TITAN: Two-dimensional lineshape analysis

Transcription

1 TITAN: Two-dimensional lineshape analysis Chris Waudby Christodoulou Group Andres Ramos Lisa Cabrita John Christodoulou

2 Inhibition of fatty acid synthesis for treatment of tularemia OH Cl O Cl Cl OH O Cl OH O OH O 2 5 OH O Inhibition of Francisella tularensis enoyl reductase (interaction of inhibitors with E NAD + complex) Lu, H. et al. ACS Chem. Biol. 4, (2) Copeland, R. A. Nat Rev Drug Discov 5, 87 5 (26)

3 Inhibition of fatty acid synthesis for treatment of tularemia OH Cl a O Cl Cl 8 OH O Cl OH O Survival (%) 6 4 OH 2 O 2 OH O K i (nm) 5 Lu, H. et al. ACS Chem. Biol. 4, (2) Copeland, R. A. Nat Rev Drug Discov 5, 87 5 (26)

4 Inhibition of fatty acid synthesis for treatment of tularemia OH b Cl Kinetics of ligand binding are O Cl crucial to in Cl 8 vivo activity OH Cl OH O O Survival (%) 6 4 OH 2 O 2 OH O Residence time (minutes) Lu, H. et al. ACS Chem. Biol. 4, (2) Copeland, R. A. Nat Rev Drug Discov 5, 87 5 (26)

5 Molecular mechanism of imatinib (Gleevec) Mechanism of binding is central to activity Agafonov, R. V., Wilson, C., Otten, R., Buosi, V. & Kern, D. Nat. Struct. Mol. Biol. 2, (24) Wilson, C. et al. Science 347, (25)

6 Rates of reactions and spectroscopic timescales UV ν NMR ν UV frequency difference ν ~ 4 Hz kex ν NMR frequency difference ν ~ Hz kex ν

7 NMR lineshapes vs exchange rates 3 s A B slow exchange % A % B frequency difference 2 s

8 NMR lineshapes vs exchange rates 3 s A B slow-intermediate exchange % A % B frequency difference 2 s

9 NMR lineshapes vs exchange rates 3 s A B fast-intermediate exchange % A % B frequency difference 2 s

10 NMR lineshapes vs exchange rates 3 s A B fast exchange % A % B frequency difference 2 s

11 NMR lineshapes vs exchange rates magnetic field strength

12 NMR titrations are information rich structure (chemical shifts) molecular weight (linewidths) free energy P binding kinetics ν ~, s => measure exchange on timescales of µs s thermodynamics PL ( G = RT ln Kd)

13 NMR lineshape analysis spin system parameters chemical shifts linewidths (relaxation rates) binding model parameters Kd, koff Bloch-McConnell equations ligand concentration P+L PL k ex = k o + k on [L] fraction bound true fraction bound chemical shift chemical shift 5 5 ligand concentration / µm

14 NMR lineshape analysis spin system parameters chemical shifts linewidths (relaxation rates) binding model parameters Kd, koff Bloch-McConnell equations ligand concentration chemical shift

15 NMR lineshape analysis example: ultrafast folding ( µs timescale) of villin headpiece Wang, M. et al. J. Am. Chem. Soc., (23)

16 From D to 2D N chemical shift / ppm H chemical shift / ppm

17 Lineshape analysis of D cross-sections O'Connor, C. & Kovrigin, E. L. Biochemistry 5, (22)

18 Why isn t lineshape analysis more common? Apathy Good enough to assume fast/slow exchange and analyse chemical shift/intensity changes? No! This risks systematic errors and ignores the richness of titration datasets Complexity of analysis? Software needs to be easy to use!

19 Problem #: Peak overlap

20 Problem #2: Normalisation D pulse-acquire: H no delay between pulse and acquisition => signal proportional to concentration 2D HSQC: signals decay during pulse program execution

21 Problem #2: Normalisation Raw intensities Normalised peak volumes H chemical shift / ppm 7.3

22 Problem #2: Normalisation Raw intensities Spectra with measurement noise Normalised peak volumes Normalised peak volumes H chemical shift / ppm H chemical shift / ppm 7.3

23 Problem #3: Differential relaxation D pulse-acquire: H no delay between pulse and acquisition => signal proportional to concentration 2D HSQC: signals decay during pulse program execution

24 Problem #4: Multiple quantum evolution [P] = µm 7 MHz 2 s ΔδH =. ppm HMQC appears to have more binding than in HSQC ligand concentration ΔδN = ppm eq.25 eq.5 eq.75 eq eq HSQC HMQC dissociation rate 2 s 2 s H chemical shift / ppm 7.7 H chemical shift / ppm s H chemical shift / ppm

25 Existing methods at best analyse 2D data using D theory There must be a better way!

26 Two-dimensional lineshape analysis spin system parameters chemical shifts linewidths (relaxation rates) binding model parameters Kd, koff QM 5 N chemical shift / ppm H chemical shift / ppm

27 Two-dimensional lineshape analysis spin system parameters chemical shifts linewidths (relaxation rates) binding model parameters Kd, koff QM 5 N chemical shift / ppm H chemical shift / ppm

28 Two-dimensional lineshape analysis pulse program +experimental settings +processing parameters spin system parameters chemical shifts linewidths (relaxation rates) binding model parameters Kd, koff QM 5 N chemical shift / ppm H chemical shift / ppm

29 Pulse programs / degrees of freedom Evolution of isolated IS spin system (without chemical exchange): 6 free parameters! How many parameters can we expect to extract from a spectrum? Helgstrand, M., Härd, T. & Allard, P. J. Biomol. NMR 8, 4 63 (2)

30 Two-dimensional lineshape analysis Helgstrand, M., Härd, T. & Allard, P. J. Biomol. NMR 8, 4 63 (2) Helgstrand, M. & Allard, P. J. Biomol. NMR 3, 7 8 (24)

31 Two-dimensional lineshape analysis No normalisation required between spectra Accounts for differential relaxation and MQ evolution during pulse sequence Can avoid regions of peak overlap or fit overlapping signals simultaneously 5 N chemical shift / ppm H chemical shift / ppm

32 Example: FIR / Nbox interaction GT rich FUSE ssdna FBP RRM Nbox RRM2 FIR

33 Lineshape analysis: FIR / FBP Nbox observed fit reported: Kd = 4 ± 8 µm fit: Kd = 22.7 ±.5 µm koff = 27 ± s observed fit RRM Nbox RRM2 FIR

34 Lineshape analysis: FIR / FBP3 Nbox reported: fit: Kd = 28 ± 36 µm Kd = 283 ± 2 µm koff = 5 ± 3 s RRM RRM2 FIR Nbox3 observed fit

35 Global fit of 8 residues:

36 3D plots for inspecting goodness of fit eq. eq.3 eq.6 eq.5 eq observed fitted eq 2.3 eq 2. eq 3.5 eq 4.6 eq eq 6. eq 8.6 eq.5 eq 4.3 eq

37 Example 2: TFP binding to (Ca 2+ )4-CaM N N N F F F S trifluoperazine

38 NMR titration 5 N chemical shift / ppm 24 S I63 F68 F4 D8 K5 D8 5 N chemical shift / ppm 24 I63 S F4 F68 D8 K5 D8 5 N chemical shift / ppm 24 I63 D8 D8 S F68 K5 F H chemical shift / ppm H chemical shift / ppm H chemical shift / ppm 5 N chemical shift / ppm 24 I63 S F4 F68 D8 K5 D8 5 N chemical shift / ppm 24 I63 D8 D8 S F68 K5 F H chemical shift / ppm H chemical shift / ppm

