solution conformation and heparin-induced dimerization of the full-length extracellular domain of...

doi:10.1016/j.jmb.2005.12.053 J. Mol. Biol. (2006) 357, 493–508

Solution Conformation and Heparin-inducedDimerization of the Full-length ExtracellularDomain of the Human Amyloid Precursor Protein

Matthias Gralle1†, Cristiano L. P. Oliveira2,3†, Luiz H. Guerreiro1

William J. McKinstry4, Denise Galatis5,6, Colin L. Masters5,6,Roberto Cappai5,6, Michael W. Parker4, Carlos H. I. Ramos3,Iris Torriani2,3 and Sergio T. Ferreira1,3*

1Instituto de Bioquımica MedicaPrograma de Bioquımica eBiofısica Celular, UniversidadeFederal do Rio de JaneiroRio de Janeiro, RJ 21944-590Brazil

2Instituto de Fısica “GlebWataghin,” UnicampCampinas, SP 13084-971Brazil

3Laboratorio Nacional de LuzSıncrotron, Campinas, SP13084-9701, Brazil

4St. Vincent’s Institute ofMedical Research, 9 Princes St.Fitzroy, Vic. 3065, Australia

5Department of PathologyCentre for NeuroscienceUniversity of MelbourneParkville, Vic. 3010, Australia

6The Mental Health ResearchInstitute of Victoria, ParkvilleVic. 3010, Australia

0022-2836/$ - see front matter q 2005 E

† M.G. and C.L.P.O. contributed eAbbreviations used: AD, Alzheim

amyloid precursor protein; sAPPa,domain of the amyloid precursor psecretase cleavage; SAXS, small-ang

E-mail address of the [email protected]

Proteolytic cleavage of the amyloid precursor protein (APP) by b andg-secretases gives rise to the b-amyloid peptide, considered to be a causalfactor in Alzheimer’s disease. Conversely, the soluble extracellular domainof APP (sAPPa), released upon its cleavage by a-secretase, plays a numberof important physiological functions. Several APP fragments have beenstructurally characterized at atomic resolution, but the structures of intactAPP and of full-length sAPPa have not been determined. Here, ab initioreconstruction of molecular models from high-resolution solution X-rayscattering (SAXS) data for the two main isoforms of sAPPa (sAPPa695 andsAPPa770) provided models of sufficiently high resolution to identifydistinct structural domains of APP. The fragments for which structuresare known at atomic resolution were fitted within the solution models offull-length sAPPa, allowing localization of important functional sites (i.e.glycosylation, protease inhibitory and heparin-binding sites). Furthermore,combined results from SAXS, analytical ultracentrifugation (AUC) andsize-exclusion chromatography (SEC) analysis indicate that both sAPPaisoforms are monomeric in solution. On the other hand, SEC, bis-ANSfluorescence, AUC and SAXS measurements showed that sAPPa forms a2:1 complex with heparin. A conformational model for the sAPPa:heparincomplex was also derived from the SAXS data. Possible implications ofsuch complex formation for the physiological dimerization of APP andbiological signaling are discussed in terms of the structural modelsproposed.

q 2005 Elsevier Ltd. All rights reserved.

Keywords: analytical ultracentrifugation; modeling; signal transduction;small-angle X-ray scattering; structural domains
*Corresponding author
Introduction

The amyloid precursor protein (APP) is linkedin two ways to Alzheimer’s disease (AD), themost widespread form of dementia and one of

lsevier Ltd. All rights reserve

qually to this work.er’s disease; APP,

soluble extracellularrotein released by a-le X-ray scattering.ing author:

the leading causes of death among the elderly. Thefirst link came from the discovery that theb-amyloid peptide (Ab), the main component ofthe extracellular plaques that characteristicallyaccumulate in the brains of AD patients,1 originatesfrom proteolytic cleavage of APP.2 Second, thefinding that mutations in APP cause hereditary,presenile AD indicated a causal connection of APPwith AD.3

APP is present in the plasma membranes of allhuman cell types investigated so far, includingneurons, and has a single membrane-spanningdomain. APP can be cleaved in vivo by two different

d.

494 Conformation and Dimerization of Human sAPPa

proteolytic pathways. a-Secretase cleaves the extra-cellular domain of APP at a distance of 12 aminoacid residues from the membrane, in the middle ofthe amino acid sequence that corresponds to Ab.4

This results in the release of a large extracellulardomain, known as sAPPa, encompassing up to 89%of the whole APP molecule. ADAM-10 has beenidentified as a protease necessary for a-secretaseactivity in vivo.5 Alternatively, and in much smalleramounts, APP can be cleaved by b-secretase (alsoknown as BACE1) at different residues upstream ofthe a-secretase cleavage site (for a review, seeVassar6). The membrane-inserted stub of APP leftafter b-secretase cleavage may be further processedby regulated intramembrane proteolysis by theg-secretase complex to release Ab and an intra-cellular fragment.7 While both Ab and the intra-cellular fragment released by g-secretase may exertsignaling functions in the healthy brain,8,9 they mayalso contribute to the pathogenesis of AD.10 As aresult, increasing a-secretase cleavage of APP to thedetriment of b-secretase cleavage has been pro-posed as a possible treatment for AD.5,11

Alongside the interest in APP generated by itsinvolvement in AD, several important physiolo-gical cellular functions of this protein have beendiscovered. Transmembrane APP may play roles inneurite outgrowth,12 neuronal adhesion,13 axonaltransport14 and copper metabolism.15 sAPPa, themain fragment of APP, may be neuroprotective,16

modulate synaptic function,17 and regulate proli-feration of neuronal progenitors18 and other celltypes.19 The primary structure of APP has beenknown for 17 years. However, in spite of the greatinterest in APP kindled by its involvement in ADand in many important cellular processes, itscomplete 3D structure and molecular mechanismsof action are still largely unknown. Very recently,there has been substantial progress in the structuralbiology of isolated APP domains,20–25 but atomic-resolution structures of full-length sAPP andAPP have not been obtained. This prompted usto pursue the characterization of sAPPa by

domain. Cyt: cytoplasmic domain. Cleavage by a-secretase (C-terminal fragment (CTFa), as indicated. Amino acid numb

spectroscopic and solution scattering tech-niques.26,27 Here, using a combination of high-resolution synchrotron small angle X-ray scattering(SAXS) and molecular modeling tools we proposestructural models for full-length sAPPa695 andsAPPa770, the main neuronal and non-neuronalisoforms of APP, respectively. These two isoformsdiffer by the presence or absence of a Kunitz-typeprotease inhibitor (KPI) domain, which is unique tothe longer isoform and necessary for some of thebiological functions of APP.28 Moreover, the state ofassociation of sAPPa in solution and its dimeri-zation induced by heparin were investigated.Possible implications of the interaction withheparin (or heparan sulfate) and dimerization ofAPP are discussed in terms of the regulation of itsbiological functions.

Results

Overall analysis of SAXS data for sAPPa

We have investigated the soluble extracellulardomains of the two main isoforms of APP, sAPPa695

and sAPPa770. APP695 is predominant in neurons,while APP770 is expressed in most non-neuronal celltypes and contains an insert of 75 amino acidresidues, corresponding to a KPI domain, in themiddle of its sequence (Figure 1). High-resolutionsmall-angle X-ray scattering data were collected forsAPPa695 and sAPPa770 (Figure 2(a) and (b),respectively) and were used to calculate thepairwise distance distribution functions by inverseFourier transform (Figure 2(c) and (d), respec-tively). These functions allow the determination ofthe maximum dimensions of the proteins, Dmax,and their radii of gyration, Rg. The maximumlengths obtained from the P(r) functions werew135 A and w150 A, respectively, for sAPPa695

and sAPPa770, while radii of gyration (Rg) of42.6(G0.3) A and 46.5(G2) A for sAPPa695

and sAPPa770, respectively, have been previously

Figure 1. Domain structure ofAPP. Grey ellipses: APP domainsconserved in evolution35 for whichcomplete or partial structures areknown (see the text). 1mwp, 1owt,1aap, 1rw6, 1tkn correspond to thePDB access codes for the fragmentsof known structure. (DE)n, RC:intervening non-conservedstretches (Asp and Glu-rich; ran-dom coil) suggested to be largelyunstructured or to have non-stan-dard secondary structures.26

HBD1, HBD2: heparin-bindingdomains. CuBD: copper-bindingdomain. Ab: amyloidogenicsequence. TM: transmembrane

vertical arrow) releases sAPPa and the membrane-boundering is for APP770.

Figure 2. Upper panels. SAXSintensity data for sAPPa695 (a) andsAPPa770 (b). Data measured atconcentrations up to 12.5 mg/ml(sAPPa695) or 9.6 mg/ml (sAPPa770)(open circles) were used to constructthe curves over the complete q rangeup to 0.8 AK1.27 These data areshown in order to allow evaluationof the quality of the fits derived fromab initio SAXS models (continuouslines) and from fitting the domains ofknown structure of APP into the abinitio SAXS models of sAPPa (brokenlines, shown up to qZ0.4 AK1). Fordescription of these models, seelegends to Figures 3 and 5. Insets:Guinier plots of the data showing themonodispersity of the samples.Lower panels, open circles: pairwisedistance distribution functions, p(r),calculated from the full range ofexperimental data for sAPPa695 (c)and sAPPa770 (d) using GNOM.57

Continuous lines: fits from ab initioSAXS models (see Figure 3).

