79

TROSY and Cross Correlated Relaxation: Transverse Relaxation can be described by energy fluctuations of the energy levels between which the coherence to be relaxed exists: Let us consider a two spin system I and S: Dipolar Relaxation

( )203 3cos 1

4I S

z zIS

D I Sr

γ γ µθ

π−

= −

038

I Sparallel

IS

Dr

γ γ µπ

−=

0316

I Sperpendicular parallel

IS

D Dr

γ γ µπ

= − =

Fig. 1: Energy level diagram for a two spin system I and S. The internuclear vector is either parallel or perpendicular to the external field (middle). The dipolar coupling through space leads to a gain or decrease of the energy of the respective levels resulting in a splitting of the S resonance into a doublet with orientation dependent splitting. Notice the shift of the two resonances by a maximum of 3Dparallel. The energy modulation of the levels leads to a reducued life time i.e. relaxation of the coherences across the levels. Single quantum coherences are relaxed whereas double quantum coherences are not. This is shown in Fig. 2.

80

2

1 02 34 5

I S cSQ

IS

Tr

γ γ µ τπ

− −=

h 1

/ 2 0DQ ZQT − =

Fig. 2: Translation of the energy modulation of the spin levels by the dipolar coupling in linewidth for single quantum coherence of S (left) and for double quantum coherence in the IS spin system (right). As one can see, the dipolar coupling modulates the energy difference between the levels |αα> and |αβ> as well as between |βα> and |ββ> that both constitute the S spin coherence leading to a line broadening. By contrast, for the double quantum coherence in the IS spin system the energy difference is not modulated leading to no relaxation of this coherence..

81

Thus for the single quantum coherence we obtain a broad peak whereas for the double quantum coherence we obtain a sharp peak. Let us compare this now with the already introduced formalism by A. Palmer: We have the following expression for the dipolar interaction between two spins:

ˆ H klDD = bkl 3

1rkl

2ˆ I k rkl( ) ˆ I lrkl( )− ˆ I k ˆ I l

= bkl Fkl(q) θkl ,φ kl( ) ˆ A kl

(q ) ˆ I k, ˆ I l( )q =−2

+2

∑

with

bkl = −µ0

γ kγ lh4πrkl

3

We are using slightly different definitions for the F and the A tensors but this or course does not change the result:

Tensor Operators in the Rotating Frame and modified Spherical Harmonics for the Dipolar and CSA Interaction. The calibration has been chosen such that

? F(q)(θ,φ)F(-q)(θ,φ)d(cosθ)dφ is independent of q.

Tensor Operators for the Dipolar Interaction

304 kl

lkkl

rb

π

γγµ

h−=

Tensor Operators for the CSA Interaction

0||31 )( Bb kk γσσ ⊥−=

Modified Spherical Harmonics

Frequency

q ˆ A kl(q )( ˆ I k, ˆ I l) ˆ A k

(q )( ˆ I k ) ),(),,( )()( φθφθ qkl

qk FF ωq

−2 38

ˆ I k− ˆ I l

− − ( )φθ i2expsin 223 + ω( ˆ I k ) + ω ( ˆ I l)

−1 38

ˆ I k ,zˆ I l

− − ( )φθθ i+expcossin6 ω( ˆ I l)

−1 38

ˆ I k− ˆ I l ,z 3

8ˆ I k

− ( )φθθ i+expcossin6 ω( ˆ I k )

0 ˆ I k ,zˆ I l, z ˆ I k ,z 1cos3 2 −θ 0

0 14

ˆ I k+ ˆ I l

− + ˆ I k− ˆ I l

+( ) − 1cos3 2 −θ ω( ˆ I k ) − ω ( ˆ I l)

+1 38

ˆ I k ,zˆ I l

+ +kI8

3 ( )φθθ i−expcossin6 ω( ˆ I k )

+1 38

ˆ I k ,zˆ I l

+ − ( )φθθ i−expcossin6 ω( ˆ I l)

+2 38

ˆ I k+ ˆ I l

+ − ( )φθ i2expsin 223 − ω( ˆ I k ) + ω ( ˆ I l)

82

The relaxation of the density matrix is given by:

( ),,

ˆˆ ˆ ˆ( ) (0)V WV W

dt

dtσ σ σ= − Γ −∑

with the relaxation superoperator given as:

ˆ ˆ Γ VWˆ σ = bVbW

ˆ A V(−q ), ˆ A W

(q), ˆ σ [ ][ ]q

∑ * jVWq (ωq )

The spectral density is given as:

jVWq (ωq ) = dτ

0

∞

∫ FV(q )(t)FW

(−q) (t + τ) exp(−iωqτ )

which has the general form for an isotropically tumbling molecule of:

( ) ( )( )

2, 2

213cos 1

5 1q cVW q V W

q c

jτ

ω θω τ

= − +

Thus we find for V=W and considering only the j(0) term the following expression for the relaxation of S spin coherence:

[ ]2 2

( ) ( ) 00 03 3

0

ˆ ˆ ˆˆ , , * ( ) , , * (0)4 4 5

q q q k l k l cVW V V V V VV q z z z z VV

q kl kl

S b b A A S j I S I S S j Sr r

γ γ γ γ τω µ µπ π

−− − − −

=

Γ = = − = − ∑ h h

For double or zero quantum coherence S I± ± , there is no relaxation since it commutes with the dipolar interaction: z zS I . Thus, we find the following:

[ ]2

( ) ( ) 00 3

0

ˆ ˆ ˆˆ , , * ( ) , , * (0) 04

q q q k lVW V V V V VV q z z z z VV

q kl

S I b b A A S I j I S I S S I jr

γ γω µπ

−± ± ± ± ± ±

=

Γ = = − = ∑ h

Thus the formal theory delivers the same result as our handwaving argument. CSA Let us consider the same two spin system consisting of I and S and now take into account chemical shift anisotropy that we consider to be axially symmetric. With the same recipe we arrive at the enery level diagram where we consider only the S spin chemical shift anisotropy for the moment (Fig. 3):

83

( )20 3cos 13 2

xx yySCSA zz z

BS

σ σγσ θ

+ Ω = − −

0

3 2xx yyS

CSA zzB σ σγ

σ+

Ω = −

0

6 2xx yyS

CSA zzB σ σγ

σ+

Ω = − −

2

1 02

42 3 5

xx yy S cSQ zz

BT

σ σ γ τσ− +

= −

Fig. 3: S spin system under the influence of the S spin chemical shift anisotropy. For the CSA parallel to the external field the chemical shift is given by σzz while for the CSA perpedicular to the external field the chemical shift is given by σxx. This shift again leads to line broadening of the S spin resonance.

Thus, both relaxation mechanisms lead to a line broadening of the S-spin lines and vice versa also the I spin lines.

