Prof. Finneyopened the discussion of Dr Smart's paper: How can you ensure that you are not cross-linking between molecules? Could not such cross-linking explain the peptide association you mention on the seventh page of your paper?Dr Smartresponded: The association you refer to occurs before the (i,i + 7)GCN4-p1 is cross-linked. We find that the circular dichroism signal at 222 nm is anomalously small in comparison with previously published values (ref. 22 and 27 of the paper) yet it shows a marked concentration dependence (Table 1 of the paper). The presence of 3 mM DTT would ensure that the sulfhydryl groups of the cysteine residues are reduced, thus precluding the formation of disulfide bonds. As set out in the Discussion and presented in Fig. 7 we believe that the effect is probably due to the mutation of a number of residues that would create a hydrophobic stripe on the surface of the intact coiled-coil.Dr Sagisaid: The cross-linked peptide has an additional spacer which links the two cysteines. Such an artificial spacer may affect the DNA binding properties of the cross linked peptide to the DNA in terms of affinity and/or the “correct” binding.Dr Smartreplied: The peptide we have described here is a variant of GCN4-p1 which lacks the DNA basic binding helix found on the longer GCN4-bZIP. A clear aim of this work is to produce a variant of GCN-bZIP whose DNA binding is under reversible photo-control. The strategy presented here is to control the dimerisation properties of the zipper part of the protein that is distant from the DNA. A clear alternative would be to try to alter the conformation of the DNA binding helix. As explained in the Introduction of the paper based on the results of Vinson and co-workers (ref. 18 of the paper) we expect control of the leucine zipper part to produce a high degree of photo-control. An advantage of controlling an area distant from the DNA is that the binding specificity of the protein is unlikely to be affected.Dr Techertsaid: 23% of the peptides do not isomerise. What are the alternative reaction channels?Dr Smartanswered: We find for the cross-linked (i,i + 7)GCN4-p1X coiled-coil peptide that illumination with 370 nm for long periods of time resulted in only 62% of the azobenzene chromophores switching to thecisstate, the other 38% remaining in atransconformation. This is a feature of using the azobenzene chromophore—the dark-adapted state is generally >99% of thetransisomer whereas the maximum percentage of thecisisomer that can be achieved by irradiation varies but is typically from 70 to 90%.1It is interesting to note that the maximum percentage conversion achievable with the cross-linker used here varies with the application it is used for. In recent work (ref. 9b in the paper) we have shown, in a peptide where the cross-linker is used to link two cysteine residues with a relative spacing of (i,i + 4), that a proportion of 90% can be switched to thecisstate by irradiation. The half-life for thermal relaxation back to thetransstate is also found to be longer than in the application presented here.1 H. Rau, inPhotochemistry and Photophysics, ed. J. F. Rabek, CRC Press, Boca Raton, 1995, p. 119.Prof. J. C. Smithasked: Can one follow the time course for helix folding in these compounds using molecular dynamics and/or fast spectroscopy? Certainly for molecular dynamics (MD) one now has sufficient computer power to sample peptide conformational equilibria and maybe combining MD with your and other experiments one might learn something about helix initiation and propagation.Dr Smartanswered: We have supplied the original cross-linked monomeric peptide (ref. 8 in the paper) to a number of groups for fast time resolved spectroscopic studies (ORD, absorption/photoacoustic spectroscopy and FTIR methods). Initial preliminary results show that it is possible to trigger photoisomerisation of the chromophore using a laser pulse and that the conformational change in the peptide that follows this is fast (<30 ns), although these results need to be confirmed. Your suggestion for molecular dynamics studies is certainly a good one. In this context Watchtveitl and co-authors have just published a study where molecular dynamics simulation is linked to femtosecond time resolved spectroscopy results for a seven residue peptide that is backbone-cyclized with an azobenzene derivative. Upon photoisomerization of the chromophore the peptide switches between two distinct coil conformations (in DMSO) and an encouraging correspondence of results between simulation and experiment is observed.11 S. Spörlen, H. Carsten, H. Satzger, C. Renner, R. Behrendt, L. Moroder, P. Tavan, W. Zinth and J. Wachtveil,Proc. Natl. Acad. Sci. USA, 2002,99, 7998.Mr Terieteasked: Can the observed difference in relaxation of the azobenzene switch be correlated to internal kinetic processes involved in the formation and folding of the helix? Has the system been investigated using NMR spectroscopic methods?Dr Smartreplied: As detailed in the response to the question of Dr Techert both the maximum percentage conversion of the azobenzene chromophore to thecisform and its thermal relaxation lifetime do depend on the ground state structure of the peptide to which it is attached. To what extent the isomerization of the linker and the refolding of the peptide are temporally coupled is presently unknown.We have been studying the original cross-linked monomeric peptide (ref. 8 in the paper) by NMR methods and will report the results soon.Dr Hädenersaid: Is it clear that the wavelengths of the light used to induce isomerisation of the azo groups will be the same for the free crosslinker as compared to the crosslinker attached to the cysteines? In the latter case the isomerisation will be sterically hindered and this may affect the energy of the photon needed to isomerise the azo groups.Dr Smartresponded: We have not observed such an effect. The absorption spectrum of the azobenzene cross-linker before reaction, once attached to a number of peptides or reacted with glutathione, is essentially identical.1As noted in the paper and in previous replies, both the maximum percentage conversion of the azobenzene chromophore to thecisform and its thermal relaxation lifetime do vary with the context. It should be noted that the absorption spectrum of azobenzene derivatives does depend on the nature of the substituent on the ring. Unmodified azobenzene in thetransform has an absorption maximum around 340 nm.21 G. A. Woolley, personal communication.2 H. Rau, inPhotochemistry and Photophysics, ed. J. F. Rabek, CRC Press, Boca Raton, 1995Prof. Wilsonsaid: I am interested in the possibility of gaining more information on the “re-coiling” of the helix in the slow 10–30 min relaxation process. Has this been examined and is there potential there to use this system (admittedly in the presence of the zipper) as a model for protein folding.Dr Smartreplied: The relaxation process that has a half-life of 10 to 30 min is the thermal revision of the azobenzene chromophore from itscisto the dark-adaptedtransform. It is quite possible to use circular dichroism to monitor the secondary structure of the peptide during this revision. However, because conformational changes of peptides occur on time scales that are very many orders of magnitude faster the results would simply reflect the particular proportions ofcisandtransforms at each particular time. It can be noted that for the three different monomeric peptides that we have studied in detail (ref. 8 and 9b in the paper) CD results show isodichroic points. This is consistent with a simple two-state equilibrium between the disordered and helical states of the proteins.Prof. Wilsonsaid: By how much can the relaxation process (fromcis-azobezene/uncoiled helix totrans-azobenzene/coiled helix) be alterede.g.by the use of laser/light flash. Further, by how much is the process speeded up if the azobenzene is removed/excised from the uncoiled protein by chemical means.Dr Smartanswered: In the coiled-coil peptide we have described we have produced a system where thetransform of the cross-linker produces a coiled-coil with a lower helicity than that produced by irradiation with 370 nm light. In monomeric peptides it is possible to initiate photo-isomerisation using a pulse of laser light (as described in a previous reply). It would be difficult to specifically remove the cross-linking compound from the peptide by chemical means without destroying the peptide. In the paper we have described in detail the properties of the GCN4-p1 derivative prior to cross-linking.Dr Hirstasked: In your paper, you report that the helicity of native GCN4 drops by about 25% when the pH changes from pH 7 to pH 5. Why might that be?Dr Smartanswered: We have indeed found this effect, in agreement with the previous results of Sosnick and co-workers (ref. 26 and 27 in the paper), but can offer no explanation. That such a subtle change in pH produces a large effect on the circular dichroism signal of a well-understood coiled-coil is of interest.Dr Grantasked: Have you ever considered using the technique of Raman optical activity (ROA) as it gives more detailed structural information than just CD, although ROA is not a time-resolved technique, if revision back totransis slow enough, it may be possible to employ this technique.Dr Smartreplied: That is a very good suggestion. We are also interested in the possibility of using synchrotron radiation circular dichroism as this offers the potential to acquire more detailed information with good time-resolution.Prof. J. C. Smithasked: A speculative question—various groups are working on using synthetic metal chelates to design sequence-specific DNA cleavage systemsi.e., artificial restriction enzymes. Is there any chance (or interest) in photoactivating such processes?Dr Smartanswered: The chemical nucleases produced by Bailly and co-workers1do not have a peptide component and so our approach is not immediately applicable. However, in this context Komiyama and co-workers have produced impressive results in oligonucleotides incorporating azobenzene in place of a DNA base.2,3Most notably they have produced an oligonucleotide whose binding to DNA can be utilized to photo-regulate the action of DNA polymerase enzyme on the complimentary strand.41 S. Routier, H. Vezin, E. Lamour, J. L. Bernier, J. P. Catteau and C. Bailly,Nucl. Acid Res., 1999,27, 4160.2 X. G. Liang, H. Asanuma and M. Komiyama,J. Am. Chem. Soc., 2002,124, 1877.3 H. Asanuma, X. G. Liang, T. Yoshida and M. Komiyama,ChemBioChem, 2001,2, 39.4 A. Yamazawa, X. G. Liang, H. Asanuma and M. Komiyama,Ang. Chem. Int. Ed., 2000,39, 2356.Prof. Moffatsaid: What would be the ideal set of properties for a light-dependent, artificial transcription factor (or DNA binding protein)?Dr Smartreplied: An ideal light-dependent transcription factor would have:(i) A complete photo-control of the DNA binding process, that is absolutely no DNA binding in the light but completely normal binding upon irradiation.