Solving NMR structures Part I:Commonly used experimentally

derived restraints

Distance restraints from crosspeak intensities in NOESY spectra; measuring and calibrating NOEs

Dihedral angle restraints from three-bond J couplings; measuring J couplings

Hydrogen bond restraints from amide hydrogen exchange protection data

Using NOESY to generate nOe distance restraints

• NOESY measurements are not steady-state nOe’s: we are not saturating one resonance with constant irradiation while observing the effects at another.

• Instead, we are pulsing all of the resonances, and then allowing nOe’s to build up through cross-relaxation during a mixing time --so nOe’s in a NOESY are kinetic: crosspeak intensities will vary with mixing time

• typical tm’s used in an NOESY will be 20-200 ms.

fromGlasel &Deutscherp. 354

basic NOESY pulse sequence

mixing time

nOe buildup in NOESY• other things being equal, the initial rate of

buildup of a NOESY crosspeak is proportional to 1/r6, where r is the distance between the two nuclei undergoing cross-relaxation.

• nOe buildup will be faster for larger proteins, which have a longer correlation time tc, and therefore more efficient zero-quantum cross-relaxation

• initially crosspeak intensity builds up linearly with time, but then levels off, and at very long mixing time will actually start to drop due to direct (not cross) relaxation.

spin diffusion

• under certain circumstances, indirect cross-relaxation pathways can be more efficient than direct ones, i.e. A to B to C more efficient than A to C. This is called spin diffusion

• when this happens the crosspeak intensity may not be a faithful reflection of the distance between the two nuclei.

Crosspeaks due to spin diffusion exhibit delayed buildup in NOESY experiments

-1

0

1

2

3

4

5

6

0 50 100 150 200 250 300

relative crosspeak intensity

mixing time

spin diffusion

direct cross-relaxation

note the delay in buildup

•spin diffusion peaksare usually observedat long mixing time, and their intensity does not reflect the initial rate of buildup

•these effects can be avoided either by sticking with short mixing times or by examining buildup curves over a range of mixing times

Other nOe caveats

• I mentioned that nOe buildup rates are faster for larger proteins because of the longer correlation time

• It’s also true that buildup rates can differ for nuclei within the same protein if different parts of the protein have different mobility (hence different correlation times)

• for parts of the protein which are relatively rigid (such as the hydrophobic core) correlation times will more or less reflect that of the whole protein molecule--nOe buildup will be fast

• disordered regions (at the N- or C-termini, for instance) may have much shorter effective correlation times and much slower nOe buildup as a consequence

• the bottom line is, the actual nOe observed between two nuclei at a given distance r is often less than that expected on the basis of the overall molecular correlation time.

The goal: translating NOESY crosspeak intensities into nOe distance restraints

• because the nOe is not always a faithful reflection of the internuclear distance, one does not, in general, precisely translate intensities into distances!

• instead, one usually creates three or four restraint classes which match a range of crosspeak intensities to a range of possible distances, e.g.

class restraint description *for protein w/Mr<20 kDa

strong 1.8-2.7 Å strong intensity in short tm (~50 ms*) NOESY

medium 1.8-3.3 Å weak intensity in short tm (~50 ms*) NOESY

weak 1.8-5.0 Å only visible in longer mixing time NOESY

• notice that the lower bound of 1.8 Å (approximately van der Waals contact) is the same in all restraint classes. This is because, for reasons stated earlier, atoms that are very close can nonetheless have very weak nOe’s, or even no visible crosspeak at all.

Calibration of nOe’s

• the crosspeak intensities are often calibrated against the crosspeak intensity of some internal standard where the internuclear distance is known. The idea of this is to figure out what the maximal nOe observable will be for a

given distance.

• ideally, one choosesan internal standardwhere the maximal nOewill be observed (i.e. something not undergoing a lot of motion)

• this calibration can then be usedto set intensity cutoffs for restraint classes, often using a 1/r6 dependence

tyrosine distancealways the same!

Coupling constants and dihedral anglesThere exist relationships between three-bond scalar coupling constants 3J and the corresponding dihedral angles , called Karplus relations. These have the general form

3J = Acos2 + Bcos + C

fromp. 30Evanstextbook

Empirical Karplus relations in proteins

• comparison of 3J values measured in solution with dihedral angles observed in crystal structures of the same protein allows one to derive empirical Karplus relations that give a good fit between the coupling constants and the angles

coupling constantsin solution vs. angles from crystal structure for BPTI

these twoquantitiesdiffer by 60°because they are defined differently from p. 167 Wuthrich textbook

Empirical Karplus relations in proteins

• Here are some empirical Karplus relations:

3JH,HN()= 6.4 cos2(- 60°) -1.4 cos(- 60°) + 1.93JH,H2()= 9.5 cos2(- 120°) -1.6 cos(- 120°) + 1.83JH,H3()= 9.5 cos2() -1.6 cos() + 1.83JN,H3()= -4.5 cos2(+ 120°) +1.2 cos(+ 120°) + 0.13JN,H2()= 4.5 cos2(- 120°) +1.2 cos(- 120°) + 0.1

• Notice that use of the relations involving the hydrogens would require that they be stereospecifically assigned (in cases where there are two hydrogens)

• note that these relations involve or 1 angles

Measuring 3JHN-H: 3D HNHA spectra

HN to Hcrosspeak

HN diagonalpeak

this is one plane of a 3D spectrum of ubiquitin. The plane corresponds to this 15N chemical shift

ratio of crosspeakto diagonal intensitiescan be related to 3JHN-H

J small J large

Archer et al. J. Magn. Reson.95, 636 (1991).