Let us consider CSA now also with the formal relaxation theory:

The Hamiltonian of the CSA in the principal axis frame of the CSA is given by:

ˆ H kCSA , PAS = γ k

σxx +σ yy +σ zz

3Bˆ I k +

+ 13 γ k σ xx −σ zz( ) 2 ˆ I k, x Bx − ˆ I k, y By − ˆ I k ,zBz[ ]+ 1

3 γ k σ yy − σzz( )2 ˆ I k ,yBy − ˆ I k,x Bx − ˆ I k, zBz[ ]

which can be brought into the canonical form in the laboratory frame in the following way:

( ) ( )1 1

, ( ) ( ) ( ) ( ), , , , , ,

1 1

ˆ ˆˆ ˆ ˆ( , ) ( , )CSALF q q q qk k z k k z k z k k k y k k y k y k k

q q

H b F A I b F A Iθ φ θ φ=− =−

= +∑ ∑

Eq. [A.1b] with 1

, 03 ( )k z zz xx kb Bσ σ γ= − and 1, 03 ( )k y yy xx kb Bσ σ γ= −

84

The transverse relaxation is then given by:

( ) ( )

( ) ( ) ( ) ( ), , , ; , , , , ; ,

0 0

( ) ( ), , , ; ,

0

221 1 10 03 3

ˆ ˆ ˆ ˆ ˆˆ , , * ( ) , , * ( )

ˆ ˆ2 , , * ( )

( ) ( )

q q q q q qVW k z k z k k k z k z q k y k y k k k y k y q

q q

q q qk z k y k k k z k y q

q

zz xx k yy xx k

S b b A A S j b b A A S j

b b A A S j

B B

ω ω

ω

σ σ γ σ σ γ

− −− − −

= =

−−

=

Γ = +

+ =

− + − −

∑ ∑

∑

( )( )10 03 3

4( ) ( )5

cyy xx k zz xx kB B τσ σ γ σ σ γ − −

This result reduces to the above formula for an axially symmetric CSA.

Cross Correlation:

However, if the two interactions are correlated then we should consider them together. This leads to the following picture in which we assume that the dipolar coupling and the chemical shift anisotropy are in parallel and that the Dparallel is σ

zz –( σ xx + σ

yy +σ zz)/3. (Fig. 4):

Fig. 4: S spin system under the influence of the S spin chemical shift anisotropy and the IS dipolar coupling. If the CSA is parallel to dipolar coupling the Iβ resonance line does not shift.

As is obvious under these cirmustances the S resonance with the I spin in the β state is not affected by the reorientation of the molecule and the resonance line is therefore not broadened whereas the S resonance with the I spin in the α state is broadened by the sum of the two interactions as depicted in Fig. 5.

85

Fig. 5: S- spectrum assuming a chemical shift anisotropy tensor that is parallel to the dipolar tensor and equal in size. Then the Iβ resonance line is sharp since it is not modulated by the reorientation of the NH vector while the Iβ resonance line is broad due to the concerted modulation by CSA and dipolar coupling. The J coupling is included in order to be able to differentiate the two resonances.

We consider this now in the relaxation formalism again: Here we obtain for a axially symmetric CSA that entertains an angle θ with the dipolar interaction the following result:

( ) [ ][ ] [ ][ ]ααα ISSbSbISSIbSIbIS zazazzDzzD−−•− += ,,,,

52 cτ

( [ ][ ] [ ][ ]αα ISSIbSbISSbSIb zzDzazazzD−− ++ ,,,, ) ( ) cτθ 21cos3

51 2 −

( )5

22/)1cos3(4/ 222 c

aDaD bbbbISτ

θα −++= −

( ) [ ][ ] [ ][ ]βββ ISSbSbISSIbSIbIS zazazzDzzD−−•− += ,,,,

52 cτ

( [ ][ ] [ ][ ]ββ ISSIbSbISSbSIb zzDzazazzD−− ++ ,,,, ) ( ) cτθ 21cos3

51 2 −

( )5

22/)1cos3(4/ 222 c

aDaD bbbbISτ

θβ −−+= −

Obviously, if ( )2/)1cos3(4/ 222 −±+ θaDaD bbbb is 0, the linewidth can be very small. For a

NH bond, the nitrogen as well as the hydrogen CSA tensor are almost exactly aligned along the NH

bond. Therefore 2/)1cos3( 2 −θ is close to 1. This means that

( ) 02/)1cos3(4/ 222 =−−+ θaDaD bbbb .

86

The correlation of CSA and dipolar coupling is not the only possibility to obtain differentially broadened lines. This can also be achieved through the interaction of two dipolar couplings on a CH2 group. This is summarized in Fig. 6.

Fig. 6: Assuming a linear ISI spin system, the outer triplet lines will be broad while the inner triplet lines are narrow. We can also consider a spin system in which the two IS vectors are orthogonal to each other. In this case we obtain the result shown in Fig. 7. Whereas in the linear spin system the outer lines of the triplett were broad, now the inner lines become broader than the outer lines. Thus there is a geometric dependence of the cross correlated relaxation effects.

87

Fig. 7: Cross correlation of two dipolar IS couplings. Here, the inner lines become broader than the outer lines. There is a geometric relationship of the differential line broadening of the multiplet lines.

88

Analyzing these two situations with the relaxation formalism we obtain:

( )

( )( )

1 2 1 1 1 1 1 2 2 2 2 2 1 2 1 1

1 1 2 2 1 2 1 1 2 2 1 2 1 2

2 2

, , , , (0)

, , , , (0)

2 3cos 15

D z z D z z D z z D z z D D

D z z D z z D z z D z z D D

cD

S I I b I S b I S S I I b I S b I S S I I j

b I S b I S S I I b I S b I S S I I j

b

α α α α α α

α α α α

τθ

•− − −

− −

= +

+ + =

+ −

The same formula is found for 1 2S I Iβ β− . For 1 2S I I

α β− and 1 2S I Iβ α− we find:

( )

( )( )

1 2 1 1 1 1 1 2 2 2 2 2 1 2 1 1

1 1 2 2 1 2 1 1 2 2 1 2 1 2 1 2

2 21 2

, , , , (0)

, , , , (0)

2 3cos 15

D z z D z z D z z D z z D D

D z z D z z D z z D z z D D

cD

S I I b I S b I S S I I b I S b I S S I I j

b I S b I S S I I b I S b I S S I I j S I I

b S I I

α α α

α α α

α

β β β

β β β

β τθ

•− − −

− − −

−

= +

+ + =

− −

It is obvious that the inner lines originating from 1 2S I Iα β− and 1 2S I I

β α− are not relaxed by the two dipolar interactions for q = 0, while the two outer lines become very broad. For θ =90, however, we find that now the outer lines are narrower than the inner lines by a ratio of 1/3. Cross correlated relaxation of double and zero quantum coherences: So far we have mainly looked at single quantum coherences and how they are affected by cross correlation of tensorial interactions. However, such interactions are not limited to single quantum coherences but can also be observed e.g. on double or zero quantum coherences. Let us assume an I1S-I2T spin system that is subject to two dipolar couplings. This is described in Fig. 8.

89

Fig. 8: Double quantum coherence of S and T spins are affected by two correlated dipolar couplings between S and I1 as well as T and I2 . When the two couplings are parallel to each other, the αα and the ββ lines are broadened.

