a new triple-stranded α-helical bundle in solution: the assembling of the cytosolic tail of...

A new triple-stranded a-helical bundle in solution: the assemblingof the cytosolic tail of MHC-associated invariant chainAndrea Motta1*, Pietro Amodeo1, Paola Fucile1, Maria A Castiglione Morelli2,Bjørn Bremnes3 and Oddmund Bakke3

Background: The invariant chain (Ii) is a transmembrane protein thatassociates with the major histocompatibility complex class II (MHC II) moleculesin the endoplasmic reticulum. The cytosolic tail of Ii contains two leucine-basedsorting motifs and is involved in sorting the MHC II molecules to the endosomalpathway where the peptide antigen is bound. This region of Ii also contributesto phenotypical changes in cells, such as the formation of large endocyticstructures.

Results: We report here the three-dimensional structure of a 27 amino acidpeptide corresponding to the cytosolic tail of Ii. The structure was determinedby nuclear magnetic resonance (NMR) spectroscopy using a computationalstrategy. At high concentration, this structure reveals a new triple-strandedα-helical bundle in which the helices, two parallel and one antiparallel, arealmost coplanar. Trimerization is mediated by electrostatic interactionsintercalated by three hydrophobic layers.

Conclusions: The new trimer fold, the first to be identified by NMR data alone,can be used to improve understanding of protein–protein interactions and tomodel multiple-helical transmembrane proteins and receptors. We suggest thatinteractions of the Ii cytosolic tails may form part of a mechanism that couldcause the endosomal retention and enlarged endosomes induced by Ii.

IntroductionAn important approach to developing an understandingof protein folding is the structural study of peptidesdesigned to fold into established protein secondary andtertiary motifs [1,2]. In recent years, a variety of differentproteins and peptides has been prepared, including four-helix bundles [3,4], β-sandwich proteins [5–7], andmixed α-β proteins [8,9]. An initial approach, by Eisen-berg and DeGrado and their coworkers towards de novosynthesis of a four-helical bundle protein [10,11], was thesynthesis of a 16-amino acid peptide (named α-1a) whichwould self-associate to form an α-helical tetramer. NMRinvestigation [12] of the solution structure of α-1, apeptide shorter than α-1a, indicated that at high peptideconcentration an oligomer forms in which the monomersare regular helices throughout the length of the peptide.Unfortunately, because of the fast exchange betweenmonomer and oligomer, the authors could not concludeanything about the symmetry or stoichiometry of thestructure of the oligomer. Lovejoy et al. [13] reported thatthe crystal structure of coil-Ser, a peptide designed tomimic the two-stranded coiled-coil conformation of tropo-myosin, is actually a triple-stranded coiled-coil formed bythree α helices forming the letter ‘N’, with the helicalaxes running up-up-down.

hIi-(1–27) peptide, the subject of this article, was origi-nally studied to understand the structure/activity relation-ship of the sorting motifs of the major histocompatibilitycomplex class II (MHC II) associated human invariantchain (hIi) protein [14]. The peptide corresponds to thefirst 27 amino acids of the N-terminal cytosolic tail of hIi,and is involved in sorting to endosomes and the formationof large endosomal structures [15–19]. The mechanismsbehind the retention of hIi and endocytosed material inearly endosomes [20] and the formation of large endo-somes are not known. It has, however, been reported thatthe N-terminal 11-residue region of hIi is essential for theformation of large vesicles and for retention in the endoso-mal pathway [16,17]. We show here that the hIi-(1–27)peptide at high concentration aggregates forming a copla-nar triple-stranded α-helical bundle. This architecture,the first to be determined in solution, well reproduces theinterhelical parameters of bacteriorhodopsin. We suggestthat this architecture can be used as a building block forthe packing of multiple-helical transmembrane proteinsand receptors, and it may also help to further improve ourunderstanding of protein–protein interactions. The struc-tural trimerization/aggregation properties of the hIi cytoso-lic tail in vitro correlate with the in vivo large endosomeformation suggesting that these two events are related.

Addresses: 1Istituto di Chimica di Molecole diInteresse Biologico del CNR (Istituto Nazionale diChimica dei Sistemi Biologici), I-80072 Arco Felice,Italy, 2Dipartimento di Chimica, Università dellaBasilicata, I-85100 Potenza, Italy and 3Departmentof Biology, Division of Molecular Cell Biology,University of Oslo, N-0316 Oslo, Norway.

*Corresponding author.E-mail: [email protected]

Key words: α-helical bundle, invariant chain, NMRstructure, protein–protein recognition, proteintrafficking

Received: 30 June 1997Revisions requested: 31 July 1997Revisions received: 10 September 1997Accepted: 24 September 1997

Structure 15 November 1997, 5:1453–1464http://biomednet.com/elecref/0969212600501453

© Current Biology Ltd ISSN 0969-2126

Research Article 1453

ResultsNMR structure determinationThe structure of the peptide hIi-(1–27) was determinedin methanolic solution at pH 7.2 by the homonuclear 1HNMR approach [21]. At a concentration of 1.2 × 10–4 Mthe peptide is monomeric [14]. Using a combination of standard scalar correlation and nuclear Overhauserenhancement spectroscopy (NOESY) experiments, com-plete sequential resonance assignments have beenobtained for the monomer. The secondary structure ofhIi-(1–27) was delineated from the qualitative patternrecognition approach of the sequential and medium-range nuclear Overhauser effects (NOEs), as derivedfrom two-dimensional (2D) experiments [22]. Figure 1summarizes the observed NOEs, the relative exchangerates of amide protons and the apparent 3JHNα couplingconstants for the monomeric hIi-(1–27) (black and emptysymbols). The fact that in the region Gln4–Leu17 theNHi-NHi+1 NOEs are intense, while the αCHi-NHi+1NOEs are much weaker, implies a generally helical struc-ture. The observation of several unambiguous αCHi-NHi+3, αCHi-βCHi+3 and two αCHi-NHi+4 cross-peaksstrongly supports the presence of a helix [22]. Furthercorroborative data come from slowly exchanging amidesand 3JHNα coupling constants. Except for Gln13, all theamide protons in the Gln4–Leu17 region are in slowexchange, and except for Met16, all residues have3JHNα < 6 Hz. As no αCHi-NHi+2 cross-peaks, suggestiveof a 310 helix, were detected in the middle region of thepeptide we conclude that an α helix is present in theGln4–Leu17 region of hIi-(1–27) with a bend at Pro14.

For the C-terminal tail, two type II β turns (Arg20–Ala23and Ala23–Ser26) are found in the region Arg20–Ser26.These data confirm the structure found for the monomerichIi-(1–27) in water-methanol (80%–20% v/v) [14].