Conformation and Dimerization of Human sAPPa 495

calculated from the same data set.27 Analysis of thedata using Guinier plots (insets to Figure 2(a) and(b)) yields Rg estimates of 46(G3) A and 47(G3) A,for sAPPa695 and sAPPa770, respectively, in goodagreement with the values obtained from P(r)distribution analysis. The similarity of molecularweights obtained from SAXS data with the valuesexpected for both isoforms27 rules out the presenceof aggregates or other oligomers in the samples. Theshapes of the P(r) functions and the ratios betweenDmax and Rg for both APP isoforms clearly indicateelongated (prolate) shapes for the proteins. Con-sidering the high extent of sequence homologybetween them (the only difference being thepresence of the KPI domain in the longer APPisoform), one would indeed expect to encountersignificant similarity in their overall three-dimen-sional structures. Interestingly, however, theshoulder observed at rw100 A in the P(r) functionfor sAPPa695 indicates that the overall shape of thatisoform is somewhat different from that ofsAPPa770.

Ab initio modeling

The choice of the right strategy for ab initiomodeling using SAXS data depends on the qualityand information content of the scattering curve.29

For momentum transfer values larger than 0.5 AK1,in addition to the determination of the overall shapeof the protein it is generally possible to obtaininformation on its fold and structural domains. Thepresent data set has an excellent signal-to-noiseratio and reaches qw0.8 AK1.27 This allowed usto analyze the experimental results employing

a procedure capable of retrieving relatively detailedstructural information on the protein. To this end,we have represented the protein as a chain ofdummy residues and carried out simulated annea-ling optimization to search for the configurationof the backbone that minimizes the discrepancybetween the scattering intensity or the P(r) functioncalculated from the ab initio model and theexperimental SAXS data.30 We have analyzed ourmodels in both real space and reciprocal space, andthe best results were obtained using real space P(r)fitting. The models generated in this way seemed torender more reliable values of radii of gyration ofthe proteins with a better fit of the low q region. Aspreviously reported, real space fitting relies less onthe high-angle scattering data than reciprocal spacefitting.31 The P(r) functions for both sAPPa isoformswere very well described by each of 20 indepen-dently calculated models that yielded the best fits tothe data. For evaluation of the quality of the fits,Figure 2 also shows the mean P(r) functions andstandard deviations (calculated using the 20 inde-pendent P(r) curves that gave best fits to the data)along with the experimental data for both sAPPaisoforms. It is apparent that the fits yielded verygood descriptions of the P(r) functions for both APPforms.

Due to the use of a high number of dummyresidues in the model calculation, as well as to thepresence of flexible domains in the structure ofAPP,27 the results of the simulation lead to therecovery of slightly different independent structuralmodels even using our high-resolution SAXS data.Consequently, in order to retrieve the most probablemodel for each isoform, an averaging procedure

Figure 3. Solution conformational models for sAPPa695 (a) and sAPPa770 (b) calculated from the experimental datausing a chain model of dummy residues (see Materials and Methods). Twenty independent models were generated foreach isoform (shown as a light grey envelope) and filtered to yield the most probable average model (dark grey spheres).The models are rotated as indicated in the Figures.


was carried out.32 In this procedure, the differentdummy residue models encountered as indepen-dent solutions to the fits are compared with eachother, and those having higher similarity areaveraged. The final 3D configuration is representedas a space-filling model with a close packing ofspheres (Figure 3, filled spheres), along with thesuperposition of all 20 independent solutions(semitransparent spheres) for each isoform.

A useful check for the validity of ab initio modelscalculated from SAXS data is to compare thevolume and molecular weight ratios obtained fortwo proteins. Hydrodynamic volumes (V) werecalculated from the models shown in Figure 3 usingHYDROPRO,33 while molecular weights (MW)were determined directly from the zero-angleintensities in the SAXS measurements. If the modelsare compatible with the experimental data, theratios between V(sAPPa695)/V(sAPPa770) andMW(sAPPa695)/MW(sAPPa770) should yield simi-lar values. For the present analysis, the valuesobtained were V(sAPPa695)/V(sAPPa770)Z0.89G0.05 and MW(sAPPa695)/MW(sAPPa770)Z0.90.The excellent agreement between those valuesindicates that the models correctly retrieve theexpected volume ratios for the two sAPPa isoforms.

Table 1. Hydrodynamic data on HC, sAPPa and sAPPa–hep

SampleBuoyant weight

(kDa)Molecular weight

(kDa)

HC n.d. n.d.sAPPa695 19 69sAPPa770 n.d. n.d.sAPPa695–heparin 44 155sAPPa770–heparin n.d. n.d.

The values of buoyant weight correspond to the best fits to equilibriumconcentrations ranging from 2.3 mM to 18.5 mM and heparin:sAPPa695

from buoyant weights using the correct partial specific volume, as dcoefficients, s (in Svedberg units), correspond to the best fits to sedimpresence of 3.5 mM heparin or for HC at 140 mM, and for serial dicoefficients calculated from the ab initio SAXS models. Maximum smolecular weights. n.d., not determined.

Calculation of sedimentation coefficients from theSAXS-derived ab initio models using HYDROPRO33

gave values of 3.77 S for sAPPa695 and 3.93 S forsAPPa770. For both sAPPa isoforms, these valuesare considerably lower than the maximum sedi-mentation coefficients calculated for sphericalproteins of equivalent molecular weights (Table 1),corroborating the notion that they exhibit highlyelongated and irregular shapes. In fact, the sedi-mentation coefficients measured by sedimentationvelocity analytical ultracentrifugation for bothisoforms were even lower (2.9 S; see below andTable 1) than the values calculated from the SAXSmodels. Consistent with our previous obser-vations,26,27 this indicates that both sAPPa isoformsexhibit considerable conformational flexibility.Thus, the filtered structural models shown inFigure 3 (dark spheres) should be considered timeand ensemble averages over the actual distributionsof possible conformations of sAPPa695 andsAPPa770 in solution.

It is important to note that the model obtainedhere for sAPPa695 is also consistent with ourpreviously reported low-resolution model for thatisoform.26 However, because of the significantlyhigher accuracy of the present SAXS data, the

arin complexes

Experimental s (S) SAXS model s (S) Maximum s (S)

1.48 1.85 2.242.9 3.77 6.302.9 3.93 6.644.8 6.21 n.d.5.4 n.d. n.d.

sedimentation data and were found to be invariant for sAPPa695

ratios ranging from 1:1 to 8:1. Molecular weights were calculatedescribed in Materials and Methods. Experimental sedimentationentation velocity data for sAPPa (7 mM) in the absence or in the

lutions of these samples. SAXS model-derived s values are thevalues are those calculated for spherical proteins of equivalent


current model is of correspondingly higher reso-lution, and it was now possible to define at least twostructural domains of sAPPa695 (Figure 3(a)), one ofwhich is quite elongated and likely corresponds tothe shoulder in the P(r) function of this isoform.Interestingly, the longer isoform, sAPPa770, containsan additional structural domain that emergeslaterally from one of the ends of the molecule(Figure 3(b)). That domain likely corresponds to theKPI domain, which is present only in the longerisoforms of APP,20 in addition to other linkerregions of the protein. The models thus indicatethat the KPI domain is freely exposed on the surfaceof APP, facilitating its interaction with targetproteases. The surface-exposed location of the KPIdomain is also in agreement with the fact that itsreplacement by a yellow fluorescent protein insertgives rise to a functional APP chimera.34

The two APP isoforms investigated here havelarge regions with identical amino acid sequences,and this high extent of homology should bereflected in similar 3D conformations. Inspectionof the primary structure of APP reveals that the KPIdomain is preceded by a sequence of 260 amino acidresidues that is identical in both isoforms. More-over, the KPI domain is followed by a 322 aminoacid residues long sequence that is also identical inboth isoforms. The latter region corresponds tomore than half of the entire sAPPa695 protein. Asexpected, the ab initio models reflect this similarityof primary structures of the two isoforms. Below,we use this fact to our advantage in identifying thelocation of APP domains of known crystallographicstructures in the 3D SAXS models.

Topology and positioning of APP domainsin the SAXS models

A number of isolated APP fragments or domainshad their structures solved in recent years.20–25

Until now, however, it has not been possible toobtain suitable crystals and solve the structure offull-length sAPPa. The evolutionarily well-con-served domains of APP,35 all of which have nowbeen structurally characterized in isolation, areseparated by less well-conserved regions (Figure 1)that appear to be quite flexible or to exhibit non-standard secondary structures in the native state.26

Those regions are easily degraded spontaneously22

or by proteases.24,25 The significant obstaclestowards obtaining crystallographic structures ofsAPP and APP are complemented by the currentlimitations of NMR methods for elucidation of thethree-dimensional structure of a protein as large assAPPa (R70 kDa). In an attempt to overcome thesedifficulties, we have performed SAXS measure-ments on a fragment of sAPPa and on the intactprotein and used the available published data toestablish topological correlations between the high-resolution structures of isolated sAPP fragmentsand our ab initio SAXS models.

The structures of the two N-terminal domains ofAPP (HBD1 and CuBD in Figure 1) have been

determined separately, while determination of thestructure of a larger N-terminal portion of APPcomprising both domains has not been possible.22,23

In order to gain insight into the interaction betweenthese two domains, we performed SAXS measure-ments on a construct of HBD1CCuBD, heredenoted HC. SAXS measurements on isolatedHBD1 and CuBD fragments yielded Rg and Dmax

values compatible with their published structures(RgZ15 A, DmaxZ55 A for HBD1; RgZ14 A, DmaxZ42 A for CuBD). Analysis of the experimental datafor HC (Figure 4(a)) yielded a molecular weight of19.1 kDa, in excellent agreement with the expectedvalue of 18.6 kDa for this fragment. On the otherhand, the values of Rg (21.3(G0.4) A) and Dmax

(68 A) obtained from a GNOM fit to these data wereconsiderably higher than one would expect if theHBD1 and CuBD domains interacted tightly to forma globular structure. In fact, the shape of thedistance distribution function (Figure 4(b)) indi-cated a prolate conformation of HC. Ab initiomodeling of the SAXS data yielded a dumb-bellshaped model of HC (Figure 4(c)). This wasconsistent with results from rigid body modelingfor the two atomic-level structures of HBD1 andCuBD connected by a flexible linker (Figure 4(c)).The program BUNCH,62 used for this purpose,optimizes the best domain spatial orientation andlinker conformation. Importantly, both ab initio andrigid body models gave excellent fits to theexperimental data on HC (Figure 4(a)).

The sedimentation coefficients calculated fromthe ab initio and rigid body models of HC were1.85 S and 1.83 S, respectively. Once again, as notedabove for the intact sAPPa isoforms, those coeffi-cients are lower than that calculated for a sphericalprotein with the equivalent mass of HC, but theyare higher than the experimentally determinedcoefficient obtained by sedimentation velocityultracentrifugation (Table 1). This suggests a highdegree of flexibility of the linker between the twodomains. For example, a model of HBD1 and CuBDconnected by a very long, friction-less linker, withcomplete hydrodynamic drain between the twodomains, would give a coefficient of 1.04 S.37 Thissuggests that the true ensemble-averaged confor-mation of HC lies between the rigid model shown inFigure 4(c) and the long linker model. Indeed,independent runs of rigid-body modeling of HCshowed slightly different orientations of CuBDrelative to HBD1, consistent with the existence ofa flexible linker connecting the two domains.

The structure of the C-terminal domain of sAPP(1RW6.PDB) shows that it comprises a highlyelongated helical-bundle with a maximal molecularlength of 100 A.25 This indicates that this domaincan only fit into the elongated stalk regionextending vertically in the sAPPa models shownin Figure 3. Moreover, we assumed that the Cterminus of this domain would face away from thetwo other N-terminal domains of APP of knownstructures. A smaller fragment of APP whosestructure was recently determined by NMR24 was

Figure 4. SAXS data on the N-terminal HC fragment of sAPP. (a) Experimental data and error bars. Continuous line:calculated from chain model fit (see (c)). Broken line: calculated from rigid body model fit (see (c)). Inset: Guinier plot ofSAXS data in the low-angle region of the curve. (b) Pair-distance distribution function. Continuous line and error barswere calculated from the GNOM fit to the data. (c) Grey spheres: low-resolution model for HC calculated from theexperimental data using a chain model of dummy residues.30 Full-color model: best rigid body model for HC, fitting theatomic-resolution structures of HBD1 and CuBD, connected by a flexible linker, to the experimental SAXS data.62


not directly employed in our modeling, since it iscontained within the larger crystallized C-terminalfragment25 and its three-dimensional structure issuperimposable onto the corresponding parts of thecrystallized fragment. The major problem for thecorrect fitting of HC and the other two knowndomains of sAPPa into the SAXS models of the full-length protein resides in the existence of severallarge gaps between the domains of known struc-ture. An initial attempt to model those gaps usingmolecular homology procedures was not fruitful asa consequence of the low sequence homology ofAPP (or even small parts of it) with proteins ofknown crystal structures. In the absence of goodmodels for those gaps, the use of a fully automatedprocedure for domain positioning guided by theSAXS data (following the steps described formodeling of the HC domain) was not feasible.Therefore, positioning of the domains was per-formed as described below, based on a combinationof: (a) biochemical data; (b) a comparison betweenthe dimensional and conformational data availablefor the crystallographic structures of isolateddomains; (c) SAXS models of HC and of the

full-length sAPPa isoforms; and (d) rigid bodydomain rotations and translations guided bySAXS data.

The structures of HC (Figure 4(c)) and of theC-terminal domain25 were initially positioned intothe sAPPa695 model (Figure 5(a)). The best orien-tation of these domains (the one that gave thelowest discrepancy between the theoretical fit andexperimental SAXS data) was obtained by rigidbody modeling using MASSHA.38 Next, a similarprocedure of rigid body optimization was appliedto position the KPI domain into the salientprotuberance of the sAPPa770 model fixing thepositions of the other two domains as in the casesAPPa695 (Figure 5(b)). The scattering intensitiescalculated for these models are shown in Figure 2(a)and (b) (broken lines) for comparison with theexperimental data. The fits are found to be quitegood in the q range conveying information on shapeand quaternary structure, i.e. up to 0.20 AK1. Asmentioned above, the existence of structural gaps inthe models leads to some deviation in the intensityfits, notably in the high q ranges (qO0.20 AK1). Inspite of these limitations, it is noted that the initial

Figure 5. Structural models of sAPPa695 (a) and sAPPa770 (b). Grey envelope: most probable SAXS model calculated asdescribed in the legend to Figure 3 (see also Results). Space-filling structures: the structures of the HC domain (yellow;see legend to Figure 4) and of fragments 1RW6 (blue, in (a) and (b)) and 1AAP (magenta, in (b)) were positioned into theSAXS models (see the text). The putative heparin-binding site in HC is indicated in orange, and the heparin binding sitein 1RW6 is shown in red. The glycosylation site in 1RW6 is indicated in green. Amino acid residues that participate in theproposed dimer interface25 are shown in cyan. The two amino acid residues that form the protease inhibitory site areshown in dark green within the magenta 1AAP domain. Model rotations are indicated in the Figures.


part of the scattering curve (low q values) was verywell approximated by the models derived for bothsAPPa isoforms, suggesting that the overall posi-tioning of the domains within the SAXS envelopeswas quite acceptable. Indeed, the fit for theneuronal isoform sAPPa695, which lacks the KPIdomain, is very good up to qZ0.35 AK1.

It is important to note that a single uniquestructure was not obtained from these simulations.Therefore, the task of building the model of the full-length protein was considered satisfactory onlyafter several reconstructions were performed, alter-natively fixing the domains and using differentstarting positions for each of them. This procedure

Figure 6. (a) Size-exclusion chro-matography of the sAPPa695–heparin complex. Black line:10 mM sAPPa695. Violet, blue,cyan, green, brown, red lines:10 mM sAPPa695 in the presence of1.25, 2.5, 5.0, 10 and 20 mM heparin,respectively. Note the disappear-ance of the sAPPa monomer peakand the appearance of a high Mr

peak corresponding to the sAPPa/heparin complex at increasing con-centrations of heparin. (b) Disap-pearance of exposed hydrophobic

surface upon formation of the sAPPa695–heparin complex monitored by bis-ANS fluorescence. All samples contained10 mM bis-ANS, 50 mM Tris–Cl (pH 7.4). Black line: 1.4 mM sAPPa695. Blue, green, red lines: 1.4 mM sAPPa695 in thepresence of 0.28, 0.7 and 1.4 mM heparin, respectively.


was useful in assessing the possibility of obtainingseveral domain configurations compatible with theexperimental data. The average position of thedomains after a number of runs to fit theexperimental data gave differences in translationalposition of w5 A, with maximum angular rotationsof w208 for the HC domain and w308 for the KPIdomain (in sAPPa770).

Positioning the high-resolution structures ofindividual domains into the overall structure ofsAPPa permits the localization of several functio-nally important sites of the protein. For example,the sole N-glycosylation site contained within thecrystallized fragments is shown in green in Figure 5.As expected, it is fully exposed to the solvent inboth isoforms, and its position provides insight intothe location of the glycan side-chains in fullyglycosylated human APP. Two proposed heparin-binding sites of APP22,25 are shown in orange andred. The N-terminal binding site (orange) islocalized in a groove between HC and 1RW6 andmay participate, along with nearby basic aminoacid residues on 1RW6, in heparin binding. On theother hand, the C-terminal heparin-binding site (inred) is fully exposed to the solvent in both isoforms.Interestingly, it has been proposed that the isolated1RW6 domain forms dimers.25 The amino acidresidues participating in the suggested hydro-phobic dimerization interface25 are shown in cyanin Figure 5. However, our structures of both sAPPaisoforms contain significant electronic density(corresponding to APP domains that are notcontained in the atomic-resolution structure of1RW6) occluding part of that putative dimerizationinterface. This may be best appreciated in thecentral panel of Figure 5 for each isoform. There-fore, the current structural models suggest that theproposed dimerization interface of the 1RW6domain is occluded in the full-length sAPPa andAPP molecules and cannot participate in suchprotein–protein interactions (see also next sectionand Discussion).