Structure of Bound Epothilone

The High-Resolution Solution Structure of Epo-thilone A Bound to Tubulin: An Understanding ofthe Structure–Activity Relationships for a Power-ful Class of Antitumor Agents**

Teresa Carlomagno,* Marcel J. J. Blommers,*Jens Meiler, Wolfgang Jahnke, Thomas Schupp,Frank Petersen, Dieter Schinzer, Karl-Heinz Altmann,and Christian Griesinger*

Most of the drugs available today for the treatment of cancerare based on the inhibition of cell proliferation and inductionof cell death by apoptosis. At the molecular level, the majorityof cytotoxic agents interfere with DNA function. However, aprominent subclass of compounds, including paclitaxel(taxol), exert their cytotoxic activity by perturbing micro-tubule dynamics. In 1993, H&fle, Reichenbach, and coworkersreported the isolation and characterization of a novel class ofcytotoxic polyketide macrolides from the myxobacteriumSorangium cellulosum, which were named epothilones (1).[1,2]

The interest in these natural products immediately soaredwhen it was found that epothilones are microtubule stabilizersand that they inhibit cell proliferation through a mechanismof action analogous to that of the renowned clinical anti-cancer drug taxol.[3] Epothilones exhibit extraordinary anti-proliferative activity in vitro and they efficiently induce celldeath in paclitaxel-resistant tumor cell lines at up to 5000-fold

[*] Dr. T. Carlomagno,++ Prof. Dr. C. Griesinger, Dr. J. Meiler+

Max-Planck-Institut f r Biophysikalische ChemieAm Fassberg 11, 37077 G,ttingen (Germany)Fax: (+49)551-201-2202E-mail: taco@nmr.mpibpc.mpg.de

cigr@nmr.mpibpc.mpg.de

Dr. M. J. J. Blommers,++ Dr. W. JahnkeCore Technologies, Novartis Pharma AGP.O. Box, 4002 Basel (Switzerland)Fax: (+41)61-324-2686E-mail: marcel_jj.blommers@pharma.novartis.com

Dr. T. Schupp, Dr. F. PetersenLead Discovery Center, Novartis Pharma AGP.O. Box, 4002 Basel (Switzerland)

Dr. K.-H. AltmannCorporate Research, Novartis Pharma AGP.O. WKL-136.5.22, 4002 Basel (Switzerland)

Prof. Dr. D. SchinzerChemisches Institut, UniversitGt MagdeburgUniversitGtsplatz 2, 39106 Magdeburg (Germany)

[þ] Present address: University of WashingtonBOX 357350, Seattle, WA 98195 (USA)

[++] Both authors have contributed equally.

[**] This work was supported by the MPG, the DFG (Gr1112/11-1, Schi230/14-1) and the Fonds der Chemischen Industrie. J.M. wassupported by a stipend by the Fonds der Chemischen Industrie. Dr.G. Rihs and H. R. Walter of Novartis Pharma AG are thanked for thestructure of epothilone A by X-ray crystallography. Dr. M. O.Steinmetz is acknowledged for his work to determine the polymer-ization state of tubulin in the NMR sample.

AngewandteChemie

2511Angew. Chem. Int. Ed. 2003, 42, 2511 – 2515 DOI: 10.1002/anie.200351276 2003 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

lower concentrations than taxol.[3–5] In addition, they are moresoluble than taxol in water,[2, 4] which allows delivery in vivowith non-chremaphor-containing formulation vehicles, thuseliminating formulation-based side effects (major hypersen-sitivity reactions (HSR)).[6] The potential clinical utility ofepothilones is supported by in vivo experiments with epothi-lone B in a variety of nude mouse human-tumor models.[4,7]

Shortly after the elucidation of the absolute stereochem-istry of epothilone B by a combination of X-ray crystallog-raphy and chemical degradation studies,[2] routes for the totalsynthesis of the epothilones were devised.[8–10] To date, morethan 20 total syntheses of epothilones A and B have beenpublished.[8–14] At the same time several hundred analogueshave been prepared and their biological activity investigated,which has led to a remarkably comprehensive understandingof the structure–activity relationships (SAR) for epothilone-derived structures.[11,12,15,16] Recently, the structure of a 2Dtubulin polymer sheet in complex with docetaxel (taxotere)was solved at 3.7-9 resolution by electron crystallography.[17]

However, in the absence of information on the bioactiveconformation of epothilones any 3D model of the tubulin–epothilone complex remains uncertain.[18–23]

Herein we describe the conformation of epothilone Abound to tubulin, as determined by NMR spectroscopy inaqueous solution. This structure provides the first piece ofexperimental information on the active conformation of thisclass of microtubule stabilizers. The conformational differencesobserved between the X-ray crystal structure and the tubulin-bound conformation of epothilone A allow the rationalizationof the biological data available on the tubulin-polymerizationactivity of several epothilone derivatives.

Both epothilones A and B, the two major naturallyoccurring epothilone variants, displace 3H-taxol from tubulin(Ki= 1.4 and 0.71 mm for epothilone A and epothilone B,respectively),[5] indicating that the tubulin-binding sites of theepothilones and taxol overlap.

The high molecular weight of the tubulin–epothilonecomplex (~ 100 kDa) renders a tight complex a difficult targetfor NMR-based structural studies. However, if the binding ofthe ligand to the receptor is weak (Kd> 10 mm, Koff> 100 Hz),transferred nuclear Overhauser enhancement (trNOE)[24] andtransferred cross-correlated relaxation (trCCR)[25,26] methodscan be employed to unravel the conformation of epothilonesin complex with tubulin. When a ligand is in fast exchangewith the receptor, its spectroscopic properties result from theweighted average of those of the free and the boundconformation. NOE interactions and CCR rates are directlyproportional to the correlation time of the molecule, and theiraverage is governed by the contribution of the boundconformation, as a result of the considerably larger correla-

tion time of the complex relative to the free ligand.Fortunately, the kinetic and thermodynamic properties ofthe complex between epothilone A and tubulin are in thedesirable range for trNOE and trCCR experiments, asindicated by the negative sign of the NOE cross-peaks in a100:1 mixture of epothilone A with tubulin. The trNOE andtrCCR data are in agreement with a Kd in the range of 10–100 mm. Evidence for specific binding of epothilone A totubulin is provided by the restoration of the NOE spectrum offree epothilone A upon addition of epothilone B to themixture, which proves the quantitative displacement ofepothilone A from tubulin by the tighter binder epothilone B.The existence of specific and transient binding of epothi-lone A to tubulin enables the structural analysis of the activeconformation of the epothilones by NMR.

A crucial provision for the structural investigation of theepothilone–tubulin complex in solution is the prevention oftubulin polymerization. Electron microscopy and gel electro-phoresis show that substitution of Mg2+ with Ca2+[27] and D2Oas solvent[28] are sufficient to prevent tubulin polymerization.Samples were stable for approximately one week.

