Protein-proteininteractions by NMR

Slow

Fast kon,off >> (νfree - νbound )

νfree νbound

A + B ABkon

koff

kon,off ~ (νfree - νbound )

kon,off << (νfree - νbound)

Measure Kd by e.g. fluorescence, ITC, Biacore, NMR

Minimize interacting region, especially of peptidese.g. limited proteolysis

Find conditions where both components are stableand interaction still occurs (e.g. salt dependence?)and exchange regime is favourable (change temperature?)

Add one component in excess to saturate the other component (excess component usually the unlabelled one)

strong

strong

weak or strong

weak or strong

weak or strong

Structures of protein complexes

Dynamics of interactions

Mapping interactions e.g. for mutagenesis studies

Extracting distance information

Docking

How do we go about studying weak interactions?

Chemical shift mapping may be possible even in intermediate exchange regime

If structures of components are known can model the complex

Distance information can be extracted using transferred NOE

Transferred NOEs

Determine the structure of a small ligand binding large molecule

Ligand is in excess over protein (improved sensitivity)

Intra-ligand NOEs are detected in the bound state by transferringthem by chemical exchange to the free state, where they can beobserved

Requirement is that off-rate is fast: the ligand must associate/dissociate afew times during the mixing time

Transferred NOEs are larger than free intra-ligand NOEs and havethe same sign as diagonal peaks

Intermolecular NOEs are low intensity due to low protein concentration

Clore & Gronenborn: J. Mag. Res. 48 402-417; J. Mag. Res. 53 423-442Ni & Scheraga Acc. Chem. Res. 27 257-264 (review)

Studying strong interactions: structures of complexes

2 types of experiment:Edited or separated: keep 1H attached to nucleus XFiltered or rejected: keep 1H NOT attached to nucleus X

Differential Labelling

13C,15N-proteinUnlabelled peptide 13C-edited NOESY

NOEs between 13C-1H and all other 1H (13C-1H, 12C-1H, 15N-1H, 14N-1H)

X-filtered, 13C-edited NOESYNOEs between 13C-1H and 12C-1H/15N-1H

Double-filtered NOESYNOEs between 12C-1H and 12C-1H

H

H

13C

13C

12C

H

12C

H

14N

H

15N

H

Considerations when working with complexes

•Multiple samples required with differential labelling patterns -cost and time implications

•Sensitivity of experiments e.g. double-filtered experiments, X-filtered, X-edited experiments

•Unlabelled peptide + large protein, peptide assignment may be problematic

•Record twice as many spectra

•Spectral simplification via labelling patterns

•Inter- vs intra-molecular NOEs

Tight Complexes: Slow Exchange

Chemical Shift Assignments

1. Backbone experiments:HSQC, HNCA, HN(CO)CA, HNCACB, CBCA(CO)NHrecorded on each component of complex

2. Sidechain experiments:15N-separated TOCSY, HCCH-TOCSYCC(CO)NH , H(CCCO)NH, HBHA(CBCA)(CO)NHon each component of the complex

3. Coupling constant experiments:HNHA on each component of the complex

Tight Complexes: Slow Exchange

NOE Assignments

4. Intra- and inter-molecular NOEs mixed:3D 13C-separated NOESY3D 15N-separated NOESY

5. Intermolecular NOEs:X-filtered experiments. 3D 13C-separated if sensitivity is good or 2D 1H/1H version

(Zhwalen et al JACS 1997 119 6711-6721)

X-filtered Experiments

Two basic types:

a) Schemes where X-attached 1H are removed directly

1/2J1H

e.g. generate MQCX

b) Schemes where interleaved experiments are recordedand have to be added or subtracted to keep or discard X-attached 1H

X-pulse present or absent inalternating experiments

1H

X

X-filter (I)Purge sequence

1H ∆1 ∆2 I = 1H attached to X nucleusH = all other 1Hy

XG2G1 G1

G

Hz + Iz -Hy - Iy cosπJIS∆1 + 2IxSzsinπJIS∆1

(∆1 = 1/2JIS) -Hy + 2IxSz-Hy - 2IxSy

90˚(1H), ∆1

90˚(X)

1. ∆1 and ∆2 can be tuned for different values of JIS e.g. 120-145 Hz (aliphatic) and 160-220 Hz (aromatic) 2. Can add spin-lock (y) on 1H to remove Ix-based antiphase that escapes X pulses3. Replaces a 1H 90˚ pulse in a sequence

X-filter (II)Refocussed half filter

Hz + Iz -Hy - Iy -Hy -IycosπJIS∆1 + 2IxSzsinπJIS∆1 (∆1 = 1/2JIS)Hy - 2IxSz Hy - Iy

OR Hy + 2IxSz Hy + Iy

1H ∆1 ∆1

X

GG1 G1

I = 1H attached to X nucleusH = all other 1H

90˚(1H)

180˚(1H,X)∆1

∆1

180˚(1H) ∆1

AB

A + B = Hy only (filtered)A - B = Iy only (edited)

All 13C-filtered experiments suffer because 1JCH varies:120-145 Hz aliphatic CH; 160-220 Hz aromatic CH∆1 delays are inevitably only tuned to one J coupling

220

200

180

160

140

12020 40 60 80 100 120

1JCH (Hz)

aliphatic

Tyr, Phe, Trp

His

δ13C (ppm)

1JCH = (0.365 ± 0.01 Hz/ppm) δC + 120.0 ± 0.5 Hz

Zwahlen et al (1997) JACS 119 6711-6721

Adiabatic pulses

Trajectory of 13C magnetization follows effective field for duration of the pulse

Carrier frequency of the pulse starts far upfield, sweeps through resonance, then

downfield

13C resonate at different frequencies so they are all inverted at different times,

depending on their frequency and the sweep rate and duration of the inversion pulse

X-filter (III)Purge scheme with adiabatic pulse

Hy + Iy Hy + IycosπJIS(∆1-2t) - 2IxSzsinπJIS(∆1-2t)Hz + IzcosπJIS(∆1-2t) - 2IxSzsinπJIS(∆1-2t)

1H

X

G

∆1/2 ∆1/2-t

G1 G1

t

G2

a b 13C spin inverted at time t (t≥0) after 1H 180˚ pulse 1JCH evolution occurs for a time ∆1-2t1JCH and t vary for each 13C nucleus

From a to b:

removed by G2IzcosπJIS(∆1-2t) = 0 for all spins∆1-2t = 1/(21JCH) (i)

Since t≥0 and 1JCH is smallest for upfield 13C (i.e. methyls), these must be inverted first. ∆1 is set close to 1/2 1JCH (methyl)

1JCH = AδC + B (ii) (A=0.365 Hz/ppm; B = 120.0 Hz)

Frequency of the transmitter δRF(t) = δC. Substitute for 1JCH in (i)

∆1-2t = 1/[2(AδRF + B)] time derivative to get sweep rate during WURST

3D 13C-filtered, 13C-edited NOESY-HSQC

Two of these modules at the start of the sequence (shorter than the original half-filter)

Suppression factors are 100-140-fold

Putting filter modules at the start of the sequence makes H2O suppression easier

Filters add to the length of the sequence, so semi-CT, concatenating 1H chemical shift evolution with the half filter

15N filter at the same time as the 13C filter

Cdc42 - 21 kDa small G protein of the Rho familyGTP cofactor replaced by GMPPNP

PAK - 5kDa fragment - can be expressed as GST-fusion

Kd ~ 30nM

Samples:

15N Cdc42 + unlabelled PAK15N,13C Cdc42 + unlabelled PAK15N PAK + unlabelled Cdc4215N,13C PAK + unlabelled Cdc42

no deuteration required - all experiments worked

3D X-filtered, 13C-edited NOESY ran on 13C,15N PAK, unlabelled Cdc42

Summary of Cdc42/PAK Intermolecular NOEs

Unambiguous Ambiguous

13C-filter/13C-edited 32 2913C-NOESY (PAK) 67 7313C-NOESY (Cdc42) 21 40

15N-NOESY (PAK) 19 515N-NOESY (Cdc42) 7 12

Total Intermolecular 146 159

Total number of NOEs = 4,000

Assignment of Unlabelled Component in Complex

Filtered/rejected experiments:to assign unlabelled peptide with labelled protein- sensitivity/water suppression poordouble filter (ω1 and ω2) required in NOESY but not inthrough-bond experiments

J(CH,NH)-separated NOESY -- better sensitivity and water behaviour(Nietlispach)

Double half-filtered NOESY(Otting & Wüthrich 1990 Q. Rev. Biophys. 23 39-96)

∆1 ∆1 t1 τmix ∆2 ∆2

Decψ1 ψ2

1H Acqψ = x: Hy - Iyψ = -x: Hy + IyX

12C -> 12C

12C -> 13C

13C -> 12C

13C -> 13C

Aψ1 ψ2x x

Bψ1 ψ2-x x

Cψ1 ψ2x -x

Dψ1 ψ2-x -x

++ +

+++

--

--

++++ --

Individual datasets can be scaled to offset effects of e.g. low labelling efficiency