The N-terminal 11-residue region of hIi is essential forthe formation of large vesicles and for retention in theendosomal pathway [16,17]. In order to investigate thetendency of hIi-(1–27) to possibly regulate vesicle forma-tion, we studied its behavior upon variation of thepeptide concentration from 1.2 × 10–4 M to 1.0 × 10–2 M.All solutions were stable for several months, but uponfurther increase of the concentration the sample rapidlyformed a gel. Concentration increase caused progressivespectral changes of backbone amide resonances. Signifi-cant variation of the chemical shift was observed for theamides of residues Asp2–Asp6 with a general broadeningof the resonances (AM, PA, PF and MACM, unpublishedresults), suggesting the presence of concentration-depen-dent self-association [12]. 2D-NMR experiments revealedadditive cross-peaks for all residues. Figure 2 comparesthe total correlation spectroscopy (TOCSY) NHi-βCHiregions of hIi-(1–27) at different concentrations. On goingfrom 1.2 × 10–4 M (Figure 2a) to 2.6 × 10–4 M (Figure 2b),extra cross-peaks of different intensity are observed forAsn10 and Asn11 as well as for the NHi-αCHi connectivi-ties of Arg5–Ser9 (not shown). The unusual observationof Met1 NH (Figure 2b) suggests that the aggregationbrings about a decrease of the exchange rate of thisproton with the solvent. This exchange mechanism isusually fast and is responsible for the non-observability

1454 Structure 1997, Vol 5 No 11

Figure 1

Sequential and secondary structuralinterresidue NOEs observed for hIi-(1–27).The data were collected in methanol at 300K(pH 7.2), for the monomer and the trimer.NOE intensities are indicated by the height ofthe bars. Empty bars refer to the effects of theHδ protons of proline residues. The couplingconstants are labeled: ●, apparent3JHNα < 5.0 Hz; ❍, measured 3JHNα > 7.0 Hz; o,6.0 Hz ≤ 3JHNα ≤ 7.0 Hz. Filled circles alsolabel residues for which slow exchange of NHprotons with deuterons was observed.Hatched bars and circles refer to extra NOEsand variation of the coupling constants,respectively, observed in the aggregate.

of the N-terminal proton in a polypeptide [21]. At1.4 × 10–3 M (Figure 2c) all the amino acids of thepolypeptide show satellite peaks. In particular, six con-nectivities were observed for amino acids in the helicalregion (compare, for example, Asp6, Asn10 and Asn11with Met1), but these reduced to three cross-peaks at6.0 × 10–3 M (Figure 2d). Further increase of the concen-tration up to 1.0 × 10–2 M always showed three peaks perresidue and no chemical shift variation. Beyond this limitthe sample formed a gel. This observation strongly sug-gests the presence of three different molecular species,

termed P1, P2 and AP, which were fully assigned byresorting to homonuclear three-dimensional (3D) experi-ments (AM, PA, PF and MACM, unpublished results).

The three cross-peaks observed in Figure 2d do not ruleout the possibility that a 3n symmetric aggregate, wheren > 1, is present in solution (e.g. n = 2, hexamer; n = 3,nonamer, etc.). To check this possibility, the apparentmolecular weight of the hIi-(1–27) at 6.0 × 10–3 M concen-tration was measured by gel filtration. At this concentra-tion hIi-(1–27) eluted with an apparent Mr of 9400 (open

Research Article A new triple-stranded a-helical bundle Motta et al. 1455

Figure 2

Comparison of the TOCSY NHi-βCHi regionsof hIi-(1–27). Data were collected in methanolat 300K (pH 7.2), at different concentrations:(a) 6.0 x 10–3 M, (b) 1.4 x 10–3 M, (c)2.6 x 10–4 M and (d) 1.2 x 10–4 M. Wherepossible, cross-peaks are labeled with theone-letter code for amino acids and thesequence number.

8.0 ppm8.48.89.2

(d)

(c)

(b)

(a)

2.8

3.0

2.8

3.0

2.8

3.0

2.8

3.0

Structure

circle in Figure 3) in close agreement with the calculatedmolecular size for a trimer (Mr = 9208). The possibility ofan equilibrium of different oligomeric states was also ruledout by examining the elution pattern obtained for the cor-responding hIi-(1–27) Ser26→Trp analogue at the aboveconcentration (data not shown). Such a derivative wasused to measure the extinction coefficient of the peptideat 280 nm (the native hIi-(1–27) lacks aromatic aminoacids) which, reported against the elution volume, appearedto be symmetric, indicating that a single species is presentin solution. 2D-NOESY spectra also ruled out possibleconformational changes for the presence of Trp26. Wealso applied electrospray ionization mass spectrometry(ESI-MS) to the analysis of the number of polypeptidechains involved in the aggregate [23] at different pHs anddeclustering potentials. The ESI spectra of the aggregateat low declustering potential (∆CS ≈ 5–10 V) displayed,among others originating from the monomer and thedimer, an intense peak for the 7+ charged molecular ion atm/z = 1316. This is diagnostic for the trimeric peptide: Mrof the monomeric hIi-(1–27) = 3069; (M3)7+ produced apeak at m/z = (3 × 3069 + 7)/7 = 1316.3. There was no indi-cation of a tetrameric structure as the possible tetramericmolecular ions (M4)7+, (M4)9+ and (M4)11+ did not appear atm/z = 1754, 1363, and 1115, respectively. The [M + 7 H]7+

ion of the trimer completely disappeared upon dissocia-tion at higher ∆CS (60 V), thus confirming the noncova-lent nature of the complex.

Structural restraints and calculationsStructural information on the aggregate was obtained using3D-NOESY data. The intramolecular NOEs observed forthe monomer were all confirmed, and three new NOEconnectivities (hatched bars in Figure 1) were detected;the coupling constants of Asp2 and Asp3 became < 6.0 Hz(hatched circles). According to these results, in the aggre-gate the chains are α helical in the region Asp2–Leu17thus including Asp2 and Asp3 within the helix. Derivationof distance restraints for calculating the NMR structure ofa homotrimer is complicated by the inability to distinguisha priori an intramolecular NOE from an intermolecularNOE. All NOEs were interpreted in a very conservativemanner, and only those that were well resolved and clearlyinconsistent with the global fold of the monomer wereassigned as intermolecular contacts. Clear examples includethe well-resolved NOEs between εCH3 of P1 Met1 andP2 Met1 with δCH3 of AP Leu14; εCH3 of AP Met1 withthe γCH3 of P2 Leu17 and with εCH3 P1 Met16 (Figure 4)as well as between the η protons of AP Arg5 and the γprotons of P1 Glu12 and P2 Glu12. In all, 40 intermolecu-lar contacts were assigned, and these were sufficiently dis-tributed to establish the antiparallel nature of the trimersymmetry with one antiparallel and two parallel helices. Itis worth noting that the trimer is well defined by 40 NOEs


Figure 4

Critical intermolecular contacts used to define the antiparallel nature ofthe hIi-(1–27) trimer. The figure shows the aliphatic region of the500 MHz 2D-NOESY spectrum of the peptide acquired with 0.040 smixing time at pH 7.2 and 300K. It is possible to see the well resolvedcross-peaks between εCH3 of P1 Met1 and P2 Met1 with δCH3 of APLeu14; εCH3 of AP Met1 with the γCH3 of P2 Leu17 and with εCH3P1 Met16.