The tubulin-bound conformation of epothilone A wascalculated from 46 interproton-distance restraints and seventorsion-angle restraints measured for a 0.5 mm solution ofepothilone A in water in the presence of 5 mm tubulin. Thedistance restraints were derived from transferred NOEexperiments. To filter out spin-diffusion-mediated peaks,only those signals with opposite sign to the diagonal peaksin a transferred ROESYexperiment were taken into account.The dihedral-angle restraints were obtained by measuringCH–CH dipolar–dipolar and CH–CO dipolar–CSA trans-ferred CCR rates for 60–70% 13C-labeled epothilone A. Thetransferred cross-correlated relaxation (trCCR) experi-ments[25,26] were indispensable to obtain a unique descriptionof the bound conformation, as more than one structure of themacrolide ring is compatible with the same HH distance set(NOE intensities). The seven trCCR rates defined the torsionangles O1-C1-C2-C3, C1-C2-C3-C4, C2-C3-C4-C5, C5-C6-C7-C8, C12-C13-C14-C15, C13-C14-C15-O1, and C14-C15-C16-C17. One particular problem in the determination of thebound conformation of epothilone was the stereospecificassignment of the two methyl groups at C4, which turned outto be opposite to the published assignment (see ExperimentalSection for details).[2]

The tubulin-bound conformation of epothilone A isshown in green in Figure 1 and is compared with the free(unbound) conformation of epothilone A determined by X-ray crystallography,[29] which is shown in gray. We chose tocompare the tubulin-bound conformation of epothilone withthe X-ray structure and not with the solution structureavailable in CD2Cl2, because of the extensive flexibility ofepothilone in solution in the absence of tubulin.[30] However,the most populated conformer in solution is indeed verysimilar to the X-ray conformer. The position of the thiazolenitrogen atom (blue in Figure 1) and of the 3-OH group (red),which are important for the delineation of a pharmocophoremodel, change significantly upon binding.

A comparison of the torsion angles of epothilone A in thetubulin-bound (green) and in the free (gray) conformation

Communications

2512 2003 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim www.angewandte.org Angew. Chem. Int. Ed. 2003, 42, 2511 – 2515

(Table 1) reveals two major changes: the first occurs in theO1–C6 region and the second affects the orientation of thethiazole ring with respect to the C16C17 double bond. In thelower region of the macrocycle, both cross-correlated relax-ation rates (see preceding paper) and NOE data define theconformational change of the dihedral angle C2-C3-C4-C5,which switches from a gaucheþ to a gauche conformation,while the variations of the dihedral angles O1-C1-C2-C3(þanti-periplanar to anti-clinal) and C3-C4-C5-C6 resultfrom the adjustment of the macrocycle to the modified torsionaround the C3C4 bond. The observed conformationalchange in the O1–C6 region of the ring primarily affects theposition of the protons at C2, and of the 3-OH group. The twoprotons at C2 move towards the inner part of the ring. At thesame time the oxygen atom attached to C3, which points tothe inside of the macrocycle in the free conformation, movestowards the exterior by 3.8 9 (i.e. it now points away from themacrocycle). Neither the C5-C6-C7-C8 dihedral angle, deter-mined by trCCR, nor the C10–C15 region exhibit anysignificant conformational change upon binding to tubulin.The latter finding corroborates a recent study in which theconformation of the macrocycle in the epoxide region isinvestigated with analogues of epothilone B and D.[31]

The second significant difference between the free andtubulin-bound conformations of epothilone occurs in the sidechain that bears the thiazole ring. The C16-C17-C18-C19dihedral angle changes from an anti-periplanar to a syn-periplanar conformation. This finding is substantiated by astrong NOE interaction between the protons of Me27 andH19 together with a significant weakening of the NOE

interaction between H17 and H19compared with that observed forepothilone A in aqueous solutionand in the absence of tubulin. Theeclipsed rotamer is energetically dis-favored as a result of the sterichindrance between Me27 and H19.However, in this conformation, thenitrogen atom of the thiazole ringbecomes more accessible for potentialhydrogen-bonding formation withfunctional groups of proteins, whichmay more than offset an intrinsicallydisfavored epothilone side-chain con-formation.

The tubulin-bound conformationof epothilone described herein isconsistent with the wealth of chemicalmodification data available for thisclass of compounds. Epothilones arequite sensitive to chemical modifica-tions in the C4–C8 region. The inver-sion of the stereochemistry at C3 alsoresults in a considerable loss ofpotency in the induction of tubulinpolymerization.[32] On the other hand,a,b-unsaturated (trans C2C3 doublebond) analogues of epothilones Aand B are almost as active as the

natural products. Avery recent report by Vite and co-workersalso showed that the replacement of the hydroxy group at C3

Table 1: Comparison of the dihedral angles of the tubulin-bound and thefree conformations of epothilone A.

Dihedral angle Tubulin-bound Freeconformation [8] conformation [8]

C1-C2-C3-C4 152.50.2 165.4C2-C3-C4-C5 51.70.1 73.0C3-C4-C5-C6 43.01.8 75.9C4-C5-C6-C7 156.40.5 145.6C5-C6-C7-C8 70.00.8 64.0C6-C7-C8-C9 74.80.3 79.2C7-C8-C9-C10 164.10.8 159.0C8-C9-C10-C11 171.90.4 176.8C9-C10-C11-C12 178.00.4 174.8C10-C11-C12-C13 129.20.5 112.8C11-C12-C13-C14 4.10.1 1.9C12-C13-C14-C15 76.30.1 93.9C13-C14-C15-O1 62.61.0 82.6C14-C15-O1-C1 179.50.5 159.6C15-O1-C1-C2 176.31.3 174.2O1-C1-C2-C3 124.31.2 156.7C14-C15-C16-C17 129.71.3 118.5C15-C16-C17-C18 180.00.2 176.0C16-C17-C18-N 180.00.3 7.6

The values given for the tubulin-bound conformation result fromaveraging over the ten lowest-energy conformations. The values given forthe free conformation result from crystal structure analysis of epothilone Ain methyl tert-butyl ether (CH3OC(CH3)3) solvate.[29] This conformation, aswell as crystal packing, is nearly identical to that of epothilone B,[2]

although the two compounds crystallize from different solvents.

Figure 1. Stereoview of a comparison of the tubulin-bound conformation of epothilone A inaqueous solution (green) with the free conformation determined by X-ray crystallography[29]

(gray). Hydrogen atoms have been removed for clarity. The conformational change that occursin the O1–C6 region is clearly visible and is associated with a significant shift in the position ofthe 3-OH group (in red). A second key feature of the bound structure is the 1808 change in theorientation of the side chain, which liberates the nitrogen atom (in blue) of the thiazole ringfrom the bulky Me27 group and makes it accessible to hydrogen-bonding donors.