39 Full binding model: 6 states, 32 equilibria 5 N chemical shift / ppm 24 S I63 F68 F4 D8 K5 D H chemical shift / ppm 5 N chemical shift / ppm 24 I63 S F4 F68 D8 K5 D8 5 N chemical shift / ppm 24 I63 S F4 F68 D8 K5 D H chemical shift / ppm H chemical shift / ppm 5 N chemical shift / ppm 24 I63 D8 D8 S F68 K5 F4 5 N chemical shift / ppm 24 I63 D8 D8 S F68 K5 F H chemical shift / ppm H chemical shift / ppm

40 Sequential binding model 5 N chemical shift / ppm 24 S I63 F68 F4 D8 K5 D H chemical shift / ppm 5 N chemical shift / ppm 24 I63 S F4 F68 D8 K5 D8 5 N chemical shift / ppm 24 I63 S F4 F68 D8 K5 D H chemical shift / ppm H chemical shift / ppm 5 N chemical shift / ppm 24 I63 D8 D8 S F68 K5 F4 5 N chemical shift / ppm 24 I63 D8 D8 S F68 K5 F H chemical shift / ppm H chemical shift / ppm

41 2D lineshape fitting to sequential binding model a CaM +TFP, 4 ± μm 227 ± s - CaM:TFP +TFP, 62 ± 2 μm ± 4 s - +TFP, 25 ± μm 27 ± 8 s CaM:TFP3 - +TFP, 4 ± mm 2 ± 3 s - D8 24 S F68 K5 F4 f 4 fitted N chemical shift / ppm I63 5 N chemical shift / ppm D8 e CaM:TFP4 4 d CaM:TFP2 b observed c H chemical shift / ppm H chemical shift / ppm based on global fit of 33 residues

42 TITAN: Easy to use 2D analysis software

43 Use a variety of build-in binding models

44 Define regions of interest to avoid overlap

45 or fit overlapping peaks directly

46 Check fit quality with 3D viewer

47 Error analysis by block residual resampling (bootstrapping)

48 Validation and comparison of simple and block resampling 5 N chemical shift / ppm K d = μm, k off =5 s, σ=.2 5 K d = μm, k off =5 s, σ=.2 5 K d = μm, k off =5 s, σ=.5 5 K d = μm, k off =5 s, σ=.5 K d = μm, k off =5 s, σ=. K d = μm, k off =5 s, σ=. K d = μm, k off =5 s, σ=.5 K d = μm, k off =5 s, σ=. K d = μm, k off =5 s, σ=.5 simulate noisy synthetic data with parameters p fit data to estimate parameters, p fit calculate bootstrap error estimates, σ z-score QQ plot: parameter z-scores vs Standard Normal Simple bootstrap (N=) Block bootstrap (N=5) H chemical shift / ppm analyse distribution of z-scores z = (p fit p ) / σ Standard Normal Quantiles Protein-ligand titrations were simulated with a fixed protein concentration of 5 µm and ligand concentrations of, 2.5, 25, 5, 62.5 and 75 µm, with Kd varied between and µm and koff between 5 and 5 s. The performance of the fitting algorithm was investigated with different levels of noise in the synthetic dataset.

49 Exchange in HSQC vs HMQC spectra [P] = µm 7 MHz 2 s HMQC appears to have more binding than in HSQC ΔδH =. ppm ligand concentration ΔδN = ppm eq.25 eq.5 eq.75 eq eq HSQC HMQC dissociation rate 2 s 2 s H chemical shift / ppm 7.7 H chemical shift / ppm s H chemical shift / ppm

50 Chemical exchange regimes Slow exchange p 2 Introduce the dimensionless parameter: =! k ex Coalescence point (5:5 equilibrium) = p 2 magnetic field strength Fast exchange p 2

51 Chemical exchange regimes in 2D experiments: I-spin only 4 S F S F! obs =! I p B S =! S k ex ξs 2 R ex =!2 I p Ap B k ex -2 S ! obs = k2 exp A p B! I I = ξ I! I k ex R ex =k AB