2.02.2

1.2

1.0

0.8AP M1ε

P1 M1ε P2 M1ε

P1 M16ε

P2 L17γ

AP L14δ

1.8ppm

Structure

Figure 3

Determination of the molecular weight of the hIi-(1–27) peptide by gelfiltration at a concentration of 6.0 × 10–3 M. The apparent molecularweight of the peptide (9400 Da) is represented by an open circle.

50000

30000

20000

10000

1000

3000

5000Mol

ecul

ar w

eigh

t

15 20 25Elution volume (ml) Structure

as in its final structure the maximum number of theoreti-cally observable NOEs is 52. A starting structure wascreated by manually positioning three monomers withhelices P1 and P2 in a parallel arrangement and AP in anantiparallel arrangement according to the N-shaped triple-stranded α-helical bundle [13]. The orientation of thethree molecules was optimized visually based on inter-molecular NOEs (Figure 4). The starting structure wasgenerated using a series of simulated-annealing (SA)runs. After analysis of energy and violations, the best 40structures were energy minimized, solvated in a periodicbox of 5120 methanol molecules and subjected to a 500ps run of restrained molecular dynamics (MD) with allintra- and interchain restraint, followed by 1 ns of MD

with interchain restraints only. Figure 5 shows a stereo-view of the best-fit superposition of the backbone atomsof 25 structures sampled every 10 ps in the last 250 ps ofthe 1 ns run: a unique backbone fold is obtained alongthe whole chain, with some lack of convergence aroundArg20. This spreading is well represented in Figure 6, inwhich the thickness of the ribbon is proportional to theroot mean square (rms) deviation of the backbone atomsduring the 1 ns simulation. The structural statistics of the25 NMR structures, as derived from the 1 ns run aresummarized in Table 1. Possible conformational biasinginduced by NOE restraints was ruled out by a 250 psunrestrained MD simulation performed on the finalrestrained 1 ns structure.


Figure 5

Figure 6

Stereoview of the calculated structures of theaverage backbone atom positions of the hIi-(1–27) trimer during the 1 ns simulation.Cylinders represent the average helicalorientations, and thickness of the ribbon isproportional to the root mean square deviationof the backbone atoms. The three monomersare colored green (P1), magenta (AP) andblue (P2). (The figure was created with theprogram MOLMOL [60].)

Stereoview of the calculated structures of thehIi-(1–27) trimer. The figure shows the best-fitsuperpositions of the backbone atoms of theselected 25 NMR-derived structures of thetrimer. The structures are superimposed usingresidues 4–15, with a root mean squaredeviation of 0.797 ± 0.179 for this region, and1.120 ± 0.320 for all residues. The threemonomers are colored green (P1), magenta(AP) and blue (P2). (The figure was createdwith the program MOLMOL [60].)

Three-dimensional structure of the trimerThe average structure exhibits an almost perfectly anti-parallel orientation between AP and P1 (central and lefthelices, respectively, in Figure 7a). Their helical axesform an angle of 5.6°, while the P2 helical axis forms anangle of 32° with both AP and P1. The distances betweenthe helical axes are 9.7 Å (AP–P1), 11.1 Å (AP–P2) and19.3 Å (P1–P2). The AP and P2 helices are almost coplanarshowing a head-to-tail arrangement with very strong inter-actions between the AP C-terminal and the P2 N-terminalregions, and between the P2 C-terminal and AP N-termi-nal regions (Figure 7b). The P1 C-terminal region (bluearrow; left helix), on the contrary, is almost perpendicularto the helical average plane, and exhibits interactions withP2 C-terminal residues (Figure 7b). The main secondarystructure elements are preserved in the MD simulation. Inparticular, α helices are observed in the AP Asp3–Leu17region, in the P1 Asp2–Leu17 region, and in the P2Asp2–Pro15 region. An analysis of average hydrogen-bondpatterns observed in the MD simulation shows some dis-tortion in the AP helix around Asp6, due to backbone–sidechain interactions. In addition, a more extended per-turbation of the hydrogen-bond network is observed for allthree helices from Ser9 to Asn11, due to the breakingeffect of Pro15 on the hydrogen-bond pattern. All threechains are distorted at the C-terminal end of the helix,with a type I β turn centered on Pro15–Met16 for AP andP2, and on Leu14-Pro15 for P1. From Gly18, all of themonomers form a bend at the end of each helical region.In addition, a type I β turn centered on Asp2-Asp3 locksthe N-terminal end of the AP helix. A stable type II β turn

centered on Pro21-Gly22 is observed for P1 and P2, whilefor AP this turn only occurs for about 4% of the simulatedtime. Another type II turn centered on Pro24-Glu25 isobserved for AP, while in the other two monomers thechain is strongly distorted at Pro24, and is found in aninverse γ-turn conformation. Finally, a short β sheet com-prising residues Gly18-Arg19-Arg20 and residues Glu25and Ser26 is observed in the P1 helix.

Salt bridges (14) and hydrogen bonds (18), are localized ina few regions. In particular, four salt bridges and fourhydrogen bonds are formed between Glu25-Ser26-Lys27of AP and the Met1–Asp6 region of P2. Five other saltbridges and one hydrogen bond join residues Asp2–Asp6of AP, to Arg19 and Arg20 of P2. The AP N-terminalregion (residues Met1–Asp6) forms one salt bridge andtwo hydrogen bonds with P2 Lys27 and P2 Ser26 andthree hydrogen bonds with the P2 central helix (Ile8,Asn11 and Glu12); AP residues Arg5 and Asp6 form onesalt bridge with P1 Glu12, and two hydrogen bonds withP1 Gln13 and P1 Ser9, respectively. The second half ofthe AP helix (residues Asn10–Gln13) interacts with boththe P1 and P2 N-terminal regions, forming two hydrogenbonds with P1, and one salt bridge and two hydrogenbonds with P2. In spite of the large distance between theP1 and P2 helical axes, strong interactions are observedfor some polar sidechains. In particular, salt bridges areobserved between P2 Arg20 and P1 Glu12, and betweenP2 Glu25 and P1 Arg20, while two hydrogen bonds join P2Glu25 to P1 Met16 and P1 Ser26. Electrostatic interac-tions are intercalated by three hydrophobic layers, whichare approximately perpendicular to three average helicalaxes (Figure 7c). These hydrophobic layers protect someof the polar intermonomer interactions from the poten-tially disrupting solvent effects. Except for the hydropho-bic areas of mutual interactions of the three chains, theoverall structure seems to lack a real inside, and somealiphatic sidechains are considerably ‘solvent exposed’.In particular, AP and P2 (magenta and blue helices,respectively, in Figure 7c) show two considerably solvent-exposed sidechains (Met16 and Pro21 for AP, Met16 andLeu17 for P2), while in P1 (green helix) six hydrophobicsidechains (Leu7, Ile8, Leu14, Pro21, Ala23 and Pro24),out of a total of ten, are solvent exposed. The solventaccessibility of aliphatic sidechains is also retained in theunrestrained MD simulations, thus ruling out the possi-bility that the accessibility is an artifact due to NMRrestraints. It is worth recalling that beyond a concentra-tion of 1.0 × 10–2 M, hIi-(1–27) rapidly formed a gel, thusthe presence of solvent exposed aliphatic sidechains canexplain the tendency of the trimer to associate into higherorder aggregates.