(ii) Use a chromophore that is altered by a long wavelength of light in comparison to azobenzene. This is important to avoid both scattering and radiation damage effects to living cells. Furthermore the chromophore should have a high quantum yield so that only short irradiation would be required.(iii) Have rapid reversible photo switching between the on and off states such that it is possible to turn on the protein for a given time and by using an alternative wavelength. This would allow the equivalent of single turnover experiments on enzymes by watching the effect in a cell of a small “dose” of the active protein.(iv) Be easily produced—most ideallyin vivoby standard molecular biology procedures with no chemical intervention so that it can be assayed in the cell in which it is expressed.Prof. Helliwellasked: What are the applications in biotechnology?Dr Smartanswered: The technology is still in a development phase. We have demonstrated how a photo-isomerisable crosslinker can be used to control the conformation of peptides and coiled-coil molecules. The next stage is to apply the method to recombinant proteins and show how activity can be controlled. The initial application will be to DNA binding proteins and work is in progress.Prof. Moffatsaid: Biological systems have evolved various forms of chemistry for absorbing light and harnessing its energy to drive certain biological processes. Can you learn from nature, to direct your chemical approaches?Dr Smartreplied: Compared to the natural biological systems you refer to our approach must be seen to be crude. The subtlety involved in the mechanisms of the rhodopsins and other photochromic proteins is impressive. Compared to a chemical approach where it is necessary to purify and modify proteinsin vitroa biologically based method of photo-control would have great advantages. For instance if it might be possible to achieve photo-control by using a fusion method where a given target was produced together with a natural photosensitive protein. This would have the advantage of being easily applied toin vivosystems.Prof. Sir John Meurig Thomasopened the discussion of Dr Techert's paper: Your interesting work on the topochemical photodimerization ofp-formyl-trans-cinnamic acid, where the crystalline monomer is converted to a polycrystalline (amorphous?) dimer, prompts me to draw to your attention a very nice example of photodimerization where a single-crystal → single-crystal transformation occurs. This (diffusionless) 2 + 2 photodimerization takes place1in the archetypal molecule, 2-benzyl-5-benzylidenecyclopentanone (BBCP) and in itsp-bromo derivative (BpBrBCP). My colleagues and I were able to record in a time-resolved fashion (on a 4-circle diffractometer) detailed crystallographic measurements on these materials at various stages of solid-state conversion at room temperatures. In particular, for the case of BBCP, the gradual formation of the cyclobutane ring between the two monomers in an incipient dimer could be directly identified2(see Fig. 1 presented here).The gradual formation of the cyclobutane ring between the two monomers of BBCP in an incipient dimer.I am also prompted to raise some more general issues concerning topochemical photo-induced transformations in organic molecular crystals. My own interest in such phenomena began when it emerged that the topochemical postulates of Schmidt and Cahen were, in some cases, vitiated. For example, anthracene may be photo-converted in the solid state to yield the diplanar-dimer, yet the structure of the stable, monoclinic form of anthracene should not permit such a transformation to occur. Even more puzzling was the observation that 9-cyanoanthracene (9CNA) yields the “wrong” dimer,i.e.thetransrather than thecisform, although the packing in the solid state of 9CNA suggests ready formation of thecisdimer, just as your cinnamic acid monomers yield the truxinic acid dimer. Why is this so? It is because crystalline defects hold sway. Excitation energy migrates through the crystal and is trapped at defects, where the local structure is what matters.At stacking faults (bounded by two partial dislocations) in 9CNA juxtaposed incipient dimers are in thetrans, not thecisconfiguration. And in anthracene, certain types of strucutal faults—like those favoured by the application of gentle pressure that facilitates the production of a metastable form of anthracene in which incipient dimers are brought within the distance (ca.4 Å) required for photodimerization—are also implicated in the photoreactivity.There are many other examples (e.g.1,8-dichloro-9-methylanthracene and 1,8-dichloro-10-methylanthracene), where my colleagues and I found that photodimerization yields the “wrong” dimer; and the explanation in each case is to be found in the role of structural faults.5With the refined techniques that Techert, Wulff and coworkers report in their paper, it would be profitable to re-investigate some of these seemingly anomalous instances of topochemical reaction.1 H. Nakanishi, W. Jones, J. M. Thomas, M. B. Hursthouse and M. Motevalli,J. Phys. Chem., 1981,85, 3636.2 J. M. Thomas,Nature, 1981,289, 633.3 M. D. Cohen, J. M. Thomas, J. O. Williams and Z. Ludmer,Proc. R. Soc. London, Ser. A, 1971,324, 459.4 J. M. Thomas, G. M. Parkinson, S. Ramdas, M. J. Gorringe, C. M. Gramaccioli, G. Filippini and M. Simonetta,Nature, 1980,284, 153; see also S. Ramdas, J. M. Thomas, J. O. Williams and G. M. Parkinson,Eighth Molecular Crystals Symp., Santa Barbara, 1977.5 J. M. Thomas,Philos. Trans. R. Soc. London, Ser. A, 1974,277, 251.Dr Techertresponded: Indeed, the suggested 2 + 2 photodimerisation you refer to in ref. 1 came to our attention recently and we decided to work on it in more detail since it is a full single-crystal–single-crystal transformation. In this context it will be particularly interesting to measure and to understand the differences in the kinetics if, as in the presented work, the product state is amorphous (or less ordered) or if the product state is fully periodically lattice ordered.9CNA and other derivatives are also classical and interesting candidates for investigations of topochemical reactions.Prof. Coppensasked: What drives the condensation of the phonon mode into the metastable state you propose. Can you reproduce the intensity changes reproduced in Fig. 12 (of the paper) with the model? What is the percent conversion by each X-ray flash that gives rise to your signal?Dr Techertanswered: In Fig. 1 (of the paper) only theeffectivereaction mechanism with the effective reaction coordinate of the dimerisation process is shown. However, if one looks more carefully into photo-induced dimerisation processes (e.g.taking the Woodward–Hoffmann rules into consideration), then one has to study in more detail the shape of the potential energy hypersurface (PES) of the ground state as well as of the electronically excited state or states in which the sample is excited by 267 nm excitation wavelength. Depending on the interactions of the different PESs of the excited states, level-crossings and adiabatic transitions between the different excited states drive the reaction to particular product states. Here, we think that the proposed metastable state is connected with a local minimum on the excited state PES decaying from there to the product state. This local minimum is populatedviaan optically allowed vibronic transition where the electronic state and one (or more) phonon modes are excited.At the moment we are performing simulations employing density functional theory of this system in order to obtain a more detailed understanding of the found transient signal changes. In this context it is necessary not only to study the geometrical aspects but also to simulate the energy levels of the (different distorted) lattices in the ground state as well as the electronically excited state PESs.Taking the preliminary state of the simulation into consideration, the intensity changes in Fig. 12 are reasonably well reproduced. However, the quality of the measurements as well as the quality of the theoretical simulations have to be improved further in order to obtain a complete and quantitative picture of the dimerisation mechanism.Due to laser damage problems, p-FCA cannot be excited with too high a laser power. From spectroscopic measurements we approximated that a maximum of 5–8% of the chromophores are excited in one laser flash. This corresponds to a relative change in X-ray signal of about 2–4% (maximum) in the reported system.Dr Nibberingasked: Are the phonons that are supposed to play a role in the reaction dynamics of the acoustic or optical type? If they are optical phonons, one could