AngewandteChemie

2513Angew. Chem. Int. Ed. 2003, 42, 2511 – 2515 www.angewandte.org 2003 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

with a cyano moiety leads to a compound that retains highbiological activity.[33]

The hydroxy group attached to C3 is one of the bio-logically most relevant sites. In the free conformation, thisOH group points towards the inner part of the ring and itsoxygen atom is not easily accessible for the formation ofexternal hydrogen bonds; conversely, in the tubulin-boundconformation, the 3-OH group points outwards, which makesthe oxygen atom a potentially better acceptor of hydrogenbonding from tubulin side chains. This observation is con-sistent with chemical modification data, which underline theimportance of the stereochemistry at C3 and therefore of theposition of the 3-OH group. On the other hand, the fact thata,b-unsaturated epothilone analogues are almost as active asthe natural compound indicates that the OH group is notcrucial for tight binding to tubulin. This finding does notnecessarily contradict the hypothesis of the existence ofhydrogen bonding between the 3-OH of epothilone andtubulin side chains, as a double bond can also function as anelectron donor.

The second profound conformational change observedupon binding of epothilone to tubulin concerns the orienta-tion of the thiazole side chain, which undergoes a 1808rotation around the C17C18 bond to liberate the nitrogenatom of the thiazole ring from the hindrance of the methylgroup C27. It might be speculated that the nitrogen atom isnow better available for hydrogen-bonding formation with atubulin side chain. The potential role of the nitrogen atom asan hydrogen-bonding acceptor is corroborated by chemicalmodification data: the tubulin polymerization activity isretained upon substitution of the thiazole ring with a 2-pyridyl moiety, whereas the incorporation of a 3-pyridylresidue causes a notable decrease in the biological activity.[21]

The recognition that the nitrogen atom of the thiazole ringmay be involved in a hydrogen bond is highly relevant, as itmight change the way of projecting new pharmacophoremodels for the orientation of epothilone in the tubulin-binding pocket. To date, all pharmacophore models are basedon the attempted overlapping of the thiazole ring ofepothilone with either the 2-OCOPh or the C3’-Ph of taxol.Nevertheless, the indubitable relevance of the nitrogen atomof the thiazole ring as an hydrogen-bonding acceptor opensthe way to new pharmacophore models in which the nitrogenatom overlaps with a hydrogen-bonding acceptor in taxol.This new working hypothesis is currently under investigationin our laboratory.

So far, the design and synthesis of new potent analogues ofepothilones was based on the available crystal structure offree epothilone. This has led to the discovery of candidateswith promising properties, which, however, are still closelyrelated to the parent compound. Potent analogues withstructures deviating more strongly from the natural productsmight have a higher potential for alterations in the overallpharmacological profile. In addition, such compounds mightbe synthetically more readily accessible than closely relatedanalogues of the natural products. The design of suchanalogues will be greatly aided by the knowledge describedherein about the conformation of epothilone in its tubulin-bound state.

Experimental SectionAll NMR experiments were performed on Bruker DRX 600 MHz orAvance 700 MHz spectrometers. The NOE constraints used tocalculate the structure of the tubulin-bound conformation of epothi-lone were obtained from “transferred-NOE” experiments for asolution of epothilone A (0.5 mm) and tubulin (5 mm) in D2O atmixing times of 100, 200, and 300 ms. The tr-ROESYexperiment hada mixing time of 60 ms with CW irradiation at a field strength of3500 Hz. CH–CH cross-correlated relaxation rates (trCCR-rates) forthe determination of the C1-C2-C3-C4, C5-C6-C7-C8, C12-C13-C14-C15, and C13-C14-C15-C16 torsion angles were measured in HCCHcorrelation experiments, which were adapted from the original pulseprogram[34] to achieve the desired selectivity on the various epothi-lone moieties. The O-C1-C2-C3 torsion angle was measured in adipole-CSA (CSA= chemical-shift anisotropy) trCCR experimentthat was adapted from that used to measure CCR rates in the proteinbackbone.[35] The C2-C3-C4-C5 and C14-C15-C16-C17 torsion angleswere measured with the help of a newly developed trCCR experi-ment.[36]

Tubulin was obtained from cytoskeleton (bovine brain, T238). Theprotein was dialyzed against calcium phosphate (3 mm), pH 7.0,lyophilized, and dissolved in D2O prior to each measurement. All thetrCCR experiments were acquired for a solution of 13C-labeledepothilone A (0.5 mm) and tubulin (5 mm) in D2O. 13C-labeledepothilone was obtained by shake flask fermentation with themyxobacterium Sorangium cellulosum So ce90 by using a high-epothilone-producing mutant, BCE99/41 (Novartis strain collection).The fermentation was performed by growing the strain in medium(1000 mL) containing 13C-labeled starch (13C6, 98%; 20 g) as carbo-hydrate source. Yields: epothilone A (20 mg) and epothilone B(15 mg), incorporation rate: 60–70% 13C.The assignment of the two diastereotopic methyl groups had to beestablished as it was essential to determine the dihedral angle aroundthe C3C4 bond by cross-correlated relaxation and consequentlydetect the large conformational change that takes place in this regionupon binding to tubulin. Owing to the paucity of protons around theC3C4 bond and to the conformational averaging present in thisregion for epothilone in solution in the absence of tubulin, thestereospecific assignment of the methyl groups could not be carriedout by a combination of NOE interactions and 3JCH couplingconstants. Therefore, we relied on the combination of NOE experi-ments, 3JC,H and 3JC,C couplings constants (3JC2,C5, 3JC2,C22, and 3JC2,C23)measured for epothilone B in the absence of tubulin at 20 8C indimethylformamide. Under these conditions, the macrocycle canassume two conformations around the C3C4 bond. Two pictures arecompatible with the measured set of 3JC,H (3JC22,H3 3JC5,H3>

3JC23,H3):1) C22=pro-R, C23=pro-S, and a mixture of gaucheþ and transconformations around the C3C4 bond; 2) proS-C22, proR-C23, anda mixture of gaucheþ and gauche conformations around the C3C4bond. The measured set of 3JC,C couplings (3JC2,C5<

3JC2,C23<3JC2,C22)

unequivocally confirms alternative 2. The resulting stereospecificassignment is opposite to that previously published.[2]

The structure of the tubulin-bound conformation of epothilone wasdetermined by using the simulated annealing protocol, includingNMR restraints, of the X-PLOR program.[37] The force field wasadapted to include atom types occurring in epothilone A. Bondlengths and angles were taken from the X-ray structure of the freeconformation. In addition, the force field included 46 NOE restraintsand seven dihedral angles derived from trCCR rates. From a randomstarting point, an initial energy minimization of 50 steps, a high-temperature phase (32.5 ps, 6500 steps, 2000 K) and two coolingphases (25.0 ps, 5000 steps, 2000 K!1000 K/10.0 ps, 2000 steps,1000 K!100 K) were performed. The closest local minimum wasreached by applying a final energy minimization of 200 steps. Thisprotocol was repeated 100 times. Within the ten lowest energystructures, all dihedral angles have a standard deviation of less than 28and the RMSD of the positions of the heavy atoms is 0.044 9.

Received: February 25, 2003 [Z51276]

Communications

2514 2003 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim www.angewandte.org Angew. Chem. Int. Ed. 2003, 42, 2511 – 2515

.Keywords: natural products · NMR spectroscopy · structureelucidation · structure–activity relationships · tubulin

[1] G. H&fle, K. Bedorf, K. Gerth, Chem. Abstr. 1993, 120, 52841.[2] G. H&fle, N. Bedorf, H. Steinmetz, D. Schomburg, K. Gerth, H.