3D HNHB experiment

• similar to HNHA but measures 3JN-H couplings instead

for =180 both 3JN ~1 Hz for =+60,-60 one is ~5, other is ~1

can’t tell the difference unless ’s are stereospecifically assigned

DeMarco, Llinas,& Wuthrich Biopolymers17, p. 2727 (1978).

3D HN(CO)HB experiment

• complementary to HNHB

• measures 3JC,H couplings

Grzesiek et al. J. Magn. Reson. 95,636 (1991).

for a particular proton,if =180, 3JC,H= ~8 Hzif =+60 or -60, 3JC,H= ~1 Hz

HNHB and HN(CO)HB together

3JC,Hsmall3JC,Hlarge3JN,Hsmall3JN,Hsmall

3JC,Hlarge3JC,Hsmall3JN,Hsmall3JN,Hlarge

3JC,Hsmall3JC,Hsmall3JN,Hlarge3JN,Hsmall

HNHB, HN(CO)HB together

•can thus get both 1 angle and stereospecific assignments for ’s from a combination of HNHB and HN(CO)HB

HNHB

HN(CO)HB

from Bax et al. Meth. Enzym. 239, 79.

Dihedral angle restraints

• derived from measured J couplings• as with nOe’s, one does not translate J directly into a

quantitative dihedral angle, rather one translates a range of J into a range of possible angles, e.g.

3JH,HN()< 6 Hz = -65° ± 25°3JH,HN()> 8 Hz = -120 ± 40°

Amide hydrogen exchange

•amide protons undergo acid- and base-catalyzed exchange with solvent protons at a rate which ranges from the second to minute time scale, depending upon solvent conditions (mostly pH)

•if a protein is placed in D2O, the amide signals due to 1H nuclei will disappear over time due to this exchange

poly-D,L-alanineexchange rate in D2Oat 20 °C. Minimum with respect to pH is due to the fact that exchange is both acid and base catalyzed

Englander & Mayne, Ann. Rev. Biophys. Biomol. Struct. (1992) 21, 243.

exchange ratehalf-life

pH but in D2O!

Amide hydrogen exchange

• when amide protons are involved in hydrogen bonds in a folded protein, they are protected from exchange with solvent and thus exchange somewhat more slowly, i.e.

faster:

N-H --> N-D

slower:

N-H....O=C --> N-D....O=C

Protection factors

• the protection factor P for a given amide proton is the intrinsic rate of exchange expected for that amide proton in an unfolded protein under a given set of solvent conditions, divided by the observed rate of exchange in the native state

P= kex(U)/kex(N)

where U is the unfolded state and N is the native state

• intrinsic rates of exchange vary with the amino acid sequence and the solvent conditions in a predictable manner--there are published tables and websites where you can look them up/calculate them.

Measuring amide exchange rates • 15N-1H HSQC spectra are a good way to monitor exchange because

individual well-resolved peaks are visible for most amides

• one might record a spectrum in H2O, freeze dry the sample, resuspend in D2O, and record a series of spectra in D2O to monitor the rate at which each amide proton disappears

picture at left:15N-1H HSQC of Arc repressor 5 min after resuspension in D2O (pH 4.7)--note that unprotected amides such as the unstructured N-terminus (1-7) have already exchanged. This is because intrinsic half-lives at this pH are in the vicinity of 10 seconds to a minute.

Burgering et al. Biopolymers (1995) 35, 217.

The plot above shows measured exchange rates from a series of 15N-1H HSQC spectra for Arc repressor. Protected amides are usually within secondary structure elements, but the presence of secondary structure doesn’t guarantee that protection will be observed (note the lack of protection for the beta-sheet from residue 9-14). Lack of protection could indicate that the secondary structure element undergoes fairly rapid local unfolding. Note the strength of NMR and HSQC in particular here--you get a separate exchange rate for every amide proton--great residue-specific dynamic information!

blue spheres: P > 4650

green spheres: 370 < P < 4650

red spheres: P < 370 Liu et al. Biochemistry (1999)38, 5362.

HX (hydrogen exchange) of Syrian hamsterprion protein

note that while all protected amides are hydrogen-bonded,not all hydrogen-bondedamides are equally wellprotected. Rates differ both within secondary structure elements (buried positions near the center usually most protected) and between different secondary structure elements

surface is less well-protected

ends of helices less well-protected

Hydrogen bond restraints

• amide protons showing significant protection are inferred to be involved in hydrogen bonds, but it is not clear from this data alone what the identity of the hydrogen bond acceptor group is.

• however, if additional information is available, one can infer what the acceptor is. For instance, d(i-3,i) and d(i-4,i) nOe’s such as are characteristic of -helices imply that the protected amide of residue i is hydrogen bonded to the carbonyl of residue i-4.

• hydrogen bond restraints are usually expressed as a pair of distance restraints, e.g dHN-O = 1.5-2.3 Å, dN-O = 2.5-3.3 Å. A pair of restraints like

this ensures reasonable good geometry for the hydrogen bond.