drive these with optical light pulses through the mechanism of impulsive stimulated Raman scattering. As a consequence, the reaction dynamics will be influenced when the phonons are light-driven, and one may hope to perform some form of optical control of the reaction dynamics.Dr Techertanswered: According to the selection rules, the proposed mechanism is based on optically active phonons, which drive the process. Concerning the energies of these phonon modes, frequencies in the THz regime are assumed—as,e.g., in solid beryllium or InSb.1Controlling the reaction dynamics by the impulsive stimulated Raman scattering technique is a very interesting approach, in particular since only a little work has been carried out concerning the vibrational spectroscopy/analysis of this type of reaction.1 D. A. Reis, M. F. DeCamp, P. H. Bucksbaum, R. Clarke, E. Dufresne, M. Hertlein, R. Merlin, R. Falcone, H. Kepteyn, M. M. Murnane, J. Larsson, Th. Misalla and J. S. Wark,Phys. Rev. Lett., 2001,86, 3072.Prof. Bürgiasked: How does the wavelength of the light used for dimerization affect the lack of crystallinity of the product?Dr Techertreplied: This is indeed a very interesting question. At the moment we do not know the answer. Using 267 nm as the excitation wavelength is surely far above the resonant state of the vibronic transition which is necessary to initiate the reaction. Part of the excess energy will be transformed in heating of the sample, which can lead to structural disorder and loosing the crystallinity of the product state. To answer this question, a systematic study of the amorphism of the product state as a function of excess excitation energy is necessary.Prof. Helliwellsaid: Please comment on the X-ray laser potentionalversusradiation damage issues for your particular compounds.Dr Techertcommented: The proposed X-ray FELs will deliver pulses of approximately 1012photons with a duration of about 100 fs that will create an ideal source for the investigation of ultra-fast processes, like structural (re)organisation. The high flux also will enable one to investigate irreversible processes much more easily as well ase.g.nonlinear properties of material. Furthermore the coherence of the XFEL radiation certainly will pave the way to new types of X-ray diffraction experiments.Also Dr Th. Tschentscher, Hamburger Synchrotronstrahlungslabor HASYLAB at the Deutsches Elektronen-Synchrotron DESY, Hamburg.It is clear, however, that sample damage due to the high number of photons must be considered carefully when defining experiments. Here one must take into account the reduced photon flux at the sample due to the throughput of mirrors and monochromator and due to collimation. This flux has to be compared to the photon flux required to measure the effect under investigation. Beam attenuation techniques (as already practised at synchrotrons of the third generation) may be applied to match both values.To estimate the sample damage one has to take into account the particular time structure of the XFEL radiation. The short duration of the pulse, and therefore the short sampling time in diffraction experiments, allows one to extract information before diffusion controlled processes can play a role.Processes occuring at the sub-picosecond timescale therefore are important to consider. If sample damage effects are found to alter the experiments single sample exchange methods can be applied, like liquid jets or sweeping beam techniques. It is also important to consider sample damage due to laser excitation. Here already sample exchange techniques are practised.Dr Techertopened the discussion of Dr Cole's paper: What is the concrete application of the rhenium carbene complex and why is it important to know the triplet structure?Dr Coleresponded: The rhenium carbene complex is one in a series of compounds that has potential as a candidate for a building block as a ‘molecular wire’, and it also has attractive prospects in catalysis (its photolysis might alter the reactivity of the carbene ligand through population of the3MLCT state).Prof. Coppenssaid: You mentioned a life time of 0.5 μs but did not mention the temperature. The life times of excited states of metalloorganic complexes are often quite temperature dependent. Did you obtain the spectra at the same temperature?Dr Colereplied: The emission spectra of the sample were measured at both room temperature and 77 K in dichloromethane solvent and a 4∶1 ethanol–methanol glass matrix, respectively. Their profiles are very similar in both cases thus indicating that the nature of the dominant3MLCT band remains the same at both temperatures,i.e.the lifetime, measured at room temperature, is not significantly altered at 77 K.11 W.-M. Xue, M. C.-W. Chan, Z.-M. Su, K. K. Cheung, S. T. Liu and C.-M. Che,Organometallics, 1998,17, 1622.Prof. Wilsonsaid: We heard that changing temperature dramatically affects the timescales of the process under study. Does it also affect the degree of transformation in the process?A related point—why not “quasi-continuously” pump to achieve higher degrees of transformation and allow higher resolution diffraction experiments (e.g.charge density). Is the problem that molecules are knocked out of the excited state by further irradiation?Dr Coleresponded: It could affect the degree of transformation, but more drastically, it could affect the entire nature of the emissive-state of the sample,e.g.a compound with a dominant3MLCT type emission at room temperature could be overridden by an intraligand (IL) state which dominates at low temperature. (See for example, compound 7 in ref. 1.)By quasi-continuously pump–probing a sample, one is likely to destroy the samplecf.a typical continuous laser provides ∼1018photons per pulse, compared with a pulsed laser of 1014–1016depending on its repetition rate.1 W.-M. Xue, M. C.-W. Chan, Z.-M. Su, K. K. Cheung, S. T. Liu and C.-M. Che,Organometallics, 1998,17, 1622.Dr Chenasked: In your paper, the excited state lifetime was measured in dichloromethane glass at 77 K. However, the diffraction experiments were carried out in the single crystal. The triplet lifetime of the molecule therefore is likely to be shortened. Have you measured the lifetime in the single crystal? With ∼20% excitation states in your paper, how do you deal with triplet–triplet annihilation?Dr Coleanswered: No. Lifetime measurements of samples on a single crystal are practically very difficult although we are presently investigating this as an option. The triplet lifetime may indeed be different in the glass compared with the single-crystal since the solvent viscosity will prevent thermal equilibrium in the rigid 77 K glass in a time comparable with the emission lifetime. Moreover, the nature of the surrounding frozen and disordered solvent environment is likely to affect the lifetime.20% is stated in the paper simply as a likely maximum level of excitation (based on observations from the literature on other samples). The level of excitation for this compound is not reported in this paper, since data refinement is still underway. Therefore, triplet-triplet annihilation is not necessarily a significant effect here.Dr Techertasked: In Fig. 2 of the paper, a laser power of about 50 μJ is used. Does one need these high energies in order to reach a sufficient excitation level in the crystal?Dr Colereplied: Tests were carried out to assess the maximum level of laser power that could be used without destroying the crystal during the experiment. As a result, we arrived at a chosen laser power of 50 μJ. The greater the laser power, the greater the level of excitation and the larger the crystal that one can use for the experiment (c.f.optical penetration depth calculations). Therefore, by using the maximum feasible laser power, we were able to maximise the intensity of X-ray diffraction, as well as the level of excitation, thereby ensuring the best possible contrast between excited and ground state.Prof. Coppensaddressed Dr Techert: You asked why Coleet al. need a power as high as 50 μJ per pulse. We have used 200 μJ per pulse on a 50 μm crystal and found this to give a marginal photon/molecule ratio, so in fact more than 50 μJ per pulse is needed. This leads to the next question. How did you get away with a lower power pulse in the DMABN experiment?Dr Techertresponded: In the experiments reported till now1,2the samples were prepared in such a way that the size of the crystalline powder sample never reached the upper limit of 1 μm guaranteeing an optical density of less than 1. Having such small samples (and calculating the number of chromophores per unit volume), higher laser power than the reported one was not necessary for an efficient excitation.1 S. Techert, F. Schotte and M. Wulff,Phys. Rev. Lett., 2001,86(10), 2030.2 G. Busse, T. Tschentscher, A. Plech, M. Wulff, B. Frederichs and S. Techert, paper presented at this meeting..Prof. Moffataddressed Dr Cole: (i) Are your crystals optically isotropic or anisotropic? (ii) What is the polarization of the incident laser beam? How do (i) and (ii) affect the ability to uniformly stimulate the molecules in the crystal, in all orientations?Dr Coleresponded: (i) They are probably anisotropic. (ii) A vertically polarised beam was used, arriving at the sample from above at an angle of 45 degrees.The level of excitation is therefore likely to vary with changing orientation of the sample, since excitation is based on the transition molecular dipole moment of the molecule. This varying level of excitation may be averaged out, but it is possible to correct for the geometrical effect if one knows the faces of the crystal at two given oscillation angles.Prof. Helliwellcommented: On your data collection strategy, atomic resolution with Laue is possible.11 See D. Gomez de Anderez, M. Helliwell, J. Habash, E. J. Dodson, J. R. Helliwell, P. D. Bailey