Combination sub-spectrum ω1/ω2(A+B) + (C+D) ω1-filter,ω2-filter unlabelled/unlabelled(A+B) - (C+D) ω1-filter,ω2-edit unlabelled/labelled(A-B) + (C-D) ω1-edit,ω2-filter labelled/unlabelled(A-B) - (C-D) ω1-edit,ω2-edit labelled/labelled

J(CH,NH)-separated NOESY(based on Melacini, JACS 122 9735-9738)

After the semi-CT period, required magnetization is:-Iy cosπJt2 cosΩHt1 modulated by J in t213C-bound 1H are at ±JCH/2; 12C-bound 1H are at J=0

J=0 plane contains intra-peptide NOEs and peptide(f1) to protein (f3) NOEsIntra-protein NOEs and protein (f1) to peptide (f3) NOEs are at ±JCH/2

Intermolecular NOEs in J=0 plane are coupled in F3Intra-peptide NOEs have a return peak in the same plane

Chromodomain complex with methylated peptide from histone H3

Problematic assignment of Me2-Lys - not J coupled to anything else

Assignment of NOEs from Me2-Lys to aromatic residues in the protein (distinguish from Me-NH NOEs by coupling constant)

Me-Me NOE in J=0 plane (confirmed assignment of Me2-Lys)

xF3:J =0 Hz

F3:J =160 Hz

Me-K9Trp 42Phe 45

Trp 42

Phe 45

Me-K9

160Hz

1H [ppm]

Structure calculation strategies

If starting from two extended chains with ideal geometry make sure they are not lying on top of each other (bias starting structures)

Assignment of intra-molecular NOEs in 13C- and 15N-edited spectra: check that they do not have any inter-molecular possibility (they may not appear in X-filter, 13C-edited experiment if they are weak)

Ambiguous NOEs that appear in 13C-edited NOESY as well as X-filter, 13C-edited NOESY should be treated in the same way as other ambiguous NOEs as they may contain intra-molecular contribution to the intensity

Ambiguous NOEs: the possibilities must be edited for the different experiment types e.g. 13C-edited NOESY: F1/F3 = 13CH only; F2 = any 1H

SH3/peptide Calculation Strategy

•All intermolecular distance restraints were ambiguous (32 from 13C-filtered/13C-edited experiment)

•NOE tables generated using AZARA ‘connect’ and then put into ARIA

•Ambiguous restraints analysed at each iteration to generate distance restraint tables for next iteration with reduced ambiguity

NOEs used in SH3/peptide Structure Calculations

Unambiguous Ambiguous

13C-filter 0 3213C-NOESY (SH3) 1026 55415N-NOESY (SH3) 395 1231H NOESYs 85 29

Totals 1506 738

At end of calculation - 66 intermolecular NOEs

Using NMR data for docking

Structures of components are known

Use NMR data to map binding contacts or determine relative orientation of components

Dynamics, allowing only interfaces to move

How do you decide which residues are in the contact site?

Pick all the peaks in free and bound and calculate combined shift difference in 15N and 1H shifts, often defined as:[(∆15N)2 x 10(∆1H)2]1/2

Define ‘significant’ chemical shift perturbation (>1 SD from average shift change) - add in the ones that have disappeared completely

Check for solvent accessibility (e.g. NACCESS): NHs that are completely buried are unlikely to be involved in the interaction but are experiencing secondary effects, unless there is a significant structural change (should be obvious from the HSQC). NACCESS cutoff: if the residue is more than 50% exposed it is available for interaction.

Using Chemical Shift Mapping Data for Docking(HADDOCK)

Take significant shift changes, screened by solvent accessibilityActive residues

Take residues close to active residues on the surfacePassive residues

Ambiguous interactive restraint (AIR)

Natoms NresB Natoms

diAB = Σ Σ ΣmiA=1 k=1 nkB=1

between any atom m of active residue i in protein A (miA) and any atom n ofboth active and passive residues k (Nres total) of protein B (nkB) and inverselyfor protein B.

For each active residue (i) in A, restraint to any active or passive residue (k) in B, over all atoms and vice versa.

1d6miAnkB

( )(-1/6)

Dominguez et al (2003) J Am Chem Soc 125 1731-1737

HADDOCK calculation strategy

(i) Randomization of orientations and rigid body energy minimization(ii) Semi-rigid simulated annealing(iii) Refinement with explicit solvent

Models clustered by interaction energies (Eelec, EvdW, EAIR) andaverage buried surface area

Lowest energy cluster with the highest buried surface area assumedto be correct

Can also add RDCs, inter-molecular NOEs and radius of gyration term to prevent expansion at the interface (Clore(2003) JACS 125 2902-12)

Cross-saturationTakahashi et al (2000) Nat. Struc. Biol. 7 220-223

-NH

-CH

-15NH

-C2H

ProteinI

ProteinII Strong binding case

R.F.Band-selective WURST saturation, followed by TROSY-HSQC

cross-saturation

Saturate the aliphatics of protein II - magnetization transferred by spin diffusion to the aromatics and amides. If saturation does not leak to H2O, it does not affect amides of protein I.

Cross-saturation to the 15NH in the interface on protein I

Measure intensity in HSQC vs time of saturation to find residues in interface

More precise than chemical shift mapping

Symmetric oligomers by NMR

NMR spectra are identical for each monomer

Cannot distinguish intra- and inter-monomer distance contacts

Breaking symmetry -mixed labelling (unfold/refold?)higher order oligomers may have to use tags to get single subunit labelledaddition of spin label to one component (Gaponenko et al 2002 J. Biomol. NMR 24 143-148)

Double half-filtered NOESY(Otting & Wüthrich 1990 Q. Rev. Biophys. 23 39-96)

∆1 ∆1 t1 τmix ∆2 ∆2

Decψ1 ψ2

1H Acqψ = x: Hy - Iyψ = -x: Hy + IyX

12C -> 12C

12C -> 13C

13C -> 12C

13C -> 13C

Aψ1 ψ2x x

Bψ1 ψ2-x x

Cψ1 ψ2x -x

Dψ1 ψ2-x -x

++ +

+++

--

--

++++ --

Mixed dimer sample:

Intramolecular: 12C -> 12C and 13C -> 13CIntermolecular: 12C -> 13C and 13C -> 12C

Combination sub-spectrum ω1/ω2(A+B) + (C+D) ω1-filter,ω2-filter unlabelled/unlabelled(A+B) - (C+D) ω1-filter,ω2-edit unlabelled/labelled(A-B) + (C-D) ω1-edit,ω2-filter labelled/unlabelled(A-B) - (C-D) ω1-edit,ω2-edit labelled/labelled

Double half-filtered NOESY(Folkers et al 1993 JACS 115 3798-3799)

∆1 ∆1 t1 τmix ∆2 ∆2

Decψ1 ψ2

1H Acqψ = x: Hy - Iyψ = -x: Hy + IyX

12C -> 12C

12C -> 13C

13C -> 12C

13C -> 13C

Aψ1 ψ2x x

Bψ1 ψ2-x x

Cψ1 ψ2x -x

Dψ1 ψ2-x -x

++ +

+++

--

--

++++ --

Intramolecular: 12C -> 12C and 13C -> 13CIntermolecular: 12C -> 13C and 13C -> 12C

Combination(A+D) 12C -> 12C and 13C -> 13C(A-D) 12C -> 13C and 13C -> 12C

Structure calculationtreat all distance restraints as ambiguous(contribution from each monomer, intra- or inter-monomer)

include non-crystallographic symmetry (NCS) restraints to keep monomers superimposable

include distance symmetry restraints to keep the interactions between monomers symmetric

A1

B1

B2

A2

NCS A1-B1 = A2-B2

Distance symmetry A1-B2 = A2-B1

Heterochromatin protein 1shadow chromo domain/CAF peptide complex

HP1 structural component of heterochromatin

Binds histone H3 methylated at Lys-9, via N-terminal chromo domain

Dimerizes via its C-terminal shadow chromo domain

Shadow domain homodimer of 2 x 70 residues

Complex with 29 residue peptide from CAF - stoichiometry is 1:2

HP1 Backbone Assignment

identical in both monomers

doubled; not assigned into A/B monomers

N N

HP1 Backbone Assignment with NOE Information

112 115

120

127133

138145

147 149

155

158

166

doubled; not assigned into A / B monomers

assigned into A and B monomers

identical in both monomers

Structure calculation strategy

With NCS restraints only on unsplit residues, one monomerwas always better defined-

all distance restraints being assigned to one monomerchemical shift dispersion between monomers not high enoughone monomer gave better spectra

NCS restraints weighted according to chemical shift degeneracy:

4 groupsAll peaks identical (2.0)NH only split (1.0)NH and HA split (0.5)NH and all of sidechain split (0.1)

Distance symmetry restraints only applied for residuesthat were identical

Decreasingweight