All the above-mentioned sites are present in themajor neuronal isoform, sAPPa695. An additionalsite of interest in the structure of sAPPa770 is the

protease inhibitory site in the KPI domain (darkgreen in Figure 5(b)), which corresponds to thescissile bond in protease substrates and is expectedto interact most strongly with the proteases that areinhibited by sAPPa770 in vivo.28 The structure showsthat this inhibitory site is fully exposed at one end ofthe sAPPa770 structure and, therefore, poised forinsertion at the active center of a target protease.

Heparin-induced dimerization of sAPPa

The well-known binding of APP to heparansulfate39 and of APP-derived peptides toheparin40,41 prompted us to investigate the inter-action of full-length sAPPa695 with heparin. Size-exclusion chromatography (SEC) analysis ofsAPPa695 showed that it elutes as a major peakcorresponding to the monomer26 and a minor peak(w2% of the total protein mass) likely correspon-ding to a dimer (Figure 6(a)). It has been suggestedthat heparin binds an APP dimer.25 Presence ofincreasing concentrations of heparin in the sampleled to a progressive increase in the higher molecularmass peak and a corresponding reduction in theAPP monomer peak (Figure 6(a)). Interestingly, at amolar ratio of 2:1 sAPPa695:heparin, the highmolecular mass peak is the major species present,with the APP monomer peak reduced to a smallshoulder (green line in Figure 6(a)).

We also investigated the effect of heparin on thebinding of the fluorescent probe bis-ANS tosAPPa695.As previously reported, bis-ANS showsrobust binding to native sAPPa, indicating theexposure of organized hydrophobic surface to theaqueous medium.27 Increasing concentrations ofheparin reduced bis-ANS binding to sAPPa695

(Figure 6(b)), suggesting the disappearance of partof the exposed hydrophobic surface, possibly viadimerization of the protein.

Analytical ultracentrifugation

The aggregation states of the HC fragment and offree or heparin-bound sAPPa were further charac-terized by analytical ultracentrifugation (AUC).

Figure 7. Sedimentation velocityanalysis of pure sAPPa, the sAPPa/heparin complex and the isolatedHC domain. The curves wereobtained employing the sedimen-tation time derivative method (g(s*)integral distribution) analysis ofsedimentation velocity AUC pro-files. Each curve derives from thecombined analysis of serial dilutionsof the indicated sample. (a) Continu-ous line: 7 mM sAPPa695. Brokenline: 7 mM sAPPa770. Dotted line:7 mM sAPPa695 in the presence of3.5 mM heparin. Dashed and dottedline: 7 mM sAPPa770 in the presenceof 3.5 mM heparin. (b) 140 mM HC.


Consistent with previously published SAXS andSEC data26,27 and with the data shown inFigure 6(a), the monomeric state of full-lengthsAPPa695 in solution was confirmed by analyticalultracentrifugation. Equilibrium sedimentationdata for pure sAPPa695 samples at several concen-trations ranging from 2.3 mM to 18.5 mM could bewell fit by a model consisting of a single idealspecies of buoyant weight MbZ19 kDa (corres-ponding to a molecular mass of 69(G5) kDa;Table 1), in excellent agreement with the molecularmass of 67 kDa of full-length sAPPa695.27 Bycontrast, equilibrium sedimentation analysis ofsAPPa695 in the presence of heparin could bewell fit by an ideal single species with MbZ44 kDa (corresponding to a molecular mass of155(G5) kDa; Table 1). Moreover, the buoyantweight of the sAPPa695/heparin complex wasfound to be independent of the concentration ofheparin ranging from 1:1 to 8:1 molar ratios ofheparin to sAPPa695 (not shown). Unbound heparincannot be observed in our experimental setup,which is based on UV absorption detection.

While the observed buoyant weight for thesAPPa695/heparin complex might be compatiblewith the binding of three heparin molecules to amonomer of sAPPa695, this hypothesis is highlyunlikely in view of the complete lack of dependenceof the buoyant weight on heparin concentrationover a large concentration range. Furthermore, itwould be impossible to observe a homogeneouspopulation of species corresponding to a 1:3sAPPa695/heparin complex in the experiments inwhich equimolar concentrations of sAPPa andheparin were employed. Moreover, SEC(Figure 6(a)) and bis-ANS fluorescence data(Figure 6(b)) indicate that the transition from freesAPPa695 to heparin-bound sAPPa695 is largelycomplete at a molar ratio of 2:1 sAPPa695/heparin.Thus, in line with the recent proposal that heparinmight bind to a dimer of sAPP fragments,25 wepropose that the buoyant weight for the complexreports the dimerization of sAPPa695 in thepresence of one molecule of heparin. We also notethat attempts to fit the experimental data with

alternative models (e.g. a complex of 2 sAPPa695C2heparin molecules) generated poor residuals (datanot shown), indicating that they were inaccuratedescriptions of the macromolecular species presentin solution.

The hydrodynamic shapes of pure and heparin-bound sAPPa were also investigated by sedimen-tation velocity experiments. Data for pure sAPPa695

and sAPPa770 could be fit by a well-defined peak at2.9 S (Table 1), though for sAPPa770 a very smallsecondary peak at 7 S was also observed, possiblycorresponding to a small percentage of higher orderaggregates (Figure 7(a)). For the isolated HCfragment, a sedimentation coefficient of 1.48 S wasobtained from sedimentation velocity data(Figure 7(b); Table 1). As noted above, thesesedimentation coefficients suggest a certain degreeof flexibility of both sAPPa isoforms in solution, aswell as for the HC fragment. In the presence of a 2:1molar ratio of sAPPa:heparin, the sedimentationcoefficients increased to 4.8 S and 5.4 S, respectively,for the complexes of sAPPa695 and sAPPa770 withheparin (Figure 7(a); Table 1). This large increasein sedimentation coefficient is compatible with theformation of a dimer of sAPPa in the presence ofone molecule of heparin.

SAXS data on heparin-bound sAPPa

The complex formed at a molar ratio of 2:1sAPPa695:heparin was structurally characterized bySAXS measurements. Figure 8(a) shows the rawdata of the sample scattering. The Guinier plot ofthe lowest-angle portion of the data indicated thatthe sample was monodisperse and confirmed thesuggested stoichiometry, with an Rg estimate of63(G3) A (Figure 8(a), inset). The pairwise distancedistribution function calculated from the completedataset (Figure 8(b)) suggests a prolate dimer with amaximum dimension of 180 A and RgZ57(G0.7) A.The latter Rg value is probably more reliable thanthe Guinier estimate because it is derived fromanalysis of the complete dataset and not only fromthe low angle portion of the scattering curve.Finally, structural models were calculated for

Figure 8. Small-angle X-rayscattering data for the sAPPa–heparin complex. (a) Experimentaldata and error bars collected at a2:1 sAPPa695:heparin ratio. Con-tinuous line: calculated fromDAMMIN model fit (see (c)).Broken line: calculated from theparallel dimer orientation model(see Figure 9(a)). Dash-dotted line:calculated from the antiparallel

dimer orientation model (see Figure 9(b)). Inset: Guinier plot of SAXS data in the low angle region of the curve.(b) Pair-distance distribution function. Continuous line and error bars were calculated from the GNOM fit to the data.(c) Average low-resolution model for the sAPPa695–heparin complex calculated from the experimental data using achain of dummy residues. Twenty independent models were generated and filtered to yield the most probable averagemodel. The model is rotated as indicated in the Figure.


heparin–APP dimers. From ab initio calculations(imposing a 2-fold symmetry axis) we obtainedinformation on the dimer shape (Figure 8(c)).Several independent calculations of the dimermodel30 suggested that two configurations werepossible, with the 2-axis parallel or perpendicular tothe longer molecular dimension. In order toinvestigate these possible conformations, we usedthe sAPPa695 monomers previously calculated togenerate the dimers and fit the models to theexperimental curve using rigid body modelingroutines to obtain the best solutions. Parallel and

Figure 9. Alternative hypotheses for the conformationconformations, derived from modeling of the SAXS data on tbetween sAPPa695 monomers (grey and red), heparin/hepintracellular and transmembrane domains of APP (yellow). Aslower resolution than those on the isolated sAPPa isoforms, woccur within each APP monomer upon dimerization, as suggThese changes may modify the conformation of sAPPa and oorientation of two APP monomers in the same cellular memsulfate chain. The juxtaposition of the two transmembrane andfor example by tyrosine phosphorylation of the cytoplasmic dcytoplasmic domain of only one APP monomer is visible (yepresent in the plasma membranes of different cells. The bridmolecular mass heparin/heparan sulfate chain, might contribu

antiparallel dimer configurations, both withapproximately the same fit qualities, were obtained(see Figure 8(a)). These results permitted theproposal of schematic representations of the poss-ible configurations of the heparan sulfate/APPcomplex in the cell membrane (Figure 9(a) and (b);see below). Qualitatively similar results wereobtained with the sAPPa770/heparan sulfatecomplex (not shown).

It is important to note that all data on sAPPa/heparin complexes were obtained using highmolecular mass heparin (w18 kDa, according to

of transmembrane APP695 dimers. Both hypotheticalhe sAPPa695/heparin complex, show the relations of scalearan sulfate (green), the cellular membrane(s), and theour SAXS data on the sAPPa695/heparin complex were ofe cannot model the conformational changes that probablyested, e.g. by the bis-ANS fluorescence data (Figure 6(b)).f heparin/heparan sulfate within the complex. (a) Parallelbrane, joined by a high molecular mass heparin/heparancytoplasmic domains might trigger intracellular signaling,omain of APP. Note that, due to the tilt of the model, thellow). (b) Antiparallel orientation of two APP monomersge formed by the two APP molecules, joined by a high-te to cell adhesion events as well as intercellular signaling.


manufacturer specifications). Both SAXS and iso-thermal titration calorimetry data failed to indicateany interaction between sAPPa695 and low (3 kDa)molecular mass heparin (data not shown). Inagreement with the proposed stoichiometry of thecomplex, these observations suggest that a singleheparin chain must extend over multiple bindingsites in both sAPPa monomers in order to inducedimerization.

Discussion

Biological functions of sAPPa

Here, a combination of structural information atseveral levels of resolution has provided a firstglimpse at the overall structure of free and heparin-bound sAPPa and insight into the spatial relation-ships between its structural domains. sAPPa is animportant growth factor in the development of thehuman brain and in adult neurogenesis,12,18

paralleling its importance in other human tis-sues.19,42 While the molecular mechanisms of itsactions in postmitotic neurons remain unclear,sAPPa has been shown to exert a memory-enhancing effect.43 Moreover, neural activity canlead to a higher secretion of sAPPa44 and sAPPa hasbeen shown to be neuroprotective against differenttypes of insults.16 In vivo experiments have led tothe hypothesis that AD may be caused in part by alack of sAPPa, in addition to the toxic effects ofAb.45 While it has been shown that sAPPa acts as anautocrine or paracrine growth or differentiationfactor18,19,42 and that it binds to defined patches onthe cell surface,46 no specific receptor for sAPPa hasbeen identified so far. The availability of structuralmodels of sAPPa may, therefore, provide tools forinvestigating the binding partners and mechanismsof action of this important APP fragment.

Location of functional domains with respectto the plasma membrane

Our model offers information on the distancesbetween different APP domains and the neuronalplasma membrane. In spite of the disorderedstructure of the juxtamembrane domain,26 it isplausible that the C terminus of sAPPa and,therefore, the peptide bond cleaved by a-secretaselies in the proximity of the C terminus of the 1RW6fragment, at the lower end of our model (Figure 5).As none of the sAPPa isoforms have beensuggested to bind lipids, it seems likely that thelong axis of sAPPa695 and the corresponding axis ofsAPPa770 extend more or less perpendicularly awayfrom the membrane (Figure 9). As a consequence,the laminin-binding and the heparin/heparansulfate binding sites localized in the helical domain(Figure 1) would lie at a distance of w5 nm fromthe membrane, restricting the range of possiblephysiological binding situations. In the same way,the N-terminal growth-factor-like domain and

the copper-binding domain would lie at a distanceof w10 nm from the membrane, while the inhibi-tory site of the KPI domain, which presumablyinteracts with the active site of the protease(s) thatare inhibited by APP, is at a distance of w12 nmfrom the membrane.

Though our model does not provide detailedinformation on the vicinity of the b-secretasecleavage sites, located six and 16 amino acid residuesupstream from the C terminus of sAPPa,6 knowledgeof the overall structure of sAPPa may be of use inmodeling the interaction of APP with b-secretase,which is also a single-pass transmembrane protein.

Signaling by APP

Full-length APP has long been suggested to be atransmembrane receptor, relaying extracellularsignals to the interior of the cell.2 Interestingly,however, while the biological functions fulfilled bytransmembrane APP have become clearer over theyears,12,13 no universally acknowledged extracellu-lar signal transduced by APP has yet been defined.Any of the several extracellular matrix moleculesshown to bind APP, including glypican, laminin,fibulin and spondin39,47–49 might trigger APP-mediated signaling, but the circumstances inwhich binding of such molecules to transmembraneAPP leads to a defined intracellular signalingcascade remain to be established.

Central to the transmission of signals by APP isthe question whether APP or any of its fragments iscapable of oligomerizing in vivo. Overexpression ofAPP in mammalian cell lines has shown that a smallproportion of transmembrane APP may exist asdimers.25,50 Along the same line, experimentswith APP fragments have indicated that thesemay exist as dimers. Scheuermann et al.50 providedchromatographic evidence that APP18–324 dimerizesin solution, while Wang & Ha25 reported thatAPP346–551 forms dimers, a conclusion that wassupported by data showing that a fraction ofAPP346–551 can be chemically cross-linked asdimers. We have now confirmed by SAXS that avery similar fragment, APP388–574, exists as amixture of monomers and dimers in solution (datanot shown), while the HC fragment (APP28–189) isstrictly monomeric in solution (Figure 4(a)). Wang& Ha25 further cite unpublished evidence that otherfragments of APP, specifically APP191–680, also formdimers. On the other hand, SAXS, size-exclusionchromatography and sedimentation equilibriumanalysis clearly show that full-length sAPPa695

and sAPPa770 are both monomeric in solution,even at relatively high concentrations26,27; thisreport). Furthermore, no fluorescence resonanceenergy transfer has been detected between twoco-localized populations of transmembrane APP770

labeled with different fluorescent tags andexpressed in human neuroglioma cells,51 indicatingthat the vast majority of transmembrane APP inthose cells exists as monomers. Our model offersa solution to this apparent paradox by suggesting


that steric hindrance between different domains offull-length sAPPa impedes the formation of adimer. Due to such steric hindrance, interactionbetween the APP domains that form the putativedimer interface is prevented in full-length APP, butis allowed in isolated APP fragments (see Figure 5).

The native state of sAPPa has been shown to bemarginally stable.27 This low stability suggests thepossibility of rearrangements of the structuraldomains of APP in different physiological condi-tions or in the presence of specific physiologicalligands.26,27 The rearrangement of domains mightcause the exposure of hydrophobic surfacepatches,27 facilitating the dimerization of sAPPand possibly of intact transmembrane APP aswell. Indeed, the present results on the influenceof heparin on the SEC, bis-ANS binding, sedimen-tation equilibrium and X-ray scattering behaviors ofsAPPa suggest that heparan sulfate-containingmolecules, such as glypican,39 may act in this wayin a physiological context. Protein oligomerizationand signaling induced by heparin/heparan sulfatebinding is a well-known phenomenon (for a recentexample, see Kwan et al.52). Furthermore, isother-mal titration calorimetry data (M.G., unpublishedobservations) show, in addition to an electrostaticcomponent expected for the binding of heparin tosAPPa, a hydrophobic component on titration ofsAPPa into heparin, which is also compatible withthe dimerization of sAPPa induced by heparin. Thus,heparin/heparan sulfate and/or other physiologicalligands of APP, as well as changes in the micro-environment in the vicinity of the plasma membrane,might tip a closely-balanced monomer–dimerequilibrium towards dimerization and initiate anintracellular signaling cascade important for physio-logical or pathological cellular events.12,53 In thisregard, two possible models for the dimerizationof transmembrane APP induced by interaction withheparin/heparan sulfate are presented in Figure 9.

In conclusion, small-angle X-ray scatteringmeasurements and ab initio model building haveallowed us to propose, for the first time, structuralmodels for full-length sAPPa695 and sAPPa770.Combined chromatographic, spectroscopic, centri-fugation and scattering data also shed light on thenature of the interactions that may be involved inthe dimerization of APP induced by heparin/heparan sulfate binding. Finally, in line withrecently published results,54 the present studyshows that high-resolution SAXS measurements,in combination with other structural and biochemi-cal data, may provide significant structural andfunctional information in cases in which fullcrystallographic models are not available.

Materials and Methods

SAXS sample preparation and data collection

Recombinant human sAPPa695 and sAPPa770 wereexpressed in Pichia pastoris and purified as described.26

The recombinant fragments HBD1 (APP28–123), CuBD(APP124–189), HC (APP28–189) and the C-terminal domain(APP388–574) were provided by Drs McKinstry, Galatis,Masters, Cappai and Parker (St. Vincent’s Institute,University of Melbourne and Mental Health ResearchInstitute of Victoria). SAXS experiments were performedat the D11A-SAXS beamline55 at the Laboratorio Nacionalde Luz Sıncrotron (LNLS, Campinas, Brazil). Measure-ments and preliminary data analysis on pure sAPPaisoforms were reported.27 We now show, in addition tothe raw data, the Guinier plots, p(r) functions, and the fitsof three-dimensional structural models to the scatteringintensities and to the p(r) functions (see Figure 2).

SAXS measurements on APP fragments were carriedout in the following buffers: 20 mM sodium citrate,250 mM NaCl (pH 5.0), 10% (v/v) glycerol for CuBD;10 mM Hepes (pH 7.0), 10% glycerol for HBD1; and20 mM Hepes (pH 7.4), 250 mM NaCl, 10% glycerol forthe HC and C-terminal domains. Sample concentrationswere l.5 mg/ml for HBD1, 1.7 mg/ml for CuBD, 7.7 mg/ml for the C-terminal domain and several concentrationsranging from 1.3 mg/ml to 13 mg/ml for HC. In all cases,the appropriate buffer signal was subtracted from theSAXS data.

Samples of sAPPa in the presence of heparin (H-3393;Sigma-Aldrich, St. Louis, MO; Mrw18 kDa according tomanufacturer) were prepared in 50 mM Tris–Cl (pH 7.4),for a molar ratio of 2:1 sAPPa/heparin. sAPPa concen-trations were varied from 2.5 mg/ml to 10 mg/ml(sAPPa695) and 5 mg/ml (sAPPa770) to investigatepossible protein concentration effects on the SAXS curves.Low (3 kDa) molecular mass heparin (H-3400, Sigma-Aldrich) was also tested in some experiments (seeResults). All samples were kept at 10 8C during themeasurements, and data acquisition was performed bytaking several successive 600 s frames of each sample,which showed that, under our experimental conditions,there was no radiation damage to the samples. Themonochromatic beam was tuned to 1.488 A to minimizeabsorption by carbon atoms. The experimental setupincluded a temperature-controlled glass capillary and aposition-sensitive detector. Several sample-detector dis-tances were employed to enable detection in the widest qrange accessible within our experimental conditions. Datatreatment was performed using the software packageTRAT1D.56 Usual corrections for detector homogeneity,incident beam intensity, sample absorption and blanksubtraction were included in this routine. The output ofthis software provides the corrected intensities and errorvalues.

Scattering data analysis and ab initio calculations

Fitting of the SAXS data was performed using GNOM,57

which also provided estimates of the radii of gyration andcalculated the pairwise distance distribution functions,p(r), for each sample. Ab initio model calculations for puresAPPa isoforms were carried out using the programpackage GASBOR in its real space version.30 As noted inResults, this version was used in order to obtain the correctvalues of the radii of gyration for the calculated models.For each sAPPa sample, 20 solution models resultingfrom independent runs of the program were compared,aligned using SUPCOMB,36 averaged and filtered usingDAMAVER32 to give the most probable final model.

Five independent runs of GASBOR for the HCfragment of APP gave very similar models, and the resultthat gave the lowest discrepancy from the experimentalSAXS data is shown in Figure 4(c). Rigid-body and


dummy chain modeling of HC was performed usingBUNCH.62 As program input, we used the atomicstructures of residues 28–189 of APP (PDB files 1MWP(amino acid residues 28–123) and 1OWT (amino acidresidues 124–189)). Residues 121–125 were removed fromthose atomic structures and replaced by dummy chainsgenerated during program optimization. Five indepen-dent runs of BUNCH also gave similar models, with onlyslight differences in the relative orientation of the twodomains. The model with the lowest c2 is shown inFigure 4(c). Ab initio model calculations for the sAPPa695/heparin complex were carried out using the programpackage DAMMIN in slow mode.29 Ten solution modelsresulting from independent runs of the program werecompared, aligned, averaged and filtered in the same wayas for pure sAPPa.

Molecular modeling

Modeling of the sAPPa695 isoform was performed byrigid body domain rotations and translations guided bySAXS data. The fits to the scattering intensities are shownas broken lines in Figure 2(a) and (b). The BUNCH modelof the HC domain and the crystallographic structure ofthe C-terminal domain 1RW625 were first manuallypositioned into the ab initio SAXS model of sAPPa695

using MOLMOL.58 Then, the positions of the domainswere optimized using MASSHA,38 which rotates andtranslates the domains searching for the best fit to thescattering data. Different starting orientations for eachdomain were tried, giving practically the same finalconfiguration, and the best result is shown in Figure 3(a).The same steps were followed to obtain the position of theKPI domain (PDB file 1AAP) in the sAPPa770 ab initioSAXS model (Figure 3(b)) using the results of the domainstructure obtained for the sAPPa695 model.

APP dimer simulation

Dimer building was carried out using DIMFOM.63 Theprogram uses the low resolution ab initio model obtainedfor sAPPa695 and, in order to generate the dimer, rolls onemonomer surface over the surface of the other monomer,searching for the arrangement of these two monomersthat gives the lowest discrepancy between the calculatedand experimental SAXS data. In order to account for thepossible steric influence of heparin, a value of 10 A wasimposed as a minimum distance between monomersurfaces. As the surfaces are represented by angularenvelope functions, F(u), this approach is restricted tolow resolution analysis. Therefore, the calculations werelimited to q values !0.2 AK1, which contain informationon shape and quaternary structure. The programgenerates 20 possible solutions ranking them by fitquality, and we show the two best solutions.

† www.jphilo.mailway.com/download.htm

Size-exclusion chromatography

Chromatography was performed on a silica-basedGPC-100 Gold column (250 mm!4.6 mm; Eichrom Tech-nologies Inc., Darien, IL) connected to an HPLCapparatus (Shimadzu, Kyoto, Japan). The column wasequilibrated and operated in buffer containing 50 mMTris–Cl (pH 6.8), 100 mM NaCl. The sample volume was100 ml and the flow rate was 0.3 ml/min. All samplescontained 10 mM sAPPa695, while heparin concentrationwas varied from 0 mM to 20 mM.

Fluorescence spectroscopy

Fluorescence emission spectra were measured at roomtemperature on a F-4500 spectrofluorometer (Hitachi,Tokyo, Japan), with excitation at 365 nm and emissionfrom 400 nm to 600 nm (5 nm bandpasses for bothexcitation and emission). All samples contained 10 mM4-4 0-dianilino-1,1 0-binaphthyl-5,5 0-disulfonic acid (bis-ANS; Molecular Probes, Eugene, OR) in 50 mM Tris–Cl(pH 7.4). The concentration of sAPPa695 was 1.4 mM, andheparin concentrations varied from 0 mM to 1.4 mM. Thespectra of the sAPPa695-containing samples were cor-rected by subtracting appropriate blanks not containingprotein.

Analytical ultracentrifugation

For sedimentation equilibrium experiments, samples ofsAPPa695 (2.3–18.5 mM) in 100 mM NaCl, 50 mM Tris–Cl(pH 7.4), were centrifuged in the absence or in thepresence of increasing concentrations of heparin rangingfrom 1:1 to 8:1 molar ratios of heparin to sAPPa695. Thesamples were centrifuged at 4 8C at successive velocitiesof 9000, 10,000 and 11,000 rpm in a Beckman XL-Aanalytical ultracentrifuge equipped with an AN60Tirotor. All scans were performed at 280 nm with a stepsize of 0.001 cm. Samples were allowed to equilibrate ateach speed for at least 10 h, and repeat scans measured1 h apart were overlaid to determine that equilibrium hadbeen reached. Data were analyzed using non-linear leastsquare fits using the ORIGIN software package providedby Beckman. Fits according to equation (1) were carriedout to determine the best-fitting buoyant molecularweight:

Ar Z exp½lnðA0ÞCMbu2=2RTÞðr2 Kr20Þ�CE (1)

Mb Z Mð1KyrÞ (2)

where Ar is the absorbance at radius r; Ao is theabsorbance at a reference radius ro, usually the meniscus;y is the partial specific volume of the particle (ml/mg); ris the density of the solvent (mg/ml); u is the angularvelocity of the rotor (radian/s); E is the baseline errorcorrection term; M is the gram molecular weight and Mb

is the buoyant molecular weight of the particle; R is theuniversal gas constant (8.314!107 erg/mol K, 1 ergZ0.1 mJ); and T is temperature (in K). The partial specificvolume of sAPPa695 was calculated to be 0.7133 ml/g at4 8C based on amino acid composition using Sednterp†,while for each fit of sAPPa695/heparin complexes thecorrect partial specific volume was calculated as a mass-weighted average of the partial specific volumes ofsAPPa695 (MZ67,000 g/mol, yZ0.7133 ml/g) andheparin (MZ18,000 g/mol, yZ0.47 ml/g).59

For sedimentation velocity experiments, samples ofsAPPa695 and sAPPa770 (7 mM) in 100 mM NaCl, 50 mMTris–Cl (pH 7.4), were centrifuged at 30,000 rpm in aBeckman XL-A analytical ultracentrifuge equipped withan AN60Ti rotor. For measurements of the sAPPa/heparin complex, samples at the same sAPPa concen-tration and in the presence of 3.5 mM heparin wereprepared and centrifuged. For measurements of theN-terminal (HC) fragment of APP, samples (18–140 mM)were centrifuged at both 30,000 rpm and 40,000 rpm. Scandata were acquired at 280 nm, and fittings were

http://www.jphilo.mailway.com/download.htm


performed using SEDFIT.60 Experimentally determinedsedimentation coefficients were compared to coefficientscalculated from SAXS models using HYDROPRO.33 Thesedimentation coefficient for a long linker model of HC(see Results) was calculated as ftotalZSfsubunit,

37 with fvalues for HBD1 (1.183 S) and CuBD (1.155 S) calculatedusing HYDROPRO.33 For comparison, sedimentationcoefficients were also calculated for spherical proteins ofequivalent molecular mass.61

Acknowledgements

We acknowledge the excellent technical assist-ance of Ms Silvia Lucas Ferreira da Silva in theanalytical ultracentrifugation measurements. M.G.was recipient of fellowships from Coordenacao deAperfeicoamento de Pessoal Docente do EnsinoSuperior (CAPES) and Fundacao de Amparo aPesquisa do Estado de Rio de Janeiro (FAPERJ).C.L.P.O. is recipient of a fellowship from Fundacaode Amparo a Pesquisa do Estado de Sao Paulo(FAPESP). M.W.P., C.L.M. and R.C. are supportedby fellowships (M.W.P. and R.C.) and grants(M.W.P., C.L.M. and R.C.) from the National Healthand Medical Research Council of Australia. S.T.F. isa Howard Hughes Medical Institute (HHMI)International Scholar. This work was supported bygrants from HHMI, CAPES, Laboratorio Nacionalde Luz Sıncrotron, FAPESP, Conselho Nacional deDesenvolvimento Cientifico e Tecnologico andFAPERJ.

References

1. Masters, C. L., Simms, G., Weinman, N. A.,Multhaup, G., McDonald, B. L. & Beyreuther, K.(1985). Amyloid plaque core protein in Alzheimer’sdisease and Down Syndrome. Proc. Natl Acad. Sci.USA, 82, 4245–4249.

2. Kang, J., Lemaire, H. G., Unterbeck, A.,Salbaum, J. M., Masters, C. L., Grzeschik, K. H. et al.(1987). The precursor of Alzheimer’s disease amyloid-A4 protein resembles a cell surface receptor. Nature,325, 733–736.

3. Goate, A., Chartier-Harlin, M. C., Mullan, M., Brown,J., Crawford, F., Fidani, L. et al. (1991). Segregation of amissense mutation in the amyloid precursor proteingene with familial Alzheimer’s disease. Nature, 349,704–706.

4. Esch, F. S., Keim, P. S., Beattie, E. C., Blacher, R. W.,Culwell, A. R., Oltersdorf, T. et al. (1990). Cleavage ofamyloid beta peptide during constitutive processingof its precursor. Science, 248, 1122–1124.

5. Postina, R., Schroeder, A., Dewachter, I., Bohl, J.,Schmitt, U., Kojro, E. et al. (2004). A disintegrin-metalloproteinase prevents amyloid plaque for-mation and hippocampal defects in an Alzheimerdisease mouse model. J. Clin. Invest. 113, 1456–1464.

6. Vassar, R. (2002). Beta-secretase (BACE) as a drugtarget for Alzheimer’s disease. Advan. Drug Deliv. Rev.54, 1589–1602.

7. Passer, B., Pellegrini, L., Russo, C., Siegel, R. M.,Lenardo, M. J., Schettini, G. et al. (2000). Generation of

an apoptotic intracellular peptide by gamma-secre-tase cleavage of Alzheimer’s amyloid beta proteinprecursor. J. Alzheimers Dis. 2, 289–301.

8. Kamenetz, F., Tomita, T., Hsieh, H., Seabrook, G.,Borchelt, D., Iwatsubo, T. et al. (2003). APP processingand synaptic function. Neuron, 37, 925–937.

9. Cao, X. & Sudhof, T. C. (2001). A transcriptionallyactive complex of APP with Fe65 and histoneacetyltransferase Tip60. Science, 293, 115–120.

10. Turner, P. R., O’Connor, K., Tate, W. P. &Abraham, W. C. (2003). Roles of amyloid precursorprotein and its fragments in regulating neural activity,plasticity and memory. Prog. Neurobiol. 70, 1–32.

11. Etcheberrigaray, R., Tan, M., Dewachter, I.,Kuiperi, C., Van der Auwera, I., Wera, S. et al. (2004).Therapeutic effects of PKC activators in Alzheimer’sdisease transgenic mice. Proc. Natl Acad. Sci. USA, 101,11141–11146.

12. Sabo, S. L., Ikin, A. F., Buxbaum, J. D. & Greengard, P.(2003). The amyloid precursor protein and itsregulatory protein, FE65, in growth cones andsynapses in vitro and in vivo. J. Neurosci. 23, 5407–5415.

13. Storey, E., Beyreuther, K. & Masters, C. L. (1996).Alzheimer’s disease amyloid precursor protein on thesurface of cortical neurons in primary culture co-localizes with adhesion patch components. Brain Res.735, 217–231.

14. Gunawardena, S. & Goldstein, L. S. (2001). Disruptionof axonal transport and neuronal viability by amyloidprecursor protein mutations in Drosophila. Neuron, 32,389–401.

15. Multhaup, G., Schlicksupp, A., Hesse, L., Beher, D.,Ruppert, T., Masters, C. L. & Beyreuther, K. (1996).The amyloid precursor protein of Alzheimer’s diseasein the reduction of copper(II) to copper(I). Science, 271,1406–1409.

16. Mattson, M. P., Cheng, B., Culwell, A. R., Esch, F. S.,Lieberburg, I. & Rydel, R. E. (1993). Evidence forexcitoprotective and intraneuronal calcium-regulat-ing roles for secreted forms of the beta-amyloidprecursor protein. Neuron, 10, 243–254.

17. Morimoto, T., Ohsawa, I., Takamura, C., Ishiguro, M.,Nakamura, Y. & Kohsaka, S. (1998). Novel domain-specific actions of amyloid precursor protein ondeveloping synapses. J. Neurosci. 18, 9386–9393.

18. Caille, I., Allinquant, B., Dupont, E., Bouillot, C.,Langer, A., Muller, U. & Prochiantz, A. (2004). Solubleform of amyloid precursor protein regulates prolifer-ation of progenitors in the adult subventricular zone.Development, 131, 2173–2181.

19. Pietrzik, C. U., Hoffmann, J., Stober, K., Chen, C. Y.,Bauer, C., Otero, D. A. C. et al. (1998). Fromdifferentiation to proliferation: the secretory amyloidprecursor protein as a local mediator of growth inthyroid epithelial cells. Proc. Natl Acad. Sci. USA, 95,1770–1775.

20. Hynes, T. R., Randal, M., Kennedy, L. A., Eigenbrot, C.& Kossiakoff, A. A. (1990). X-ray crystal structure ofthe protease inhibitor domain of Alzheimer’s amyloidbeta-protein precursor. Biochemistry, 29, 10018–10022.

21. Kroenke, C. D., Ziemnicka-Kotula, D., Xu, J., Kotula,L. & Palmer, A. G., 3rd (1997). Solution conformationsof a peptide containing the cytoplasmic domainsequence of the beta amyloid precursor protein.Biochemistry, 36, 8145–8152.

22. Rossjohn, J., Cappai, R., Feil, S. C., Henry, A.,McKinstry, W. J., Galatis, D. et al. (1999). Crystal


structure of the N-terminal, growth factor-likedomain of Alzheimer amyloid precursor protein.Nature Struct. Biol. 6, 327–331.

23. Barnham, K. J., McKinstry, W. J., Multhaup, G.,Galatis, D., Morton, C. J., Curtain, C. C. et al. (2003).Structure of the Alzheimer’s disease amyloid pre-cursor protein copper binding domain—a regulator ofneuronal copper homeostasis. J. Biol. Chem. 278,17401–17407.

24. Dulubova, I., Ho, A., Huryeva, I., Sudhof, T. C. &Rizo, J. (2004). Three-dimensional structure of anindependently folded extracellular domain of humanamyloid-beta precursor protein. Biochemistry, 43,9583–9588.

25. Wang, Y. & Ha, Y. (2004). The X-ray structure of anantiparallel dimer of the human amyloid precursorprotein E2 domain. Mol. Cell. 15, 343–353.

26. Gralle, M., Botelho, M. M., de Oliveira, C. L. P.,Torriani, I. & Ferreira, S. T. (2002). Solution studiesand structural model of the extracellular domain ofthe human amyloid precursor protein. Biophys. J. 83,3513–3524.

27. Botelho, M. G., Gralle, M., Oliveira, C. L. P., Torriani, I.& Ferreira, S. T. (2003). Folding and stability of theextracellular domain of the human amyloid precursorprotein. J. Biol. Chem. 278, 34259–34267.

28. Zhang, Y., Scandura, J. M., Van Nostrand, W. E. &Walsh, P. N. (1997). The mechanism by whichheparin promotes the inhibition of coagulationfactor XIa by protease nexin-2. J. Biol. Chem. 272,26139–26144.

29. Svergun, D. I. (2000). Advanced solution scatteringdata analysis methods and their applications. J. Appl.Crystallog. 33, 530–534.

30. Svergun, D. I., Petoukhov, M. V. & Koch, M. H. J.(2001). Determination of domain structure of proteinsfrom X-ray solution scattering. Biophys. J. 80,2946–2953.

31. Petoukhov, M. V. & Svergun, D. I. (2003). Newmethods for domain structure determination ofproteins from solution scattering data. J. Appl. Crystal-log. 36, 540–544.

32. Volkov, V. V. & Svergun, D. I. (2003). Uniqueness of abinitio shape determination in small-angle scattering.J. Appl. Crystallog. 36, 860–864.

33. de la Torre, J. G., Huertas, M. L. & Carrasco, B. (2000).Calculation of hydrodynamic properties of globularproteins from their atomic-level structure. Biophys. J.78, 719–730.

34. Ehehalt, R., Keller, P., Haass, C., Thiele, C. & Simons,K. (2003). Amyloidogenic processing of the Alzheimerbeta-amyloid precursor protein depends on lipidrafts. J. Cell Biol. 160, 113–123.

35. Daigle, I. & Li, C. (1993). apl-1, a Caenorhabditis elegansgene encoding a protein related to the human beta-amyloid protein precursor. Proc. Natl Acad. Sci. USA,90, 12045–12049.

36. Kozin, M. B. & Svergun, D. I. (2001). Automatedmatching of high- and low-resolution structuralmodels. J. Appl. Crystallog. 34, 33–41.

37. Cantor, C. R. & Schimmel, P. R. (1980). Ultracentrifu-gation. Biophysical Chemistry. Part II: Techniques for theStudy of Biological Structure and Function. 1st edit., pp.591–642. W. H. Freeman, New York.

38. Konarev, P. V., Petoukhov, M. V. & Svergun, D. I.(2001). MASSHA—a graphics system for rigid-bodymodelling of macromolecular complexes againstsolution scattering data. J. Appl. Crystallog. 34,527–532.

39. Williamson, T. G., Mok, S. S., Henry, A., Cappai, R.,Lander, A. D., Nurcombe, V. et al. (1996). Secretedglypican binds to the amyloid precursor protein ofAlzheimer’s disease (APP) and inhibits APP-inducedneurite outgrowth. J. Biol. Chem. 271, 31215–31221.

40. Clarris, H. J., Cappai, R., Heffernan, D.,Beyreuther, K., Masters, C. L. & Small, D. H. (1997).Identification of heparin-binding domains in theamyloid precursor protein of Alzheimer’s disease bydeletion mutagenesis and peptide mapping.J. Neurochem. 68, 1164–1172.

41. Mok, S. S., Sberna, G., Heffernan, D., Cappai, R.,Galatis, D., Clarris, H. J. et al. (1997). Expression andanalysis of heparin-binding regions of the amyloidprecursor protein of Alzheimer’s disease. FEBSLetters, 415, 303–307.

42. Quast, T., Wehner, S., Kirfel, G., Jaeger, K., De Luca, M.& Herzog, V. (2003). sAPP as a regulator of dendritemotility and melanin release in epidermal melano-cytes and melanoma cells. FASEB J. 17, 1739–1741.

43. Meziane, H., Dodart, J. C., Mathis, C., Little, S.,Clemens, J., Paul, S. M. & Ungerer, A. (1998).Memory-enhancing effects of secreted forms of thebeta-amyloid precursor protein in normal andamnestic mice. Proc. Natl Acad. Sci. USA, 95,12683–12688.

44. Farber, S. A., Nitsch, R. M., Schulz, J. G. &Wurtman, R. J. (1995). Regulated secretion of beta-amyloid precursor protein in rat brain. J. Neurosci. 15,7442–7451.

45. Stein, T. D. & Johnson, J. A. (2002). Lack ofneurodegeneration in transgenic mice overexpressingmutant amyloid precursor protein is associated withincreased levels of transthyretin and the activation ofcell survival pathways. J. Neurosci. 22, 7380–7388.

46. Hoffmann, J., Pietrzik, C. U., Kummer, M. P.,Twiesselmann, C., Bauer, C. & Herzog, V. (1999).Binding and selective detection of the secretoryN-terminal domain of the Alzheimer amyloidprecursor protein on cell surfaces. J. Histochem.Cytochem. 47, 373–382.

47. Kibbey, M. C., Jucker, M., Weeks, B. S., Neve, R. L.,Van Nostrand, W. E. & Kleinman, H. K. (1993). Beta-amyloid precursor protein binds to the neurite-promoting IKVAV site of laminin. Proc. Natl Acad.Sci. USA, 90, 10150–10153.

48. Ohsawa, I., Takamura, C. & Kohsaka, S. (2001).Fibulin-1 binds the amino-terminal head of beta-amyloid precursor protein and modulates its physio-logical function. J. Neurochem. 76, 1411–1420.

49. Ho, A. & Sudhof, T. C. (2004). Binding of F-spondin toamyloid-beta precursor protein: a candidate amyloid-beta precursor protein ligand that modulatesamyloid-beta precursor protein cleavage. Proc. NatlAcad. Sci. USA, 101, 2548–2553.

50. Scheuermann, S., Hambsch, B., Hesse, L., Stumm, J.,Schmidt, C., Beher, D. et al. (2001). Homodimerizationof amyloid precursor protein and its implication inthe amyloidogenic pathway of Alzheimer’s disease.J. Biol. Chem. 276, 33923–33929.

51. Kinoshita, A., Fukumoto, H., Shah, T., Whelan, C. M.,Irizarry, M. C. & Hyman, B. T. (2003). Demonstrationby FRET of BACE interaction with the amyloidprecursor protein at the cell surface and in earlyendosomes. J. Cell Sci. 116, 3339–3346.

52. Kwan, C. P., Venkatamaran, G., Shriver, Z., Raman, R.,Liu, D., Qi, Y. et al. (2001). Probing fibroblast growth


factor dimerization and role of heparin-like glycosa-minoglycans in modulating dimerization and signal-ing. J. Biol. Chem. 276, 23421–23429.

53. Hashimoto, Y., Niikura, T., Chiba, T., Tsukamoto, E.,Kadowaki, H., Nishitoh, H. et al. (2003). Thecytoplasmic domain of Alzheimer’s amyloid-betaprotein precursor causes sustained apoptosis signal-regulating kinase 1/c-Jun NH2-terminal kinase-mediated neurotoxic signal via dimerization.J. Pharm. Expt. Ther. 306, 889–902.

54. Marquez, J. A., Smith, C. I. E., Petoukhov, M. V., LoSurdo, P., Mattsson, P. T., Knekt, M. et al. (2003).Conformation of full-length Bruton tyrosine kinase(Btk) from synchrotron X-ray solution scattering.EMBO J. 22, 4616–4624.

55. Kellermann, G., Vicentin, F., Tamura, E., Rocha, M.,Tolentino, H., Barbosa, A. et al. (1997). The small-angleX-ray scattering beamline of the Brazilian Synchro-tron Light Laboratory. J. Appl. Crystallog. 30, 880–883.

56. Oliveira, C. L. P., dos Santos, D. R., Kellermann, G.,Plivelic, T. & Torriani, I. L. (1997). Data treatmentprogram for the SAXS beamline LNLS Technical Note.Laboratorio Nacional de Luz Sıncrotron, Campinas.

57. Semenyuk, A. V. & Svergun, D. I. (1991). GNOM—aprogram package for small-angle scattering dataprocessing. J. Appl. Crystallog. 24, 537–540.

58. Koradi, R., Billeter, M. & Wuthrich, K. (1996).MOLMOL: a program for display and analysisof macromolecular structures. J. Mol. Graph. 14,51–55.

59. Zhuang, P., Chen, A. I. & Peterson, C. B. (1997).Native and multimeric vitronectin exhibit similaraffinity for heparin—differences in heparin bindingproperties induced upon denaturation are due to self-association into a multivalent form. J. Biol. Chem. 272,6858–6867.

60. Schuck, P. (2000). Size-distribution analysis of macro-molecules by sedimentation velocity ultracentrifuga-tion and Lamm equation modeling. Biophys. J. 78,1606–1619.

61. Lebowitz, J., Lewis, M. S. & Schuck, P. (2002). Modernanalytical ultracentrifugation in protein science: atutorial review. Protein Sci. 11, 2067–2079.

62. Petoukhov, M. V. & Svergun, D. I. (2005). Global rigidbody modelling of macromolecular complexesagainst small-angle scattering data. Biophys. J. 89,1237–1250.

63. Svergun, D., Barberato, C. & Koch, M. H. J. (1995).CRYSOL—a program to evaluate X-ray solutionscattering of biological macromolecules from atomiccoordinates. J. Appl. Crystallog. 28, 768–773.

Edited by P. T. Lansbury Jr

(Received 20 May 2005; received in revised form 19 September 2005; accepted 14 December 2005)Available online 3 January 2006

Note added in proof: While this paper was under review, Soba and co-workers (Soba, P., Eggert, S., Wagner, K.,Zentgraf, H., Siehl, K., Kreger, S. et al. (2005). Homo- and heterodimerization of APP family memberspromotes intercellular adhesion. EMBO J. 24, 3624–3634) presented evidence suggesting that APP and theamyloid precursor-like proteins are capable of forming homo- and heterodimers that may be implicated intrans-cellular adhesion. Such dimers may be similar to our model depicted in Figure 9(b), though Soba andco-workers did not investigate the possible presence of other molecules (e.g. heparan sulfate) in the APPdimers.

solution conformation and heparin-induced dimerization of the full-length extracellular domain of...

Documents