and R. E. Gammon,Acta Crystallogr., Sect. B, 1989,45, 482–488; M. Helliwell, D. Gomez de Anderez, J. Habash, J. R. Helliwell and J. Vernon,Acta Crystallogr., Sect. B, 1989,45, 591–596.Dr Coleresponded: Yes, atomic resolution is of course possible for small molecule compounds. However, better resolution is generally achieved using the monochromatic oscillatory method, since this is void of any problems of wavelength overlap and the more restricted coverage of the Ewald sphere, common in Laue experiments.Prof. Helliwellsaid: The monochromatic rotating single crystal method will be very sensitive to unit cell changes, unlike Laue geometry, and thus could provoke intensity change errors. (I agree of course that the stroboscopic approach creates a stable transient species for which the conventional monochromatic approach is definitely suited.)Dr Techertsaid: Three years ago we tried to investigate at ID09b of the ESRF time-resolved Laue diffraction on small molecules (in collaboration with Dr S. Arzt). However, at that time with the old set-up, the collected Laue pattern (unit cell size of the system of investigation: about 20 × 30 × 20 Å3) contained 10–20 diffraction spots and we were not able to index the Laue pattern with the available software. Today, the beamline as well as the software have been improved and further attempts would be useful.Prof. Coppenssaid: I would like to clear up a possible misconception about the effect of a cell dimension change on the accuracy of the monochromatic stroboscopic method. In fact two diffraction patterns are measured by alternate light-on light-off measurements; each can be integrated separately with the appropriate cell dimension, if the change is indeed sufficiently large to warrant such a treatment. So there is no reduction in accuracy of the results associated with the change in cell dimensions. It is of course necessary to take a change in cell dimensions into account when transferring ground state molecular geometry from a dark experiment to the light-on structure.Prof. Cernikasked: Are third generation SR sources the best way to achieve short bunch structure as the pulse length is limited by the RF structure? What are the possibilities of using FELs or the lower stimulated femtosecond X-ray1at ALS?1 For example, R. W. Schoenlein, S. Chattopadhyay, H. H. W. Chong, T. E. Glover, P. A. Heimann, W. P. Leemans, C. V. Shank, A. Zholents and M. Zolotorev,Appl. Phys. B, 2000,71, 1–10.Dr Colereplied: The use of FELs are likely to provide X-rays that are too damaging for the crystal, and the ALS developments for achieving femtosecond X-rays are said to give too low flux. I refer to the paper of Moffat presented in this Discussion Meeting where the viability of these options are specifically discussed [see the section ‘Ultra-fast time-resolved X-ray experiments’, part (a)].Prof. Moffatsaid: Femtosecond laser based “slicing” of 100 fs synchrotron X-ray pulses1is indeed feasible; but the number of X-rays emitted by the perturbed “slice” is small; too small for all but experiments that have very strong scatterers or absorbers. The fs free electron laser looks like a more promising route to short X-ray pulses (which are also extremely intense, with full temporal and spatial coherence).1 R. W. Schoenlein, W. P. Leemans, A. H. Chin, P. Volfbeyn, T. E. Glover, P. Balling, M. Zolotorev, K.-J. Kim, S. Chattopadhyay and C. V. Shank,Science, 1996,274, 236–238.Prof. Moffatopened the discussion of Prof. Helliwell's paper: With the benefit of hindsight, this is a remarkably complicated reaction with numerous chemically-distinct microstates. Nevertheless, the crystals are active. Please comment on the overall suitability and/or desirability of this system for time-resolved studies.Prof. Helliwellresponded: I agree that this is a complicated reaction but hydroxymethylbilane synthase (HMBS, EC 4.3.1.8) is a very important enzyme1,2because of its fundamental position in the metabolism of all organisms, except viruses, in the production of haem, vitamin B12, chlorophylletc. The complexity of the reaction constitutes a challenge, to all available techniques, to examine the enzyme's structure and function if we are to understand it.Time-resolved crystallographic studies are a valuable option for the following reasons: (i) HMBS is a rather slow enzyme (kcat = 0.1 s−1for HMBS fromE. coliunder optimal conditions) and can therefore readily be examined under pre-steady-state conditions; (ii) the substrate can gain access to the crystalline enzyme's active site which is open to the solvent channels of the crystal; (iii) the crystal habit has a favourable shape (thin plates) for quick diffusion of small molecules into the crystal so that the flow-cell method of reaction initiation would work; (iv) the time-dependent changes of electron density are expected to be substantial, as movements of a number of aromatic rings are involved during a single, polymerisation-like, catalytic cycle (see Fig. 1 of our paper); (v) the crystals are relatively resistant to white radiation and the crystal mosaicity remains small also even after substrate diffusion and colour transformation from colourless to red have occurred; (vi) several mutant variants of HMBS have been constructed that exhibit altered reaction kinetics; among these is the Lys59Glu mutant known to accumulate ES2as a predominant intermediate (after 2 min for Lys59Gln HMBS immobilised on an anion-exchange column).3These are technical feasibility reasons of choice. In addition it is worth emphasising that there are no good competitive inhibitors of HMBS known that could allow traditional inhibitor soak experiments followed by crystallographic analysis to determine the location of the inhibitor in the active site in three-dimensional space; a new approach was needed.The fruits of the time-resolved Laue protein crystallography approach thus far are: (i) the three-dimensional binding site for substrate and/or product within the enzyme active site has been experimentally located; (ii) protein engineering results, namely that mutations of Asp 84, Arg 149, Arg 155,i.e.residues proximal to the elongated electron density, which are known to affect the various stages of the enzyme reaction, are supported; (iii) the present results now guide future experiments, involvinge.g.when to freeze-trap the reaction in the crystal, and provide the basis for further efforts to determine the function of the mobile loop of residues 43 to 59. Overall the HMBS enzyme structure–function relationship, including the correlation of electron density changes with the colour transformation from colourless to red in the crystal, has been successfully probed by direct experiment.1 A. R. Battersby, C. J. R. Fookes, G. W. J. Matcham and E. McDonald,Nature, 1980,285, 17–21.2 A. R. Battersby and F. J. Leeper,Chem. Rev., 1990,90, 1261–1274.3 A. C. Niemann, P. K. Matzinger and A. Hädener,Helv. Chim. Acta, 1994,77, 1791–1809.Prof. Wilsonasked: What is the rationale for the difference in timescales for the build-up of enzyme–substrate complexes (2 h in the diffractioncf.2–10 min expected)? Is this a solid state effect (e.g.compared with solution)?Prof. Helliwellanswered: There may be several reasons as to why the elongated electron density builds up slowly in the enzyme active site: (i) the crystal lattice may restrict conformational change(s) for optimal enzyme reaction; (ii) the pH at which the crystals are grown is non-optimal for the enzyme reaction; (iii) the delivery of the substrate to the enzyme active site through the solvent channels of the crystal may cause delay (however, as Helliwellet al.1analyse, the HMBS crystal habit of thin parallelepipeds, 50 μm thick, and the substrate size are estimated to facilitate diffusion of PBG into the crystal in seconds. The flow delivers approximately one substrate molecule per enzyme molecule per second which is far in excess of what Lys59Gln HMBS, even if it were to be saturated with substrate, can turn over at room temperature and pH 5.3). Reason (ii) above is probably the most likely but since the crystals are not stable at the enzyme's optimum pH this is not testable.1 J. R. Helliwell, Y.-P. Nieh, J. Raftery, A. Cassetta, J. Habash, P. D. Carr, T. Ursby, M. Wulff, A. W. Thompson, A. C. Niemann and A. Hädener,J. Chem. Soc., Faraday Trans., 1998,94, 2615–2622.Prof. J. C. Smithasked: Is the reason that the ‘missing loop’ remained undetected that there is static disorder in the crystal at low temperaturesi.e., the loop is trapped in different energy minima, and, if so, might there be a diffuse scattering signature for this disorder?Prof. Helliwellreplied: This is possible but the specific-to-this-loop diffuse scattering signature is likely to be difficult to extricate from other contributors to the X-ray background notably from solvent and the rest of the protein diffuse scattering. Nevertheless it is an interesting suggestion especially if it were possible to make some of the residues on the loop selenomethionine variants (i.e.those that are hydrophobic could be substituted by selenomethionine). Then the diffraction recorded at two wavelengths, at thef′ dip of the seleniumKedge and a 'remote' wavelength, would allow a difference, Δf′, diffuse scattering pattern to be measured. The precise localization of the loop from such measurements would be difficult but could involve matching to computational models to ascertain where the loop might be. This is an interesting suggestion. Thank you.Mr Terietesaid: On the flexible loop, are there any glycines or prolines on dominant start and end positions, where the glycines could work as hinges or the proline cause an alteration between two conformations, throughcis–transisomerisation?Prof. Helliwellreplied: TheE. coliloop amino acid sequence is, starting from residue 43, Thr-Arg-Gly-Asp-Val-Ile-Leu-Asp-Thr-Pro-Leu-Ala-Lys-Val-Gly-Gly-Lys,i.e.finishing with residue 59. So, ‘yes’ there are glycines located at putative hinge positions to facilitate conformational change. Regarding prolines, it is established that these (staying intrans-configuration) tend to restrict conformational change of a polypeptide.1The proline near the centre of the mobile loop may therefore restrict conformational freedom near this part of the loop.Cis–transisomerisation, however, as you suggested, seems unlikely to me, sincecis-prolines are quite rarely found in proteins (within a database of 571 representative proteins, only 5,2% of the Xaa-Pro peptide bonds are incis-, the rest are in thetrans-configuration.2It is notable also that the conserved residues between HMBS fromE. coliandM. tuberculosishere are Thr 43, Gly 45, Asp 46, Pro 52, Ala 54, and Gly 57 (E. colinumbering).1 C. Branden and J. Tooze,Introduction to Protein Structure, Garland Publishing, New York, 1991, p. 259.2 M. S. Weiss, A. Jabs and R. Hilgenfeld,Nature Struct. Biol., 1998,5, 676.Prof. J. C. Smithasked: Is time-resolved solution X-ray scattering sensitive enough to resolve the inter-domain changes you are looking for?Prof. Helliwellanswered: Time-resolved solution small angle X-ray scattering (SAXS), involving synchrotron radiation, could resolve changes in radius of gyration of around 1 Å. This might not be enough but perhaps more helpful is the current practice in SAXS of recording as high angle diffraction as possible. Thus the shape of the protein is estimated.1Inter-domain movements could lead to significant changes of the HMBS protein shape. It is nevertheless a challenging experiment.1 D. I. Svergun,J. Appl. Cryst., 2000,33, 530–534.Prof. Mooresaid: I am interested in the influence of pH in this system because the cofactor has four carboxylic acid groups and the substrate and product also have carboxylic acid groups. (i) Is the enzyme activity pH dependent? (ii) What are the pKavalues of the cofactor, substrate and product substituents? I imagine they will be 5–5.6 in the absence of carboxylate–carboxylate electrostatic interactions (which could be substantial). (iii) Was the pH of the crystallisation solution 5.3? (iv) Do you have a mixture of forms in the crystal, and could this contribute to disorder thus making the ‘loop’ invisible in the electron density map?Dr Hädenerreplied: (i) At 37 °C, the optimal pH for wild-type HMBS fromE. coliis 7.4, with values forkcatandKmbeing 0.1 s−1and 7 μM, respectively, but the enzyme still exhibitsca.20% of its maximal activity at pH 6.2 and at pH 8.7.1,2At pH 7.4, the Lys59Gln mutation does not affectkcatbut it increasesKmapproximately 30-fold.2At room temperature and pH 5.3, the kinetic parameters of crystalline Lys59Gln HMBS can be estimated to beKm ⩾ 1.9 mM andkcat ⩽ 1.9 × 10−3s−1.3(ii) The pKavalues of the substrate (PBG) are known. They are 10.1 (aminomethyl substituent), 4.95 (propionic acid substituent), and 3.70 (acetic acid substituent).4The formation of a favourable ion pair between the negatively charged acetate substituent and the protonated aminomethyl side chain5is probably the reason for the latter low value. The pKavalues of the substituents of the cofactor and of the product (HMB) have not been determined as far as I know. The formation of intramolecular ion pairs is not possible in the HMB case, so you would expect the pKavalues to be around 5 including those of the acetic acid substituents, in the absence of the enzyme and in the absence, as you said, of carboxylate–carboxylate electrostatic interactions. In the case of the cofactor and enzyme-bound PBG and HMB, however, ion-pair formation between the carboxylate groups and positively charged side chains of amino acids may have a substantial influence on individual pKavalues.(iii) This is correct.(iv) This may well be the case. There are a number of charged residue side chains in the mobile loop (aspartates and lysines, two each, and one arginine). They are disordered in the structure of the holoenzyme but during catalysis they may find a suitable charged group for ion pairing among those of the growing oligopyrrole or of the side chains of amino acids.1 G. J. Hart, C. Abell and A. R. Battersby,Biochem. J., 1986,240, 273–276.2 A. Hädener, P. R. Alefounder, G. J. Hart, C. Abell and A. R. Battersby,Biochem. J., 1990,271, 487–491.3 J. R. Helliwell, Y.-P. Nieh, J. Raftery, A. Cassetta, J. Habash, P. D. Carr, T. Ursby, M. Wulff, A. W. Thompson, A. C. Niemann and A. Hädener,J. Chem. Soc., Faraday Trans., 1998,94, 2615–2622.4 S. Granick and L. Bogorad,J. Am. Chem. Soc., 1953,75, 3610.5 P. M. Jordan, inThe Biosynthesis of the Tetrapyrrole Pigments, ed. D. J. Chadwick and K. Ackrill, 1994, Ciba Foundation Symposium, vol. 180, pp. 70–89, Wiley, Chichester.Dr Techertasked: What is the biological reason for the unusually lowkcatvalue of 0.1 s−1?Dr Hädeneranswered: The product of the reaction, hydroxymethylbilane (HMB), is not stable at physiological conditions. With release of H2O, it cyclises to uroporphyrinogen I with a half-life of less than 5 min.1Uroporphyrinogen I, however, is not an intermediate of the biosynthetic pathway leading to the tetrapyrrolic pigments. An isomer of it in which ring D (see Fig. 1 of our paper) is rearranged, uroporphyrinogen III, is the true intermediate.2A companion enzyme, formerly called ‘cosynthetase’, now uroporphyrinogen III synthase, has evolved to take charge of HMB by ring-closing it at high rate to the correct isomer uroporphyrinogen III with release of H2O. Uroporphyrinogen III synthases exhibit relatively highkcatvalues such as 500 s−1for the enzyme fromE. coli.3The evolutionary aspects of the biosynthetic pathway of which this enzyme system is part have been thoroughly discussed by Eschenmoser.41 A. R. Battersby, C. J. R. Fookes, K. E. Gustafson-Potter, E. McDonald and G. W. J. Matcham,J. Chem. Soc., Perkin Trans. I, 1982, 2427–2444.2 A. R. Battersby, C. J. R. Fookes, G. W. J. Matcham and E. McDonald,Nature, 1980,285, 17–21.3 A. F. Alwan, B. I. A. Mgbeje and P. M. Jordan,Biochem. J., 1989,264, 397.4 A. Eschenmoser,Angew. Chem. Int. Ed. Engl., 1988,27, 5–39.Prof. Wattsasked: So the HMBS constitutes the rate limiting and regulating step for the much faster subsequent reactions?Dr Hädenerreplied: This is certainly true regarding the conversion of PBG to uroporphyrinogen III. There are, however, preceding reactions that are precisely regulated to control the entire pathway leading to the tetrapyrroles by preventing even the formation of PBG if it is not needed. This regulation affectse.g.porphobilinogen synthase, 5-aminolevulinate synthase, and, in plants and many anaerobic bacteria, glutamyl-tRNAGlureductase and glutamate-tRNA ligase, which are all inhibited by haem.1A reason for the prevention of PBG being accumulatedin vivomight be its intrinsic lability with respect to oligomerisation.2,31 G. Michal,Biochemical Pathways, Spektrum Akademischer Verlag, Heidelberg, 1999.2 G. H. Cookson and C. Rimington,Biochem. J., 1954,57, 476–484.3 D. Mauzerall,J. Am. Chem. Soc., 1960,82, 2605–2609.Prof. Wattsopened the discussion of Prof. Moore's paper: In Fig. 4 of your paper, you indicate far less α-helix structure from the NMR data, than shown in the modes or crystal structures of related colicins. Why?Prof. Mooreresponded: The only relevant X-ray structure for Fig. 4 is the structure in Fig. 1 of the paper. This shows a high content of helix in the min-R region of the protein. Fig. 4 is the output of the chemical shift index, which is a conservative empirical procedure for obtaining information on secondary structures from NMR chemical shifts. This procedure requires a run of NMR data for four or more sequential residues to all indicate a helix before one is indicated. Because of the incomplete resonance assignment for min-R we cannot provide a complete secondary structure map for it. However, what information the CSI does provide is fully consistent with the X-ray structure of Fig. 1; and with the shape analyses and relaxation time study later in the paper that provides strong confirmation that min-R has a helix–loop–helix structure.Prof. Finneyasked: What relationship do you expect between the two conformers and the structure determined crystallographically? Are the structural differences between the two conformers such that you might expect both of them to be accommodated within the crystal, or force the selection of one of them?Prof. Moorereplied: In connection with the DNase domain of colicin E9 referred to in the Introduction to the paper, one possibility is that there is only a small difference between the two conformers in solution,1with both co-existing in the crystal though the quality of the crystallographic electron-density map is not sufficient to resolve the different conformers. A second possibility is that the crystallisation process selects only one of the two conformers. At present we don't know which happens but we do know from NMR data that the difference between the conformers is very small and so I favour the first possibility. Another possibility is that the X-ray structure was determined at 100 K (ref. 2) while the NMR data were measured at room temperature. Perhaps at 100 K the protein preferentially populates the lowest energy conformer?The story with the conformers on the min-R protein that is the subject of the paper is different. The X-ray structure is only at 3 Å resolution, and it is for a full-length protein while the NMR study was carried out on a fragment of the structure. So the conformational exchange we detect by NMR may be a consequence of min-R being a sub-domain of the coiled-coil or it may be a feature of the coiled-coil itself with the X-ray structure not revealing it because it has a relatively low resolution. We plan to carry out NMR studies of the full-length coiled-coil domain and these should show whether the two conformers are an intrinsic property of it.1 S. B.