kinking the coiled coil negatively charged residues at the

doi:10.1016/j.jmb.2006.11.083 J. Mol. Biol. (2007) 366, 1232–1242

Kinking the Coiled Coil – Negatively Charged Residuesat the Coiled-coil Interface

Ravid Straussman1, Ami Ben-Ya’acov1, Derek N. Woolfson2,3

and Shoshana Ravid1⁎
1Department of Biochemistry,Faculty of Medicine, HebrewUniversity, Jerusalem 91120,Israel2School of Chemistry,University of Bristol, Cantock’sClose, Bristol BS8 1TS, UK3Department of Biochemistry,School of Medical Sciences,University of Bristol, BristolBS8 1TD, UK
Present addresses: R. StraussmanCellular Biochemistry and Human GMedicine, Hebrew University, JerusYa’acov, Liver Unit, Department ofHebrew University Medical Center,Abbreviations used: ACD, assemb

MHC, myosin II heavy chain; a.a., aE-mail address of the correspondi

[email protected]

0022-2836/$ - see front matter © 2006 E

The coiled coil is one of the most common protein-structure motifs. It isbelieved to be adopted by 3–5% of all amino acids in proteins. Itcomprises two or more α-helical chains wrapped around one another. Thesequences of most coiled coils are characterized by a seven-residue(heptad) repeat, denoted (abcdefg)n. Residues at the a and d positionsdefine the helical interface (core) and are usually hydrophobic, thoughabout 20% are polar or charged. We show that parallel coiled-coils have aunique pattern of their negatively charged residues at the core positions:aspartic acid is excluded from these positions while glutamic acid is not.In contrast the antiparallel structures are more permissive in their aminoacid usage. We show further, and for the first time, that incorporation ofAsp but not Glu into the a positions of a parallel coiled coil creates aflexible hinge and that the maximal hinge angle is being directly related tothe number of incorporated mutations. These new computational andexperimental observations will be of use in improving protein-structurepredictions, and as rules to guide rational design of novel coiled-coilmotifs and coiled coil-based materials.

© 2006 Elsevier Ltd. All rights reserved.

Keywords: coiled coil; amino acid profiles; fibrous proteins; myosin II;protein–structure prediction
*Corresponding author
Introduction

The α-helical coiled coil was first proposed byCrick in 1953.1 Crick’s model describes polypeptidechains configured into right-handed α-heliceswrapping around each other slightly offset toform left-handed rope-like structures. The helix–helix interactions are cemented by tight interdigita-tion of side-chains at the helical interface, whichCrick termed “knobs-into-holes” packing. A heptadsequence repeat, in which hydrophobic residues thatfill the interface are spaced alternately three and four

, Department ofenetics, Faculty of

alem, Israel; A. Ben-Medicine, HadassahJerusalem, Israel.ly competent domain;mino acid residue(s).ng author:

lsevier Ltd. All rights reserve

residues apart, specifies this structural motif. Thesestructural and sequence features have subsequentlybeen verified by numerous experimental studies.2

The coiled coil is in fact one of the most commonprotein-structural motifs: it is believed to be adoptedby approximately 3–5% of all amino acids inproteins.3 In effect, the coiled coil is a motif forprotein–protein interactions, or protein oligomeri-zation. Typically α-helical coiled coils comprise twoto five α-helices. The sequences of most, but notall,4–6 coiled coils are characterized by the afore-mentioned seven-residue (heptad) repeat, which isoften denoted (abcdefg)n (Figure 1(a)). It is the a andd positions that are predominantly hydrophobic andare found at the interface between the α-helices;positions e and g flank this core, and tend to beoppositely charged residues; and the remainingpositions, b, c and f, lie away from the coiled-coilinterface, and tend to be occupied by polarresidues.7 The alternate spacing of a and d of threeand four residues along the chain averages hydro-phobic residues to approximately one every 3.5residues. As this falls short of the 3.6 residues perturn of the α-helix, the resulting a/d hydrophobic

d.

mailto:[email protected]

†http://www.biols.susx.ac.uk/coiledcoils

Figure 1. The structure of a coiled coil and of myosinII. (a) The helical wheel representation of the coiled-coilheptad repeat showing how the positions a–g appearwhen viewed from the C-terminal end of the parallelhelices. (b) Schematic diagram of the myosin II molecule.

1233Negative Residues in the Coiled-coils Interface

seam winds around the surface of the helix in theopposite sense to the direction of the polypeptidebackbone. Thus, in order for the hydrophobic seamsof the helices to remain in contact over the length ofthe coiled coil they must wrap or supercoil aroundeach other, and, as predicted by Crick,1 heptad-based coiled coils have a left-handed supercoil.Based on the above characteristic repeating pat-

terns, coiled coils are particularly amenable to com-puter-based protein-structure prediction. Indeed,reasonable sequence-based predictors of coiled-coilmotifs are available.3,8–10 Unfortunately, these dosuffer to differing degrees from false-negative andfalse-positive predictions.8,11 For instance, evenwhen asking “simply” where coiled-coil motifsoccur in linear sequences, there is disagreementbetween results from the above predictors.11 Thus,an important objective in current coiled-coil researchis to improve and expand coiled-coil predictionmethods. Clearly, this effort is hampered, or at leastsubject to certain degree of inaccuracy, with onlycoiled-coil databases garnered on the basis ofsequence data alone.With a view to tackling this problem, Walshaw &

Woolfson created SOCKET, which is a program thatidentifies and analyses coiled-coil motifs withinprotein structures.11 Unlike the above predictors,which search solely for sequence features to predictcoiled-coil motifs, the SOCKET algorithm uses thecharacteristic knobs-into-holes side-chain packing ofcoiled coils to identify coiled-coil motifs fromprotein-structure coordinate files. SOCKET definescoiled-coil helix boundaries, oligomerization stateand helix orientation, it also assigns heptad registersto the identified sequences. SOCKET has beenapplied to gather a set of unambiguous coiled-coilstructures from the Brookhaven Protein Data

Bank†.11 We used this database to investigate theoccurrence of charged amino acids in the hydro-phobic cores of coiled coils.Many studies demonstrate the importance of the

hydrophobic nature of the core a and d positions incoiled-coil folding and stability.12–19 Polar and chargedresidues are also found at these positions, however,where they account for up to 20% of the residues,and are, in many cases, highly conserved.3,8–10 Insome respects, the roles of these inclusions areunderstood. Such inclusions help specify oligomerstates (i.e. they select between dimers, trimers andtetramers), and strand orientation (i.e. parallel versusantiparallel arrangements) of coiled coils. Theyappear to act by destabilizing alternative folds morethan the native structures.18–21Aside from the general requirement for a and d to

be hydrophobic, the precise nature of these sitesdiffer markedly both in terms of coiled-coil sequenceand structure. Although this had been hinted atbefore,22 Harbury and colleagues were the first toformalize the structural differences at the twosites.23 For instance, in parallel dimers the packinggeometries of the knobs-into-holes interactions at aand d are different and referred to as parallel andperpendicular, respectively; in parallel tetramersthese geometries are reversed. In trimers thegeometries at a and d are similar and intermediatebetween parallel and perpendicular packing, re-ferred to as acute. The core-packing geometries inantiparallel coiled coils are slightly more compli-cated.11 These packing differences are reflected inamino acid preferences observed at the two sites indifferent oligomer states both of engineered,23 andnatural sequences.10 For example, Leu is the mostfavoured residue at perpendicular sites, hence thestrong preference for this residue at the d sites ofparallel dimers.10,11,23

Regarding the inclusion of polar residues at theinterfaces of parallel coiled coils, which is the subjectof this paper, there is a clear preference for certainpolar residues at the a sites. It appears that theparallel packing geometry at these sites is moretolerant of different amino acid side-chains ingeneral.11,23,24 Having said this, only certain polaramino acids, namely, Asn, Lys and Arg, appear toconfer dimer specificity.10,20,21,25 High-resolutionstructure studies reveal that these particular resi-dues either make specific side-chain–side-chaininteractions, or simply “escape” the core at the asite; whereas, neither could be achieved by otherpolar residues or at the d positions.21

Firstly, here we used the SOCKET-derived data-base to examine the occurrence of charged aminoacids at the a and d positions of known coiled-coilstructures. We show that the inclusion of asparticacid (Asp) at these sites differs between parallel andantiparallel coiled coils as it is totally excluded onlyfrom the core positions of parallel coiled coils. Incontrast, glutamic acid (Glu) can be found at the core

http://www.biols.susx.ac.uk/coiledcoils

Table 1. Frequency of charged amino acids in coiled-coil“core” positions

Parallel

a d

A. Five human myosin II heavy chain subtypes (Supplementary Data,Table 1)Asp 0 0.05Glu 0 1.5Lys 2.3 0.4Arg 2.3 0.2

Parallel Antiparallel

a d a d

B. Coiled coils that were tested positive for coiled coil motifs in theSOCKET analysis of the PDBAsp 0 0 0.05 0.3Glu 0.08 0.5 0.3 0.6Lys 0.2 0.3 0.5 0.4Arg 0.4 0.3 0.8 0.2

The data were normalized for the frequency of occurrence of eachamino acid in SWISSPROT.

1234 Negative Residues in the Coiled-coils Interface

positions of both parallel and antiparallel coiledcoils. We also show that both negatively chargedamino acids prefer the d over the a positions.We then used the human non-muscle myosin II

heavy chain (MHC) to examine the impact ofintroducing a negatively charged amino acid intothe a position of a long parallel homo-dimeric coiledcoil. Myosin II is a hexamer composed of two heavychains of ∼200 kDa and two pairs of light chains.26The heavy chains are organized into two globularhead regions containing the force generating, actin-binding and ATPase functions, and a rod regioncomposed of a long (>1000 residue) α-helical coiledcoil (Figure 1(b)). Myosin II forms filaments throughself-association of its rod domain.27 The protein ismonomeric at high ionic strengths (>300 mMNaCl),but assembles into filaments in low salt.28 A specificregion near the C-terminal end of the rod, dubbedthe “assembly competent domain” (ACD), is neces-sary for filament assembly.28–30 We mutated severala positions of the myosin II rod (N-terminal to theACD) replacing them with Asp, Glu or Leu. Weshow that all of the Asp mutations introducedflexible kinks in the coiled-coil rod at positionscorresponding to the mutation site, while Glu or Leumutations had no such effect.Combined, our sequence analysis and experimen-

tal studies strongly indicate that Asp residuesdisrupt the assembly of parallel coiled-coils, whenplaced at the a position of the coiled-coil interface.This observation should help improve coiled-coilstructure predictions from sequences, and couldprovide useful in de novo protein design: it could beused as a negative design rule to terminate coiled-coil motifs, and as a positive design rule forintroducing kinks and hinges into novel coiledcoil-based nanostructures and biomaterials.31

‡http://www.biols.susx.ac.uk/coiledcoils

Results

A unique pattern of charged residues at the coresites of MHC coiled coils

While working with the long, parallel homo-dimeric coiled-coil rods of MHC subtypes, wenoticed that there were virtually no Asp at eitherthe a or d core sites, and no Glu at the a sites, whilepositively charged residues (Lys and Arg) could befound at both core positions. McLachlan & Karndescribed the sequence of the nematodeMHC rod ashighly repetitive with features typical of an α-helicalcoiled coil.32 They note that Asp and Glu areexcluded from the a positions of the nematodeMHC rod, and are rare at d. Since then, many moreMHC sequences have been determined. From these,we selected 14 rod sequences to represent thedifferent subtypes of MHC from different speciesand, within these we assigned the coiled-coil regions(Supplementary Data, Table S1). Despite a total of15,489 coiled-coil residues with more than 2000 asites, there were no negatively charged amino acid

residues in this position (p<0.001) and there was avirtual absence of Asp from the d position (p<0.001).In contrast, Glu was slightly favored at the dposition (p<0.001). To avoid bias resulting fromincluding more than one representative structure ofeach MHC subtypes, only the five known humanMHC subtypes were used in the analysis presentedin Table 1A and Supplementary Data, Table S2.Consistent with foregoing analyses, the positivelycharged amino acids, Arg and Lys, are favoured ata and disfavored at d (p<0.001).10,33,34

Aspartic acids are excluded from the core sitesof all parallel coiled coils

To test if the above observation for MHCsequences extended to coiled coils in general, weturned to the sequences of structurally verifiedcoiled coils lodged in the SOCKET-derived CC+database‡. Unlike sequence-based predictors ofcoiled-coil motifs3,8–10 which search solely for se-quence features to predict coiled-coil motifs,8,11 theSOCKET algorithm uses the characteristic knobs-into-holes side-chain packing of coiled coils toidentify coiled-coil motifs in the RCSB ProteinData Bank. Table 1B gives data culled from thecoiled-coil structures in the currently available ver-sion of the CC+ database. The data in Table 1B andin Supplementary Data, Tables S3–S4 show thatindeed there are no Asp residues in the core sites ofall identified parallel coiled coils (p<0.001) whileGlu can be found at both core sites, though it ismuch more abundant in the d position (p<0.001).Indeed, the only incidence of a parallel coiled-coilwith Glu at an a position (out of 205 verified paralleldimeric, trimeric and tetrameric coiled coils) was thec-myc-max heterodimer35 (PDB identifier 1a93). Theantiparallel structures were generally more permis-

http://www.biols.susx.ac.uk/coiledcoils


sive in their use of these side-chains at both positionsbut here again, in the a positions Asp or Glu wereless abundant compared to the d positions(p<0.001). Like MHC subtypes, the positivelycharged amino acids, Lys and Arg, occur morefrequently than Asp and Glu in the cores of coiledcoils as commented (p<0.001).10,21,25,33,34 However,there are no strong preferences for Lys or Arg atthese positions in any of the structures (Table 1B).Thus, the high incidences of these residues at thepredicted a sites of the MHC sequences is possibly aunique and another interesting feature of theseproteins (Table 1A).

Introducing aspartic acid to a positions of theMHC rod causes local kinking

We then used the human non-muscle MHC (NM-MHC) to test the effect of inserting a negativelycharged amino acid at the a sites of a parallel homo-dimeric coiled coil. An expression plasmid wasconstructed to yield a protein similar to the C-ter-minal 640 residues of the NM-MHC, with anadditional methionine and alanine residues at theN terminus (NM-MHC-642). Similar NM-MHCconstructs have been shown to form coiled-coilfilaments similar to wild-type NM-MHC.28,30 One,two or four a positions along the rod were replacedby Asp or Glu. The same positions were alsoconverted to Leu, which is the most abundantamino acid at the a sites of coiled coils. The resulting

Figure 2. Rotary shadowing of the wild-type myosin II anshow a clear kink in the rod. The bar represents 100 nM.

proteins were named according to the site of themutations, counted as the number of amino acidsfrom the N-terminal end of the construct. To verifythat the expressed proteins formed rods, molecular-length measurements of rotary shadowed prepara-tions were made taking full-length chicken-gizzardsmooth-muscle myosin as a standard (Figures 2 and3). All of the mutants formed rods of the expectedlength (Table 2, and data not shown). Moreover,there were clear kinks in the Asp-mutated rods butnot in theGlu or Leu-mutated rods (Figures 2–4). Thelow baseline level of kinks observed in all mutantsshown in Figure 4 is probably due to a technicalartifact resulting from the force applied on the rodswhen spreading them on the grids. Indeed, most ofthe kink sites in the Asp-mutated rods correlatedwell with the position of the mutations (Table 2)while the rare kinks in all other rodswere distributedalong the rod (data not shown). Ranges of kinkangles were observed for each Asp mutation (Figure2), with about 20% of the rods showing a maximalkink. It is evident that the maximal kink is related tothe number of mutations in each protein. Since thetwo mutations are only one heptad apart, twoseparate kinks cannot be resolved (Figure 3).

Kinks in the coiled-coil interfere with itsassembly properties

The coiled-coil structure is well known as one ofthe most common motifs for protein–protein inter-

d some of the MHC mutants. Note that only Asp mutants

Figure 3. Representation ofthe maximal kink angles in theAsp-MHC mutants. One, two orfour a positions along the parentrod (NM-MHC-642) were selectedfor replacement by Asp. Usingrotary shadowing, a kink at aposition that correlates to themutation site (Table 2) can beclearly seen. Ranges of kink angleswere observed for each mutation,with only about 20% of the rodsshowing the maximal kink angle.Only rods having the maximalkink angle were selected for theFigure. The rods were selectedfrom 10–15 images as thoseshown in Figure 2. The bar repre-sents 100 nM.


actions.36 As Asp in the a position causes a localkink in the coiled coil, it might interfere badly withits assembly properties. To check if this is indeed thecase we took advantage of the fact that our modelprotein, myosin, can form highly ordered filamentsthrough interactions between its coiled-coil motifsin vivo37 and when dialyzed against low ionicstrength buffer in vitro.29,30,38 We have thus carriedout a sedimentation assay in which monomericmyosin II coiled coils in high ionic strength bufferare dialyzed against a range of salt concentrationsand then subjected to centrifugation: monomersremain in the supernatant, while multimeric struc-tures pellet and samples of both fractions wereevaluated by SDS–PAGE and densitometry todetermine the percentage of soluble myosin. Figure5 shows the sedimentation curves of the parent rodand the mutant myosin II rods in positions 453 and

460. The parent rod NM-MHC-642 and the Leumutant were completely insoluble below 200 mMsalt, and became more soluble at salt concentrationsup to 300 mM, at which point they were almostcompletely in the soluble fraction. In contrast, whena negatively charged amino acid was introduced tothese a positions the mutated myosin II rodsbecame soluble at much lower salt concentrations,indicating that these mutations indeed interferedbadly with the assembly properties of the mutatedcoiled coil.Another way to evaluate the ability of the mono-

meric coiled coils to form ordered filaments is todialyze them against low salt buffer and examine theresulting samples by negative-staining electron mi-croscopy (see Figure 6 for representative images). Theparent rod, NM-MHC-642, formed long, orderedfilaments with the normal striation pattern typical

Table 2. Measured total lengths and kink sites of rotaryshadowed proteins

Total length ina.a.⁎ (±1 SD)

Kink site ina.a.⁎ (±1 SD)

Smooth muscle chickengizzard rod (1130 a.a.)

1148.9±62.8

NM-MHC-642 641.1±46.9NM-MHC-642 K(200)D 643.6±47.9 202.3±27.5NM-MHC-642 F,M(235,242)D 651.4±30.7 237±26.9NM-MHC-642 K(390,397)D 634.1±46 409.3±25.5NM-MHC-642 R,N(453,460)D 635.6±56.9 466.9±19.1NM-MHC-642 K,K,R,N

(390,397,453,460)D638.8±47.2 436.5±29.6

The measured distances were converted to a.a.⁎ by the formula:1a.a.⁎=0.14 nm. SD, standard deviation.

Figure 5. Solubility of the purified MHC rod proteinsas a function of salt concentration. MHC proteins(0.5 mg/ml) were dialyzed against a range of saltconcentrations (50 mM–300 mM NaCl) in 10 mM phos-phate buffer (pH 7.5) and 2 mM MgCl2 for 4 h at 4 °C andassayed for solubility by centrifugation of 100,000g for 1 h.The supernatant and pellet fractions were dissolved onSDS–PAGE and the percentage of soluble MHC wascalculated. It is evident that Asp and Glu mutants sufferfor some degree of filament destabilization and thusbecome soluble at lower ionic strength while Leu mutantshave a solubility curve similar to the wild-type rod.


for myosin filaments29,30,38 (Figure 6(a) and (f)). Incontrast, all of the Asp mutated myosin II rods hadaberrant morphologies. It is also interesting to notethat in contrast to the parent monomeric coiled coils,which all assembled into filaments (Figure 6(a)),only fractions of the mutant coiled coils partici-pated in filament formation and the remainderformed non-specific aggregates (Figure 6(b)–(e)),indicating their inability to foster normal protein–protein interactions.

Discussion

In summary, by examining the sequences of allcoiled coils identified with the SOCKET algorithmfrom the RCSB Protein Data Bank, we found thatAsp is completely excluded from the a and d coresites of parallel coiled coils in general, while Glu canappear in both sites. We also show that theantiparallel coiled-coil structures are more permis-

Figure 4. The percent of kinked MHC rods in thedifferent mutants. Sixty-six to 252 (average–145) MHCrods from rotary shadowing images were counted andeach MHC rod was marked as straight or kinked. As canbe seen, only Asp mutants have a high percentage ofkinked rods that is above the noise base line. 1, NM-MHC-642; 2, 3 and 4, NM-MHC-642 M,F(235,242) to D,E,L respectively; 5, 6 and 7, NM-MHC-642 K(390,397) to D,E,L respectively; 8, 9 and 10, NM-MHC-642 R,N(453,460)to D,E,L respectively; 11, NM-MHC-642 K,K,R,N(390,397,453,460)D; 12, NM-MHC-642 K(200)D.

sive in their amino acid selection in both core sitesand that the d position is more permissive tonegatively charged amino acids than the a position.To test the effect of introducing a negatively chargedresidue into the a core site of a parallel coiled coil,we made a series of mutated MHCs. The mutantMHCs had one to four mutations introduced at the asites along the rod at which the original amino acidresidues were replaced with Asp, Glu or Leu. Thewild-type and mutant proteins were assessed forfolding and assembly using a sedimentation assayand imaging by electron microscopy. We show thatall the Asp-mutated rods had clear kinks around thepoints of mutation, while the Glu and Leu mutantshad no such kinks. We also show that the assemblyproperties of the Asp/Glu-mutated rods wereimpaired compared to the wild-type and Leumutants, and that the Asp mutants exhibit assemblyproperties that were more impaired than the Glumutants.

Positively versus negatively charged aminoacids in the core positions

The hydrophobic interactions between non-polarresidues are thought to be among the mostimportant in determining the three-dimensionalstructures of proteins.16,39 Indeed, the importanceof the hydrophobic nature of the core positions tothe folding and stability of the coiled coil is welldocumented.12–19 Thus, all polar and chargedresidues, in general terms, would not be consideredas good candidates for the a and d sites. This is ageneralization, however, and does not apply equallyto all polar residues. For instance, Asn is actuallyfavoured at certain a sites of parallel dimeric coiledcoils such as the leucine zipper.20,21 The inclusion of

Figure 6. Negative staining of the different MHC rod proteins. MHC proteins, in high salt buffer, were dialyzedagainst low salt buffer for 16 h and then negatively stained with uranyl acetate and examined on electron microscope (seeMaterials and Methods). (a)–(e) Magnification 8800x, the bar represents 1 μm; (f)–(j) Magnification 88,000x, the barrepresents 100 nm. The parent rod, NM-MHC-642, formed long, ordered filaments with the expected striation pattern of14.5 nm ((a) and (f)). NM-MHC-642 K(200)D formed very short filaments ((b) and (g)). NM-MHC-642 K(390,397)D formedvery long filaments with long external grooves ((c) and (h)); NM-MHC-642 R,N(453,460)D filaments had manyprotrusions extending from the filament borders ((d) and (i)); and NM-MHC-642 K,K,R,N(390,397,453,460)D filamentshad thin and uneven width ((e) and (j)).


this residue is well understood: two Asn residuesat the same a layer in a coiled-coil dimer can forma side-chain–side-chain hydrogen bond that speci-fies coiled-coil oligomer state, orientation andregister, as this specific interaction is only possiblein in-register, parallel dimers. Returning to Aspand Glu versus Lys and Arg: firstly, these chargedside-chains are not accommodated within thehydrophobic cores of parallel coiled coils as readilyas Asn, and they must escape the core to have theirterminal charged groups solvated. In this respectLys and Arg side-chains are amphipathic withaliphatic alkyl chains terminating with positivelycharged amino and guanidino groups, respectively.The aliphatic component may add to the hydro-phobic interface between the two α-helical chains,or act as a spacer to help remove the chargedtermini to solvent21. In contrast, the side-chains ofAsp and Glu are shorter, and might be expected tobe less tolerated at core sites.32,33 Indeed, Hodgesand colleagues showed that introducing a nega-tively charged amino acid at these positions of a denovo designed coiled-coil peptide drastically desta-bilizes its structure.18,40,41

Aspartic versus glutamic acid in the corepositions

As shown in Table 1 and Figures 2, 3 and 4, Asp isby far less tolerated in the core positions ascompared to Glu. This is probably due to the factthat Asp has the shortest side-chain of all of thepolar charged residues.33 Furthermore, Asp, like the

other short polar side-chains, Asn, Ser and Thr, isconsidered an α-helix breaker. This is related to theability of their side-chains to interact with the main-chain and compete with the normal α-helicalhydrogen-bonding pattern.42 These two points,namely, the requirement to solvate the short side-chain of Asp, and the ability of the residue to formside-chain–main-chain hydrogen bonds, are thelikely causes of the exclusion of Asp from thehydrophobic core sites of parallel coiled coils (Table1B), and the sharp kinking in the rods that weobserve in the Asp-containing myosin mutants(Figures 2–4).Note, that as the crystallographic structures of the

MHC rods have not been determined, we cannot besure that the two rare incidences of Asp in positiond of the MHC rods (Table 1A; NM-MHC-B, heptad78 and skeletal MHC, heptad 146) do not causesome local disruption of the coiled-coil structure.

The a versus d core positions

As shown, side-chains in the a position show astructural geometry that is different from that ofside-chains at d.22 For two stranded coiled-coilsthese structural geometries were described asparallel and perpendicular, respectively. This is thebasis for the work of Harbury and colleaguesregarding the influence of these different packingangles on the amino acid selection at the coresites.11,23 Indeed, even though both a and d sites liein the hydrophobic core of the coiled coil and bothare considered to be mainly hydrophobic, there is a


clear difference in the amino acid selection betweenthese two positions (Table 1).10,11,33,43

A computer model for the packing of side-chainsin the inner core of coiled coils has shown that Gluat the d position can change its conformation bychanges in the plane of the COOH group and thuscontact different partners, and a hydrogen bondmay form between the COOH group and a carbonylgroup from the main-chain of the other strand.43 Asproposed,33 based on sequence analysis of repre-sentative coiled coils, the virtual absence of Glufrom the a positions suggests that the local structureof the coiled coil around this position does not infact allow Glu to be solvated, whereas in the dposition this apparently does become possible. Sincethis analysis was based on the sequences of onlyfour representative coiled coils, all of them ofhaving parallel conformation, we can now statemore accurately that this rule does not generallyapply to antiparallel coiled coils (Table 1B). Indeed,it has been shown that the parallel structures showgreater discrimination of amino acid residues be-tween the two core sites as compared to antiparallelstructures.11

Aspartic acid in an a position kinks the coiledcoil and affects filament assembly

Introducing Asp into several a positions along theMHC coiled coil caused a clear kink at the mutationsite (Figures 2–4; Table 2). The maximal angle of thekink was directly related to the number of mutationsthat were introduced (Figure 3). However, there wasa great variation of the kink angles between rodswith the same mutation, with only about 20% of therods showing the maximal kink angle. We suggestthat the mutations cause local disruptions of thecoiled coils, which then serve as flexible linkersbetween the coiled-coil regions on either side. Thenumber of these local disruptions probably deter-mines the maximal kink angle. To our knowledgethis is the first report of such an effect, resultingsimply from introducing Asp into a positions.The existence of a kink at the sites of mutation and

the maximal kink angle were not affected by thesequence of the amino acids flanking the differentmutation sites. Thus, creating an Asp-in-a-positionsituation is probably enough to kink a coiled-coilstructure, regardless of the sequence of amino acidsflanking the mutation site.Since the Asp mutations described above affected

the MHC rod structure, it was not surprising thatthey also affect its ability to from filaments as well.All of the Asp-mutated MHC form filamentsdifferent from the MHC wild-type filaments (Figure6, and data not shown) and they all became solublein lower salt concentrations compared to wild-typeMHC (Figure 5, and data not shown), indicating thatthe inter-coiled coil interactions were also badlyaffected by these mutations. As Glu at the a sitebadly affects the inter-coiled coils interactions aswell, it was not surprising to find that all the Glumutants badly affected filament formation, but to a

smaller extent compared to the Asp mutants (Figure5, and data not shown).

Summary and conclusions

The absence of Asp and Glu from the a positionsof four representative coiled coils has been com-mented before33 as well as the fact that introducing anegatively charged amino acid at the core positionsof a de novo designed coiled-coil peptide destabilizesits structure.18,40,41 Here we show for the first time aclear difference between Asp and Glu: only Asp isincompatible with the coiled-coil structure whenintroduced to the a position of a homodimer.Indeed, we have shown that Asp, and not Glu, isabsent from the core positions of all parallel coiledcoils gathered from the Protein Data Bank using theSOCKET algorithm; while Glu can be found at thesepositions. We also show for the first time thatintroducing Asp, but not Glu, into the a position of aparallel coiled coil, leads to the formation of kinks inthe coiled-coil structure, and that the maximal kinkangle is directly related to the number of mutationsthat were introduced. Furthermore, we demonstratethat the introduction of a kink along the coiled coilinterferes with its ability to play its critical role infostering protein–protein interactions and forminghigher-ordered structures.These new observations should help achieve

better predictions of coiled-coil motifs from se-quence analysis: for instance, in helping to delimitcoiled-coil regions and possibly even to help resolveconflicts in oligomer-state and topology predictionsof coiled-coil structure. They could also be used as astrong negative-design rule when planning de novocoiled-coil designs. We also believe that the de-scribed flexible hinge/kink in the coiled-coil can beused as a positive design rule. Considerable effort inprotein engineering has been devoted to introducingvariations into otherwise natural protein sequencesto achieve new or altered functions or properties.44,45

Several research groups have begun to take proteindesign to the next level of organization by designingself-assembling systems. Most of these studies havefocused on particularly simple protein-folding pat-terns, especially, some simple coiled-coil arrange-ments. Recently, new strategies for designing morecomplex protein assemblies, such as filaments, cagesand extended ordered arrays have been develo-ped.24,31,44 The strategic introduction of Asp at an asite of an otherwise canonical coiled coil thatassembled to a fibrous biomaterial could be used tokink the structures.

Materials and Methods

Myosin heavy chain sequences

We have selected 14 MHC rods (15,489 amino acidresidue(s) (a.a.)) that represent the different subtypes ofMHC in different species and assigned them to theircorrect heptad register (Supplementary Data, Table S1).


The first and last heptads of the rod region were notincluded in the database because of uncertainties as to thecoiled-coil boundaries.8 The ten residues before and aftereach skip residue were not included because of uncertain-ties as to where the skips were;8 that is, the entire helicalsequence of the coil can be mapped out as a cyclic patternof 28 amino acid residues that is interrupted at three orfour positions by the insertion of one extra residue that istermed a “skip residue”. To reduce bias resulting fromincluding more then one representative structure of eachMHC subtypes, only the five human MHC subtypes wereused for the statistical analysis (Table 1A). We compiled atable for each myosin rod that represents the number oftimes that each amino acid is found at each of the heptadpositions. Then we summed all five tables and compiled atable that represents the frequency of each of the 20 aminoacid residues at each heptad position (20/7 frequencytable). The resulting profiles were then normalized bydividing that frequency by the relative frequency of eachamino acid in SWISSPROT46 (Table 1A; SupplementaryData, Table S2).

SOCKET derived coiled-coil database

The SOCKET database11 contains all 561 structures thatwere identified by SOCKET to have a coiled-coil regionout of 9255 structures in release 87 of the RCSB ProteinData Bank. To reduce bias resulting from duplication ofproteins or their homologues in the RCSB Protein DataBank, the positive coiled-coil structures were grouped into92 sequence-based homologous families and only repre-sentative structures of each family were included in thestatistical analysis. We created an Excel macro programthat uses the database to compile a table that representsthe frequency of each of the 20 amino acids at each heptadposition (20/7 frequency table) of parallel and antiparallelcoiled-coils and of parallel homo-oligomeric coiled-coils.The resulting profile was then normalized by dividingthat frequency by the relative frequency of each aminoacid in SWISSPROT46 (Table 1B; Supplementary Data,Tables S3–4).

Cloning and site directed mutagenesis of NM-MHC-B642 a.a. proteins

To create the NM-MHC-642 protein we used thedescribed47 pET21-Rod vector that contains the C-termi-nal 657 a.a. of human NM-MHC-B in pET21c. We inserteda stop codon just after the last NM-MHC-B amino acidusing the primers 5′-GAT CCC GGG CCC GCG TTA CTCTGA CTG GGG and 5′-CCC CAG TCA GAG TAA CGCGGG CCC GGG ATC. All site-directed mutageneses weredone with ExSite™ PCR-Based Site-Directed MutagenesisKit (200502; Stratagene). We then cut the rod sequenceusing BamHI-BamHI restriction enzyme and prepared anempty pET21c vector by cutting it with NheI-BamHIrestriction enzymes. Both insert and vector were treatedwith Klenow polymerase to fill-in the overhang arms andthe insert was then sub-cloned into the empty pET21cvector. The resulting expression vector was fully se-quenced and was shown to code for a protein withmethionine and alanine at the 5′ and then the C-terminal640 a.a. of human non-muscle MHC-B (accession no.A59252). We then used the site-directed mutagenesis kit tochange several amino acids at an a position into Asp, Gluor Leu,. Here are the primers that were used to create theAsp mutants (mutagenesis sites are underlined): For NM-MHC-642 K(200)D: 5′-C GAA CTT GAA AAATCC GAC

CGG GCC CTA GAG CAG CAG GTG G and 5′-C CACCTG CTG CTC TAG GGC CCG GTC GGATTT TTC AAGTTCG. ForNM-MHC-642R,N(453,460)D: 5′-GTGACAATGCA GAC CAG CAA CTG GAG CGG CAG GAC AAGGAG CTG and CAG CTC CTT GTC CTG CCG CTC CAGTTG CTG GTC TGC ATT GTC AC. For NM-MHC-642 K(390,397)D: 5′- GC GCC TCT GGC GAC TCC GCG CTGCTG GAT GAG GAC CGG CGT CTG G and 5′- C CAGACGCCGGTCCTCATCCAGCAGCGCGGAGTCGCCAGA GGC GC. NM-MHC-642 M,F(235,242)D: 5′ G GAGGTC AAC GAC CAG GCC ATG AAG GCG CAG GACGAG AGA GAC C and 5′ – G GTC TCT CTC GTC CTGCGC CTT CAT GGC CTG GTC GTT GAC CTC C.

Bacterial expression and purification of NM-MHC-B642 a.a. proteins

Expression and purification of myosin proteins wascarried out essentially as described48 with only a fewchanges. Bacteria containing the expressed vectors weregrown at 37 °C in 500 ml of 2xYT+Amp to an A595 nm of0.5 , then IPTG was added to 0.5 mM and the cells weregrown for additional 2 h. The bacteria were pelleted,washed with cold phosphate-buffered saline (PBS),pelleted again and frozen at −20 °C overnight Thebacterial pellet was suspended in 15 ml of lysis buffer B(50 mM Tris–HCl (pH 7.5), 0.8 M NaCl, 10 mM EDTA,10 mM EGTA, 1 mM DTT, 1 mg/ml of lysozyme, 0.5 mMPMSF), lysed using sonication and the cell debris removedby centrifugation. Polyethylenimine (PEI) (250 μl) (Sigma,P-3143) was slowly added from a 10% stock solution inorder to clear the solution from all nucleic acids.49 After10min of stirring on ice the PEI–nucleic acids complex wasprecipitated with centrifugation (Sorvall SS-34 rotor,5000 rpm at 4 °C for 10 min). The remaining proteinsolution was boiled for 15 min and cleared with cen-trifugation (Beckman Ti-60 rotor, 45,000 rpm at 4 °C for1 h). To get rid of any trace of PEI, the purifiedmyosin rodswere precipitated twice with ammonium sulphate (1.5 mlsaturated ammonium sulphate/ml myosin solution),suspended with 8 ml of buffer G (20 mM Tris–HCl (pH7.5), 600 mMNaCl, 5 mMEDTA, 1mMDTT) and dialyzedovernight against buffer G. The purified myosin rods werestored at −20 °C at concentrations of 0.5–4 mg/ml. Theconcentrations of the myosin proteins were determinedby comparing the densitometry of the band of a sampleof each protein on Coomassie-stained PAGE to a knownseries of bovine serum albumin (BSA) samples on the samegel.

Sedimentation analysis

Samples of the expressed myosin rod proteins werediluted with buffer G to 0.5 mg/ml. A 100 μl of the dilutedsamples of each protein were dialyzed in small dialysistubes (mini GeBAflex tubes; catalog no. DO70-6-30)against a range of salt concentrations (50 mM–300 mMNaCl) in 10 mM phosphate buffer (pH 7.5) and 2 mMMgCl2 for 4 h at 4 °C. Then 80 μl of each dialyzed samplewere transferred to centrifugation tubes and centrifuged at100,000g for 1 h (Beckman TLA-55 rotor, 45,000 rpm at,4 °C). The top 50% of each supernatant was taken foranalysis by SDS–PAGE. The reminder of the supernatantwas removed from the tube and the pellet was dissolvedin 360 μl of buffer G. Samples of the dissolved pellet weretaken for SDS–PAGE. The solubility of the myosin rodproteins (expressed as percentage of the fragment in thesupernatant) were determined by measuring the relative


amount of each fragment in the pellet and in thesupernatant fractions by densitometry of the Coomasie-stained PAGE bands using Fujifilm image analyser LAS-1000plus and the densitometry program Fujifilm ImageGauge version 3.4.

Rotary shadowing

Samples were diluted to 50 μg/ml in 600 mM ammo-nium acetate, 50% (v/v) glycerol (pH 7.0) and sprayedonto freshly cleaved mica, then rotary shadowed withplatinum at a 6° angle and the replicas picked up on 400mesh hex grids. The replicas were imaged on a Philips CM10 transmission electron microscope using an acceleratingvoltage of 80 kV.50

Negative staining

Expressed myosin rods in buffer G were diluted withbuffer G to 0.5 mg/ml and then dialyzed against buffer Q(10 mM Tris–HCl (pH 7.5), 50 mM NaCl, 20 mM MgCl2,25 mM CaCl2) for 16 h at 4 °C in small dialysis tubes (miniGeBAflex tubes; catalog no. DO70-6-30). The 7 μl sampleswere applied onto formvar-carbon coated 400 mesh grids(SPI Supplies, Westchester, PA, USA; 3440c) and left for2 min and then negatively stained with 2% (w/v) uranylacetate for 15 s. The negatively stained samples wereexamined on a transmission electron microscope (PhilipsCM12). Striation periods were measured at X88K magni-fication using Fujifilm Image Gauge version 3.4.

Acknowledgements

We thank Dr Ezra Rachamin (Hebrew University,Faculty of Medicine), Mr Oliver Testa and Dr JohnWalshaw for advice on using the CC+ database, andDr Howard Cedar for critical reading of the manu-script. We also thank S.R. laboratory members fortheir support and assistance.

Supplementary Data

Supplementary data associated with this articlecan be found, in the online version, at doi:10.1016/j.jmb.2006.11.083

References

1. Crick, F. H. C. (1953). The packing of alpha-helices:simple coiled-coils. Acta Crystallog. 6, 689–697.

2. Lupas, A. N. & Gruber, M. (2005). The structure ofalpha-helical coiled coils. Adv. Protein Chem. 70,37–78.

3. Wolf, E., Kim, P. S. & Berger, B. (1997). MultiCoil: aprogram for predicting two- and three-stranded coiledcoils. Protein Sci. 6, 1179–1189.

4. Brown, J. H., Cohen, C. & Parry, D. A. (1996). Heptadbreaks in alpha-helical coiled coils: stutters andstammers. Proteins: Struct.Funct. Genet. 26, 134–145.

5. Hicks, M. R., Holberton, D. V., Kowalczyk, C. &Woolfson, D. N. (1997). Coiled-coil assembly by

peptides with non-heptad sequence motifs. Fold. Des.2, 149–158.

6. Lupas, A. (1996). Prediction and analysis of coiled-coilstructures. Methods Enzy. 266, 513–525.

7. McLachlan, A. D. & Stewart, M. (1975). Tropomyosincoiled-coil interactions: evidence for an unstaggeredstructure. J. Mol. Biol. 98, 293–304.

8. Berger, B., Wilson, D. B., Wolf, E., Tonchev, T., Milla,M. & Kim, P. S. (1995). Predicting coiled coils by use ofpairwise residue correlations. Proc. Natl Acad. Sci.USA, 92, 8259–8263.

9. Lupas, A., Van Dyke, M. & Stock, J. (1991). Predictingcoiled coils from protein sequences. Science, 252,1162–1164.

10. Woolfson, D. N. & Alber, T. (1995). Predictingoligomerization states of coiled coils. Protein Sci. 4,1596–1607.

11. Walshaw, J. & Woolfson, D. N. (2001). Socket: aprogram for identifying and analysing coiled-coilmotifs within protein structures. J. Mol. Biol. 307,1427–1450.

12. Hodges, R. S., Saund, A. K., Chong, P. C., St-Pierre,S. A. & Reid, R. E. (1981). Synthetic model for two-stranded alpha-helical coiled-coils. Design, synthesis,and characterization of an 86-residue analog oftropomyosin. J. Biol. Chem. 256, 1214–1224.

13. Hodges, R. S., Zhou, N. E., Kay, C. M. & Semchuk,P. D. (1990). Synthetic model proteins: contributionof hydrophobic residues and disulfide bonds toprotein stability. Pept. Res. 3, 123–137.

14. Hu, J. C., O’Shea, E. K., Kim, P. S. & Sauer, R. T. (1990).Sequence requirements for coiled-coils: analysis withlambda repressor-GCN4 leucine zipper fusions. Sci-ence, 250, 1400–1403.

15. Zhu, B. Y., Zhou, N. E., Kay, C. M. & Hodges, R. S.(1993). Packing and hydrophobicity effects on proteinfolding and stability: effects of beta-branched aminoacids, valine and isoleucine, on the formation andstability of two-stranded alpha-helical coiled coils/leucine zippers. Protein Sci. 2, 383–394.

16. Zhou, N. E., Kay, C. M. & Hodges, R. S. (1992).Synthetic model proteins. Positional effects of inter-chain hydrophobic interactions on stability of two-stranded alpha-helical coiled-coils. J. Biol. Chem. 267,2664–2670.

17. Zhou, N. E., Kay, C. M. & Hodges, R. S. (1992).Synthetic model proteins: the relative contribution ofleucine residues at the non-equivalent positions of the3-4 hydrophobic repeat to the stability of the two-stranded alpha-helical coiled-coil. Biochemistry, 31,5739–5746.

18. Wagschal, K., Tripet, B., Lavigne, P., Mant, C. &Hodges, R. S. (1999). The role of position a indetermining the stability and oligomerization stateof alpha-helical coiled coils: 20 amino acid stabilitycoefficients in the hydrophobic core of proteins.Protein Sci. 8, 2312–2329.

19. Akey, D. L., Malashkevich, V. N. & Kim, P. S. (2001).Buried polar residues in coiled-coil interfaces. Bio-chemistry, 40, 6352–6360.

20. Lumb, K. J. & Kim, P. S. (1995). A buried polarinteraction imparts structural uniqueness in adesigned heterodimeric coiled coil. Biochemistry, 34,8642–8648.

21. Gonzalez, L., Jr, Woolfson, D. N. & Alber, T. (1996).Buried polar residues and structural specificity in theGCN4 leucine zipper. Nature Struct. Biol. 3, 1011–1018.

22. O’Shea, E. K., Klemm, J. D., Kim, P. S. & Alber, T.(1991). X-ray structure of the GCN4 leucine zipper,


a two-stranded, parallel coiled coil. Science, 254,539–544.

23. Harbury, P. B., Zhang, T., Kim, P. S. & Alber, T. (1993).A switch between two-, three-, and four-strandedcoiled coils in GCN4 leucine zipper mutants. Science,262, 1401–1407.

24. Woolfson, D. N. (2001). Core-directed protein design.Curr. Opin. Struct. Biol. 11, 464–471.

25. Lumb, K. J. & Kim, P. S. (1998). A buried polarinteraction imparts structural uniqueness in a de-signed heterodimeric coiled coil. Biochemistry, 37;13042–13042.

26. Sellers, J. R. (2000). Myosins: a diverse superfamily.Biochim. Biophys. Acta, 1496, 3–22.

27. Cripps, R. M., Suggs, J. A. & Bernstein, S. I. (1999).Assembly of thick filaments and myofibrils occurs inthe absence of the myosin head. EMBO J. 18, 1793–1804.

28. Hodge, T. P., Cross, R. & Kendrick-Jones, J. (1992).Role of the COOH-terminal nonhelical tailpiece inthe assembly of a vertebrate nonmuscle myosin rod.J. Cell Biol. 118, 1085–1095.

29. Ikebe, M., Komatsu, S., Woodhead, J. L., Mabuchi, K.,Ikebe, R., Saito, J. et al. (2001). The tip of the coiled-coilrod determines the filament formation of smoothmuscle and nonmuscle myosin. J. Biol. Chem. 276,30293–30300.

30. Sohn, R. L., Vikstrom, K. L., Strauss, M., Cohen, C.,Szent-Gyorgyi, A. G. & Leinwand, L. A. (1997). A 29residue region of the sarcomeric myosin rod isnecessary for filament formation. J. Mol. Biol. 266,317–330.

31. Ryadnov, M. G. &Woolfson, D. N. (2003). Introducingbranches into a self-assembling peptide fiber. AngewChem. Int. Ed. Engl. 42, 3021–3023.

32. McLachlan, A. D. & Karn, J. (1982). Periodic chargedistributions in the myosin rod amino acid sequencematch cross-bridge spacings in muscle. Nature, 299,226–231.

33. Parry, D. A. (1982). Coiled-coils in alpha-helix-containing proteins: analysis of the residue typeswithin the heptad repeat and the use of these data inthe prediction of coiled-coils in other proteins. Biosci.Rep. 2, 1017–1024.

34. Conway, J. F. & Parry, D. A. (1990). Structural featuresin the heptad substructure and longer range repeats oftwo-stranded alpha-fibrous proteins. Int. J. Biol. Macro-mol. 12, 328–334.

35. Lavigne, P., Crump, M. P., Gagne, S. M., Hodges, R. S.,Kay, C. M. & Sykes, B. D. (1998). Insights into themechanism of heterodimerization from the 1H-NMRsolution structure of the c-Myc-Max heterodimericleucine zipper. J. Mol. Biol. 281, 165–181.

36. Cohen, C. & Parry, D. A. (1990). Alpha-helical coiled

coils and bundles: how to design an alpha-helicalprotein. Proteins: Struct. Funct. Genet. 7, 1–15.

37. Shoffner, J. D. & De Lozanne, A. (1996). Sequences inthe myosin II tail required for self-association.Biochem. Biophys. Res. Commun. 218, 860–864.

38. Philpott, D. E. & Szent-Gyorgyi, A. G. (1954). Thestructure of light-meromyosin: an electron microscop-ic study. Biochim. Biophys. Acta, 15, 165–173.

39. Kauzmann, W. (1959). Some factors in the interpreta-tion of protein denaturation. Adv. Protein Chem. 14,1–63.

40. Wagschal, K., Tripet, B. & Hodges, R. S. (1999). Denovo design of a model peptide sequence to examinethe effects of single amino acid substitutions in thehydrophobic core on both stability and oligomeriza-tion state of coiled-coils. J. Mol. Biol. 285, 785–803.

41. Tripet, B., Wagschal, K., Lavigne, P., Mant, C. T. &Hodges, R. S. (2000). Effects of side-chain character-istics on stability and oligomerization state of a denovo-designed model coiled-coil: 20 amino acidsubstitutions in position “d”. J. Mol. Biol. 300, 377–402.

42. Vijayakumar, M., Qian, H. & Zhou, H. X. (1999).Hydrogen bonds between short polar side chains andpeptide backbone: prevalence in proteins and effectson helix-forming propensities. Proteins: Struct. Funct.Genet. 34, 497–507.

43. Offer, G. & Sessions, R. (1995). Computer modelling ofthe alpha-helical coiled coil: packing of side-chains inthe inner core. J. Mol. Biol. 249, 967–987.

44. Fairman, R. & Akerfeldt, K. S. (2005). Peptides asnovel smart materials. Curr. Opin. Struct. Biol. 15,453–463.

45. Yeates, T. O. & Padilla, J. E. (2002). Designingsupramolecular protein assemblies. Curr. Opin. Struct.Biol. 12, 464–470.

46. Bairoch, A. & Apweiler, R. (2000). The SWISS-PROTprotein sequence database and its supplementTrEMBL in 2000. Nucl. Acids Res. 28, 45–48.

47. Ben-Ya’acov, A. & Ravid, S. (2003). Epidermal growthfactor-mediated transient phosphorylation and mem-brane localization of myosin II-B are required forefficient chemotaxis. J. Biol. Chem. 278, 40032–40040.

48. McNally, E., Sohn, R., Frankel, S. & Leinwand, L.(1991). Expression of myosin and actin in Escherichiacoli. Methods Enzymol. 196, 368–389.

49. Burgess, R. R. (1991). Use of polyethyleneimine inpurification of DNA-binding proteins. Methods Enzy-mol. 208, 3–10.

50. Stafford, W. F., Jacobsen, M. P., Woodhead, J.,Craig, R., O’Neall-Hennessey, E. & Szent-Gyorgyi,A. G. (2001). Calcium-dependent structural changesin scallop heavy meromyosin. J. Mol. Biol. 307,137–147.

Edited by M. Moody

(Received 5 September 2006; received in revised form 26 November 2006; accepted 29 November 2006)
Available online 2 December 2006

kinking the coiled coil negatively charged residues at the

Documents