Reichenbach, Angew. Chem. 1996, 108, 1671 – 1672; Angew.Chem. Int. Ed. Engl. 1996, 35, 1567 – 1569.

[3] D. M. Bollag, P. A. McQueney, J. Zhu, O. Hensens, L. Koupal, J.Liesch, M. Goetz, E. Lazarides, C. M. Woods, Cancer Res. 1995,55, 2325 – 2333.

[4] K.-H. Altmann, M. Wartmann, T. O'Reilly, Biochim. Biophys.Acta 2000, 1470, M79-M91.

[5] J. Kowalski, P. Giannakakou, E. Hamel, J. Biol. Chem. 1997, 272,2534 – 2541.

[6] E. K. Rowinsky, Annu. Rev. Med. 1997, 48, 353 – 374.[7] K.-H. Altmann, G. Bold, G. Caravatti, N. End, A. Florsheimer,

V. Guagnano, T. O'Reilly, M. Wartmann, Chimia 2000, 54, 612 –621.

[8] A. Balog, D. Meng, T. Kamenecka, P. Bertinato, D.-S. Su, E. J.Sorensen, S. J. Danishefsky, Angew. Chem. 1996, 108, 2976 –2978; Angew. Chem. Int. Ed. Engl. 1996, 35, 2801 – 2803.

[9] Z. Yang, Y. He, D. Vourloumis, H. Vallberg, K. C. Nicolaou,Angew. Chem. 1997, 109, 170 – 173; Angew. Chem. Int. Ed. Engl.1997, 36, 166 – 168.

[10] D. Schinzer, A. Limberg, A. Bauer, O. M. B&hm, M. Cordes,Angew. Chem. 1997, 109, 543 – 544; Angew. Chem. Int. Ed. Engl.1997, 36, 523 – 524.

[11] K. C. Nicolaou, F. Roschangar, D. Vourloumis, Angew. Chem.1998, 110, 2120-2153; Angew. Chem. Int. Ed. 1998, 37, 2014 –2045.

[12] C. R. Harris, S. J Danishefsky, J. Org. Chem. 1999, 64, 8434 –8456.

[13] J. Mulzer, H. J. Martin, M. Berger, J. Heterocycl. Chem. 1999, 36,1421 – 1436.

[14] K. C. Nicolaou, A. Ritzen, K. Namoto, J. Chem. Soc. Chem.Commun. 2001, 1523 – 1535.

[15] K.-H. Altmann, M. Wartmann, T. O'Reilly, Biochim. Biophys.Acta 2000, 1470, M79-M91.

[16] M. Wartmann, K.-H. Altmann, Curr. Med. Chem. - Anti-cancerAgents 2002, 2, 123 – 148.

[17] E. Nogales, S. G. Wolf, K. H. Downing, Nature 1998, 391, 199 –203.

[18] J. D. Winkler, P. H. Axelsen, Bioorg. Med. Chem. Lett. 1996, 6,2963 – 2966.

[19] M. Wang, X. Xia, Y. Kim, D. Hwang, J. M. Jansen, M. Botta,D. C. Liotta, J. P. Snyder, Org. Lett. 1999, 1, 43 – 46.

[20] I. Ojima, S. Chakravarty, T. Inoue, S. Lin, L. He, S. B. Horwitz,S. D. Kuduk, S. J. Danishefsky, Proc. Natl. Acad. Sci. USA 1999,96, 4256 – 4261.

[21] K. C. Nicolaou, R. Scarpelli, B. Bollbuck, B. Werschkun,M. M. A. Pereira, M. Wartmann, K. H. Altmann, D. Zaharevitz,R. Gussio, P. Giannakakou, Chem. Biol. 2000, 7, 593 – 599.

[22] P. Giannakakou, R. Gussio, E. Nogales, K. H. Downing, D.Zaharevitz, B. Bollbuck, G. Poy, D. Sackett, K. C. Nicolaou, T.Fojo, Proc. Natl. Acad. Sci. USA 2000, 97, 2904 – 2909.

[23] L. He, P. G. Jagtap, D. G. I. Kingston, H.-J. Shen, G. A. Orr, S. B.Horwitz, Biochemistry 2000, 39, 3972 – 3978.

[24] F. Ni, Prog. Nucl. Magn. Reson. Spectrosc. 1994, 26, 517 – 606.[25] T. Carlomagno, I. C. Felli, M. Czech, R. Fischer, M. Sprinzl, C.

Griesinger, J. Am. Chem. Soc. 1999, 121, 1945 – 1948.[26] M. J. J. Blommers, W. Stark, C. E. Jones, D. Head, C. E. Owen,

W. Jahnke, J. Am. Chem. Soc. 1999, 121, 1949 – 1953.[27] F. Solomon, Biochemistry 1977, 16, 358 – 363.[28] G. Chakrabarti, S. Kim, M. L. Gupta, Jr., J. S. Barton, R. H.

Himes, Biochemistry 1999, 38, 3067 – 3072.[29] G. Rihs, H. R. Walter, unpublished results.

[30] R. E. Taylor, J. Zajicek, J. Org. Chem. 1999, 64, 7224 – 7228.[31] R. E. Taylor, Y. Chen, A. Beatty, D. C. Myles, Y. Zhou, J. Am.

Chem. Soc. 2003, 125, 26 – 27.[32] K. C. Nicolaou, D. Vourloumis, T. Li, J. Pastor, N. Winssinger, Y.

He, S. Ninkovic, F. Sarabia, H. Vallberg, F. Roschangar, N. P.King, M. Ray, V. Finlay, P. Giannakakou, P. Verdier-Pinard, E.Hamel,Angew. Chem. 1997, 109, 2181 – 2187; Angew. Chem. Int.Ed. Engl. 1997, 36, 2097 – 2103.

[33] A. Regueiro-Ren, K. Leavitt, S.-H. Kim, G. H&fle, M. Kiffe, J. Z.Gougoutas, J. D. DiMarco, F. Y. F. Lee, C. R. Fairchild, B. H.Long, G. D. Vite, Org. Lett. 2002, 4, 3815 – 3818.

[34] I. C. Felli, C. Richter, C. Griesinger, H. Schwalbe, J. Am. Chem.Soc. 1999, 121, 1956.

[35] D. W. Yang, R. Konrat, L. E. Kay, J. Am. Chem. Soc. 1999, 121,11938 – 11940.

[36] See following manuscript in this issue: T. Carlomagno, V. M.Sanchez, M. J. J. Blommers, C. Griesinger, Angew. Chem. 2003,115, 2619 – 2621; Angew. Chem. Int. Ed. 2003, 42, 2515 – 2517.

[37] A. T. Bruenger, X-PLOR A System for X-ray Crystallographyand NMR (Yale University Press, New Haven), 1992.