-M. Whittaker, M. Czisch, R. Wechselberger, R. Kaptein, A. M. Hemmings, R. James, C. Kleanthous and G. R. Moore,Protein Sci., 2000,9, 713.2 C. Kleanthous, U. C. Kühlmann, A. J. Pommer, N. Ferguson, S. E. Radford, G. R. Moore, R. James and A. M. Hemmings,Nature Struct. Biol., 1999,6, 243.Prof. J. C. Smithasked: Could you comment on ways of analysing the contribution of internal motion to NMR spin relaxation in terms of collective dynamical variables, such as that being developed by Brueschweileret al.—this may be a way of linking the relaxation results to the interdomain dynamics you appear to have identified here.Prof. Mooreanswered: I can't give a full answer to your question, partly because the approach advocated by Prompers and Brüschweiler has only recently appeared in the literature.1Nuclear spin relaxation results from the modulation of inter-nuclear interactions by motions of the nuclei, and relevant motional modes are the overall tumbling of the protein and independent nuclear motion, usually described as internal motions. In order to define the internal motions of amino acid residues, which are really the dynamic information we seek, we have to be able to separate the contributions these two factors make to the experimentally determined relaxation parameters. There are a number of approaches for analysing the contribution of internal motions to NMR spin relaxation, with the Lipari–Szabo ‘model-free’ formalism2,3most widely used. The usual starting point for model-free analysis is to decide which residues have relaxation characteristics determined by the overall protein tumbling rate and using these to determine the rotational correlation time.4Those residues with significant internal motion can then be characterised. This benchmarking process is a key stage of the analysis and if it looks as if all the residues have internal motion then alternative procedures, such as reduced spectral density mapping,1,5is employed. This is less powerful than the model-free approach in the details it provides, but for disordered proteins in particular it is the most widely used approach. A key feature of the method described by Prompers and Brüschweiler1is that it does not require separation of the residues into those whose relaxation parameters are affected only by the overall tumbling of the molecule and those with significant additional internal motions. This seems to offer a substantial advance over other methods. However, it does involve molecular dynamics simulations in the analysis so it is not as straight-forward to apply as the model-free or reduced spectral density mapping approaches.1 J. J. Pompers and R. Brüschweiler,J. Am. Chem. Soc., 2002,124, 4522.2 G. Lipari and A. Szabo,J. Am. Chem. Soc., 1982,104, 4546.3 For example, A. G. Palmer III,Curr. Opin. Struct. Biol., 1997,7, 732; V. A. Feher and J. Cavanagh,Nature, 1999,400, 289.4 N. Tjandra, S. E. Feller, R. W. Pastor and A. Bax,J. Am. Chem. Soc., 1995,117,12 562.5 J. F. Lefevre, K. T. Daye, J. W. Peng and G. Wagner,Biochemistry, 1996,35, 2674.Dr Sagisaid: One of the main goals in biocatalysis or bio-molecular recognition processes is to correlate dynamics with the actual biological process (e.g.catalysis, molecular recognition). However, NMR probes only equilibrium states therefore one cannot really probe a real molecular recognition process or catalysis and correlate then with dynamics. How much can we stretch NMR to give direct information about correlating catalysis with dynamics?Prof. Moorereplied: You are right to say that NMR is looking at equilibrium states—to use Prof. Helliwell's analogy, the cliff edge and the beach. NMR is too slow to measure what happens to a body after it has jumped off the cliff until it lands on the beach, but what it can do is to characterise—sometimes in considerable kinetic and structural detail—dynamic processes taking place before the jump. With some proteins such dynamic processes are extremely significant for determining molecular recognition events, and in some enzymes a relationship between such dynamics and catalysis has been suggested. The clearest examples of dynamics affecting molecular recognition involve proteins that only fold into a globular form when they bind to their partner1—so-called intrinsically disordered2or natively unfolded proteins.3Coupling of protein folding to intermolecular complex formation has only been discovered to be a widespread phenomenon with the development of NMR spectroscopy for studying proteins and with estimates that as many as 50% of the proteins in humans contain at least one domain that is intrinsically disordered2this is clearly an important field. The way in which a newly synthesized protein folds from its unfolded state to its folded form is related to intermolecular interactions involving at least one disordered protein—in this case the interacting surface is intramolecular—and again NMR is one of the major experimental methods for determining what is happening to the polypeptide chain at key stages of the folding process. Of course, protein folding is generally very rapid so NMR is limited in the states it can inform on, but it is already clear from NMR studies of unfolded protein chains that some are not random coils, even at low pH4or in high urea of guanidine·HCl concentrations.5Colicin E9 is a member of a related class of natively unfolded protein that interacts with other proteins in that its unfolded domain contains some ordered non-random structure that is highly dynamic and forms the binding epitope for its partner protein.6Our NMR studies have shown that even when a ternary protein complex of 114 kDa is formed that anchors the previously dynamic binding epitope, the residues upstream and downstream of it retain considerable flexibility.6We have speculated that this is because this region is like a fishing line, with many hooks on it to bind to specific proteins in constructing a complex of many proteins; however, the biophysical characterization has moved ahead of the biological studies and additional interaction partners have not yet been identified. Even when the binding of a protein to a partner molecule does not involve folding of the protein, NMR relaxation studies are providing new insights into the molecular recognition events,7which include not only identification of residues that interact with the partner molecule but also information on changes in conformational entropy of the protein on binding obtained in a residue-specific fashion. Investigating catalytic events themselves is as difficult with NMR as it is with any structural technique, and perhaps even more so. However, as indicated above, NMR is giving unique information on the dynamics of enzyme interactions with inhibitors and substrate analogues,7and coupled with other data, including that obtained with site-directed mutagenesis, I believe that NMR will turn out to be valuable for characterising catalytic events. Crucially, as my answer to Prof. Smith indicated, this application of NMR will benefit considerably from being combined with theoretical studies of protein dynamics, which themselves may be tested by spectroscopic examination of mutants.1 H. J. Dyson and P. E. Wright,Curr. Opin. Struct. Biol., 200212, 54.2 A. K. Dunker, C. J. Brown, J. D. Lawson, L. M. Iakoucheva and Z. Obradović,Biochemistry, 2002,41, 6573.3 V. N. Uversky,Eur. J. Biochem., 2002,269, 2.4 J. Yao, J. Chung, D. Eliezer, P. E. Wright and H. J. Dyson,Biochemistry, 2001,40, 3561.5 J. Klein-Seetharaman, M. Oikawa, S. B. Grimshaw, J. Wirmer, E. Duchardt, T. Ueda, T. Imoto, L. J. Smith, C. M. Dobson and H. Schwalbe,Science, 2002,295, 1719.6 E. S. Collins, S. B.-M. Whittaker, K. Tozawa, C. MacDonald, R. Boetzel, C. N. Penfold, A. Reilly, N. J. Clayden, M. J. Osborne, A. M. Hemmings, C. Kleanthous, R. James and G. R. Moore,J. Mol. Biol., 2002,318, 787.7 V. A. Feher and J. Cavanagh,Nature, 1999,400, 289; J. J. A. Huntley, S. D. B. Scrofani, M. J. Osborne, P. E. Wright and H. J. Dyson,Biochemistry, 2000,39, 13 356; R. Ishima and D. A. Torchia,Nature Struct. Biol., 2000,7, 74; A. J. Wand,Nature Struct. Biol., 2001,8, 926; F. A. A. Mulder, A. Mittermaier, B. Hon, F. W. Dahlquist and L. E. Kay,Nature Struct. Biol., 2001,8, 932.Prof. Wilsonsaid: You mentioned that the function in this system involves allowing a 100 Å coiled coil to traverse a 150 Å cell membrane. How are the dynamics you have elucidated related to this function?Prof. Moorereplied: How the dynamics of the helix–loop–helix region of the receptor binding domain we have observed relate to the activity of the colicin is not clear at present. The relatively high degree of flexibility of the inter-helix loop compared to the remainder of the coiled-coil might be connected to the loop being part of the binding site for the BtuB receptor on the outer membrane of the target cell, as indicated by site-directed mutagenesis studies.1,2As we have discussed already (in the discussion with Dr Sagi), flexibility in an intermolecular interaction site may help in the recognition events leading to the complex formation.Perhaps it is more interesting to consider whether the coiled-coil unfurls. This is our current hypothesis3, which assumes that the helices remain largely intact, partly because they are interacting with other proteins in the system (including the outer membrane receptor BtuB and periplasmic Tol proteins). If the helices of the now unfurled coiled-coil remain helices, and one of them extends into the periplasm from the loop of the min-R region attached to BtuB it will not be long enough to cross to the inner membrane unless the periplasm is reduced from its normal size. This occurs at contact points. Alternatively, the helix may partly or completely unwind, which would give it