The 1 ns restrained MD simulation shows similar intra-molecular and peptide–solvent interaction energies for APand P2, and both exhibit a strong interchain stabilization


Table 1

Structural statistics of the 25 NMR structures of trimeric hIi-(1–27) peptide.

Backbone rms deviations (Å)all (243 atoms)* 1.120 ± 0.320residues 4–15 (108 atoms)† 0.797 ± 0.179

Rms deviations from 40 intermoleculardistance restraints (Å)

lower limits 0.008 ± 0.050upper limits 0.019 ± 0.097total 0.028 ± 0.108sum of all violations 1.100

(φ,ψ) Rms from average values (°)Whole chain(s)

AP/ P1/ P2/ All 7.714/ 6.949/ 7.490/ 7.378Residues 4–15

AP/ P1/ P2/ All 5.186/ 4.977/ 4.827/ 4.997

Angular order parameter S‡

Whole chain(s)AP/ P1/ P2/ All 0.989/ 0.992/ 0.989/ 0.990

Residues 4–15AP/ P1/ P2/ All 0.996/ 0.996/ 0.996/ 0.996

*The maximum deviation was 1.684 at 0.558 min. †The maximumdeviation was 1.122 at 0.447 min. ‡See [61].

(AP–P2 energy is lower than AP–solvent and P2–solventcontributions). In contrast, P1 has more favorable intra-molecular and peptide–solvent interaction energies, but aconsiderably weaker interaction with both the other twochains. Therefore, from an energetic point of view, a‘dimer-plus-one’ picture of this trimer emerges. The for-mation of the trimer, however, involves a rearrangementof the monomers, and not a simple addition of a monomerto a dimer. MD simulations on isolated AP and P1 initiallyset in a conformation corresponding to that assumed in thetrimer, have shown substantial differences with the trimerstructure (AM, PA, PF and MACM, unpublished results).The first four residues and the region following Pro15differ significantly in the monomeric and trimeric forms ofboth simulated structures. In particular, the monomeric P1chain shows a helical region (Arg5–Leu17) shorter thanthe corresponding chain in the trimer, whose helix starts atAsp2. In addition, a completely different orientation of theC terminus is observed, owing to substantial differences inthe bend after Pro15. The short β sheet observed in the C-terminal domain of trimeric P1 (blue arrow; left helix inFigures 7a and b) is absent in the monomer, while theremaining secondary structure elements are substantiallypreserved. These results suggest that the structure of theP1 chain in the trimer is strongly induced by aggregation,

and that the formation of the trimer involves a rearrange-ment of each chain. The AP–P1 interactions are preva-lently hydrophobic in nature, while those between AP andP2 are mainly based upon electrostatic forces amongcharged and polar groups. These structural data suggestthat the stability of the trimer depends upon a balance ofpolar and hydrophobic interactions.

DiscussionStructural implicationsWe report here the structure of a new triple-strandedα-helical bundle in solution. The trimer is formed bythree α helices running up-up-down and located almostcoplanarly. To the best of our knowledge, this is the firsttime that the structure of a homotrimer has been solved byresorting only to NMR data and calculations. The onlyhomotrimer to be previously described is the crystal struc-ture of coil-Ser [13], in which the three α helices runningup-up-down form the letter ‘N’. As shown by the bac-teriorhodopsin structure, the coplanar and the N-shapedtrimers appear to be complementary and both can be rel-evant as building blocks for modeling multiple-helicaltransmembrane proteins and receptors. In the structureof bacteriorhodopsin [24], six out of seven helices arearranged into two quasi-parallel coplanar trimers (helices


Figure 7

Different views of the hIi-(1–27) trimer asderived from 1 ns restrained moleculardynamics calculations. The figure showsviews which are (a) perpendicular to and (b)parallel to the average helical plane. The Nterminus of each monomer is in red and the Cterminus is in blue. Large ribbons, arrows andtubes correspond to helical regions, β sheetsand other conformations, respectively. Theexternal monomers correspond to P1 (left)and P2 (right). (c) Hydrophobic surfaces(orange dots) formed within the trimer; thecorresponding hydrophobic residues areshown in yellow. The P1, AP and P2monomers are shown in green, magenta andblue, respectively. (The figure was drawn withthe program MidasPlus 2.0.)

B, C and D; and helices E, F and G), similar to thearrangement found for the hIi-(1–27) peptide. The twotrimers of bacteriorhodopsin are then locked by helix A,which faces helices B and D, forming with them an N-shaped structure. Figure 8 shows a superposition of P1, APand P2 with the bacteriorhodopsin helices D, C and B,respectively, showing an impressive fitting of the interhe-lical parameters. This observation suggests that the twotypes of homotrimer represent two possible ways of arrang-ing three helices. Both helical arrangements would certainlyhelp in predicting the packing of rhodopsin-like G protein-coupled receptors, a widespread and rapidly growing familyof proteins that mediate signal transduction through adiverse series of activating ligands. Members of this familyof proteins share a basic structural framework, believed toconsist of seven transmembrane helices [25]. The coplanartrimeric folding motif has also been found in the structureof aquaporin-1 [26,27], a member of the major intrinsicprotein superfamily of integral membrane proteins. Thestructure of aquaporin-1 is composed of six membrane-spanning α helices which appear to pack in three two-helix pairs. This arrangement resembles the dimer-plus-onepicture found for the hIi-(1–27) peptide (vide supra). Ourresults obtained for hIi-(1–27), therefore, establish a newmotif for understanding the topology and design of mem-brane protein channels.

The hIi-(1–27) trimer is not related to the coiled-coilfamily: it does not show the standard repeat (a b c d e f g)nwith hydrophobic residues at positions a and d and polarresidues elsewhere [28], and no superhelical parameterscould be obtained [29]. The three-dimensional structure

of hIi-(1–27) is determined largely by the pattern of polarand hydrophobic residues and appears to be independentof the geometric properties of the amino acid sidechainsinvolved in the trimerization. This is in line with recentstudies which report that the tertiary folds of globular pro-teins, and not those of proteins of the coiled-coil family[30], are insensitive to the details of packing, they aredetermined instead by hydrophobic, polar and secondarystructure patterns in the amino acid sequence [31–34].