AngewandteChemie

2515Angew. Chem. Int. Ed. 2003, 42, 2511 – 2515 www.angewandte.org 2003 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Torsion Angles of Epothilones

Derivation of Dihedral Angles from CH–CHDipolar–Dipolar Cross-Correlated RelaxationRates: A C–C Torsion Involving a QuaternaryCarbon Atom in Epothilone A Bound toTubulin**

Teresa Carlomagno,* Vctor M. Snchez,Marcel J. J. Blommers, and Christian Griesinger*

Cross-correlated relaxation (CCR) rates constitute a validalternative to coupling constants for the determination ofdihedral angles.[1] For high-molecular-weight systems themeasurement of scalar coupling constants poses seriousproblems caused by large linewidths and by differentialrelaxation effects.[2] Conversely, CCR measurements arefeasible also for molecules of large size and have beensuccessfully employed in the past years to determine dihedralangles in biomolecules.[3–6] We have recently reported theusage of cross-correlated relaxation rates to derive angular

[*] Dr. T. Carlomagno, Prof. Dr. C. Griesinger, Dr. V. M. SnchezMax-Planck-Institut f#r Biophysikalische ChemieAm Fassberg 11, 37077 G.ttingen (Germany)Fax: (+49)551-201-2201E-mail: taco@nmr.mpibpc.mpg.de

cigr@nmr.mpibpc.mpg.de

Dr. M. J. J. BlommersNovartis Pharma AGP.O. Box, 4002 Basel (Switzerland)

[**] This work was supported by the MPG, the DFG and the Fonds derChemischen Industrie. V.M.S. is supported by the EU through aMarie Curie fellowship. All measurements were performed at theLarge Scale Facility for Biomolecular NMR in Frankfurt (HPRI-1999-CT-00014). The authors thank T. Schupp and F. Petersen of NovartisPharma AG for the preparation of 13C-labeled epothilone.

AngewandteChemie

2515Angew. Chem. Int. Ed. 2003, 42, 2515 – 2517 DOI: 10.1002/anie.200350950 2003 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

structural information for a small ligand weakly bound to amacromolecule.[7, 8] Similarly to “transferred-NOEs”,[9,10]

“transferred-CCR rates” are informative of the bound con-formation of the ligand when the complex is in the fast-exchange regime on the NMR timescale. Information on thebound conformation of the ligand cannot be obtained fromcoupling constants, because their value is independent of themolecular correlation time.[7]

Herein we introduce a pulse sequence that measures CH–CH cross-correlated relaxation in a HaCa–Cb(CH3)c moiety, inwhich the second C atom is quaternary. This rate allows thecalculation of the dihedral angle about the CaCb bond.

We have applied the pulse sequence to a 100:1 mixture of13C-labeled epothilone A (1) and tubulin to determine thedihedral angle about the C3C4 bond of epothilone in its

bound conformation. Epo-thilones are very potent anti-tumor drugs that exert cyto-toxic activity by binding tothe 100-kDa tubulin a/b het-erodimers.[11a] Their confor-mation in the bound state isnot known, whereas theunbound conformation waspreviously determined by X-

ray crystallography.[11b] The equilibrium and kinetic constantsof the epothilone A–tubulin complex (KD= 10–100 mm)render it suitable for “transferred” CCR experiments. Fur-thermore, it could be shown by competition experiments withthe strong specific binder epothilone B that there is only onespecific binding site for epothilone A. A sample of 13C-labeledepothilone A was obtained by growing the myxobacteriumSorangium cellulosum on 13C-rich medium.

The dihedral angle about the C3C4 bond of epothilone isaccessible by measuring the dipolar–dipolar CCR ratesbetween the C3–H vector and the C22–H or C23–H vectors.The direction of the three CH bonds in each methyl group isaveraged, as a consequence of the fast rotation around theC4C22 and C4C23 bonds. Therefore, the C22–H and C23–H vectors assume an average orientation that is collinear withthe C4C22 and C4C23 bonds, respectively.

The pulse sequence used for the measurement of theGCH,CH rate between the C3–H vector and the (C22/23–H)av.vectors is shown in Figure 1. Two experiments are recorded,yielding a cross and a reference spectrum (quantitative G

approach).[12,13] The CCR rate is extracted from the ratio ofpeak intensities in the cross and reference experiments. Thepulse sequence is optimized for maximum sensitivity of thecross experiment: the 8Hz

22/23Cy22/23Cz

4Cy3 coherence present

at point c is transformed by GC23–H,C3–H into 8Cx22/23Cz

4Cx3Hz

3

with an efficiency equal to 3sinh(GC22/23–H,C3–HTrel)cosh2(GC22/23–

H,C3–HTrel). The 8Cx22/23Cz

4Cx3Hz

3 term is then transformed intoHx

3 for detection. The delays are optimized for optimalrefocusing and defocusing of the couplings. Selective p pulseson the carbon atoms prevent loss of magnetization to otherpassively coupled carbons. The details of the coherencetransfers are provided in Figure 1 and its legend.

The cross and reference spectra are shown in Figure 2.Two peaks are observed in the reference spectrum at the

chemical shifts of C22 and C23 in w1 and the chemical shift ofH3 in w2. In the cross-spectrum, the C23(w1)–H3(w2) peak ismissing, indicating that the dipolar–dipolar CCR between the(C23–H)av. vector and the C3–H vector is close to zero. Thequantitative value for the two CCR rates can be extractedaccording to [Eq. (1)] (Icross and Iref are the intensities of thepeaks in the cross and reference experiment, respectively andG indicates the cross-correlated relaxation rate).

Icross=Iref ¼ tanhðGTrelÞ=½sin2ðpJC;HD=2Þcos2ðpJC;HD=2Þ ð1Þ

The measured CCR rates between the C3–H vector andthe (C23–H)av. and (C22–H)av. vectors are 0.1 0.1 and1.40.4, respectively. The error was calculated applying errorpropagation theory, assuming that the noise represents theexperimental uncertainty of the peak intensities.

The dependence of the two CCR rates on the C2-C3-C4-C5 dihedral angle q is given in [Eq. (2)] and [Eq. (3)].

GC23H;C3H ¼ 0:3296k S2tcP2½cos2 109:7o sin2 109:7o cosðqÞ ð2Þ

GC22H;C3H ¼ 0:3296k S2tcP2½cos2 109:7o sin2 109:7o cosðq120oÞð3Þ

Figure 1. Pulse sequence for the measurement of dipolar–dipolarcross-correlated relaxation between the (C22/23–H)av. vector and theC3–H vector in epothilone. D=1/(21JC,H); T=1/(41JC,C)=7 ms; Tc=1/(41JC,C)=7 ms; T1=1/(81JC,C)=3.5 ms; T1c=Tc. Delays 1–4 are equal toTrel/4 (Trel=1/1JC,C=28 ms) in the cross-experiment; in the referenceexperiment, delay 1=Trel/4þD/8=delay 3 and delay 2=Trel/4D/8=delay 4. f1= y ; f2= x, x ; f3=2(y), 2(y); f5=4(x), 4(x);f6=8(y), 8(y); f8= x ; f10= y ; f11=2(x,x), 4(x,x), 2(x,x);f4=f7=f9= x. The pulses with phase f4 and f9 are Q3 pulses of768 ms duration, centered at d=63 ppm, to invert C3 and C4 selec-tively (C3 resonates at 70.9 ppm and C4 at 53.0 ppm); the pulse withphase f7 is a Q3 pulse with a duration of 512 ms centered atd=37 ppm, to invert the C22/23 and C4 selectively (C22 resonates atd=19.1 ppm and C23 at d=21.7 ppm). The proton carrier was atd=4.78 ppm and the carbon carrier at d=37 ppm; spectral widthswere 4807.69 Hz for 1H and 8333.33 Hz for 13C. Carbon decoupling inacquisition consisted of a GARP modulated pulse train at 2.38 kHzfield strength.

