22,23

and ConSurf.24 All three rely on multiple sequencealignments (MSAs) and phylogenetic trees to identifystructurally clustered functional residues. Evolution-ary trace has been applied to Gβ and the twopredicted interfaces correlated with those of Gα andGγ.25 To date, none of these three methods haspredicted previously uncharacterized Gβ interfaces.We took advantage of both the rich source of

divergent Gβ subunit sequences and the differentialevolutionary emergences of known Gβ interactorsto understand not only how to predict novelinteraction interfaces on the Gβ surface in theabsence of crystallized complexes but also howthese new interactions gave rise to new bindingsurfaces. These results can be utilized for moretargeted and informed biochemical studies. Basedon the hypothesis that the acquisition of mammali-an-like sequence identity on the surface of Gβreflects the utilization of new binding interfaces,we applied a suite of bioinformatic and phylogenetictechniques to follow the shift in patterns of aminoacid conservation in Gβs, concentrating on changesbetween distinct points in the evolutionary historyof the Gβ subunit represented by five reference

1046 Novel Gβ Interfaces

species. By placing this conservation in a structuralcontext, we predicted regions of interest (ROIs) thatare composed of adjacent surface residues thatsimultaneously evolved to residues conserved withmammals. We identified a novel PLC-β2 interfaceby demonstrating that at least one ROI that becameconserved in C. elegans is involved in PLC-β2activation by Gβ. Similarly, we propose that theremaining ROIs also compose at least portions ofbinding interfaces.

Results and Discussion

Gβ proteins fall into two major classes: Gβ1-likeand Gβ5-like

Since most extant plant and fungal species have asingle Gβ, while nematodes and later metazoanshave at least two Gβ subunits, the first Gβ geneduplication occurred between the splitting offungi and C. elegans from the mammalian lineage.To compare preduplication plant and fungalGβ sequences with extant postduplication Gβsequences from metazoans, we reconstituted ances-tral sequences to plants and to fungi from MSAs(Figs. S1 and S2, respectively) of extant plant andfungal Gβ sequences using the Bayesian ancestralreconstruction implemented in MrBayes (see Mate-rials and Methods). These inferred preduplicationancestors were important as they, and not individ-ual plant or fungal species, reflect sequence con-straints common to all plants or fungi and thereforemore closely reflect the predecessor of all extantpostduplication Gβ genes.An MSA containing Gβ sequences from the plant

ancestor, fungal ancestor, and extant C. elegans,Drosophila melanogaster, and human Gβ sequenceswas created (Fig. S3) for comparison of preduplica-tion and postduplication Gβ genes. This MSA wasused to generate a Bayesian phylogenetic tree(Fig. S4) that elucidates the evolutionary relation-ship between the different Gβ proteins. HumanGβ1, Gβ2, Gβ3, and Gβ4 sequences formed amonophyletic Gβ1-like clade also containing D.melanogaster and C. elegans Gβ1 sequences. HumanGβ5 formed a Gβ5-like monophyletic clade contain-ing D. melanogaster Gβ5 and C. elegans Gβ2sequences. Both plant and fungal ancestors wereoutside of these two clades.Based on the phylogenetic data (Fig. S4) and

corroborating previous observations that Gβ5 oftenbehaves differently from the other four mammalianGβ proteins,5 metazoan Gβ proteins can be groupedinto two major classes, Gβ1-like and Gβ5-like. Theinitial Gβ gene duplication between fungi and C.elegans gave rise to the Gβ5-like family since C.elegans is the earliest species examined to contain aGβ5-like protein. The ancestral plant and fungal Gβproteins each contains characteristics of both majorclasses, but neither strictly belongs to either class.Finally, the phylogeny revealed that, after duplica-tion, the Gβ5-like genes diverged from the ancestor

more than the Gβ1-like genes, while the Gβ1-likegenes maintained more ancestral characteristics.

Interaction interfaces identified from Gβcomplex structures co-evolved with interactors

Complexes containing mammalian Gβ1–Gα,Gβ1–GRK2, Gβ1–phosducin, and Gβ5–RGS9 areshown in Fig. 1 (Gβ in spheres and interactingprotein in ribbons). While these structures provide athree-dimensional view of Gβ–protein complexes,the interfaces between the proteins must still bedefined.26 To define sites of interaction, we calculat-ed the solvent accessibility of each residue in the fourcrystal structures using Naccess.27 Those Gβ resi-dues whose side chains have lower relative solventaccessibility (RSA) by more than 5% between themonomer and the protein complex, including thosethat hydrogen bond with the binding partner, weredefined to comprise the interface and were noted onthe structure (see Materials and Methods) (Fig. 1).These four Gβ interacting proteins (1) cover threemajor areas on the Gβ surface, (2) all partiallyoverlap with one another, and (3) appeared atdifferent times over Gβ evolution.In order to examine the evolution of these known

Gβ–protein interfaces, we compared the patterns ofamino acid conservation within these regions. Foreach interface residue in the MSA, comparisonswere made between each of the four referencesequences and the corresponding mammalian resi-due. For Gβ1 complexes (Gα, GRK2, and phosdu-cin), the plant ancestor, fungal ancestor, C. elegansGβ1, and D. melanogaster Gβ1 were compared withbovine Gβ1. Identical interface residues weremapped onto the bovine Gβ1 structure (1got.pdb).For the RGS9 complex, the plant ancestor, fungalancestor, C. elegans Gβ5, and D. melanogaster Gβ5were compared with mouse Gβ5 and identities werenoted on the mouse Gβ5 structure (2pbi.pdb). Asexpected, conservation within known binding areascorrelated with the known utilization of theseregions (Fig. 2). The complexes subsequently dis-cussed were observed in the four crystal structurecomplexes.

Gβ1–Gα

The Gβ–Gα interaction is the most ancient; thus, itwould be expected that the Gα–Gβ interface is wellconserved in all organisms studied. As expected, theGα binding interface was highly conserved (77%–100% identity) in all sequences analyzed (Fig. 2a, toprow) as illustrated by the predominantly orangecoloration of the interface.

Gβ1–GRK2

The GRK2 interface (Fig. 2a, second row) wascompletely conserved in C. elegans and D. melanoga-ster, correlating with the emergence of GRK2between fungi and C. elegans. Although a largeportion of the interface was also conserved in the

Fig. 1. Binding interfaces of four Gβ interacting proteins as determined by crystal structures. Top, bottom, and sideviews of Gβ or the Gβγ dimer (Gβ in light gray spheres and Gγ in dark gray spheres) bound to four interacting proteins(ribbon): Gβ1–Gα in green (1got.pdb), Gβ1–GRK2 in magenta (1omw.pdb), Gβ1–phosducin in blue (N-terminal domainis dark and C-terminal domain is light; 1a0r.pdb), and Gβ5–RGS9 in red (2pbi.pdb). Gβ residues that contact eachinteracting protein are colored accordingly. Binding contacts were determined by evaluating the solvent accessibilitydifference between single molecules and those in complex.

1047Novel Gβ Interfaces

plant (62%) and fungal (92%) ancestors, this can beattributed to the fact that the interface largelyoverlaps that of Gα.

Gβ1–phosducin

Plants contain several phosducin-like sequences,28

and our yeast three-hybrid data suggest that at leastone phosducin in Arabidopsis thaliana interacts withthe A. thalianaGβ (Fig. 3). The phosducin interface ishighly conserved in C. elegans and D. melanogaster(94% identity); however, it is only partly conservedin plant and fungal ancestors (55% and 64%,

respectively). Mammalian phosducin has twodomains: a helical N-terminal domain and a C-terminal domain (see Fig. 1 for locations of N- andC-terminal binding interfaces), both of which bindmammalian Gβ.29 The plant and fungal ancestorsshow high conservation in the N-terminal bindingarea (68% and 76%, respectively) but not in the C-terminal binding area (25% and 38%, respectively)(Fig. 2a, third and fourth rows). Two possibleexplanations for these results are that plants andfungi bind only one phosducin domain and not thesecond and that the second domain binds in aspecies-specific manner.

Fig. 2. Conservation within known binding interfaces based on bovine Gβ1 crystal structures. (a and b) Conservationof the (a) Gα, GRK2, and phosducin binding interfaces on Gβ1 (1got.pdb) and (b) RGS9 binding interface on Gβ5 (2pbi.pdb) in four reference organisms. While the interface is composed of all colored residues, conserved residues are coloredlight orange and nonconserved residues are colored green (Gα), blue (phosducin), magenta (GRK2), or red (RGS9). Onlyresidues in the binding area were analyzed.

1048 Novel Gβ Interfaces

Gβ5–RGS9

The RGS9 binding interface (Fig. 2b) is only 42%identical in the plant ancestor and 50% identical in thefungal ancestor. Correlating with the genesis of theRGS9 protein, conservation in the binding area risesto 67% and 76% in C. elegans and D. melanogaster,respectively. Although this is markedly less than the90%–100% identity seen in corresponding Gβ1binding areas, it is important to note the size of theRGS9 interface. Encompassing 103 residues, theRGS9interface is over three times larger than the Gα (31residues) andGRK2 (26 residues) interfaces and twice

the size of the phosducin interface (47 residues). Thus,we expect that the RGS9 interface could tolerate moresubstitutions than its smaller counterparts and thatonly those residues making energetically criticalcontacts in the interaction interface are conserved.

Newly conserved regions of adjacent surfaceresidues arose over time and likely contributeto Gβ–protein interfaces

As conservation within known Gβ–protein inter-faces correlated with the emergence of proteins

Fig. 3. An Arabidopsis phosducinand Gβ interact physically. Growthof yeast strain AH109 containing thegenes indicated (AtGβ1γ1 alone orAtGβ1γ1 and phosducin) on yeastdropout medium. Medium missingtryptophan (−W) selects for theAtGβ1γ1 vector, resulting in posi-tive growth for both genotypes.Mediummissing leucine (−L) selectsfor the phosducin vector, resultingin no growth for the strain lackingphosducin and positive growth forthe strain containing phosducin.Medium missing tryptophan, leu-cine, and histidine (−WLH) selectsfor a positive interaction betweenthe two genes, resulting in nogrowth for the strain containingAtGβ1γ1 alone and positive growthfor the strain containing bothAtGβ1γ1 and phosducin. The lattergrowth indicates that the two genesphysically interact.

1049Novel Gβ Interfaces

utilizing these interfaces, we next used a similarmethod of analysis to predict novel binding inter-faces. This unbiased comparative evolutionaryanalysis was not restricted to predetermined bind-ing interfaces and was performed on the entiremolecule, but it was otherwise similar to themethods used in the validation stage described inthe previous section. Bovine Gβ1 was compared ateach residue in the MSA with the plant ancestor,fungal ancestor, C. elegans Gβ1, and D. melanogasterGβ1 (Fig. 4a). Identical residues between the plantancestor and bovine Gβ1 were located on thesurface of Gβ1 and colored dark green. These 162residues form region zero (ROI0) and represent theprimordial function of the Gβ molecule, since theplant molecule is the most ancestral-like of the Gβfamily. For all other reference species, conservedresidues that continued to be conserved from theprevious species were colored light green (indicat-ing the persistence of an existing function), whilenewly conserved residues were colored dark green(indicating the emergence of potential novel func-tion). Similarly, mouse Gβ5 was compared with theplant ancestor, fungal ancestor, C. elegans Gβ5, andD. melanogaster Gβ5 and conserved residues weremapped onto the structure (Fig. 4b).Due to its primordial nature, ROI0 contains

residues conserved both for structural maintenanceand for interaction surfaces. ROI0 contains theresidues (e.g., W99, M101 K57, Y59, L117, D186,D228, and W332) previously indicated as interact-ing with multiple effectors and that also madeenergetically critical contacts in various interactioninterfaces.9,14 Subsequent non-ancestral ROIs ineach organism (for both Gβ1 and Gβ5) were

defined as clusters of three or more structurallyadjacent (within 5 Å30) newly conserved surfaceresidues and are encircled on the structure (Fig. 4;for residue designations, see Table 1). It is impor-tant to note that although all surface-exposedresidues could form part of an interface, we chosea larger cluster size (containing three or more newlyfixed residues) in order to predict clusters thatlikely make significant energetic contributions tobinding. We also ignored several large clusters nearthe Gγ binding area due to their likely involvementin the Gβ–Gγ interface.Five of the 13 ROIs (ROI7, ROI10, ROI11, ROI12,

and ROI13) could be explained by existing struc-tural data. ROI7 lies in the phosducin bindingregion on Gβ1 (Fig. 1). Portions of ROI10–ROI13localize to the RGS9 binding interface on Gβ5(Fig. 1). Additionally, a portion of ROI2 was shownto interact with Gα, adenylyl cyclase 2, calciumchannels, potassium channels, and PLC-β2 inmutational studies.9 These data support the hy-pothesis that regions of adjacent surface residuesthat become conserved to the mammalian state atthe same time are likely sites of Gβ–proteininteractions. Similarly, we hypothesize that theremaining ROIs are binding interfaces with yet-to-be identified interacting proteins or additional,undescribed binding interfaces with previouslyidentified interacting proteins.

Several ROIs contain Gβ residues that activatePLC-β2

To test our hypothesis that ROIs represent Gβ–protein interaction interfaces, we mutated residues

Fig. 4. Gβ ROIs as determined by a comparative evolutionary analysis. Three views (top, bottom, and side) of theevolution of conserved regions in the plant ancestor, fungal ancestor, C. elegansGβ, andD. melanogasterGβ [(a) Gβ1 (1got.pdb) and (b) Gβ5 (2pbi.pdb)]. The dark green regions of the plant ancestor highlight primordial function and form ROI0.In all other organisms, newly conserved residues (those that matched themammalian value) are colored dark green, whileconserved residues present in a previous organism are colored light green. Thus, dark green patches represent acquisitionof new function. ROIs are indicated and corresponding residues are listed in Table 1.

1050 Novel Gβ Interfaces

likely to interact with PLC-β2. Despite the absenceof a crystal structure of the Gβγ–PLC-β2 complex,several studies implicated multiple potential inter-faces. Peptide-binding assays implicated a region atthe base of Gβ's N-terminal helix.18,20 A patch ofresidues near this region was implicated as aninterface when it was revealed that the C-terminaltail of Gγ was responsible for the activation of PLC-β2.19 Therefore, ROI4, ROI6, and ROI7 were chosenfor mutational studies due to their proximity to thegeneral region implicated in these peptide-bindingand domain-swapping experiments. Additionally,all three regions became conserved between thefungi and C. elegans split, corresponding with theemergence of PLC-β proteins.In order to evaluate the extent of PLC-β2

activation, we transfected COS-7 cells with wild-type and mutant forms of Gβ1 and wild-type Gγ2in the presence or absence of PLC-β2. Equalexpression of all Gβ1, Gγ2, and PLC-β2 constructswas confirmed by immunoblot analysis. Addition-ally, equivalent expression of Gγ2 was observed inthe presence of all forms of Gβ1. The location of

each mutation on the surface of Gβ1 is depicted inFig. 5. The construct containing the doublemutation T65A_D322A, which is located outsideof the predicted binding region and lacks sequenceconservation with bovine Gβ1, activates PLC-β2 tothe same extent as wild-type Gβγ. Conversely, twosingle mutations, W99A and D228A, abolish PLC-β2 activation as previously shown.9 Within thenewly predicted binding areas, the mutations thatmost significantly reduced PLC-β2 activation werefound in ROI4 and ROI6 (Figs. 4a and 5).R52S_F335S slightly reduced PLC-β2 activation,while K127S_R129S_E130S nearly abolished PLC-β2 activation. In summary, when residues withinthe new binding areas predicted by our de novoanalysis are mutated, there is a concordant drop inactivity; conversely, when residues we identifiedas not part of binding surfaces are mutated, thereis no change in activation. These results indicatethat our evolutionary structural analysis predictsthe nonoverlapping portions of interaction inter-faces and can be used to further dissect the PLC-β2 interface on Gβ1.

Table 1. ROI residue numbers

ROI Residues First conserved in

1 R96, S97, S98, I120, R134, E138, E172 Fungi Gβ12 L55, A56, S334 Fungi Gβ13 D66, R68, Y85, N88, V90, Y105 Fungi Gβ14 K127, R129, E130, N132 C. elegans Gβ15 T143, T173, Q175, Q176, T181,

T184, M217C. elegans Gβ1

6 R46, T47, R52, D312, F335 C. elegans Gβ17 R42, Q44, D267, N268, I269,

I270, C271, G272, I273, D290,D291, N293, N295, V307, A309

C. elegans Gβ1

8 L152, D153, N155, D195, R197, L210 C. elegans Gβ19 D20, A24 D. melanogaster Gβ110 T102, P104, T106, N141, K146, N154 C. elegans Gβ511 E43, K279, E280, S281 C. elegans Gβ512 F284, N303, Y305 D. melanogaster Gβ513 N163, L203, P205, E207 D. melanogaster Gβ5

Shown are mammalian residue positions for each Gβ residuewithin the ROIs highlighted in Fig. 4.

1051Novel Gβ Interfaces

Molecular evolution can be used as a tool topredict novel binding interfaces

Conservation is a hallmark of a residue's func-tional importance,23 and as the plant ancestor ismost similar to the most ancient Gβ molecule,primordial Gβ function can be inferred to correlatewith those residues that were conserved in the plantancestor (ROI0) (Fig. 4). These primordial functionalregions comprise both conserved protein-bindinginterfaces and regions of the protein important forstructural integrity. With each comparative step(e.g., between plants and fungi, fungi and C. elegans,

etc.), new functional regions became conserved.These regions may represent sites of novel Gβ–protein interaction or new interaction sites withexisting proteins (co-evolution). Lineage-specificprotein interactions will however not be revealedby this type of analysis. Lineage-specific interactionswould be indicated by conservation of a residuewithin a particular organismal group (e.g., within allplants or within all fungi) that is not conservedoutside of that group (e.g., between plants andfungi). All interfaces identified in our analysis areones that are maintained in the mammalian lineage.Although species-specific interactions are also ofgreat interest, a different approach would benecessary to reveal those regions.The Gβ gene family was produced by a series of

gene duplications. After these events, the new Gβparalogs likely evolved novel functions as a result ofrelaxed functional constraint. Eventually, as thesenew Gβ functions become critical to the organism,these new functional regions become highly con-served regions of the Gβ structure and include theROIs we identified. We hypothesize that these ROIsrepresent regions of newly acquired function in theorganism in which they first appeared. For instance,portions of six ROIs presented in Fig. 4 (ROI2, ROI7,and ROI10–ROI13) contain residues that likely bindinteractors as previously identified in mutational orstructural studies. The remaining ROIs in Fig. 4 arein regions that lie outside of known regionspresented thus far. Although interactions withmany proteins have been determined by mutationalstudies, many studies focus on known interfacesand therefore indirectly introduce bias. For example,

Fig. 5. Regions of Gβ involvedin PLC-β2 activation. Groups ofresidues mutated on the Gβ surface(1got.pdb) are colored and coordi-nate with colors in the bar graphand associated blots. Wild-type andmutant Gβ1 subunits were tested inthe absence (−) or presence (+) ofPLC-β2 for their ability to activatePLC-β2 (⁎pb0.01 and ⁎⁎pb0.005;error bars represent standard error).Immunoblot analysis confirmedequal expression of all proteinsutilized.

1052 Novel Gβ Interfaces

since the Gα binding area is utilized in many Gβ–protein interactions, studies historically focused onthese residues. Our de novo analysis eliminates thisbias, generating a set of potential binding interfacesthat are yet to be explored by mutational studies. Aswith the residues implicated in the PLC-β2 bindinginterface, these ROIs would make ideal targets forsite-directed mutagenesis. In this manner, we canidentify new Gβ interacting proteins and furthercharacterize interactions with existing proteins.Ultimately, this will allow us to characterize newGβ-mediated signaling pathways.

Materials and Methods

Ancestral reconstruction

To reconstruct plant and fungal ancestors, we queriedsequence databases with known plant and mammalianGβ sequences for full-length plant and fungal Gβhomologs. Additionally, Gβ sequences were also retrievedfor two outgroups: Entamoeba (Entamoeba dispar) anddiatom (Thalassiosira pseudonana). MSAs were generated inClustalX.31 Expasy and Genbank IDs for each gene used inthe alignment are denoted in Figs. S1 and S2. For eachMSA generated, all gaps except those found exclusively inthe two outgroups were removed. From the resultingNEXUS file, ancestors were generated with MrBayes32,33using the fixed equalin model, using the inverse gammarate, and sampling 100,000 generations at a frequency of100. In order to eliminate ancestral residue values chosenwith low confidence, we accepted and included in theancestral sequence only those residues predicted N90% ofthe time with a maximum value N0.8. All other residueswere assigned a value of “X” as they were too variable tobe called with confidence.

Sequence collection, alignment, and phylogenygeneration

All Gβ sequences from D. melanogaster, C. elegans, andhuman (Expasy and Genbank sequence IDs are denotedin Fig. S3) were combined with the plant and fungalancestral sequences and the diatom and Entamoebaoutgroups to create an MSA (see previous section foralignment and NEXUS file generation). The alignmentswere made robust by structural comparisons betweenbovine Gβ1 and mouse Gβ5. MrBayes32,33 was runusing a fixed equalin model, using the inverse gammarate, and sampling 1 million generations at a frequencyof 100 for three independent runs with a burn-in of250,000 generations to generate a consensus phyloge-netic tree.

Interface determination

In order to identify binding interfaces from structuresof Gβ in complex with interacting proteins, we usedNaccess27 to calculate the RSA of all Gβ residue sidechains both as a monomer and in complex with otherproteins. All residue side chains with an RSA decreaseN5% (RSA of Gβ monomer–RSA of complex, N5)between monomer and complex were denoted as beingpresent in the interface. This group of residues alsoincluded all Gβ residues that formed hydrogen bonds

with the interacting protein as determined in the PyMOLmolecular modeling system.34 The 5% ΔRSA value waschosen because it is more stringent than the commonlyutilized 1-Å2 change in absolute solvent accessibility26,35

but still included all residues that form hydrogen bonds.Additionally, a residue with ≥5% RSA is often desig-nated as a surface residue, while a residue with b5%RSA is designated as an interior residue.26,35

Yeast three-hybrid protein interaction

A. thaliana Gβ (AGB1) and Gγ1 (AGG1) were clonedinto the pBridge vector (Clontech, Palo Alto, CA).A. thaliana phosducin (At5g14240) was cloned into thep-ENTR/D-TOPO vector (Invitrogen, Carlsbad, CA) andthen recombined into the pACTGW-attR Gatewayvector,36 which contains an activation domain and iscompatible with the pBridge vector. The prey (phosducin)was transformed into yeast strain AH109, which hadpreviously been transformed with the bait (AGB1/AGG1). Both strains (containing the bait alone andcontaining both bait and prey) were grown on nutrition-ally selective media. Presence of the bait and that of theprey were confirmed by the expression of nutritionalmarkers (positive growth on media lacking tryptophanand leucine, respectively). Interaction was confirmed bythe expression of an additional nutritional marker(positive growth on medium lacking histidine).

Transfection of COS-7 cells with Gβ1, Gγ2, andPLC-β2

The accumulation of [3H]inositol phosphates wasmeasured in transiently transfected COS-7 cells aspreviously described.37 Briefly, COS-7 cells were platedin 12-well dishes at a cell density of 60,000 cells per wellin Dulbecco's modified Eagle's medium supplementedwith 10% fetal bovine serum, 10 U/ml of penicillin, and10 U/ml of streptomycin. Following incubation for 24 hat 37 °C in an atmosphere of 95% air/5% CO2, cells weretransfected with 200 or 300 ng each of the indicatedhuman Gβ1 and Gγ2 DNA in the presence and in theabsence of 30 ng of PLC-β2 and empty vector, for a totalof 700 ng of DNA per well. DNA was complexed withFuGENE 6 (Roche Applied Sciences, Indianapolis, IN) asper the manufacturer's protocol prior to transfection.Twenty-four hours posttransfection, the medium wasaspirated and cells were metabolically labeled with1 μCi/well of myo-[2-3H(N)]inositol (American Radiola-beled Chemicals, St. Louis, MO) in inositol-free Dulbec-co's modified Eagle's medium for 12–16 h. Subsequently,10 mM LiCl was added. One hour after incubation withLiCl, reactions were stopped by aspiration of themedium and addition of 50 mM formic acid. Sampleswere neutralized by the addition of 150 mM NH4OH,and [3H]inositol phosphates were quantified by Dowexchromatography. Statistical significance for comparisonsbetween the activation of PLC-β2 by wild-type Gβ1γ2and that by each mutant Gβγ was determined byperforming Student's t test assuming equal variancebetween the log-transformed number of [3H]inositolphosphates for each construct. Western blotting wasperformed to confirm equal expression of each constructin COS-7 cells using an antibody directed toward thec-Myc epitope (Invitrogen) on human Gβ1, the hemag-glutinin epitope (Roche Applied Sciences) on Gγ2, and amonoclonal antibody (Santa Cruz Laboratories, SantaCruz, CA) against PLC-β2.

1053Novel Gβ Interfaces

Acknowledgements

Work in A.M.J.'s laboratory on the ArabidopsisG-proteins was supported by the National Instituteof General Medical Sciences (GM065989-01), the U.S.Department of Energy (DE-FG02-05ER15671), andthe National Science Foundation (MCB-0718202and MCB-0723515). Work in J.S.'s laboratory wassupported by the National Institutes of Health(5-50921-2311).

Supplementary Data

Supplementary data associated with this articlecan be found, in the online version, at doi:10.1016/j.jmb.2009.07.076

References

1. Ohno, S. (1970). Evolution by Gene Duplication. Springer,Berlin, Germany.

2. Jones, A. M. & Assmann, S. M. (2004). Plants: the latestmodel system for G-protein research. EMBO Rep. 5,572–578.

3. McCudden, C. R., Hains, M. D., Kimple, R. J.,Siderovski, D. P. & Willard, F. S. (2005). G-proteinsignaling: back to the future. Cell. Mol. Life Sci. 62,551–577.

4. Milligan, G. & Kostenis, E. (2006). Heterotrimeric G-proteins: a short history. Br. J. Pharmacol. 147(Suppl. 1),S46–S55.

5. Cabrera-Vera, T. M., Vanhauwe, J., Thomas, T. O.,Medkova, M., Preininger, A., Mazzoni, M. R. &Hamm, H. E. (2003). Insights into G protein structure,function, and regulation. Endocr. Rev. 24, 765–781.

6. van der Linden, A. M., Simmer, F., Cuppen, E. &Plasterk, R. H. (2001). The G-protein beta-subunitGPB-2 in Caenorhabditis elegans regulates the Goα–Gqαsignaling network through interactions with theregulator of G-protein signaling proteins EGL-10 andEAT-16. Genetics, 158, 221–235.

7. Miller, K. G., Emerson, M. D. & Rand, J. B. (1999). Goαand diacylglycerol kinase negatively regulate the Gqαpathway in C. elegans. Neuron, 24, 323–333.

8. Fukuto, H. S., Ferkey, D. M., Apicella, A. J., Lans, H.,Sharmeen, T., Chen,W. et al. (2004). G protein-coupledreceptor kinase function is essential for chemosensa-tion in C. elegans. Neuron, 42, 581–593.

9. Ford, C. E., Skiba, N. P., Bae, H., Daaka, Y., Reuveny,E., Shekter, L. R. et al. (1998). Molecular basis forinteractions of G protein βγ subunits with effectors.Science, 280, 1271–1274.

10. Lambright, D. G., Sondek, J., Bohm, A., Skiba, N. P.,Hamm, H. E. & Sigler, P. B. (1996). The 2.0 Å crystalstructure of a heterotrimeric G protein. Nature, 379,311–319.

11. Lodowski, D. T., Pitcher, J. A., Capel,W.D., Lefkowitz,R. J. & Tesmer, J. J. (2003). Keeping G proteins at bay: acomplex between G protein-coupled receptor kinase 2and Gβγ. Science, 300, 1256–1262.

12. Loew, A., Ho, Y. K., Blundell, T. & Bax, B. (1998).Phosducin induces a structural change in transducinβ γ. Structure, 6, 1007–1019.

13. Cheever, M. L., Snyder, J. T., Gershburg, S., Siderovski,D. P., Harden, T. K. & Sondek, J. (2008). Crystal

structure of the multifunctional Gβ5–RGS9 complex.Nat. Struct. Mol. Biol. 15, 155–162.

14. Davis, T. L., Bonacci, T.M., Sprang, S. R.& Smrcka,A.V.(2005). Structural and molecular characterization of apreferred protein interaction surface on G protein β γsubunits. Biochemistry, 44, 10593–10604.

15. Johnston, C. A., Kimple, A. J., Giguere, P. M. &Siderovski, D. P. (2008). Structure of the parathyroidhormone receptor C terminus bound to the G-proteindimer Gβ1γ2. Structure, 16, 1086–1094.

16. Li, Y., Sternweis, P. M., Charnecki, S., Smith, T. F.,Gilman, A. G., Neer, E. J. & Kozasa, T. (1998). Sites forGα binding on the G protein β subunit overlap withsites for regulation of phospholipase Cβ and adenylylcyclase. J. Biol. Chem. 273, 16265–16272.

17. Panchenko, M. P., Saxena, K., Li, Y., Charnecki, S.,Sternweis, P. M., Smith, T. F. et al. (1998). Sitesimportant for PLCβ2 activation by the G protein βγsubunit map to the sides of the β propeller structure.J. Biol. Chem. 273, 28298–28304.

18. Bonacci, T. M., Ghosh, M., Malik, S. & Smrcka, A. V.(2005). Regulatory interactions between the aminoterminus of G-protein βγ subunits and the catalyticdomain of phospholipase Cβ2. J. Biol. Chem. 280,10174–10181.

19. Myung, C. S., Lim,W. K., DeFilippo, J. M., Yasuda, H.,Neubig, R. R. & Garrison, J. C. (2006). Regions in the Gprotein γ subunit important for interaction withreceptors and effectors. Mol. Pharmacol. 69, 877–887.

20. Yoshikawa, D. M., Bresciano, K., Hatwar, M. &Smrcka, A. V. (2001). Characterization of a phospho-lipase C β 2-binding site near the amino-terminalcoiled-coil of G protein β γ subunits. J. Biol. Chem. 276,11246–11251.

21. Lichtarge, O., Bourne, H. R. & Cohen, F. E. (1996). Anevolutionary trace method defines binding surfacescommon to protein families. J. Mol. Biol. 257, 342–358.

22. Gu, X. & Vander Velden, K. (2002). DIVERGE:phylogeny-based analysis for functional–structural di-vergence of a protein family. Bioinformatics, 18, 500–501.

23. Zheng, Y., Xu, D. & Gu, X. (2007). Functionaldivergence after gene duplication and sequence–structure relationship: a case study of G-protein alphasubunits. J. Exp. Zool. B: Mol. Dev. Evol. 308, 85–96.

24. Armon, A., Graur, D. & Ben-Tal, N. (2001). ConSurf:an algorithmic tool for the identification of functionalregions in proteins by surface mapping of phyloge-netic information. J. Mol. Biol. 307, 447–463.

25. Lichtarge, O., Bourne, H. R. & Cohen, F. E. (1996).Evolutionarily conserved Gαβγ binding surfacessupport a model of the G protein–receptor complex.Proc. Natl Acad. Sci. USA, 93, 7507–7511.

26. Jones, S. & Thornton, J. M. (1996). Principles ofprotein–protein interactions. Proc. Natl. Acad. Sci.USA, 93, 13–20.

27. Hubbard, S. J. & Thornton, J. M. (1993). ‘NACCESS,’computer program. Department of Biochemistry andMolecular Biology, University College London.

28. Blaauw,M.,Knol, J. C., Kortholt, A., Roelofs, J., Ruchira,Postma, M. et al. (2003). Phosducin-like proteins inDictyosteliumdiscoideum: implications for the phosducinfamily of proteins. EMBO J. 22, 5047–5057.

29. Gaudet, R., Bohm, A. & Sigler, P. B. (1996). Crystalstructure at 2.4 angstroms resolution of the complex oftransducin βγ and its regulator, phosducin. Cell, 87,577–588.

30. Madabushi, S., Yao, H., Marsh, M., Kristensen, D. M.,Philippi, A., Sowa, M. E. & Lichtarge, O. (2002).Structural clusters of evolutionary trace residues are

1054 Novel Gβ Interfaces

statistically significant and common in proteins. J. Mol.Biol. 316, 139–154.

31. Thompson, J. D., Gibson, T. J., Plewniak, F., Jeanmou-gin, F. & Higgins, D. G. (1997). The CLUSTAL_Xwindows interface: flexible strategies for multiplesequence alignment aided by quality analysis tools.Nucleic Acids Res. 25, 4876–4882.

32. Huelsenbeck, J. P., Ronquist, F., Nielsen, R. &Bollback, J. P. (2001). Bayesian inference of phylogenyand its impact on evolutionary biology. Science, 294,2310–2314.

33. Ronquist, F. & Huelsenbeck, J. P. (2003). MrBayes 3:Bayesian phylogenetic inference under mixed models.Bioinformatics, 19, 1572–1574.

34. DeLano, W. L. (2002). The PyMOL Molecular GraphicsSystem. DeLano Scientific, San Carlos, CA.

35. Jones, S., Marin, A. & Thornton, J. M. (2000). Proteindomain interfaces: characterization and comparisonwith oligomeric protein interfaces. Protein Eng. 13,77–82.

36. Nakayama, M., Kikuno, R. & Ohara, O. (2002).Protein–protein interactions between large proteins:two-hybrid screening using a functionally classifiedlibrary composed of long cDNAs. Genome Res. 12,1773–1784.

37. Wing, M. R., Snyder, J. T., Sondek, J. & Harden, T. K.(2003). Direct activation of phospholipase C-ε by Rho.J. Biol. Chem. 278, 41253–41258.