Wild type hIi is found to induce enlarged endosomal vesi-cles and alter the characteristics of the endosomal pathway[16–19]. The requirements within the essential membranedistal part of hIi for altering the endosomes has been inves-tigated by exchanging each residue within the region 2–11to alanine and thereafter transfecting the simian COS cellline (BB and OB, unpublished data). Such changes have novisible influence on the formation of the large endosomalstructures. This agrees with the structural finding that glob-ular proteins are able to accommodate multiple hydropho-bic substitutions at the core positions without adopting anew fold [31–34]. MD simulations of the trimer with mul-tiple alanine substitutions in the region 2–11 show a morecompact helical region and no variation of the hydropho-bic layers in the aggregate. Thus, rearrangement of thehydrophobic core in response to the mutations allows thetertiary fold to be conserved.

On the contrary, changing the polar pattern of theAsp2–Asp6 region showed a dramatic change: the largeendosomes were not induced although the sorting signalswere intact. In particular, the single mutation of


Figure 8

Stereoview superposition of the hIi-(1–27)trimer chains P1 (green), P2 (dark blue) andAP (magenta) with the D (orange), C (pink)and B (cyan) helices of bacteriorhodopsin,respectively. (The figure was created with theprogram MOLMOL [60].)

Asp6→Arg, is sufficient to abrogate the effect (BB andOB, unpublished data). A 250 ps MD simulation usingthe Arg6 mutant, indicates an opening of the trimericstructure (PA, PF, MACM and AM, unpublishedresults). In general, shorter and distorted helical seg-ments are observed for P1 (Asp3–Asn10 region) and P2(Asp2–Leu14 region), with an almost complete unwindingof the AP helix. Furthermore, energetic analysis showsvery low energy interactions between monomers. Arginineis able to modify the intra- and interchain pattern of saltbridges and hydrogen bonds due to the nature of itscharge, and its longer sidechain hampers the overallpacking of the trimer.

The overall three-dimensional structure of the trimer andthe presence of solvent exposed hydrophobic sidechainssuggest a possible explanation for hIi-(1–27) gelation. Acomputer model of the gel presents layers of differenttrimers held together by strong interactions betweenC-terminal regions, with some penetration between adja-cent hydrophobic regions and some noncovalent cross-linking between them. The alternating hydrophobic andhydrophilic layers of the trimer are fully preserved andextended into three dimensions. Within the limitations ofthe model, the formation of the gel network is mediatedby exposed hydrophobic interactions which stabilizemutual interactions between trimers.

Possible implications for the Ii proteinIn the absence of association with the MHC class II mol-ecules, hIi forms a trimer in the endoplasmic reticulum[35]. A fraction of the hIi trimers is transported to theGolgi complex and is found in enlarged endosomal vesi-cles. We postulate here that the aggregational propertiesshown by the cytosolic tail of hIi may shed some light onthe formation of such vesicles. A model for retention ofmolecules in the Golgi stack suggests that they formaggregates which are too large to enter into forward-moving transport vesicles [36,37]. Data showing that Golgienzymes interact and, in doing so retain each other, ledNilsson et al. [38,39] to propose that these enzymes existin linear hetero-oligomers which are held together byspecific interactions. The interaction site has been local-ized to the luminal stalk region of the proteins close tothe membrane. In analogy with this, the retention of hIiin endosomes may also result from the formation ofoligomers, a property which might be linked to the in vitroaggregation and gel formation shown here. hIi is found toform trimers and the trimerization domain is believed tobe located within the luminal domain of hIi [40,41]. As theoligomerization of the cytosolic tail requires an up-down-up orientation, this may imply that different trimers tothose seen in vitro are involved in vivo. Higher order mul-timers, at least hexamers, have been observed during thepurification of hIi and by gel filtration [35]. The retentionof hIi by oligomerization may cause large endosomes to

form as a result of the assembly of hIi ‘rafts’, which inhibitfission or budding of the hIi-containing vesicles. On thecontrary, an interaction of the cytosolic tail may also leadto the fusion of different hIi-containing endosomal vesi-cles forming larger units (i.e., one hIi tail interacts withtwo tails from another vesicle). Immunofluorescence [16]and immunoelectron microscopy [42] studies have shownthat the large vesicular structures exist after proteolyticprocessing of the luminal domain of hIi. This observationindicates that the cytosolic tail is important for the mainte-nance of these large endosomes at later, more proteolyticstages of the endocytic pathway.

Large endosomes are induced by hIi protein and thesevesicles are reduced in size when the cells are transientlycoexpressing the associated α and β chain of MHC II aftertransient transfection [16]. In stable cell lines generated indifferent species and cell types, we have noted that thelarge hIi induced endosomes are much reduced in numberwhen cells are supertransfected with the MHC class IImolecules, although the level of hIi expression is regu-lated to be the same by an inducible promoter (S Kjølsrud,T Nordeng and OB, unpublished results). This mightimply that the α and β chain shield the formation of largehIi aggregates.

We have shown that in vitro the cytosolic tail of hIi assem-bles into triple-stranded α-helical bundles which form agel network. At this stage, however, we have no firm evi-dence of a direct correlation between this property and theobserved endosomal retention and enlarged endosomesinduced by hIi. Nevertheless, the data presented heremay imply that the interactions between the cytosolic tailsof hIi form part of a mechanism that could cause theobserved biological effects.

Biological implicationsWe have determined the solution structure of hIi-(1–27),a peptide corresponding to the first 27 amino acids ofthe cytosolic tail of human invariant chain (hIi), a trans-membrane protein that associates with class II MHCmolecules. In the aggregated form studied here, thepeptide contains structural determinants relevant to theunderstanding of protein folding. At high concentrationthe peptide forms a new triple-stranded α-helical bundlein which the helices, two parallel and one antiparallel,are almost coplanar. The structure is largely stabilizedby the pattern of polar and hydrophobic residues andappears to be independent of the geometric properties ofthe amino acid sidechains involved in the trimerization.

The helical arrangement of the coplanar trimer describedhere is reproduced within two transmembrane proteinsof known structure, bacteriorhodopsin and aquaporin-1.Thus, this new structure will help in predicting thepacking of rhodopsin-like G protein-coupled receptors,


which share the structural framework of seven trans-membrane helices. Furthermore, the structure establishesa new motif which may help our understanding of thetopology and the design of membrane protein channels.

Regarding the biological functions of hIi, our findingssuggest that it is the interactions between cytosolicdomains that modulate the nature of the endosomalpathway and thereby may modulate the antigen pro-cessing compartments. This effect might be via anincreased fusion of endocytic vesicles when the tailsinteract, or the cytosolic domain interaction may leadto inhibition of the process whereby vesicles bud off.This may explain biological data showing that hIiretards the endocytic process and the finding that over-expression of hIi leads to the formation of large endo-somal vesicles.

Materials and methodsSample preparationThe peptides (MDDQR5DLISN10NEQLP15MLGRR20PGAPE25SK andits W26 derivative) were synthesized as already described [14].Samples were prepared by dissolving appropriate amounts of thepeptide in 0.4 x 10–3 l of 1H2O buffer (20 mM phosphate and150 mM NaCl, pH 7.2), with the concentration ranging from0.92 x 10–4 M to 1.0 x 10–2 M. The samples were lyophilized andredissolved in C2H3O1H. The temperature was 300K. Partially deu-teriated methanol was obtained from Cambridge Isotope Laborato-ries (Woburn, MA, USA).

NMR spectroscopyNMR spectra were recorded on a Bruker DRX-500 spectrometerequipped with a triple resonance probehead and pulsed field gradients.Proton chemical shift assignments were performed using conventional2D and 3D homonuclear techniques. 2D-NOESY [43] and TOCSY [44]were recorded in the phase-sensitive mode with quadrature detection inω1 provided by the time-proportional phase incrementation scheme[45]. 16 transients of 2048 points were collected for each of the 512 t1increments, with a spectral width of 6024 Hz. Time-domain data matri-ces were all zero-filled in both dimension to 4K, thus yielding a digitalresolution of 2.94 Hz/point. NOESY experiments were obtained withmixing times of 0.040, 0.080 and 0.16 s with no random variation.Clean version [46] of the TOCSY experiments, were recorded with spin-lock periods of 0.046 and 0.086 s, achieved with the MLEV-17 pulsesequence [47]. The 3D 1H-1H NOESY-TOCSY [48,49] and NOESY-NOESY [50,51] experiments were acquired with mixing times of0.080 s for both mixing periods in the first experiment, and 0.050 and0.100 s for the second, with eight scans of 1024 points. This wasrepeated for 256 t2 values and 256 t1. Time-domain data were zero-filled once in t2 and t1 and cosine-bell windows were used in all threetime domains. An automatic third-order polynomial baseline correctionwas applied on the three frequency domains. Processing was per-formed using TRIAD (Tripos Inc., St Louis) and Aurelia 2.0 (Bruker Ana-lytische) running on Silicon Graphics Indigo2 R4400 computers.

Structure calculationDistance restraints for structure calculations were obtained by integrat-ing the volumes of NOE peaks at different mixing times and fitting thebuild-up of the NOEs to a second-order polynomial. Using an r–6

dependence of the initial slope, the corresponding interproton dis-tances were calculated. In those cases where this procedure failed infitting the build-up, the distance restraints were set to a maximum dis-tance of 4.5 Å. Peaks that appeared to be overlapped in the 2D-NOESY spectra and peaks picked from the 3D experiments were also

set to the maximum distance restraints of 4.5 Å. An upper bound for alldistance restraints was set to 10% above the calculated distance whilethe lower bound was conservatively set to 2.5 Å, representing a valueclose to the van der Waals contact. For the methyl protons a correctionof 0.3 Å was made [52]. φ and ψ dihedral angle restraints were derivedfrom 3JNHα coupling constants. For the α-helical region, values of –60°± 30° and –40° ± 30° were used for the φ and ψ angles, respectively;for the residue in position 2, φ =–60° ± 20° and ψ = 120° ± 20°; andfor the residue in position 3 in the type II β turns φ = 80 ± 20° and ψ =0° ± 20°. Hydrogen-bond restraints were obtained by analyzing the1H/2H exchange rate and the NOE patterns characteristic of α helices.Two distance restraints, rNH–O (1.8–2.3 Å) and rN–O (2.3–3.3 Å), wereused for each hydrogen bond. The final structure of the monomer alonewas calculated with 326 NOEs (101 intraresidue, 123 sequential[αCHi-NHi+1 and NHi-NHi+1], and 102 medium range [αCHi-NHi+n,n ≥ 2 and αCHi-βCHi+3]). For each monomer within the aggregate weadded three new NOEs (two sequential and one medium-range). Thefinal structure of the trimer was calculated with a total of 1027 NOEs,which included 40 intermolecular NOEs defining the mutual orientationof single chains.

Model building and preliminary calculations were performed on an UnitedAtom (UA) model [53] with the SYBYL 6.2 package (Tripos Inc., StLouis). The solvation effects were roughly included using a distance-dependent dielectric constant ε = r. A cut-off radius of 8 Å for nonbondedinteractions, with a residue-based pairlist generation routine, was appliedto all calculations. In the UA model, distance restraints were included asC–C or C–N distance increasing the upper limits calculated for interpro-ton distances of 1Å. Semiparabolic penalty functions were used withforce constants in energy minimization (EM) and room temperature MD of5 kcal mol–1rad–2 for dihedral restraints and 5kcal mol–1Å–2 for distancerestraints. In SA and preliminary MD simulations a time step of 1 fs, withno restraint on bond length applied, was used. SA was performed withdifferent lengths (from 10 to 250ps), temperature (maximum values from500 to 800K, applied for 25% to 75% of the total time, cooling ratesfrom 1 to 5K ps–1) and restraint force constant time profiles. EM and MDsimulations with solvent were performed with the SANDER module of theAMBER 4.1 package using the AMBER/OPLS force field [54,55]. A timestep of 2 fs, with rigid restraining of all the bond lengths (SHAKE algo-rithm) [56,57] and periodic boundary conditions were applied. All MDsimulations were performed in the isothermal-isobaric ensemble (300K,1 atm), with a solvent box which initially extended 8 Å from the mostexternal solute atom on each side of the box. The final analyses were per-formed with specific modules of the AMBER 4.1, SYBYL 6.2, andWHAT IF 4.99 [58] packages. For molecular graphics we used the pro-grams SYBYL (6.2 Tripos Inc., St. Louis) and MidasPlus 2.0 [59] fromthe Computer Graphics Laboratory, University of California, San Fran-cisco (supported by NIH RR-01081), and MOLMOL [60] rendered withthe POV-ray software.

Molecular weight characterization of the trimerGel filtration of the hIi-(1–27) peptide was carried out using a 1 x 50 cmcolumn of Sephadex LH-20 eluted with methanol, pH 7.2, at a rate of8 ml h–1. 0.5 ml of each solution corresponding to those used for NMRspectroscopy were applied to the column, and 1.0 ml fractions werecollected. Proteins used as molecular weight markers were from Sigma(α-chymotrypsinogen, RNase S protein and aprotinin) and BoehringerMannheim (insulin B-chain).

ESI-MS measurements were performed with a BIO-Q triple quadrupolemass spectrometer (Micromass, Manchester, UK). The ion spray inter-face temperature was kept at approximately 313 K for all measure-ments. The quadupole mass analyzer, with a nominal m/z range of2000, was operated at unit resolution. Mass calibration was performedby means of multiple charged molecular ions from a separate injectionof myoglobin. Peptide solutions were prepared in 2.0 x 10–3 M ammo-nium acetate:methanol (9:1), and the pH adjusted with acetic acid andaqueous ammonia.


Accession numbersCoordinates will be deposited with the Protein Data Bank, BrookhavenNational Laboratory. In the meantime they may be obtained via elec-tronic mail from [email protected].

AcknowledgementsWe thank WF DeGrado (University of Pennsylvania School of Medicine,Philadelphia, PA) and DC Wiley (Harvard University, Cambridge, MA) forcritical reading of the manuscript. Special thanks are due to A Scaloni(IABBAM del CNR, Napoli) for skillful assistance with mass spectrometry.Support was provided by EC TMR grant ERBFMRXCT960069 (AM andOB), the Norwegian Research Council (OB) and the Norwegian CancerSociety (BB and OB).

References1. Richardson, J.S. & Richardson, D.C. (1989). The de novo design of

protein structure. Trends Biochem. Sci. 14, 304–309.2. DeGrado, W.F., Raleigh, D.P. & Handel, T.M. (1991). De novo protein

design: what are we learning? Curr. Opin. Struct. Biol. 1, 984–993.3. Handel, T.M., Williams, S.A. & DeGrado, W.F. (1993). Metal ion-

dependent modulation of a designed protein. Science 241, 879–885.4. Choma, C.T., Lear, J.D., Nelson, M.J., Dutton, P.L., Robertson, D.E. &

DeGrado, W.F. (1994). Design of a heme-binding four-helix bundle. J. Am. Chem. Soc. 116, 856–865.

5. Quinn, T.P., Tweedy, N.B., Williams, R.W., Richardson, J.S. &Richardson, D.C. (1994). Betadoublet: de novo design, synthesis andcharacterization of a β-sandwich protein. Proc. Natl. Acad. Sci. USA91, 8747–8751.

6. Yan, Y. & Erickson, B.W. (1994). Engineering of betabellin 14D:disulfide-induced folding of a β-sheet protein. Protein Sci.3, 1069–1073.

7. Pessi, A., Bianchi, E., Crameri, A., Venturini, S., Tramontano, A. &Sollazzo, M. (1993). A designed metal binding protein with a novelfold. Nature 362, 367–369.

8. Beauregard, M., et al., & Martial, J.A. (1991). Spectroscopicinvestigation of structure in octarellin (a de novo protein designed toadopt the α/β-barrel packing). Protein Eng. 4, 745–749.

9. Tanaka, T., Kuroda, Y., Kimura, H., Kidokoro, S. & Nakamura, H.(1994). Cooperative deformation of a de novo designed protein.Protein Eng. 7, 969–976.

10. Regan, L. & DeGrado, W.F. (1988). Characterization of a helicalprotein designed from first principles. Science 241, 976–978.

11. Eisenberg, D., Wilcox, W., Eshita, S.M., Pryciak, P.M., Ho, S.P. &DeGrado, W.F. (1986). The design, synthesis and crystallization of analpha-helical peptide. Proteins 1, 16–22.

12. Ciesla, D.J., Gilbert, D.E. & Feigon, J. (1991). Secondary structure ofthe designed peptide alpha-1 determined by nuclear magneticresonance spectroscopy. J. Am. Chem. Soc. 113, 3957–3961.

13. Lovejoy, B., et al., & Eisenberg, D. (1993). Crystal structure of asynthetic triple-stranded α-helical bundle. Science 259,1288–1293.

14. Motta, A., Bremnes, B., Castiglione Morelli, M.A., Frank, R.W.,Saviano, G. & Bakke, O. (1995). Structure-activity relationship of theleucine-based sorting motifs in the cytosolic tail of the MHC-associated invariant chain. J. Biol. Chem. 270, 27165–27171.

15. Bakke, O. & Dobberstein, B. (1990). MHC class II-associated invariantchain contains a strong signal for endosomal compartments. Cell63, 707–716.

16. Romagnoli, P., Layet, C., Yewdell, J., Bakke, O. & Germain, R.N.(1993). Relationship between invariant chain expression and majorhistocompatibility complex class II transport into early and lateendocytic compartments. J. Exp. Med. 177, 583–596.

17. Pieters, J., Bakke, O. & Dobberstein, B. (1993). The role of thecytoplasmic tail of MHC class II associated invariant chain in targetingto and retention in endocytic compartments. J. Cell Sci.106, 831–846.

18. Bremnes, B., Madsen, T., Gedde-Dahl, M. & Bakke, O. (1994). A LIand a ML motif in the cytoplasmic tail of the MHC-associated invariantchain mediate rapid internalisation. J. Cell Biol. 107, 2021–2032.

19. Pond, L., et al., & Peterson, P.A. (1995). A role for acidic residues indi-leucine motif-based targeting to endocytic pathway. J. Biol. Chem.270, 19989–19997.

20. Gorvel, J.-P., Escola, J.-M., Stang, E. & Bakke, O. (1995). Invariantchain induces a delayed transport from early to late endosomes. J. Biol. Chem. 270, 2741–2746.

21. Wüthrich, K. (1986). NMR of Proteins and Nucleic Acids. J Wiley andSons, Inc., New York, NY.

22. Wagner, G., Neuhaus, D., Wörgötter, E., Vasák, M., Kägi, J.R.H. &Wüthrich, K. (1986). Nuclear magnetic resonance identification of“half-turn” and 310-helix secondary structure in rabbit livermetallothionein-2. J. Mol. Biol. 187, 131–135.

23. Przybylski, M. & Glocker, M.O. (1996). Electrospray massspectrometry of biomacromolecular complexes with noncovalentinteractions — new analytical perspectives for supramolecularchemistry and molecular recognition processes. Angew. Chem. Int.Ed. Engl. 35, 806–826.

24. Grigorieff, N., Ceska, T.A., Downing, K.H., Baldwin, J.M. & Henderson,R. (1996). Electron-crystallographic refinement of the structure ofbacteriorhodopsin. J. Mol. Biol. 259, 393–421.

25. Attwood, T.K. & Findlay, J.B.C. (1994). Fingerprinting G-protein-coupled receptors. Protein Eng. 7, 195–203.

26. Walz, T., et al., & Engel, A. (1997). The three dimensional structure ofaquaporin-1. Nature 387, 624–627.

27. Cheng, A., van Hoek, A.N., Yeager, M., Verkman, A.S. & Mitra, A.K.(1997). Three-dimensional organization of a human water channel.Nature 387, 627–630.

28. McLachlan, A.D. & Stewart, M. (1975). Tropomyosin coiled-coilinteractions: evidence for an unstaggered structure. J. Mol. Biol.98, 293–299.