Communications

2516 2003 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim www.angewandte.org Angew. Chem. Int. Ed. 2003, 42, 2515 – 2517

The (C22–H)av. and (C23–H)av. vectors lie along thedirection defined by the C4C22 and C4C23 bonds,respectively. The CCR rates GC22–H,C3–H and GC23–H,C3–H arescaled by P2[cos109.78]=

1=2[3cos2(109.78)1]=0.3296 as aresult of the averaging of the C–H directions in each methylgroup, (k= (gCgHhm0/8p2)2, tc is the effective correlation timeof the molecule, q is the torsion angle C2-C3-C4-C5 and S2 is theorder parameter accounting for internal motions.)

A graphical representation of the dependence of the twoCCR rates on the torsion angle q is shown in Figure 3 for a tcof 0.74 ns, which is the effective tc of epothilone in thepresence of 5 mm tubulin, as estimated from other CCRmeasurements, and an order parameter S2= 0.8 to account for

internal motions. The stereospecific assignment of the methylgroups (C22 in pro-S and C23 in pro-R) was obtained by thecombination of NOE interactions and 3JH,C and 3JC,C coupling-constant data for epothilone B in the absence of tubulin and isopposite to that reported in the literature.[11b,14]

The dihedral angle q that complies with the ratio betweenthe two CCR rates is 4558. We used the ratio of the twoCCR rates, instead of the rates themselves, to determine thedihedral angle because this approach is independent of both thecorrelation time and the internal motion order parameter S2, ifequal internal reorientation of all the C–H vectors is assumed.

The change in the q dihedral angle from the gauche+conformation in the free form to the gauche conformationin the bound form is supported by transferred-NOE data. Thestructure of the bound form of epothilone determined bytransferred-NOE and transferred-CCR was reported in theprevious communication.[14]

The new experiment allows the determination of X-Ca-Cb-Y dihedral angles by CCR rates when X is a proton and Y is amethyl group. The method is broadly applicable to any HC-CCH3 moiety, and therefore to a number of amino acid sidechains in proteins.

Received: January 16, 2003 [Z50950]

.Keywords: conformation analysis · dihedral angle · naturalproducts · NMR spectroscopy · tubulin

[1] B. Reif, M. Hennig, C. Griesinger, Science 1997, 276, 1230 – 1233.[2] a) G. S. Harbison, J. Am. Chem. Soc. 1993, 115, 3026 – 3027;

b) T. J. Norwood, J. Magn. Reson. Ser. A 1993, 104, 106.[3] D. Yang, K. H. Gardner, L. E. Kay, J. Biomol. NMR 1998, 11,

213 – 220.[4] D. Yang, R. Konrat, L. E. Kay, J. Am. Chem. Soc. 1997, 119,

11938 – 11940.[5] E. Chiarparin, P. Pelupessy, R. Ghose, G. Bodenhausen, J. Am.

Chem. Soc. 2000, 122, 1758 – 1761.[6] P. Pelupessy, E. Chiarparin, R. Ghose, G. Bodenhausen, J.

Biomol. NMR 1999, 14, 277 – 280.[7] T. Carlomagno, I. C. Felli, M. Czech, R. Fischer, M. Sprinzl, C.

Griesinger, J. Am. Chem. Soc. 1999, 121, 1945 – 1948.[8] M. J. J. Blommers, W. Stark, C. E. Jones, D. Head, C. E. Owen,

W. Jahnke, J. Am. Chem. Soc. 1999, 121, 1949 – 1953.[9] a) F. Ni, Prog. Nucl. Magn. Reson. Spectrosc. 1994, 26, 517;

b) L. Y. Lian, I. L. Barsukov, M. J. Sutcliffe, K. H. Sze, G. C. K.Roberts, Methods Enzymol. 1994, 239, 657 – 707.

[10] a) G. M. Clore, A. M. Gronenborn, J. Magn. Reson. 1982, 48,402 – 417; b) G. M. Clore, A. M. Gronenborn, J. Magn. Reson.1982, 53, 423 – 442.

[11] a) G. HHfle, N. Bedorf, K. Gerth, Chem. Abstr. 1993, 120, 52841;b) G. HHfle, N. Bedorf, H. Steinmetz, D. Schomburg, K. Gerth,H. Reichenbach, Angew. Chem. Int. Ed. Engl. 1996, 35, 1567 –1569; Angew. Chem. 1996, 108, 1671 – 1673.

[12] I. C. Felli, C. Richter, C. Griesinger, H. Schwalbe, J. Am. Chem.Soc. 1999, 121, 1956 – 1957.

[13] T. Carlomagno, M. J. J. Blommers, J. Meiler, B. Cuenoud, C.Griesinger, J. Am. Chem. Soc. 2001, 123, 7364 – 7370.

[14] See previous manuscript in this issue: T. Carlomagno, M. J. J.Blommers, J. Meiler, W. Jahnke, T. Schupp, F. Petersen, D.Schinzer, K.-H. Altmann, C. Griesinger, Angew. Chem. 2003,115, 2511 – 2515; Angew. Chem. Int. Ed. 2003, 42, 2615 – 2619.

Figure 2. Reference (left) and cross-spectra (right) acquired with thesequence of Figure 1 on a Bruker DRX 600 MHz spectrometer. Thepeaks visible in the reference spectrum are at the chemical shift of C22(d=19.1 ppm) and C23 (d=21.7 ppm) in the 13C dimension and atthe chemical shift of H3 (d=4.27 ppm) in the 1H dimension. Thecross-experiment was four times longer than the reference experiment.During the experiment, 1024 and 64 complex points were acquired int2 and t1, respectively. The final matrices were 2048K128 points. Thegray lines represent negative contours. The sample contained 0.5 mm

epothilone A and 5 mm tubulin dissolved in D2O.

Figure 3. Dependence of the two cross-correlated relaxation rates GC22–

H,C3–H (in black) and GC23–H,C3–H (in gray) on the dihedral angle q(C2-C3-C4-C5). The two gray shadowed regions represent the value of q thatsatisfy the measured CCR rates. The dashed curve represents the ratioGC23–H,C3–H/GC23–H,C3–H.

AngewandteChemie

2517Angew. Chem. Int. Ed. 2003, 42, 2515 – 2517 www.angewandte.org 2003 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim