Structural Basis for Mobility in the 1.1Å Crystal Structure of the NG Domain of Thermus aquaticus Ffh

Structural Basis for Mobility in the 1.1 A CrystalStructure of the NG Domain of Thermus aquaticus Ffh

Ursula D. Ramirez1†, George Minasov1†, Pamela J. Focia1

Robert M. Stroud2, Peter Walter2, Peter Kuhn3 andDouglas M. Freymann1*

1Department of MolecularPharmacology and BiologicalChemistry, NorthwesternUniversity Medical SchoolChicago, IL 60611, USA

2Department of Biochemistryand Biophysics, Universityof California, San FranciscoSan Francisco, CA 94143, USA

3Stanford SynchrotronRadiation LaboratoryStanford UniversityStanford, CA 94309, USA

The NG domain of the prokaryotic signal recognition protein Ffh is a two-domain GTPase that comprises part of the prokaryotic signal recognitionparticle (SRP) that functions in co-translational targeting of proteins tothe membrane. The interface between the N and G domains includes twohighly conserved sequence motifs and is adjacent in sequence and struc-ture to one of the conserved GTPase signature motifs. Previous structuralstudies have shown that the relative orientation of the two domains isdynamic. The N domain of Ffh has been proposed to function in regu-lating the nucleotide-binding interactions of the G domain. However, bio-chemical studies suggest a more complex role for the domain inintegrating communication between signal sequence recognition andinteraction with receptor. Here, we report the structure of the apo NGGTPase of Ffh from Thermus aquaticus refined at 1.10 A resolution.Although the G domain is very well ordered in this structure, the Ndomain is less well ordered, reflecting the dynamic relationship betweenthe two domains previously inferred. We demonstrate that the anisotropicdisplacement parameters directly visualize the underlying mobilitybetween the two domains, and present a detailed structural analysis ofthe packing of the residues, including the critical a4 helix, that comprisethe interface. Our data allows us to propose a structural explanation forthe functional significance of sequence elements conserved at the N/Ginterface.

q 2002 Elsevier Science Ltd. All rights reserved

Keywords: ultrahigh resolution; SRP; Ffh; GTPase; X-ray crystallography*Corresponding author

Introduction

The signal recognition particle (SRP) is a phylo-genetically conserved ribonucleoprotein thatmediates co-translational targeting of nascentproteins to the membrane. SRP-mediated targetingis dependent on GTP binding and hydrolysis bytwo components of the pathway, SRP54 and SR,its membrane receptor. Homologs of these twoproteins are found throughout evolution, and inprokaryotes they are termed Ffh and FtsY, respec-tively. Remarkably, the two proteins, the SRP

GTPase and its receptor, each contains a struc-turally homologous two-domain module, the NGdomain, which defines the SRP subfamily ofGTPases. The structures of the NG domains fromThermus aquaticus and Acidianus ambivalens Ffh,and Escherichia coli FtsY, have been determined.1 – 3

The Ffh NG comprises a four a-helix N domainand a G domain, which has similarity toother members of the GTPase superfamily ofproteins (Figure 1). The N domain abuts the Gdomain across the a4 helix of that domain. (Thea4 helix, so named with reference to the structureof small GTPases, is termed the a6 helix in otherstudies.3). The a4 helix is adjacent to one of thecharacteristic GTPase recognition motifs, termedmotif IV.4

The structures of the complex of the T. aquaticusNG domain with GDP revealed that the relation-ship between the two domains is dynamic.5 In the

0022-2836/02/$ - see front matter q 2002 Elsevier Science Ltd. All rights reserved

† These authors contributed equally to this work.

E-mail address of the corresponding author:[email protected]

Abbreviations used: ADP, anisotropic displacementparameter; SRP, signal recognition particle; TLS analysis,translation, liberation, and screw analysis.

doi:10.1016/S0022-2836(02)00476-X available online at http://www.idealibrary.com onBw

J. Mol. Biol. (2002) 320, 783–799

apo protein, the N domain is mobile, as evidencedby the fact that it can be trapped in differentconformations under different crystallizationconditions. The interface is closely associated withthe position of the guanine recognition motif IV,so that its position is coupled to the position ofthat motif,1 and, therefore, the relationshipbetween the N domain and the G domain in thenucleotide-bound state is somewhat restricted.This coupling between the two domains, and thefact that the interface comprises sequence elementsthat are highly conserved between SRP GTPases(Figure 1), suggests that the N domain plays a rolein the regulation of the nucleotide-binding state ofthe protein.

Although biochemical studies have elucidated,to some extent, the function of GTP binding andhydrolysis in SRP-mediated targeting, the roleplayed by the N domain in this process remainsunknown. The interface has been the focus ofmutagenesis studies,6,7 which suggest that thisregion plays a role in two different functionalitiesof the SRP GTPase; signal peptide recognition andinteraction with the SRP receptor. In the firststudy, hydrophobic residues of the conservedsequence motif of the N domain part of theinterface were changed to alanine, and it wasfound that the mutations caused defects in signal

sequence recognition.7 In the second study, aconserved glycine residue of the G domain motifwas mutated, and it was found that this caused adefect in the interaction between SRP and receptor,but had little effect on GTP binding or hydrolysis.6

These data are consistent with the N domain play-ing a central role in SRP-mediated targeting, asthe NG interface is involved directly or indirectlyin both the initial recognition of the signalsequence and the subsequent interaction with themembrane receptor.

Crystals of the apo NG domain of Ffh fromT. aquaticus diffract to 1.0 A resolution,1 allowingus the opportunity to understand the protein struc-ture in detail at ultrahigh resolution. The approxi-mately tenfold increase in diffraction data relativeto the previous report at 2.0 A resolution1 enablesrefinement of anisotropic displacement parameters(ADPs) and provides well-defined positions formost of the hydrogen atoms. Both allow us todefine better the structural relationships that deter-mine the function of the protein. The anisotropygives information about the directionality ofatomic and domain motions, and we find that itreflects directly the underlying mobility betweenthe N and G domains. And, because tight proteinpacking is dependent on proper orientations ofthe small but numerous hydrogen atoms,8 packinganalyses of this structure reveals details thatcan explain the roles of the particular conservedhydrophobic side-chains at the N/G interface.Therefore, the increase in resolution between thisstructure and the previous one is significant, as itrepresents the difference between a model thatcan describe the protein packing only at the levelof the heavy atoms of the structure, and onethat defines the myriad of interactions betweenthe hydrogen atoms of the structure that actuallydetermine the fold. (In this context, it is worthnoting that the difference is analogous to thetransition that distinguishes the protein foldingendgame.9)

Our goal is to understand the structural basis forfunction in the SRP GTPases. Here, we focus on theinterface between the N and G domains, takingadvantage of the ultrahigh-resolution data to visu-alize directly the distribution of orientations of thetwo domains in the crystal. Packing analysis allowsus to present a structural rationale for the conser-vation of residues central to the interface that isconsistent with the functional roles suggested bybiochemical studies. This work thus complementsthose studies. The a4 helix at the N/G interface islikely to play a central role in SRP GTPase function.In the T. aquaticus Ffh NG domain, the structuralbasis for its mobility is the persistence of tightpacking interactions with the N domain, and loosepacking with the underlying b-sheet that allow itto adjust to different conformations of the motif IVloop. Differences in the structures of the T. aquati-cus Ffh NG and the available structures of otherSRP GTPases can be understood in terms of thismodel.

Figure 1. Two-domain structure of the SRP GTPase.The N domain, a four a-helix bundle, interacts with theG domain across the a4 helix of the G domain. Theposition of the GTP-binding site, on the far side of theplane of the image, is indicated. The a4 helix couplesthe position of the N domain to structural features ofthe nucleotide binding site. The highly conservedALLEADV and DARGG sequence motifs contribute tothe interaction across the N/G interface. The arrowspoint to the regions of the motifs from their respectivesequence boxes. The ALLEADV motif includes theC-terminal end of aN2 and the turn between aN2 andaN3 (top arrow); the DARGG motif comprises the looppreceding the a4 helix and much of the helix itself(bottom arrow).