52 Chemical exchange regimes in 2D experiments: HSQC 4 S S F S F! obs =! S p B S =! S k ex ξs 2 F R ex =!2 S p Ap B k ex -2 S -4 S ! obs = k2 exp A p B! S I = ξ I! I k ex R ex =k AB

53 Chemical exchange regimes in 2D experiments: HZQC 4 S F F! obs =(! S! I )p B 2 R ex = (! S! I ) 2 p A p B k ex S =! S k ex ξs -2 S I = ξ I! I k ex S! obs = k2 exp A p B! S! I R ex =k AB

54 Chemical exchange regimes in 2D experiments: HDQC 4 F S F! obs =(! S +! I )p B 2 R ex = (! S +! I ) 2 p A p B k ex S =! S k ex ξs fast exchange (I-spin SQ) fast exchange (S-spin DQ) -2 S -4 S I = ξ I! I k ex! obs = k2 exp A p B! S +! I R ex =k AB

55 Chemical exchange regimes in 2D experiments: HMQC 4 2 SS + FZFD! obs =! S p B R ex = (!2 S +!2 I )p Ap B k ex S =! S FF SS k ex ξs -2-4 SZSD! obs = k2 exp A p B! S! 2 S! 2 I R ex =k AB ξ I I =! I k ex Skrynnikov (22)

56 Example: fast/fast exchange FZFD MQ 5N lineshape SQ 5N lineshape! obs =! S p B 8 7 k ex s w N 2 s w H 2 s 8 7 R ex = (!2 S +!2 I )p Ap B k ex δn / ppm δn / ppm ξs ξ I

57 Example: slow/slow exchange SZSD MQ 5N lineshape k ex s w N s w H 75 s SQ 5N lineshape! obs = k2 exp A p B! S! 2 S! 2 I R ex =k AB δn / ppm δn / ppm ξs ξ I

58 MQ coalescence point 4 2 S =! S k ex ξs fast exchange (I-spin SQ) fast exchange (S-spin MQ) ξ I I =! I k ex

59 Summary: 2D fast exchange regimes 3 HSQC 3 HMQC 2 2 ξs, ξs, ξ I ξ I 3 HZQC 3 HDQC 3 SIM-H[Z/D]QC ξs, ξs, ξs ξ I ξ I ξ I

60 The SOFAST-SIM-H[Z/D]QC experiment H degree excitation degree excitation x, x.2 5 N δ t δ decouple Sensitivity G z g g2 g Acquisition + recycle delay / ms Separate HZQC and HDQC by postacquisition receiver phase cycling Sensitivity each experiment = / 2 of HSQC

61 The SIM-H[Z/D]QC experiment: cyclophilin HZQC HDQC H chemical shift / ppm H chemical shift / ppm

62 Summary: 2D fast exchange regimes 3 HSQC 3 HMQC 2 2 ξs, ξs, ξ I ξ I 3 HZQC 3 HDQC 3 SIM-H[Z/D]QC ξs, ξs, ξs ξ I ξ I ξ I

63 Paramagnetic longitudinal relaxation enhancement δ N / ppm 5 R,p R 2,p = 2S(S + )µ2 B g2 5r 6 SNR t / s -½ (T e c ) 2 T e a b c mm NiDO2A 4 mm NiDO2A [NiDO2A] / mm SNR t / s -½ δ N d δ H / ppm mm NiDO2A SNR t e T rec / s.. Cai et al. JACS (26) 28, Chan, Waudby et al. J Biomol NMR (25) SNR t -½ f

64 Summary NMR titrations a powerful tool for characterising structural, mechanistic, thermodynamic and kinetic aspects of macromolecular interactions Introduction of 2D lineshape analysis more accurate (differential relaxation, normalisation) more convenient (peak overlap) block residual resampling method for error analysis Applications to simple ligand binding but also more complex models Comparison of SQ, MQ, ZQ and DQ experiments provides additional and complementary information on exchange phenomena Download now for Mac and Linux! Waudby, Ramos, Cabrita & Christodoulou. Sci Rep 6, (26)