sufficient length to cross the periplasm. This is something we are studying with a combined luminescence spectroscopy and protein engineering approach. In the NMR studies reported in this paper we have detected that the two helices of the min-R helix–loop–helix sub-domain move relative to each other in a way that retains the helix–loop–helix structure; and possibly the two helices slide across the surfaces of each other. Such motion may be related to events that lead to unfurling of the coiled-coil in the colicin, though in our model binding to a receptor is the key event that triggers this unfurling and it is not an intrinsic feature of the unbound colicin.1 S. Soelaiman, K. Jakes, N. Wu, C. Li and M. Shoham,Mol. Cell, 2001,8, 1053.2 C. N. Penfold, C. Garinot-Schneider, A. M. Hemmings, G. R. Moore, C. Kleanthous and R. JamesMol. Microbiol., 2000,38, 639.3 R. James, C. N. Penfold, G. R. Moore and C. Kleanthous,Biochimie, 2002, in press.Mr Terieteasked: Has dynamic data been gathered on the loop, which is believed to be involved in the activity, by using RDCs?Prof. Mooreanswered: Not yet.Prof. Wattsopened the discussion of Prof. Finney's paper: All co-operative transitions have an on-set and completion point. Yours are defined by the dynamic range of the method in some cases, in others, only a single transition is observed. But there must be a tool where the whole dynamic range is concerned and onset and completion are encompassed for this 220 K transition, or perhaps not.Prof. Finneyresponded: We are not aware of any experimental technique that can give such information across the whole dynamic range. Encompassing the whole dynamic range implies, in neutron scattering parlance, elucidation of the elastic incoherent scattering function which is determined by the infinite time limit of the van Hove single particle space–time distribution function. Accurate determination of this is possible in some systems where the relaxation times are well within reach of the instruments concerned. For protein solutions the range of motions is such that it has not yet been possible to reach this limit with neutron techniques alone. A range of experiments will be required.Dr Schmidtsaid: Data shown trace Doster and Settles:1On IN13 and IN6 at the ILL, all data show the same transition temperature despite different time scales. The Mossbauer data from Parak and Knapp2show the same transition point at 180–190 K. What type of motions do you expect, which account for your transitions? Are these transitions motions of the entire molecule, or are these motions of domains of the molecule?1 W. Doster and M. Settles, The dynamical transition in proteins: the role of hydrogen bonds, 1998; Les Houches Lectures: workshop on hydration processes in biology: theoretical and experimental approaches, ed. M. C. Bellisent-Funel and J. Teixera, IOS Press.2 F. Parak and E. W. Knapp,Proc. Natl. Acad. Sci. USA, 1984,81, 7088–7092.Prof. FinneyandProf. J. C. Smithresponded: IN6 and IN13 measure motions on very similar time scales, so you would not expect to get a large change in the apparent temperature of the dynamical transition. The measurements you mention are taken on samples at low hydration, while the ones discussed in our paper are in solution, where the behaviour is likely to be different (as we find it is experimentally). Moreover, the method used to determine the mean-square displacements in the paper cited is unclear. IN13 in principle works only at relatively highQ, where the Gaussian approximation for individual atoms fails—this is required for a model-free determination of mean-square displacements. Concerning the kinds of motions involved, determining these is one of the objects of the work. The experimental technique measures motions of all the non-exchanged hydrogen atoms of the enzymes, and hence averages over the whole molecule.Prof. Sir John Meurig Thomassaid: Your arguments, which seem to me to be convincing, would become irrefutable if you (or others) were to try and do measurements of the onset of anharmonicity and of the change of enzymatic activity underin situconditions. This is obviously not an easy thing to do; but one of the reasons for discussion such as this is to identify crucial (time-resolved) experiments. Could you comment on that?Prof. Finneyreplied: We have tried this. The ‘fastest’ enzyme known is catalase. We have measured its activity down to 173 K, and see no deviation from that expected from Arrhenius behaviour. 173 K does appear to be well below the onset of anharmonicity for all proteins looked at so far. However, the assay at 173 K requires ∼10 h to obtain measurable product. Nanosecond assays appear to be quite out of reach with current technology.Also Prof. R. M. Daniel, University of Waikato, New Zealand.In these circumstances, where it is not possible to measure both dynamics and activity at the same time, it is crucial that the neutron measurements are made under the same conditions as those under which the activity is measured. This was an important aspect of our experimental protocol: the solvents in which the neutron measurements were made were those in which activity was measured at these temperatures. To ensure that the enzymes were still active after the neutron experiments, the samples were recovered and assayed subsequently, confirming that they retained their activity.Dr Colesaid: The error bar at 200 K on Fig. 1 is notably larger than any other in the graph. Since this temperature pertains to the vicinity of the transition, is the size of the error bar a physical consequence of the transition is some way, or does it show the model breaking down here, or does it just emanate from an experimental effect?Prof. Finneyanswered: The error bars on the points relate to those estimated by the extrapolation procedure described in the paper, and for the point you mention, the errors were larger for experimental reasons. This larger error may relate to the phase behaviour of the solvent, though this is not definite. However, the determination of the temperature of the so-called transition does not really depend on this one point. As is seen from a number of papers in which a transition temperature is assigned, it is obtained rather by drawing straight lines to the data above and below the transition, estimating the transition temperature as that at which these lines cross. The points in the vicinity of the transition have little effect on this geometrical construction.Prof. Wattsaddressed Dr Cole: In any transition phenomenon, there are highly disordered intermediates formed which are neither initial nor final state related. These are often very small in population compared with the majority and can account for significant errors in determination of experimental parameters.Prof. Helliwelladdressed Prof. Finney: (i) Re Fig. 1 of the paper: Isn't the functional assay resting on the last data point (that the enzyme behaviour is Arrhenius over the full temperature range)?(ii) Is the xylanase also from a thermophile (i.e.Zaccai is saying (personal communication) that the stiffness of proteins in extremophiles is higher and therefore you could look at non-extremophiles too)?Prof. FinneyandProf. J. C. Smithresponded: (i) I don't think so. Close examination of the figure shows that there are several data points at temperatures lower than the 220 K transition. These show no deviation from linearity. If there were a deviation from Arrhenius behaviour in this region, these results would have shown it.Also Prof. R. M. Daniel, University of Waikato, New Zealand.(ii) The xylanase was from a thermophile. I would point out however that we started out using thermophiles for these measurements as it was thought that their expected greater ‘stiffness’ might shift the transition temperature upwards, which would ease the enzymology. However, where we have taken measurements on equivalent thermophilic and mesophilic enzymes, we find no measurable difference in their dynamic behaviours as measured by these techniques. Thus we find no evidence for greater ‘stiffness’ as measured by transition temperature of the thermophiles. Furthermore, other comparative activity measurements on non-thermophile enzymes give similar results.1,21 J. M. Bragger, R. V. Dunn and R. M. Daniel,Biochim. Biophys. Acta, 2000,1480, 278–282.2 N. More, R. M. Daniel and H. Petach,Biochem. J.1995,305, 17–20.Prof. Mooresaid: (i) Dehydrating proteins can perturb properties even though they may appear to be active. I recall that Kaminskyet al.1reported freeze-dried cytochromecon dissolving in water undergoes a structural change detected spectroscopically, even though the dehydrated protein retains electron-transfer capability. (ii) In your 70% CH3OH/30% H2O solutions how is the protein solvated? Does it have a water shell surrounded by H2O/CH3OH?1 S. Kaminsky, K. M. Ivanetich and T. E. King,Biochemistry, 1974,13, 4866.Prof. Finneyreplied: (i) There is a significant amount of data looking at changes in protein structure on dehydration. In summary, it seems that these are generally quite small (seee.g.ref. 1 and 2), though they may have functional significance.(ii) This is an interesting observation, and is discussed in an earlier paper on the solvent dependence of the enzyme dynamics.3Interestingly, we found that the dynamics did not vary significantly with methanol concentration except when going to pure D2O. One possible interpretation of this observation is that the solvent in the near neighbourhood of the protein is relatively unperturbed by the change in concentration of the bulk, but there may well be other interpretations of the observation.1 P. L. Poole and J. L. Finney,Biopolymers, 1983,22, 255.2 N.