784 1.1 A X-ray Structure of T. aquaticus Ffh NG

Results and Discussion

Quality of the refined model

The structure of the Ffh NG domain was refinedat 1.1 A resolution using data measured from asingle crystal at SSRL BL 9-1. Particular care wastaken to ensure that the data were very complete,and the resulting dataset has excellent statistics(Table 1). Crystallographic refinement was carriedout in several stages. The starting model for therefinement was the structure of the apo Ffh NGdetermined at 2.0 A resolution (PDB ID 1ffh).Initial refinement was carried out using X-PLOR,10

and the structure was subsequently refined usingSHELXL-9711 in order to introduce ADPs. Thedrop in Rfree on adding anisotropy was ,4%,which is typical for structures at this resolution.12,13

Following refinement with SHELXL, Rcryst was13.3%; however, Rfree remained relatively high,18.7%, and a number of loops of the N domainwere poorly ordered but could not be modeled inalternative conformations. Furthermore, althoughwe had made a point of measuring complete datafrom the low to high-resolution limits, we wereunable to use low-resolution data beyond 10 Aresolution because of inadequacies of the Babinetsolvent model implemented in SHELXL. Therefore,the final stage of the refinement was carried outusing the program REFMAC,14 which implements

a solvent model similar to that used in X-PLOR.15

This stage resulted in little change in Rcryst,but yielded a significant 2% drop in Rfree for alldata over the full 50.0–1.10 A resolution range(Table 2).

The current model comprises 289 amino acidresidues and 324 solvent molecules, with a crystal-lographic R value of 13.5%, and an Rfree of 16.9%.The same subset of reflections was used for Rfree

throughout the refinement. The geometry of themodel is quite good (Table 2) and the electrondensity map in the well-ordered core of the Gdomain is remarkably clear (Figure 2(a)). All resi-dues are within the allowed (94.4%) or additionalallowed (5.6%) regions defined by PROCHECK.16

There is only one striking deviation in the main-chain geometry, which occurs at the Phe102/Leu103 peptide bond in the interior of the Gdomain. There, the omega torsion is 1678, and theconformations of the two residues, which occur inthe interior of the protein, are very well-defined inthe electron density map.

Thirty residues exhibit clearly defined multipleconformations in the electron density map (Figure2(b)). Most occur on the surface and are poorlyconstrained but, interestingly, a number occur inthe interior of the protein. Nine leucine residues,including the well-buried Leu103, exhibit discretedisorder; both Leu118 and Leu202 exhibit x1rotamer flips, three others exhibit x2 flips, and theremainder undergo small positional shifts. In theN domain, 13 residues are in alternative confor-mations, with over half of those (Leu45, Arg49,Arg35, Met39, Lys83, Thr77, and Glu80) clusteredacross a crystallographic 2-fold screw axis thatrelates adjacent N domains. The multiple confor-mations of these residues reflect accommodationof different conformational substates that arise asa consequence of mobility between the N and Gdomains (see below).

A number of loops are poorly ordered in thestructure and cannot be modeled as discretealternative conformations. The electron density ofthe distal loops of the N domain, in particular, isremarkably smeared, presumably due to convolu-tion of intrinsic disorder of the loops with thedistribution of orientations of the N domain inthe crystal. Poorly defined electron density in theregion of the DARGG motif loop (Figure 1) thatlinks motif IV with the a4 helix1 may be function-ally significant. Although the overall conformationof the region is discerned readily and is similar tothat in previous apo and GDP structures, in severalrecent structures of the nucleotide-bound protein,the loop (residues 250–252) has been observed tobe displaced due to crystal packing interactions, ina position that uncouples the N and G domains(P.J.F., H. Alam & D.M.F., unpublished results).Finally, more than 10% of the solvent atoms havebeen built into well-defined but “peanut-like” den-sity features. These are consistent with, and havebeen modeled as, discrete and mutually exclusivesolvent sites; however, we cannot exclude the

Table 1. Data collection statistics

Space group C2

Unit cella, b, c (A), b (deg)

a ¼ 99.73, b ¼ 53.67,c ¼ 57.84, b ¼ 119.92

Resolution (A) 50.0–1.10Observations 694,235Unique reflections 102,368

50.0–1.10 A 1.12–1.10 A

Rsyma 0.037 0.324

Completeness (%) 95.4 89.9Redundancy 4.27 2.96Average I/s 37.8 3.2

a Rsym ¼P

lIh 2 kIhll=P

Ih; where, kIhl is the average intensityover symmetry equivalents.

Table 2. Refinement statistics

Resolution (A) 50.0–1.10No. reflections 92,430Rcryst

a (%) 13.5Rfree (%) 16.9Protein atoms 2513Alt. conformations 30Solvent atoms 326

Average Biso (A2)Protein 22.6Solvent 39.9

rms bond (A) 0.024rms angle (deg.) 2.086

a Rcryst ¼P

lFo 2 Fcl=P

Fo; Rfree was calculated for a test set ofreflections (8%) omitted from the refinement.

1.1 A X-ray Structure of T. aquaticus Ffh NG 785

possibility that they represent an uncharacterizedsolvent component.

Although the electron density map is excellentoverall, its quality varies tremendously. Despitediffraction to 1.0 A resolution, the equivalentisotropic B factors average ,25 A2, which isrelatively high for a structure at this resolution.17

This arises because two extensive crystal packinginteractions stabilize the position of the G domain,but less well-defined interactions allow the Ndomain to occupy a range of conformationalsubstates in the crystal. This is reflected as well ina skewed distribution of temperature factorsbetween the N and G domains (see Figure 4), andit explains the relatively high crystallographic Rand B factors in this structure. What is intriguingabout the ultrahigh-resolution structure, however,is that it allows us to determine the directionalityof the intrinsic intramolecular motion and propose

a structural explanation for the functional mobilityat the N/G domain junction.

Analysis of anisotropicdisplacement parameters

Following refinement of the ADPs, the averageanisotropy (defined as the ratio of the minimumand maximum eigenvalue of the anisotropic dis-placement matrix) was found to be 0.55 ðs ¼ 0:13Þ;which is similar to the distribution seen in otherproteins.18 The mean value of the solvent aniso-tropy is ,0.40. We examined the distribution ofanisotropy more closely and found that overmuch of the N domain, but not the remainder ofthe protein, the ellipsoids of motion of the alphacarbon atoms were clearly correlated (Figure 3).This suggested that there was a coherent motionof the domain, and was analyzed by applying a

Figure 2. Electron density at 1.1 A resolution. (a) Electron density in the well-ordered core of the G domain. The2Fo 2 Fc map calculated using data from 50 to 1.1 A resolution is contoured at 5s. Superimposed is a hydrogen atomomit difference map, calculated using data from 5.0 to 1.1 A resolution and contoured at 1.8s. The positions of manyhydrogen atoms are well defined in the well-ordered core of the G-domain (note the disordered methyl groups andbackbone amide hydrogen atoms). (b) Residues in multiple conformations. In the first panel, the Cg of proline occupiestwo positions; interestingly, at lower resolution in X-PLOR this residue refines as “flat”. The final panel illustrates atypical leucine x1 rotamer flip. The 2Fo 2 Fc electron density is contoured at 0.8s.


rigid-body analysis of the domain movement interms of a translation, liberation, and screw (TLS)model.13 Because rigid-body motion is not con-sidered explicitly during the refinement, an under-lying anisotropic domain shift will be reflected inthe atomic ADPs, and a description of the move-ment can be extracted from those ADPs.19,20 Thereare two caveats to bear in mind: first, the diffrac-tion data cannot themselves distinguish betweencorrelated and uncorrelated motion,20 and thisinference has to be imposed during our interpret-ation of the data. However, as we discuss below,domain motion at the interface is well established,based on the comparison of the apo and GDP-bound structures, so this assumption is well justi-fied in the structural analysis here. Second, theTLS parameters tend to over-fit the underlyingmotion, which may or may not be due to domainmotions, and so provide only an upper limit onthe rigid-body contribution.13

To carry out the analysis, we considered twodomains of the protein as rigid bodies. The firstcomprised the helices of the N domain, residues2–14, 26–59, and 71–88 (thus excluding the distalloops of the N domain) and helix a4 of the Gdomain, residues 253–262. The second comprisedthe rigid core of the G domain (residues 99–219).This is certainly a simplification of the dynamicsof the protein; however, it works well for ourpurpose here, to characterize and explain theimplicit motion of the N domain (which presu-mably arises as a series of conformationalsubstates21) in the crystal. We used the programANISOANL22 to obtain the parameters of motion

(translation, libration, and screw) for a rigid bodycomprising each domain from the distribution ofatomic ADPs of the main-chain atoms. The derivedTLS tensors were then analyzed using TLSANL.23

That the TLS model accounts for much of theanisotropy observed in the structure is shown bythe good fit between the model and the observedanisotropy and equivalent isotropic B values(Figure 4). The correlation coefficient over theequivalent Biso values for both domains is ,0.85.The distal loops of the N domain, which are dis-ordered, were not included in the TLS model andthey degraded the fit when they were included.

The anisotropy observed in the structure appearsto be fit particularly well by the TLS model (Figure4(b)), with a correlation coefficient over both the Nand G domains of 0.78. The mean translational dis-placement of the TLS component for the twodomains was 0.20 A and 0.16 A, respectively, andthe mean-square angles of libration13 of the twodomains were 1.7882 and 1.2882, respectively.However, while the mean values were not dis-similar, the distribution of the displacementsaround the libration axes was quite different.Thus, for the G domain they range from 0.97 to1.8382, while for the N domain, they range from0.60 to 3.5982, a range of ,sixfold that is charac-teristic of anisotropic domain motion.13,22 Themajor axis of libration of the N domain is approxi-mately aligned with the aN4 helix (Figure 5(a)).This implies that the anisotropy of the domainarises predominantly from a longitudinal rotationof the N-domain helical bundle, and is remarkablyreminiscent of the domain motion inferred fromcomparison of the apo and GDP-bound structures.5

To compare this explicitly, we carried out ananalysis of the conformational change between theapo and ligand-bound states of Ffh NG in termsof a hinge rotation between the two domainsusing the model-independent analysis imple-mented in DYNDOM.24 The resulting inter-domainrotation axis is consistent with and has similardirectionality to the predominant libration axis ofthe N domain (Figure 5(b)). That they are notidentical is, of course, expected, because both areidealizations of what is certainly a more complexstructural change, and one represents a singlehinge for rotation between the two conformationalstates, while the other represents a decompositionof the motion of the one domain. The scales of thetwo motions are also quite different (the meanrotation around the major libration axis in thisstructure is only 3.5982, while the hinge motionbetween the two domains that accompanies theconformational change on binding GDP is ,98.Nevertheless, their approximate coincidence con-firms that the directional preference for motion ofthe N domain is consistent with the conformationalchange between the apo and GDP-boundstates, and this motion is, therefore, likely to befunctionally significant.

It appears, therefore, that the relatively hightemperature factors for the N domain are not due

Figure 3. Anisotropic temperature factors reflectcoherent motion of the N-domain. The anisotropicellipsoids of motion at the 90% probability level for thea-carbon atoms of helices aN1 and aN2 highlight themobility of the main-chain atoms of the N-domain. Theprincipal axes of vibration along both of the helices arelargely coherent (,up and down; compare with theellipsoids near residue 1). The direction of this motion(presumably trapped substates in the crystal) is similarto that seen between the apo and GDP complex crystalstructures.


to random displacements in the crystal, but insteadarise from correlated motion of the N domain thatis accommodated within the crystal packing (asevidenced, for example, by clusters of side-chainsin multiple conformations) to yield a distributionof conformational substates. However, because itreveals directly a biologically-relevant structuralvariation, and because the protein model is verywell refined with an Rfree of ,17%, the structureaffords the opportunity to examine in unprece-dented detail the structural elements, primarily

contributed by two highly conserved sequencemotifs in the hinge region (see Figure 1), thatenable motion across the N/G domain interface.

Analysis of the N/G domain interface

The N/G domain interface is unique to theSRP GTPase subfamily, and is characterized by anumber of conserved sequence motifs (Figure 1).The interface between the two domains comprisesthe ALLEADV motif of the N domain,7 elements

Figure 4. A rigid-body modelaccounts for much of the mobilityof the N and G domains. TLS para-meters were derived for rigidbodies taken as (i) the N domainand a4 helix (residues 2–88 and253–262) and (ii) the core of the Gdomain (residues 99–219). In theupper plot, (a), the observedequivalent isotropic B values foreach group are compared to theisotropic displacement parameterspredicted from the rigid-bodymodel (shown as the diffuseyellow-green line superimposed onthe graph). The horizontal linesindicate the average Biso for each ofthe domains. The loops of the Ndomain, which are poorly fit,exhibit very high temperaturefactors and are disordered in theelectron density maps. In the lowerplot, (b), the anisotropy in therefined structure is compared tothe anisotropy predicted from therigid-body model (shown as the dif-fuse yellow-green line). Note thatanisotropy is defined such that iso-tropic 1.0 ! 0.0 highly anisotropic.


of helix a3 of the G domain, the DARGG motif andthe a4 helix that follows it,1 and a number ofwater molecules. The functions of the solvent-exposed residues of the interface motifs remain

largely unknown, as only a few interactions aredefined in the available crystal structures2,5 andthere is little biochemical evidence for their func-tion. Instead, it is the residues that contribute tothe hydrophobic core of the interface that havebeen shown to be sensitive to mutation.6,7 Muta-genesis of the leucine, valine and glycine residuesof the ALLEADV and DARGG motifs producereadily detectable phenotypes, while mutations ofthe hydrophilic residues do not. Furthermore,the effects of these mutations are functionally far-reaching. Although they do not affect the intrinsicGTP-binding and hydrolysis activity of theproteins, different mutations cause defects in therecognition of the signal sequence of the nascentpolypeptide, which is mediated primarily by theC-terminal M-domain,25,26 and in the formation ofthe SRP/SR complex. The data imply that the junc-tion between the domains and, presumably, theirrelative orientation is of central significance to thefunction of the SRP GTPase.

As a first step towards understanding themolecular architecture of the interface in atomicdetail, we carried out packing analysis using anumber of computational tools that quantify andvisualize the molecular goodness-of-fit.8,27,28 Thesetools, implemented in the program PROBE,explicitly evaluate hydrogen atom contacts, and soare particularly effective when used with struc-tures refined at ultrahigh resolution.27,29 The hydro-phobic nature of a number of the functionallysignificant residues of the N/G interface makesthe interface particularly accessible to this type ofanalysis. Thus, by taking advantage of thewell-refined structure of the Ffh NG domain, wecan clearly understand the packing interactions ofthese residues, and this, because their interactionsare primarily hydrophobic, provides us withinformation about their functional roles.

The analysis carried out using PROBE allowed adirect visualization of the packing interactions ofeach atom as well as a quantitative measure ofpacking density in terms of a “score” that reflectedthe fraction of the available atomic surface areaparticipating in packing interactions. A qualitativeindication of the extent of atomic packing inter-actions over the entire structure is provided by acontour density plot,28 and in this representation anumber of features of the protein packing arehighlighted (Figure 6). Of course, the core of theprotein exhibits higher packing density than thesurface, but a number of other features, such asintrahelical packing, stand out as having high“density” as well. In the structure of the Ffh NGdomain, however, this representation reveals anadditional and unexpected feature—a packingdensity hole that occurs adjacent to the N/Gdomain interface (Figure 6(a)). This region of lowpacking density does not occur between the twodomains, but instead occurs between the a4 helixof the G-domain and the underlying b-sheet ofthe G-domain. The feature can be understood asarising from a particular pattern of residue

Figure 5. TLS analysis of the anisotropic displacementparameters reveals a prominent N-domain libration.(a) The libration axes of the TLS matrices derived sepa-rately for the N and G domains are plotted and super-imposed on a ribbon representation of the structure.The length of each axis is proportional to its mean-squarelibration. The most striking feature is a predominantlibration axis that lies approximately along the aN4helix of the N domain. (b) The vector of rotation obtainedfollowing model-independent analysis of the confor-mational change between the structures of the apo andMg2þGDP-bound Ffh NG.5 The vector generated byDYNDOM24 is superimposed on the structure color-coded to distinguish a “hinge” region (in yellow) fromthe two domains (taken as mobile and fixed). Note thatthe a4 helix of the G domain is paired with the Ndomain in this model-independent analysis. Note alsothat the hinge axis is not coincident with, but is consist-ent with, the major axis of libration in the TLS model.However, the scales of the two motions are quitedifferent.


conservation (see below). It occurs just aboveWat423, which can, therefore, be considered toplug the hole (Figure 6(b)).

A packing cavity at the N/G interface is intri-guing, of course, as it suggests an underlyingdesign that facilitates movement of the N domainand helix a4 against the G domain consistent withthat inferred previously from the comparison ofthe apo and GDP-bound structures (see Figure5(b)).5 Indeed, visualization of the packing inter-actions between the a4 helix and the N and Gdomains separately (Figure 7) reveals a clearclustering such that the a4 helix is packed tightly(i.e. has many interactions) with the conservedALLEADV motif of the N domain. It interacts alsowith the a3 helix of the G domain, which has beenshown to move in concert with it.5 In contrast, thehelix has remarkably few packing interactionswith the underlying b-sheet of the G domain(Figure 7(b)). The helix, therefore, can be con-

sidered to be packed with the N domain ratherthan the G domain from which it arises, and wepropose that this “unpacking” from the G domainoccurs in the SRP GTPase so that the helix canfacilitate the relative motion of the two domains.

Packing interactions of conserved residues atthe interface

The residues of the ALLEADV motif occur at theturn between helices aN2 and aN3 of the Ndomain (Figures 1 and 6(b)). The motif showssome variation in sequence between differentspecies (in the T. aquaticus Ffh, the sequence of themotif is ALMDADV); however, the conformationof the motif is well-conserved.1,2 The solvent-exposed residues Met39 and Asp40 have noobvious interactions at the interface, and it hasbeen speculated that they, along with conservedresidues of the DARGG motif and the a3 helix of

Figure 6. Contact density analysis reveals a loosely packed hinge region. (a) A contour representation of the packingdensity is superimposed on the a-carbon backbone of the NG domain. The large hole between the two domains repre-sents a region of low packing density between the a4 helix of the G domain and the underlying b-sheet that may acti-vate it to shift across the b-sheet during binding and release of nucleotide. The contour level corresponds to an atomicpacking density of approximately 20% of the total available. (b) A stereo overview of the residues at the N/G interface.In this view, the ALLEADV motif (residues 37–43) is on the far side the a4 helix (orange). The N domain is on the left,and the G domain on the right. Residues discussed in the text are labeled, and the water molecules discussed in thetext are highlighted. Three strands of the b-sheet of the G domain are indicated, in the right foreground.


the G domain, contribute to a functionally impor-tant intermolecular interaction surface.2,5 Asp42reaches across the interface toward the G domain,where it contributes a hydrogen bond at theamino terminus of helix a3.5 The remaining

residues of the motif, all hydrophobic, contributeto the core packing of the N domain and the N/Ginterface. Significantly, many of their packing inter-actions are with other highly conserved residues(Figure 6(b), and see Figure 8). Thus, Ala37 packsagainst the aliphatic chain of Arg8 and againstLeu5, and Ala41, completely buried, contributesextensive interactions to the interface between thealiphatic chain of Arg252, the N domain, andLeu5. Val43 seals off the hydrophobic core of theN domain, as it rests against a4 of the G domain,and is situated adjacent to Leu257 of the a4 helix.Mutation of the two hydrophobic residues of themotif corresponding to Leu38 and Met 39 inT. aquaticus in the eukaryotic SRP54 to either valineor alanine causes defects in signal sequence recog-nition in in vitro translation/targeting assays.7

Leu5, which is involved in packing interactionswith a number of residues of the motif, is veryhighly conserved in the SRP GTPases.30 Althoughits function has previously been obscure, the inter-actions identified above suggest that Leu5 contri-butes important packing interactions to the resi-dues that form the interface between the twodomains, and so can be considered part of theN/G interface of Ffh.

The conserved DARGG motif of the G-domaincomprises the residues Asp250 through Ser258,and includes the loop that precedes helix a4 andmost of the helix itself (Figures 1 and 6(b)). Thefirst three residues are solvent-exposed, and theirfunction is defined only partly. Asp250 hydrogenbonds to Wat481, which helps initiate the a4 helixand both Ala251 and Arg252 are somewhat dis-ordered, with the position of the arginine side-chain very poorly defined at the surface. Theremainder of the motif is buried, however, andpacking analysis illuminates a number of function-ally significant features. Both the glycine residues,in particular, are very tightly packed at the N/Ginterface (see Table 3). Gly253, highly evolution-arily conserved, is situated between the side-chainsof Phe2 and Asp250 such that the alpha carbonhydrogen atoms are in extensive van der Waalscontact with the phenyl group, on the one side,and with Asp250 and Wat481 at the amino termi-nus of helix a4, on the other (Figure 8(a)). Gly254,completely conserved in the SRP GTPases, plays acentral role in the interface between the twodomains. Its alpha carbon atom is nestled againstthe 42/43 peptide bond, with extensive packinginteractions against the main-chain and side-chainatoms of both residues. The carbonyl oxygen atomof residue 41 is adjacent, and is in position to forma long hydrogen bond to the backbone amidegroup of the glycine residue. Not surprisingly, thepacking density scores for Gly253 and Gly254 areapproximately double that of average glycine resi-dues in this structure, with Gly253 having contactsover 42% of its possible contact surface area andGly254 having contacts over 34% of its possiblecontact surface area (Table 3). The two glycineresidues clearly function here to allow both tight

Figure 7. PROBE contact dots on the N domain side,but not the G domain side, of helix a4 indicate tightpacking. Small probe contact dots8 are shown in blackbetween the a4 helix and surrounding areas of the Nand G domains. (a) Contact dots from residues of the a4helix to only the N domain. The N terminus (residues1–4), the loop between helices aN2 and aN3 (ALLEADVmotif), and the C-terminal end of aN4 exhibit numerousinteractions with the a4 helix. (b) Contact dots fromresidues of the a4 helix to the G domain. Interactionsare primarily with residues 220–224 of helix a-3. (c) Asimilar analysis of the packing of helix a1 of the Gdomain. This helix, which contributes the phosphatebinding P-loop, motif I, is tightly packed relative to theother helices of the NG domain (see Table 3).


packing of the a4 helix and the close apposition ofthe N and G domains. It is mutation of the secondof these glycine residues that has been shown tocause a functional defect in the E. coli Ffh.6

In contrast to the well-packed glycine residues ofthe interface, the conserved Ala255 that followsthem appears to be much less constrained. Theside-chain of the alanine residue is directedtowards the b5-a3 loop of the G domain and,while the residue is completely buried (Table 3), ithas remarkably few interactions with the remain-der of the polypeptide. Its packing interactions arelimited to Wat481 and Wat482, which initiate thehelix a4, and to the carbonyl group of Gly223.Significantly, the alanine residue is directedtowards the packing hole between the a4 helixand the b-sheet of the G-domain. In Ffh andSRP54, the residue at the corresponding positionoccurs only as alanine in prokaryotes and only as

glycine in eukaryotes. Ala255 therefore plays acentral, and presumably functionally significant,role in creating the packing hole below the a4helix.

The side-chain of Leu257 is directed from the a4helix into the center of the N-domain helix bundle.There, it is situated by packing interactions withsurprisingly few residues: Leu5, Leu38, Val43,Leu82 and Leu86 (Figure 8(b)). Indeed, Leu257 hasan average packing density score compared toother leucine residues (Table 3), substantially lessthan the maximum observed in this structure(41%). However, of the five residues that theleucine packs against, two, Val43 and Leu82, areconserved completely, and two, Leu5 and Leu38,are highly conserved in the SRP GTPase subfamily.This implies that although the packing interactionsare relatively sparse, they maintain a precise struc-tural relationship; one that would be disrupted,

Figure 8. Conserved residues atthe NG interface contribute to awell-packed interface. Small probecontact dots8 are shown at a dotdensity of 50 dots/A2. The color ofthe dots indicates the types of inter-actions, with blue dots indicatingwide contacts, greater than 0.25 A,green indicating close contactsbetween 0 A and 0.25 A, yellowindicating small overlaps of lessthan 0.2 A, and purple indicatinghydrogen bonds. Hydrogen atomsare included in the analysis, butare not shown in the Figure. (a) Thepacking density around Gly253 andGly254 (at the center) is especiallytight for glycine residues and thusexplains conservation of these tworesidues. (b) Leu257 appears tohave relatively few interactions;however, those are with highly con-served residues, including Leu5,Val43, Leu38, and Leu82. Note theinteraction between Leu5 andLeu257, which involves methylhydrogen atoms of the two resi-dues. (c) Interactions with Ser258,which is a completely conservedresidue in the SRP GTPases. Boththe Gly253/Gly254 and Ser258interactions serve as the “glue”across the N/G interaction surface.


for example, by the substitution of leucine forvaline or vice versa. Their highly conserved inter-actions, therefore, may be functionally significant,and we speculate that they serve to facilitate oraccommodate the mobility of the N/G domaininterface.

The residue that follows, Ser258, is, like theglycine residues of the motif, particularly well-packed in the T. aquaticus Ffh NG, with a packingdensity 1.3 s above the average for serine residues(Table 3). The serine hydroxyl group accepts ahydrogen bond from the amide nitrogen atom ofAsn44 (which follows the ALLEADV loop) andinteracts with weakly bound Wat597. It alsodonates a hydrogen bond back to the a4 helixmain chain. The Cb atom is positioned by packinginteractions with Val47, Leu227 and Val262, whichfunction, therefore, to bridge the N and G domainsnear the C-terminal end of the a4 helix.

The key conclusion from these observations isthat highly conserved residues of the two sequencemotifs contribute to specific packing interactions atthe N/G interface that incorporate a number ofadditional highly conserved residues to assemblea specific structural relationship between the twodomains. There is tight packing at the interface

between the N domain and helix a4, and there isloose packing between the helix and the under-lying b-sheet. This supports the notion that the a4helix is designed to move in concert with the Ndomain instead of being a fixed part of the Gdomain, and is consistent with the observation ofintrinsic anisotropy of the N domain arising fromthe loose fit at the junction.

Three well-ordered water molecules also contri-bute to the N/G interface; their positions are well-defined in the 1.1 A resolution structure, and theirinteractions appear to be preserved between thedifferent structures of the apo, Mg2þGDP-bound,and GMPPNP-bound Ffh NG domain.1,5,31 Two ofthe water molecules (Wat426 and Wat439) areassociated with the packing between a4 and the Ndomain, and one (Wat423) occurs between thehelices a3 and a4 and the underlying b-sheet(Figure 6(b)). Wat426 provides an importantadditional contact bridging the a4 helix of the Gdomain and aN4 of the N domain by hydrogenbonding the carbonyl oxygen atom of Ala85 andLeu257 and the amide nitrogen atom of His261.The water is particularly well-ordered in the dif-ferent structures of the NG domain; in the 1.1 Astructure its equivalent Biso value is 16.8 A2.Wat439, also well ordered, is linked to Wat426 byhydrogen bonding the intervening Wat440. Wat439hydrogen bonds the carbonyl group of Val43 ofthe ALLEADV motif, and packs tightly against theside-chain of Val47 and the main chain of Phe51.Finally, Wat423 is positioned in a hydrophobicpocket between the C-terminal ends of the a3 anda4 helices by hydrogen bonds to the carbonylgroup of Ala230 and to an adjacent Wat424. Thiswater molecule effectively closes the divergentend of the packing hole that occurs between thea4 helix and the b-sheet. The conserved positionsof these water molecules in this variety of struc-tures suggests that they have a structural role; inparticular, all can be considered to fill gaps in thefold by both hydrogen bonding and packinginteractions, and they may function to maintainthe ability of the two domains to move relative toone another, thus acting as “lubricants” for thepolypeptide interface.

The design of the N/G interface

We identify the following features of the inter-face between the N and G domains of Ffh: first, onone side of the interface a well-defined network ofside-chain and main-chain packing interactionscontributes to the interface, and these interactionsin the interior of the N-domain provide the firstexplanation for why specific hydrophobic residuesare sparsely distributed but highly conserved inthe N domain sequence. Second, on the other sideof the interface, a packing defect is maintained byconservation of a number of small hydrophobicresidues that enable mobility of the a4 helixrelative to the remainder of the G domain. Thefact that many of the residues involved are

Table 3. Packing scores and solvent accessibilities

Density scorea Deviationb Accessibility (A2)

A. ALLEADV 37–43Ala37 28 0.7 0.3Leu38 35 þ1.2 1.6Met39 14 0.1 77.1Asp40 13 20.5 77.2Ala41 35 þ1.4 0.1Asp42 18 0.1 51.6Val43 31 0.4 2.1

B. DARGG 250–258Asp250 18 0.1 11.7Ala251 10 21.1 40.8Arg252 16 0.1 124.3Gly253 42 þ3.1 0.0Gly254 34 þ2.4 1.0Ala255 16 20.5 0.1Ala256 24 0.3 0.1Leu257 24 0.1 2.3Ser258 43 þ1.3 0.0

C. G-domain helicesc

Density scorea Deviationb Contextd

a1 23 1.4 P-loopa1a 11 20.9 IBDa1b 12 20.7 IBDa2 12 20.7a3 16 0.2a4e 8 21.5 N/G interface

a Density score: % of total possible van der Waals contactsurface, evaluated using at 1.0 A PROBE sphere.

b Deviation from mean in standard deviations for thatresidue type (residues) or all helices.

c Scores are evaluated for the helix packing against theremainder of the G-domain.

d Context: identifying context for that helix.e The a4 helix when considered in the context of both the N

and G domains has a density score of 22.


evolutionarily conserved implies that these struc-tural features are functionally important. Becauseit has been shown that the hydrophobic residuesof the N/G interface can critically affect function,6,7

the interactions we identify are, therefore, likely tobe key to understanding the biological function ofthe protein. Thus, for example, the glycine residuesallow close approach of the backbone of theALLEADV motif to the loops of the G-domainbecause both are highly packed. The role of theglycine residues here is reminiscent of a similarrole for glycine residues in the tight self-associationof the glycophorin transmembrane helix.32 Leu257,in contrast, although completely solvent-inac-cessible, is relatively loosely packed. Nevertheless,its interactions are functionally significant, as theyare largely with other highly conserved residuesof the N domain. Thus, instead of providing a“glue” for association, as apparently in the case ofthe two glycine residues, the leucine residue mayprovide a loosely constructed joint that contributesto the conformational freedom of the domain inter-face. Finally, the alanine residues of the a4 motifare directed towards the underlying b-sheet, andhave relatively few interactions. These define oneedge of the functionally significant packing holediscussed above.

The a4 helix shifts over the underlying b-sheeton the transition between the apo and Mg2þGDPcomplexes of the NG domain.5 The movement istangential to the b-sheet and results in displace-ment of the helix by ,1.5 A along its axis.5 Theprecisely located interactions we identify and thepreferential directionality of those packing inter-actions provide a means to understand andpredict, at the structural level, the dynamics ofthis interface. As we do not yet have structures ofthe NG nucleotide complexes beyond ,2.0 A reso-lution, we cannot provide a definitive analysis ofthe shift between the two structural elements thataccompanies nucleotide binding. What is clear,however, is that there is no large conformationalchange or structural reorganization. There are noresidues in alternate conformations in the interiorof the interface; only one residue bordering thepacking hole, Leu244, undergoes a rotamer shiftbetween the apo and Mg2þGDP complex struc-tures. This shift, a relatively minor x2 side-chainreorientation, is relatively isosteric, and does notappear to change the packing requirements signifi-cantly. Leu244 occurs as valine, isoleucine orleucine in the SRP GTPases, and is directedtowards the hole so that it is packed relativelyloosely. Thus, it appears that the side-chains buriedin the interface are accommodated by relativelynon-specific restraints on their position along thepacking defect and the mobility of the a4 helixrelative to the G domain is maintained.

Interestingly, the packing density of only two ofthe helices of the G domain varies significantlyfrom the average (Table 3). Helix a4 is the mostloosely associated, with only 8% of the availablesurface contributing to packing interactions with

the G-domain (compared to an average of ,15%).In contrast, helix a1 of the G domain is exception-ally well-packed against the core protein structure(Figure 7(c)), with 23% of its surface in packinginteractions (2s above average). It can, therefore,be considered to be relatively rigid, consistentwith its central position providing the immobileplatform of the phosphate-binding P-loop in theGTPase active site. The remaining helices of the Gdomain (a1a, a1b, a2, and a3) exhibit lower pack-ing densities than helix a1, but are more tightlyassociated than helix a4. It is intriguing to proposethat, perhaps in contrast to the mobile glide surfaceprovided by helix a4,33 these helices switchbetween specific conformational states. It is stillnot known whether the helices of the IBD (a1a,a1b) undergo conformational change, but con-formational change of the a3 helix has beenobserved directly in the comparison of the apoand GDP-bound structures, and it can be inferredfor helix a2 by analogy to the behavior of the helixin other GTPases during GTP binding andhydrolysis.

Comparison with other NG domain structures

Unfortunately, two structures of the NG domainsof Ffh and FtsY from different species reveal anumber of conformational differences that limitthe generalization of our analysis. First, thestructure of the E. coli FtsY NG domain reveals anirregular conformation of the region correspondingto aN1 in Ffh such that the N terminus extendsbeyond the helical bundle of the N domain.3 Moresignificantly, the published structure of theA. ambivalens Ffh NG domain2 differs remarkablyfrom structure of the T. aquaticus Ffh NG domainin precisely the region of the N/G interface dis-cussed above. However, the constructs used forthe latter two structures differ in significant ways.The A. ambivalens construct is a C-terminalHis6-tagged protein in which the tag is insertedjust before a completely conserved leucineresidue30 that precedes a poorly ordered linkerpeptide to the C-terminal M-domain. In the struc-ture of the T. aquaticus Ffh NG (and in the structureof E. coli FtsY NG), two hydrophobic residues ofthe final turn of the C-terminal helix of the Gdomain pack against the open face of the N-termi-nal domain hydrophobic core.1 In the A. ambivalensFfh NG, only one hydrophobic residue, Leu293, isavailable, and this may have consequences forboth the stability of the N terminus of the protein(the first two residues of the aN1 helix are dis-ordered) and for the position of the N domain.Thus, relative to the G domain, the N domain isshifted by ,3 to 4 A from its position in the NGdomain of T. aquaticus Ffh. Interestingly, the confor-mation of the ALLEADV motif is similar in the twostructures (despite sequence differences2), and theloops and beta-strands of the G domain that pre-cede and follow helix a4 are in similar positions.Remarkably however, helix a4 itself is found to be


out of phase by one residue between the two struc-tures. Superimposition of the G domains of the twoproteins overlays (within 1 A on alpha carbonatoms) Arg252 with Lys254 (of A. ambivalens ) atthe N terminal end of the helix, and Thr263 withThr265 (of A. ambivalens ) at the C terminal end.Different loop conformations, however, displacethe ten intervening residues such that, althoughthe helix is clearly in a similar overall position(with corresponding alpha carbon atoms 0.7–2.3 Aapart), the sequence is shifted so that each residueat position N in T. aquaticus is at the position ofthe residue N þ 1 in the A. ambivalens structure.Thus the glycine residues discussed above nolonger pack against the ALLEADV motif in theA. ambivalens structure, and the leucine residue(Leu257 in T. aquaticus ) is not directed to thehydrophobic core of the N domain. Indeed, theserine residue is shifted so that it is directedinstead in that direction.

Given the observation that the availableA. ambivalens Ffh and E. coli FtsY structures differfrom the structure of the T. aquaticus Ffh, whatconfidence can we have in our analysis? First, theresidues of the interface are highly conserved inthe SRP GTPase family, and the structure of theT. aquaticus Ffh, but not that of the A. ambivalensFfh, allows us to rationalize this sequence conser-vation as reflecting a role in the functionallyimportant interface between the two domains. Theresidues have been demonstrated in biochemicalstudies to be important for function,6,7 and areseen in our structure to be involved in a largenumber of specific packing interactions at theinterface. The conformation observed in theA. ambivalens structure is less consistent withthe biochemical studies,6 and, we argue, makesless structural sense than does the conformationobserved in the structures of the T. aquaticus NGdomain. For example, in the A. ambivalens struc-ture, the glycine residue corresponding to Gly254is directed to a large space between the N terminusand motif IV that would easily accommodate alarger side-chain. Further structural studies will berequired to resolve this discrepancy.

Alternatively, that the other structures differ maysuggest a conformational complexity to the inter-face we have yet to understand. Perhaps thedifferences hint that the N domain structurechanges during the functional cycle, perhaps byunfolding so that it does not function as afour-helix bundle. Indeed, recent studies of theT. aquaticus Ffh and FtsY that suggest a specificconformational change of the N domain of FtsYoccurs during nucleotide-dependent formation ofthe targeting complex.34

Conclusion

GTPases undergo conformational changes onbinding and hydrolyzing GTP, and an understand-ing of their molecular mechanisms is important.

The structures of many nucleotide complexes ofsignal transduction and translation GTPases haverevealed a number of large conformational changesthat accompany binding. These are generally loca-lized to a set of structurally homologous “switch”regions that surround the active site and exhibitstructural properties that enable them to undergoconformational changes during binding andhydrolysis of GTP. The a2 helix of the canonicalGTPase fold, for example, can be considered tohave, at least, two different stable packing arrange-ments that function during the GTPase cycle.35

The SRP GTPases, while members of the GTPasesuperfamily, are distinct. As yet, the nature of theconformational changes that accompany GTP acti-vation and hydrolysis have not been determined.31

The N domain, unique to the SRP GTPases, isdistant from the switch regions that surround theactive site; however, it now appears that it plays arole in integrating the distinct signal transductionevents that occur during co-translational proteintargeting.6 One feature of its function is that it isin dynamic relation to the G domain. Here, wetake advantage of X-ray diffraction data to 1.1 Aresolution to demonstrate that under the crystalli-zation conditions the position of the N domainvaries in a way consistent with the motion inferredfrom comparison of the structures of the apo andGDP bound complexes. Further, we show by pack-ing analysis that the a4 helix of the G domain is akey player in this movement, as, on the one hand,it is coupled tightly to the N domain through inter-actions across a conserved interface and, on theother, it has few packing constraints against theunderlying b-sheet of the G domain. Thus the a4helix couples the N domain to the G domain, andfacilitates its motion. The high-resolution packinganalysis further explains cryptic patterns ofsequence conservation in the SRP GTPases, mostnotably that of hydrophobic residues at theproximal core of the N domain that pack againstthe conserved leucine side-chain of the a4 helix.

What our data do not explain, however, is whythe different structures of SRP GTPases from otherspecies have, so far, not revealed a consistentarrangement of the two domains. The possibilityis raised that the large deviations betweenmembers of the SRP GTPase family actually reflectexploration of a large conformational space acces-sible to the N domain sequence during SRPtargeting. That is, the structures hint that the foura-helix bundle of the N domain is not static, apossibility that may itself be consistent with loosepacking of the domain relative to other four-helixbundles. However, the well-defined structure ofthe domain interface in the ultrahigh-resolutionstructure of the T. aquaticus Ffh NG is consistentwith the results of mutational studies, and there-fore, if it defines only one endpoint of a range ofmobility, it defines a significant one.

The 1.1 A resolution structure of T. aquaticus FfhNG domain expands the basis for future studiesof the structural mechanism of the SRP GTPases.


The structure reported here directly visualizes afunctionally significant motion that occurs at theinterface of the N and G domains of the SRPGTPase, Ffh. It will be of interest to see, as thestructures of the nucleotide bound complexes aredetermined at ultrahigh resolution, whether wecan develop predictive algorithms that, based onpacking analysis and an understanding of theanisotropic motion in each structure, can explainthe intrinsic motions of this family of proteins.Mutations that affect the packing (and thus, pre-sumably, the relative orientation) of the twodomains produces striking defects, implyingthat the interface between the two domains isexquisitely designed for specific function. Becausewe identify a “neighborhood” of conservedresidues that contribute to packing relationshipson the tight side and on the hole side of the inter-face, our data help explain why the mutations ofhydrophobic residues have such a significanteffect. The intimate relationship between the Ndomain and the a4 helix of the G domain, and thepacking interactions that facilitate mobilitybetween the N and G domains, are therefore, likelyto be important in understanding the structuralmechanisms by which Ffh and SRP54 integratesignal sequence recognition, GTP binding andhydrolysis, and targeting of nascent polypeptidesto their membrane-associated receptor.

Materials and Methods

Crystallization and data collection

The NG domain of T. aquaticus Ffh was expressed inE. coli, and purified as described.1 Protein was concen-trated to 33 mg/ml in water, and was crystallized bysitting-drop, vapor-diffusion over 30% PEG MME 550,200 mM MgCl2, 50 mM N-tris[hydroxymethyl]methyl-3-aminopropanesulfonic acid (Taps) (pH 9.0) at roomtemperature. Large bipyramidal crystals grew overseveral months to ,800 mm maximum dimension. Thecrystal was mounted on the surface of a ,700 mm nylonloop, so that it was surrounded by relatively littleresidual mother liquor and frozen immediately in the2170 8C N2 cryostream.36

Data were measured at SSRL BL 9-1 on a MAR 300image plate detector using a wavelength of 0.98 A. Initialultrahigh-resolution data collection experiments carriedout on SSRL BL 7-1 established that these crystalsexhibited a decay in the ultrahigh-resolution diffractionsignal over a relatively short period of time. Our goalwas to obtain complete, well-measured data from asingle crystal; therefore, data were measured in threeresolution segments, and the high-resolution datameasured at four different translations of the crystal inthe beam. Low-resolution datasets were obtained at thestart and end of data collection to monitor overall decay.Low and medium-resolution data were collected withconstant exposure time, the high-resolution data withconstant dose. Data were integrated using DENZO andthe seven individual datasets were scaled together usingSCALEPACK37 with a 23s cutoff. The overall Rsym was3.7% over the resolution range 50–1.10 A, and was

32.4% in the high-resolution shell (1.12–1.10 A). Thedata quality statistics are summarized in Table 1.

Refinement

The crystallographic refinement was begun using themodel of the apo NG domain (PDB ID 1ffh) determinedat 2.0 A resolution.1 The unit cells of the crystals usedfor the 2.0 A and 1.10 A datasets were similar (spacegroup C2, cell parameters a ¼ 99:90 �A; b ¼ 53:91 �A; c ¼57:36 �A; b ¼ 119:88 and a ¼ 99:73 �A; b ¼ 53:67 �A; c ¼57:84 �A; b ¼ 119:98; respectively) so that it was straight-forward to begin several cycles of conventionalpositional and simulated-annealing refinement withX-PLOR 3.851.10 2Fo 2 Fc and Fo 2 Fc electron densitymaps were used for manual corrections of the refinedmodel using the graphics program O.38 The refinementusing isotropic temperature factors converged at an Rcryst

value of 23.3% (with an Rfree value of 27.2%) and at thisstage the model included three residues with alternateconformations, 120 water molecules and one hydratedmagnesium ion at a crystal lattice contact. The crystallo-graphic R at this stage was somewhat higher thanexpected, although not unprecedented.39,40 The positionof the disordered closing loop noted in the 2.0 A resolu-tion structure was not resolved in the higher-resolutionelectron density map.

