history

nature structural biology • volume 8 number 3 • march 2001 201

Every first year biochem-istry student learns aboutthe planarity of the pep-tide bond. Planarity, theresult of ∼ 40% N–C′ dou-ble bond character arisingfrom two dominant reso-nance structures, allowsfor a great simplificationin the understanding ofprotein structures. Of thethree repeating proteinbackbone dihedral angles,only φ and ψ need to beconsidered when assign-ing secondary structure,because the peptide bonddihedral angle ω is gener-ally considered fixed at180°. Linus Pauling’s pre-diction of the α-helix, oneof the greatest achieve-ments in structural biolo-gy, was made by assuming(i) that the peptide bondis planar, (ii) that all amino acid residuesare equivalent with respect to backboneconformations and (iii) that each amideproton is hydrogen bonded to an oxygenatom of another residue with an N–Odistance of 2.72 Å (ref. 1). Similar criteriaallowed Pauling to predict several otherprotein secondary structural elementsincluding the γ-helix and parallel andantiparallel β-sheets, accomplishmentsthat resulted in his 1954 Nobel Prize inChemistry.

In 1968, Ramachandran recognized theneed for nonplanar peptide bonds incyclic peptides2. In later work, the samegroup argued that discrepancies in calcu-lated values of 3JHN–Hα scalar couplingconstants in peptides with bulky sidechains showed the need to allow for non-planar peptide bonds in both cyclic andlinear peptides3. They also noted thatmost of calculations up to that time used“the standard, completely planar, transpeptide unit of dimensions given byPauling and Corey4 as early as 1951”3.Much later and with relatively high levelHartree-Fock extended basis set ab initiocalculations, Pople’s group showed thatthe peptide bond in the gas phase has sig-nificant flexibility and rotates as much as40° with little energetic cost5, a resultsubsequently verified by Edison and

coworkers6. Simple geometric drawingswill show that deviations of this sizeresult from relatively small changes ininteratomic distances.

In a statistical survey of peptide andprotein databases, Thornton’s group con-vincingly showed that experimentallyderived peptide and protein structureshave significant deviations from planarpeptide bonds7. Using data from 492 pep-tide bonds in the Cambridge StructuralDatabase and 45,851 trans peptide bondsfrom 187 different protein structures inthe Brookhaven Protein Data Bank (nowthe Research Collaboratory for StructuralBioinformatics Protein Data Bank), theywere able to estimate the energies of pep-tide bond rotation using Maxwell-Boltzmann statistics (recreated in Fig. 1)7.Fig. 1 also provides updated values toThornton’s survey that are in qualitativeagreement with the results from 1996.Interestingly, in both the original studyand updated values, there is a small butsignificant tendency toward angles <180°.Thus, both theory and experiment onboth small peptides and proteins suggestthat Linus Pauling’s famous planar pep-tide bond is overly simplistic.

Could Linus Pauling, one of theworld’s greatest chemists, have failed tounderstand that some flexibility is clearly

a real (and quite likelyimportant) property ofthe peptide bond? Severalintroductory and summa-ry statements made byPauling show the impor-tance of peptide bond pla-narity: “The normalcoplanarity of the atomsof [the peptide bond] isthe result of resonancewhich gives rise to partialdouble bond character ofthe N–C′ peptide bond.Rotation about this bondis, in general, severelyrestricted.” (abstract ofref. 8); “The normal pla-narity of the amide groupis established on bothexperimental and theoret-ical grounds as a soundstructural principle. Astructure in which theatoms of the amide group

are not approximately coplanar shouldbe regarded with scepticism until its rela-tively unstable configuration has beenadequately confirmed.” (conclusion ofref. 8); “The N–C bond has about 40 per-cent double-bond character (bond length1.32 Å). The group is planar, and it hasbeen found to have the trans configura-tion in all substances studied except thecyclic peptides (diketopiperazine).” (page498 of ref. 9).

Pauling, like any great chemist, clearlysaw the importance of generalization. Hewas a master at reducing complicatedproblems to their essence and stressingwhat he knew to be the most importantpoints, which are emphasized as themajor conclusions of his work. A morethorough reading, however, shows that hewas keenly aware of the flexibility of thepeptide bond: “About 40% double-bondcharacter appears to be associated withthe C′–N peptide bond. We may thereforeestimate the strain energy involved inrotation around this bond. If the planes ofthe two ends of the amide group form adihedral angle δ, and if A is the amide res-onance energy for the planar configura-tion, the strain energy may be taken equalto A sin2δ. A reasonable value for A isabout 30 kcal/mole. From this we can calculate strain energies of about

Linus Pauling and the planar peptide bondArthur S. Edison

Fig. 1 Peptide bond rotational energies and angular frequency distribution. Theblack and gray points are from ref. 7 and were derived from Maxwell-Boltzmannrelations for proteins and peptides, respectively. The histogram represents theangular frequency distribution of 237,807 values of ω from coiled regions of 3,938proteins in the current release of the protein data base (January, 2001) with resolu-tion ≤ 2.0 Å and R-factor ≤ 20%. The red points are energies derived from Maxwell-Boltzmann relations of the current values of ω, as described in ref. 7. The line is thefunction A sin2ω for A = 30 kcal mol–1, as derived by Corey and Pauling in ref. 8.

©20

01 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://s

tru

ctb

io.n

atu

re.c

om

© 2001 Nature Publishing Group http://structbio.nature.com

history

picture story

0.9 kcal/mole for 10 degree distortion ofthe amide group, 3.5 kcal/mole for 20degree distortion, and so on.” (page 14 ofref. 8).

A plot of Corey and Pauling’s 1953energy function for rotation of the pep-tide bond is shown alongside the 1996and most recent estimates based upon thestatistical survey of the peptide and pro-tein data bases. Nearly 50 years ago andwithout the benefit of a single high reso-lution protein structure or supercomput-er, Pauling got it right.

Tension can be a killer. This is especiallytrue when pressure builds up in single cellorganisms such as bacteria. To deal withrapid changes in osmotic pressure, bacte-ria have evolved safety valves in the form ofmechanosensitive channels. In manyorganisms such channels convert mechan-ical strain induced by sound, touch, gravi-ty or pressure into an electrochemicalresponse. When tension within the cellincreases, these channels open to releasepressure and osmolytes quickly and keepthe bacterium alive. While this channel is alifesaver in times of stress, at other times itneeds to be completely closed to maintainthe electrical integrity of the membrane. Aleaky channel would eventually kill thebacterium.

To gain an understanding of the gatingmechanism of such mechanosensitivechannels, Sukharev and coworkers (Nature409, 720–724 (2001)) developed a numberof structural models based on the previous-ly solved crystal structure of themechanosensitive channel of large conduc-tance, MscL, from Mycobacterium tubercu-losis (the closed structure was solved to 3.5 Å resolution) and previous patch clampexperiments on the Escherichia coli MscLchannel.

Features of the crystal structure and theenergetic parameters of the bacterial chan-nel were incorporated into a series of mole-cular models. The crystal structure showedthat the channel is composed of five identi-cal subunits (each one is a different color).Within each subunit, helical segments M1,M2 and S3 (shown as cylinders) wereobserved. The structure suggested that theM1 helices formed the gate.

Thermodynamic analyses of the MscL

Leak proof channel

channel predicted that the channel opensthrough pre-expanded intermediate states.The closed conformation (left) expands totwo-thirds of its open size to form the so-called closed-expanded conformation(middle) before it actually opens (right).Channel opening could occur by tilting ofthe M1 helices away from the center of thepore like the iris of an eye. Moreover, theability of the channel to go from theclosed-expanded conformation to theopen conformation without a substantialchange in the overall size of the channelsuggested the presence of a second gate.Given that S2 is poorly conserved and S3 isdispensable, the best candidate for a sec-ond gate was S1, which was disordered inthe crystal structure. Sukharev andcoworkers modeled the S1 segments asshort helices that block the pore when thechannel is in the closed conformation.

To test their model for the role of S1, theysubstituted cysteines for specific S1 residues

and asked whether these residues wereclose together in neighboring subunits inthe closed conformation and whethercrosslinking of S1 prevented channel open-ing. They found that S1 segments form abundle when the channel is closed andcrosslinking between S1 segments preventsopening. They also found that S1 segmentscrosslink with an outer helix (M2) whenthe channel is open, and this impedeschannel closing.

Thus, the MscL channel appears to havetwo gates in series to ensure that there is noleakage when the channel is closed. Thecytoplasmic (S1) gate ‘feels’ the tensiononly when the barrel is critically distortedand the flexible linkers connecting S1 toM1 are in the extended conformation. Inthis way, these channels serve as ‘emergencyrelease valves’ when the going gets toughand remain absolutely closed when life isstress free.

Boyana Konforti

AcknowledgmentI thank H. Weissig of the San Diego SupercomputerCenter and Protein Data Bank for providing thedihedral angles used in Fig. 1. D. Purich, R. McKenna, M. Agbandje-McKenna, and members of mylaboratory provided helpful and stimulatingdiscussions.

Arthur S. Edison is in the Department ofBiochemistry & Molecular Biology, Universityof Florida, Box 100245, Gainesville, Florida32610-0245. email: art@ascaris.ufbi.ufl.edu

Received 7 December, 2000; accepted 31January, 2001.

202 nature structural biology • volume 8 number 3 • march 2001

1. Pauling, L., Corey, R.B., and Branson, H.R. Proc. Natl.Acad. Sci. USA 37, 205–211 (1951).

2. Ramachandran, G.N. Biopolymers 6, 1494–1496 (1968).3. Ramachandran, G.N., Lakshminarayanan, A.V., and

Kolaskar, A.S. Biochim. Biophys. Acta 303, 8–13 (1973).4. Pauling, L., and Corey, R.B. Proc. Natl. Acad. Sci. USA

37, 235–240 (1951).5. Head-Gordon, T., Head-Gordon, M., Frisch, M. J.,

Brooks III, C.L., and Pople, J.A. J. Am. Chem. Soc. 113,5989–5997 (1991).

6. Edison, A.S., Markley, J.L., and Weinhold, F. J. Biomol.NMR, 4, 519–542 (1994).

7. MacArthur, M.W. and Thornton, J.M. J. Mol. Biol. 264,1180–1195 (1996).

8. Corey, R.B. and Pauling, L. Proc. Roy. Soc. B 141, 10–20(1953).

9. Pauling, L. The Nature of the Chemical Bond. (CornellUniversity Press, Ithaca, New York; 1960).

©20

01 N

atu

re P

ub

lish

ing

Gro

up

h

ttp

://s

tru

ctb

io.n

atu

re.c

om

© 2001 Nature Publishing Group http://structbio.nature.com