29. Crick, F.H.C. (1953). The packing of α-helices: simple coiled-coils.Acta Cryst. 6, 689–697.

30. Harbury, P.B., Zhang, T., Kim, P.S. & Alber, T. (1993). A switchbetween two-, three-, and four-stranded coiled coils in GCN4 leucinezipper mutants. Science 262, 1401–1407.

31. Lim, W.A. & Sauer, R.T. (1991). The role of internal packinginteractions in determining the structure and stability of a protein. J. Mol. Biol. 219, 359–365.

32. Hurley, J.H., Baase, W.A. & Matthews, B.W. (1992). Design andstructural analysis of alternative hydrophobic core packing arrangementin bacteriophage T4 lysozyme. J. Mol. Biol. 224, 1143–1159.

33. Stites, W.E., Gittis, A.G., Lattman, E.E. & Shortle, D. (1991). In astaphylococcal nuclease mutant the side-chain of a lysine replacingvaline 66 is fully buried in the hydrophobic core. J. Mol. Biol.221, 7–14.

34. Bowie, J.U., Luthy, R. & Eisenberg, D. (1991). A method to identifyprotein sequences that fold into a known three-dimensional structure.Science 253, 164–170.

35. Marks, M.S., Blum, J.S. & Cresswell, P. (1990). Invariant chain trimersare sequestered in the rough endoplasmic reticulum in the absence ofassociation with HLA class II antigen. J. Cell. Biol. 111, 839–855.

36. Pfeffer, S.R. & Rothman, J.E. (1987). Biosynthetic protein transportand sorting by the endoplasmic reticulum and Golgi. Annu. Rev.Biochem. 56, 892–905.

37. Weisz, O.A., Swift, A.M. & Mackhamer, C.E. (1993). Oligomerizationof a membrane protein correlates with its retention in the Golgicomplex. J. Cell Biol. 122, 1185–1196.

38. Nilsson, T., et al., & Warren, G. (1994). Kin recognition betweenmedial Golgi enzymes in HeLa cells. EMBO J. 13, 562–574.

39. Nilsson, T., Rabouille, C., Hui, N., Watson, R. & Warren, G. (1996).The role of the membrane-spanning domain and stalk region of N-acetylglucosaminyltransferase I in retention, kin recognition andstructural maintenance of the Golgi apparatus in HeLa cells. J.CellSci. 109, 1975–1989.

40. Bijlmakers, M.E., Benaroch, P. & Ploegh, H.L. (1994). Mapping thefunctional regions in the lumenal domain of the class II-associatedinvariant chain. J. Exp. Med. 180, 623–629.

41. Gedde-Dahl, M., Freisewinkel, I., Staschewski, M., Schenck, K., Koch,N. & Bakke, O. (1997). Exon 6 is essential for invariant chaintrimerization and induction of large endosomal structures. J Biol.Chem. 272, 8281–8287.

42. Stang, E. & Bakke, O. (1997). MHC class II associated invariant chaininduced enlarged endosomal structures. A morphological study. Exptl.Cell Res. 235, 79-92.

43. Jeener, J., Meier, B.H., Bachmann, P. & Ernst, R.R. (1979).Investigation of exchange processes by two-dimensional NMRspectroscopy. J. Chem. Phys. 71, 4546–4553.

44. Braunschweiler, L. & Ernst, R.R. (1983). Coherence transfer byisotropic mixing: application of proton correlation spectroscopy. J. Magn. Reson. 53, 521–528.

45. Drobny, G., Pines, A., Sinton, S., Weitkamp, D. & Wemmer, D. (1979).Fourier transform multiple quantum nuclear magnetic resonance.Faraday Div. Chem. Soc. Symp. 13, 49–55.


46. Griesinger, C., Otting, G., Wüthrich, K. & Ernst, R.R. (1988). CleanTOCSY for 1H spin system identification in macromolecules. J. Am.Chem. Soc. 110, 7870–7872.

47. Bax, A. & Davis, D.G. (1985). MLEV-17-based two-dimensionalhomonuclear magnetization transfer spectroscopy. J. Magn. Reson.65, 355–366.

48. Vuister, G.W., Boelens, R. & Kaptein, R. (1988). Nonselective three-dimensional NMR spectroscopy. The 3D NOE-HOHAHA experiment.J. Magn. Reson. 80, 176–183.

49. Oschkinat, H., et al., & Clore, G.M. (1988). Three-dimensional NMRspectroscopy of a protein in solution. Nature 332, 374–376.

50. Boelens, R., Vuister, G.W., Koning, T.M.G. & Kaptein, R. (1989).Observation of spin diffusion in biomolecules by three-dimensionalNOE-NOE spectroscopy. J. Am. Chem. Soc. 111, 8525–8526.

51. Breg, J., Boelens, R., Vuister, G.W. & Kaptein, R. (1990). 3D NOE-NOE spectroscopy of proteins. Observation of sequential 3D NOEcross peaks in arc repressor. J. Magn. Reson. 87, 646–651.

52. Koning, T.M.G., Boelens, R. & Kaptein, R. (1990). Calculation of thenuclear Overhauser effect and the determination of proton–protondistances in the presence of internal motions. J. Magn. Reson.90, 111–123.

53. Weiner, S.J., et al., & Weiner P.K. (1984). A new force field formolecular mechanical simulation of nucleic acids and proteins. J. Am.Chem. Soc. 106, 765–784.

54. Jorgensen, W.L. (1986). Optimized intermolecular potential functionsfor liquid alcohols. J. Phys. Chem. 90, 1276–1284.

55. Jorgensen, W.L. & Tirado-Rives, J. (1988). The OPLS potentialfunctions for proteins. Energy minimizations for crystals of cyclicpeptides and crambin. J. Am. Chem. Soc. 110, 1657–1666.

56. Ryckaert, J.P., Ciccotti, G. & Berendsen, H.J.C. (1977). Numericalintegration of the Cartesian equations of motions of a system withconstraints: molecular dynamics of n-alcanes. J. Comp. Phys.23, 327–341.

57. van Gunsteren, W.F. & Berendsen, H.J.C. (1977). Algorithms formacromolecular dynamics and constraint dynamics. Mol. Phys.34, 1311–1327.

58. Vriend, G. (1990). WHAT IF: a molecular modeling and drug designprogram. J. Mol. Graph. 8, 52–56.

59. Ferrin, T.E., Huang, C.C., Jarvis, L.E. & Langridge, R. (1988). TheMIDAS display system. J. Mol. Graph. 9, 13–27.

60. Koradi, R., Billeter, M. & Wüthrich, K. (1996). MOLMOL: a program fordisplay and analysis of macromolecular structures. J. Mol. Graph.14, 51–55.

61. Hyberts, S.G., Goldberg, M.S., Havel, T.F. & Wagner, G. (1992). Thesolution structure of eglin c based on measurements of many NOEsand coupling constants and its comparison with X-ray structures.Protein Sci. 1, 736–751.


a new triple-stranded α-helical bundle in solution: the assembling of the cytosolic tail of...

Documents