Subsequent introduction of ADPs in SHELXL resultedin a relatively large decrease of over 4% in both Rcryst

(to 18%) and Rfree (to 21%). At this point, the ratio ofobservations to parameters was about 4.2. Test refine-ments were carried out to optimize the stereochemicalweighting factors by monitoring Rcryst and Rfree, and theanisotropic restraints were optimized by validating thedistribution of ADPs with PARVATI.18 Occupancies ofalternative conformations were refined using the FVARinstruction. After each manual rebuilding, the correctedparts of the model were refined with isotropic displace-ment parameters for five to ten cycles before the intro-duction of individual ADPs. Introduction of explicithydrogen atoms in the latter stages of refinement causeda further drop of ,1% in Rcryst. The programREFMAC14,41 was used during the final stages of therefinement in order to apply an empirical solventmodel, which allowed us to include all low-resolutiondata to 50 A resolution. The Rcryst value changed little atthis stage, but we observed a significant ,2% drop inRfree (from 18.7 to 16.9%). The resulting electron densitymaps were somewhat better defined than those pro-duced after SHELXL refinement (which included datafrom 10–1.1 A only). Statistics from the final REFMACrefinement are summarized in Table 2. Multiconformerrefinement42 was not carried out.

Analysis

The ADPs were validated using PARVATI,18 as notedabove. We found that as restraints towards isotropywere released, the average anisotropy moved towards0.5; further weakening of the restraints increased thenumber of highly anisotropic outliers but did not shiftthe overall average significantly. The atomic ADPs werethen analyzed in terms of a TLS model for domainmotion19,20,43 using the programs ANISOANL22 andTLSANL.23 The first TLS domain was defined to includethe four a-helices of the N domain and the a4 helix ofthe G domain; the second comprised residues 99–217 ofthe G domain. Following least-squares refinement of the


TLS parameters, the R values for the fit were 0.23 and0.24, and the goodness of fit parameters were 18.18 and17.72, respectively, indicating that the two TLS domainswere modeled equally well. The correlation coefficientover the equivalent B values for the N domain was 0.88,and for the G domain it was 0.84.

Atomic packing analysis including explicit hydrogenatoms was carried out using the programs PROBE andREDUCE.8 The PROBE algorithm simulates rolling a0.25 A radius sphere along the van der Waals surfaces;where the probe sphere contacts two surfaces, each ismarked with an indication (a dot) that classifies whetherthe surfaces are in wide contact (.0.25 A apart), closecontact (0.25–0.0 A), overlapped (interpenetrating to20.25 A), or clashing (.20.25 A). Bad contacts, orclashes, can be considered to indicate inaccuracies in themodel and are found to increase with decreasing crystal-lographic resolution.8 Therefore, to normalize for thelower effective resolution of the N domain, packing den-sity scores were calculated throughout using an averageof the score including bad overlap indicators and thescore without them. The explicit hydrogen atoms usedwere placed (and in some cases refined) by SHELXLand/or REFMAC; their positions were essentially thesame as those generated by REDUCE. The distributionof contact dots was contoured and displayed usingScoreDotsAtAtom and Kin3Dcont.44 For Table 3, a pack-ing density score was calculated for each residue byfirst determining the van der Waals surface area acces-sible to a 1.0 A radius probe in the context of the tripep-tide in which it occurred in the structure. Sincehydrogen is the smallest atom involved in packing con-tacts, the 1.0 A probe, corresponding to the van derWaals sphere for hydrogen used in PROBE,8 providedan estimate of the maximal possible contact surfacearea. The packing interactions of each residue were thenevaluated using a 0.25 A probe, and the fraction of thevan der Waals area involved in packing interactions wastabulated. The ratio of the two was taken as a packingdensity score. Averages were calculated for each residuetype in the structure; for glycine (23 residues) the aver-age was 9.6%, and for leucine (32 residues) the averagewas 23.3%, with a maximum of 41.2% and a minimumof 1.4%. For scoring the packing of individual helices,the van der Waals surface of the helix as a whole wasevaluated similarly in the context of the remainder ofthe structure. The contour level that illustrates the pack-ing hole between helix a4 and the b-sheet of the Gdomain in Figure 6 corresponds to an atomic packingdensity of approximately 20% of the maximum. Surfaceaccessibilities were calculated using AREAIMOL.45

The domain motion observed between the structuresof the apo and Mg2þGDP-bound NG domain (PDB ID1ng1) was analyzed using DYNDOM.46 The model-freemethod implemented in that program was found to berelatively insensitive to the input parameters, althoughthe limits of the domains so identified varied somewhat.The a4 helix of the G domain was always observed tobe grouped with the helices of the N domain, consistentwith the model presented here. Least-squares super-position of structures was carried out using LSQMAN.47

Figures were generated using MOLSCRIPT,48 Raster3Dand RASTEP,49 MAGE50 and O.38

Protein Data Bank accession codes

The atomic coordinates and structure factors have beendeposited with the RCSB PDB, with accession code 1LS1.

Acknowledgments

We thank J. Brunzelle for interesting and helpfuldiscussions, and we thank Garib Murshudov for discus-sions and assistance in implementing our refinementstrategy. This work was supported by grant GM-58500from the NIH. The data presented here were measuredwhile D.M.F. was a postdoctoral fellow in the labora-tories of P.W. and R.M.S. (UCSF). SSRL is supported bythe DOE Office of Basic Energy Sciences. The SSRL Bio-technology Program is supported by the NIH, NationalCenter for Research Resources, Biomedical TechnologyProgram, and by the DOE Office of Biological andEnvironmental Research.

References

1. Freymann, D. M., Keenan, R. J., Stroud, R. M. &Walter, P. (1997). Structure of the conserved GTPasedomain of the signal recognition particle. Nature,385, 361–364.

2. Montoya, G., Kaat, K. t., Moll, R., Schafer, G. &Sinning, I. (2000). The crystal structure of the con-served GTPase of SRP54 from the archeon Acidianusambivalens and its comparison with related structuressuggests a model for the SRP–SRP receptor complex.Structure, 8, 515–525.

3. Montoya, G., Svensson, C., Luirink, J. & Sinning, I.(1997). Crystal structure of the NG domain from thesignal-recognition particle receptor FtsY. Nature, 385,365–369.

4. Bourne, H. R., Sanders, D. A. & McCormick, F.(1991). The GTPase superfamily: conserved structureand molecular mechanism. Nature, 349, 117–127.

5. Freymann, D. M., Keenan, R. J., Stroud, R. M. &Walter, P. (1999). Functional changes in the structureof the SRP GTPase on binding GDP and Mg2þGDP.Nature Struct. Biol. 6, 793–801.

6. Lu, Y., Qi, H.-Y., Hyndman, J. B., Ulbrandt, N. D.,Teplyakov, A., Tomasevic, N. & Bernstein, H. D.(2001). Evidence for a novel GTPase priming step inthe SRP protein targeting pathway. EMBO J. 20,6724–6734.

7. Newitt, J. A. & Bernstein, H. D. (1997). TheN-domain of the signal recognition particle 54 kDasubunit promotes efficient signal sequence binding.Eur. J. Biochem. 245, 720–729.

8. Word, J. M., Lovell, S. C., LaBean, T. H., Taylor, H. C.,Zalis, M. E., Presley, B. K. et al. (1999). Visualizingand quantifying molecular goodness-of-fit: small-probe contact dots with explicit hydrogen atoms.J. Mol. Biol. 285, 1711–1733.

9. Levitt, M., Gerstein, M., Huang, E., Subbiah, S. &Tsai, J. (1997). Protein folding: the endgame. Annu.Rev. Biochem. 66, 549–579.

10. Brunger, A. T. (1992). X-PLOR: A System for X-RayCrystallography and NMR, Yale University Press,New Haven, CT.

11. Sheldrick, G. & Schneider, T. (1997). SHELXL: higherresolution refinement. Methods Enzymol. 277,319–343.

12. Walsh, M. A., Schneider, T. R., Sieker, L. C., Dauter,Z., Lamzin, V. S. & Wilson, K. S. (1998). Refinementof triclinic hen egg-white lysozyme at atomic reso-lution. Acta Crystallog. sect. D, 54, 522–546.

13. Schneider, T. R. (1996). What can we learn fromanisotropic temperature factors? Proceedings of theCCP4 Study Weekend (Dodson, E., Moore, M., Ralph,


A. & Bailey, S., eds), pp. 133–144, SERC DaresburyLaboratory, Daresbury.

14. Murshudov, G. N., Vagin, A. A. & Dodson, E. J.(1997). Refinement of macromolecular structures bythe maximum-likelihood method. Acta Crystallog.sect. D, 53, 240–255.

15. Jiang, J.-S. & Brunger, A. T. (1994). Protein hydrationobserved by X-ray diffraction. J. Mol. Biol. 243,100–115.

16. Laskowski, R. A., MacArthur, M. W., Moss, D. S. &Thornton, J. M. (1993). PROCHECK: a program tocheck the stereochemical quality of protein struc-tures. J. Appl. Crystallog. 26, 283–291.

17. Dauter, Z., Lamzin, V. S. & Wilson, K. S. (1995).Proteins at atomic resolution. Curr. Opin. Struct. Biol.5, 784–790.

18. Merritt, E. A. (1999). Expanding the model: aniso-tropic displacement parameters in protein structurerefinement. Acta Crystallog. sect. D, 55, 1109–1117.

19. Schomaker, V. & Trueblood, K. N. (1968). On therigid-body motion of molecules in crystals. ActaCrystallog. sect. B, 24, 63–76.

20. Dunitz, J. D., Schomaker, V. & Trueblood, K. N.(1988). Interpretation of atomic displacement para-meters from diffraction studies of crystals. J. Phys.Chem. 92, 856–867.

21. Rader, S. D. & Agard, D. A. (1997). Conformationalsubstates in enzyme mechanism: the 120 K structureof alpha-lytic protease at 1.5 A resolution. ProteinSci. 6, 1375–1386.

22. Winn, M. D., Isupov, M. N. & Murshudov, G. N.(2001). Use of TLS parameters to model anisotropicdisplacements in macromolecular refinement. ActaCrystallog. sect. D, 57, 122–133.

23. Howlin, B., Butler, S. A., Moss, D. S., Harris, G. W. &Driessen, H. P. C. (1993). TLSANL: TLS parameteranalysis program for segmented anisotropic refine-ment of macromolecular structures. J. Appl. Crystal-log. 26, 622–624.

24. Hayward, S. & Berendsen, H. J. (1998). Systematicanalysis of domain motions in proteins from confor-mational change: new results on citrate synthaseand T4 lysozyme. Proteins: Struct. Funct. Genet. 30,144–154.

25. Lutcke, H., High, S., Romisch, K., Ashford, A. J. &Dobberstein, B. (1992). The methionine-rich domainof the 54 kDa subunit of signal recognition particleis sufficient for the interaction with signal sequences.EMBO J. 11, 1543–1551.

26. Zopf, D., Bernstein, H. D., Johnson, A. E. & Walter, P.(1990). The methionine-rich domain of the 54 kDaprotein subunit of the signal recognition particlecontains an RNA binding site and can be crosslinkedto a signal sequence. EMBO J. 9, 4511–4517.

27. Word, J. M., Lovell, S. C., Richardson, J. S. &Richardson, D. C. (1999). Asparagine and gluta-mine: using hydrogen atom contacts in the choice ofside-chain amide orientation. J. Mol. Biol. 285,1735–1747.

28. Word, J. M., Bateman, R. C., Jr, Presley, B. K., Lovell,S. C. & Richardson, D. C. (2000). Exploring stericconstraints on protein mutations using MAGE/PROBE. Protein Sci. 9, 2251–2259.

29. Lovell, S. C., Word, J. M., Richardson, J. S. & Richard-son, D. C. (1999). Asparagine and glutaminerotamers: B-factor cutoff and correction of amideflips yield distinct clustering. Proc. Natl Acad. Sci.USA, 96, 400–405.

30. Gorodkin, J., Knudsen, B., Zwieb, C. & Samuelsson,T. (2001). SRPDB (signal recognition particle data-base). Nucl. Acids Res. 29, 169–170.

31. Padmanabhan, S. & Freymann, D. M. (2001). Theconformation of bound GMPPNP suggests a mecha-nism for gating the active site of the SRP GTPase.Structure, 9, 859–867.

32. MacKenzie, K. R., Prestegard, J. H. & Engelman,D. M. (1997). A transmembrane helix dimer: struc-ture and implications. Science, 276, 131–133.

33. Gerstein, M., Lesk, A. M. & Chothia, C. (1994).Structural mechanisms for domain movements inproteins. Biochemistry, 33, 6739–6749.

34. Shepotinovskaya, I. V. & Freymann, D. M. (2002).Conformational change of the N-domain on for-mation of the complex between the GTPase domainsof Thermus aquaticus Ffh and FtsY. Biochim. Biophys.Acta, 1597, 92–99.

35. Sprang, S. R. (1997). G protein mechanisms: insightsfrom structural analysis. Annu. Rev. Biochem. 66,639–678.

36. Bellamy, H. D., Phizackerly, R. P., Soltis, S. M. &Hope, H. (1994). An open-flow cryogenic coolerfor single-crystal diffraction experiments. J. Appl.Crystallog. 27, 967–970.

37. Otwinowski, Z. (1993). Oscillation data reductionprogram. In Data Collection and Processing (Sawyer,L., Isaacs, N. W. & Bailey, S., eds), pp. 55–62, SERCDaresbury Laboratory, Warrington.

38. Jones, T. A., Zou, J. Y., Cowan, S. W. & Kjeldgaard,M. (1991). Improved methods for building proteinmodels in electron density maps and the location oferrors in these models. Acta Crystallog. sect. A, 47,110–119.

39. Deacon, A., Gleichmann, T., Kalb, A. J., Price, H.,Raftery, J., Bradbrook, G. et al. (1997). The structureof concanavalin A and its bound solvent determinedwith small-molecule accuracy at 0.94 A resolution.J. Chem. Soc., Faraday Trans. 93, 4305–4312.

40. Tong, L., Warren, T. C., King, J., Betageri, R., Rose, J.& Jakes, S. (1996). Crystal structures of the humanp56lck SH2 domain in complex with two shortphosphotyrosyl peptides at 1.0 A and 1.8 A reso-lution. J. Mol. Biol. 256, 601–610.

41. Murshudov, G. N., Vagin, A. A., Lebedev, A., Wilson,K. S. & Dodson, E. J. (1999). Efficient anisotropicrefinement of macromolecular structures using FFT.Acta Crystallog. sect. D, 55, 247–255.

42. Wilson, M. A. & Brunger, A. T. (2000). The 1.0 Acrystal structure of Ca(2 þ )-bound calmodulin: ananalysis of disorder and implications for functionallyrelevant plasticity. J. Mol. Biol. 301, 1237–1256.

43. Schomaker, V. & Trueblood, K. N. (1998). Correlationof internal torsional motion with overall molecularmotion in crystals. Acta Crystallog. sect. B, 54,507–514.

44. Word, J. (2000). All-atom small-probe contact surfaceanalysis: an information-rich description of molecu-lar goodness-of-fit. PhD Thesis, Duke University

45. Collaborative Computational Project, No. 4 (1994).The CCP4 suite: programs for protein crystal-lography. Acta Crystallog. sect. D, 50, 760–763.

46. Hayward, S., Kitao, A. & Berendsen, H. J. (1997).Model-free methods of analyzing domain motionsin proteins from simulation: a comparison of normalmode analysis and molecular dynamics simulationof lysozyme. Proteins: Struct. Funct. Genet. 27,425–437.


47. Kleywegt, G. J. & Jones, T. A. (1997). Detecting fold-ing motifs and similarities in protein structures.Methods Enzymol. 277, 525–545.

48. Kraulis, P. J. (1991). MOLSCRIPT—a program to pro-duce both detailed and schematic plots of proteinstructures. J. Appl. Crystallog. 24, 946–950.

49. Merritt, E. A. & Bacon, D. J. (1997). Raster3D photo-realistic molecular graphics. Methods Enzymol. 277,505–524.

50. Richardson, D. C. & Richardson, J. S. (1992). Thekinemage: a tool for scientific communication.Protein Sci. 1, 3–9.

Edited by D. Rees

(Received 22 January 2002; received in revised form 8 May 2002; accepted 14 May 2002)


Documents

Structural Basis for Mobility in the 1.1Å Crystal Structure of the NG Domain of Thermus aquaticus Ffh