-T. Yu and B. H. Jo,Arch. Biochem. Biophys., 1973,156, 469.3 V. Réat, R. Dunn, M. Ferrand, J. L. Finney, R. M. Daniel and J. C. Smith.Proc. Natl Acad. Sci. USA, 2000,97, 9961.Prof. Moffatsaid: The techniques for determining MSDs are as you state—sensitive to time scale. The activities presented by you are essentially on a single time scale for various enzyme systems. Some proteins exhibit “activity” over a very wide range of time scales, fs,μs → s,e.g.the individual steps in the photocycles of lR or photoactive yellow protein. Could you envisage parallel activity and dynamic measurements on such systems?Prof. Finneyreplied: Enzyme activity measurements are confined to the rate limiting step (as defined by the turnover number), and these certainly vary over at least seven orders of magnitude depending upon the enzyme, and vary by another several orders of magnitude with conditions such as temperature. All other events, including all the other reaction steps, must take place over the same or shorter time scales. Ideally we need to correlate activity with a full dynamic ‘spectrum’ of motions, but all techniques have either time scale windows (often defined at only one end) or are time averaged. But the good news is that activity varies predictably and smoothly (in an Arrhenius sense) with temperature, suggesting that the real need may be to identify the groups/motions involved in activity rather than their time scales.Prof. Wattssaid: Lets try and find out what each approach and method of the papers by Prof. Helliwell, Prof. Moore and Prof. Finney can bring to the systems we have heard about. Loops are ill defined in Prof. Helliwell's crystal structure whereas Prof. Moore defines loops, both of which are functionally important. Prof. Helliwell, what can you do for Prof. Moore's system andvice versa, and Prof. Finney finally, where do you see global dynamics fitting into each of these systems?Prof. Mooreresponded: I have been asked how NMR could help clarify what is happening with the ‘missing’ loop in the X-ray data of HMBS. The first experiment would be to look at15N labelled HMBS. I would expect sharp NH1H–15N HSQC peaks from the loop and broader peaks from the remainder of the protein. The cofactor NH signals may also be sharp. Peaks of the loop could be assigned from triple-resonance spectra of the13C/15N labelled material (2H labelling may also be needed given the 34 kDa size of HMBS—the loop residues should not need this); and backbone relaxation parameters measured for the loop with the15N labelled samples. Monitoring the loop signals during the reaction cycle, and as functions of pH and temperature, should establish whether they are playing a role in the reaction. The pKavalues of the carboxylic acid groups of the cofactor and its adducts might be determined with specifically labelled samples.Prof. Finneycommented: In general terms, the ‘elastic scan’ neutron scattering measurements we used in the work reported here give information on the global dynamics of the protein through the average mean square displacements. As the signal is dominated by the scattering from the hydrogen atoms, it is the motions of the non-exchangeable hydrogens on the protein that will be largely reported. As the contributions of deuterium atoms to the measured signal are much less than those of hydrogens, if the protein were selectively deuterated, then we could focus on the motions of those regions that remained hydrogenated. The time scales of the motions that can be accessed depend on the instrument used, and can broadly range from picoseconds to tens of nanoseconds. Furthermore, by making quasielastic and inelastic scattering measurements, we can also begin to obtain more detailed information on the natures of the motions involved and the density of states.Prof. Wattssaid: New genetically engineered labelling methods permit protein domains to be labelled. The case of PBDG seems an ideal one where NMR visible isotopes placed specifically in the ‘invisible’ loop, could be studied by NMR (solution or solid state in an amorphous or semi-crystalline state) to resolve structures, or an ensemble of structures. Ligand labelling too, could help here.Prof. Moorereplied: Probably yes, a set of atomic coordinates can be determined. However, I wonder if this is what is really needed. Given the probable dynamic nature of the loop there are likely to be large uncertainties in an ensemble of structures for this. I would put the effort into the relaxation characteristics of the loop under various conditions. Perhaps the NMR and X-ray data combined might be data used in MD calculations rather than as an attempt to define the “structure” of the complete enzyme as a static entity.Prof. Helliwellcommented: A simulation approach perhaps could be the best benefit of NMR information on the loop,i.e.to resolve where the HMBS loop is, and its excursions. Also, the hypothesis of the role of inter-domain movements of the 3-subunit protein in the enzyme reaction could be assessed. At present, it seems, simulation cannot readily approach longer time scales than ∼10 ns simulations and such a loop might be expected to have slowly moving aspects.Prof. Helliwelladdressed Prof. Moore: Re: Fig. 11 and the 3 Å X-ray structure: crystal annealing could improve the resolution of the protein crystallographic study andBfactor estimates would be much better (to compare with yourT1/T2ratios).Prof. Moorereplied: We clearly need a better X-ray structure and more complete NMR data to fully characterise the structural dynamics of these colicins, but the comparison of X-rayBfactors andT1/T2ratios is suggestive. Remember that the X-ray data is for a full-length protein whilst the NMR data is for a fragment. The reduction in theT1/T2ratio for residues 4 and 5, and beyond residue 70, of min-R (Fig. 11(a)) are because they are its N and C termini while the corresponding residues of colicin E3 (Fig. 11(b)) are in the middle of the coiled-coil helices!Prof. Finneysaid: Neutron scattering measurements are essentially limited to probing global dynamical processes on time scales of ∼1 ns or faster. Yet processes relating to the enzyme activity are likely to be occurring on slower time scales. Mössbauer offers timescales of ∼10−7s, but requires the presence of Fe and cannot be effectively used under solution conditions. NMR would seem to offer a way forward to both (a) probing longer time scales under realistic solution conditions and (b) (through selectively focussing on particular nuclei) probing motions locally through the enzyme. How realistic are such measurements, and how long a programme of work would be needed to characterise fully the dynamics of an enzyme of, say 50 kDa?Prof. Mooreresponded: Let me rephrase your question into 3 parts and then answer them in turn. (i) Can NMR fully characterise the motions in small proteins (say up to 20 kDa)? (ii) How long would it take to do so? (iii) How feasible is it for an enzyme of 50 kDa mass?I assume you will want to do this over the temperature range used in your neutron scattering studies to look at the dynamic transition. The first point to note is that the NMR approach I described is for solutions; solid-state NMR of proteins is a rapidly developing field but cannot yet provide a global description of protein structure and dynamics in the way that you want. The reason size of the protein is important is that the overall tumbling time of the molecule in solution is a key determinant of the nuclear relaxation properties, with slow tumbling producing efficient relaxation and broad lines. Along with the size of the protein, the viscosity of the solvent is important. If the water–alcohol mixtures you use have a viscosity at 200 K and above that allows relatively sharp NMR lines to be observed, you will be set to get the data you need with a small enzyme. This will have to be produced by an expression system that allows isotopic labelling with13C and15N; most NMR groups employ bacterial systems growing on minimal media for this. Starting from scratch with13C/15N labelled samples, it will take 6–8 weeks to collect and analyse spectra that allow the NH resonances observed in the1H–15N HSQC spectrum to be assigned. A further 2 weeks data collection with15N labelled samples allows the backbone15NT1andT2and1H–15N nuclear Overhauser enhancements1to be collected at one field strength. It is likely that the level of detail you want will require the relaxation parameters to be determined with at least one other spectrometer operating at a different field strength to the first. This should also take 2 weeks. The dynamics of the side chains can be determined by13C relaxation times,1,2and also with selective2H labelling,1and both of these require samples with different isotope labelling patterns to those needed for the assignments and for15N relaxation measurements. However, provided the samples are sufficiently stable once you have them you can readily carry out a temperature dependence study. Provided the protein does not undergo substantial temperature-induced conformation changes it is unlikely that you will have to repeat much of the assignment work with13C/15N labelled samples at each temperature. So, assuming a stable protein of less than 20 kDa at a concentration of about 1 mM it may take a year to measure the temperature dependence of the15N relaxation properties.2H and13C relaxation analyses will take longer. Probably you should concentrate on the backbone characteristics first to determine if the dynamic transition is reflected in these. If you pick a protein that is already well-studied by NMR then you will save time; for example, the 14 kDa hen egg white lysozyme may be suitable.An enzyme of 50 kDa presents major problems because of slow tumbling in solution3. Proteins of this size have been studied by NMR but in general, in order to obtain full assignment of the NH1H–15N HSQC spectrum the non-exchangeable protons of the protein have to be replaced by2H. NMR experiments to measure the relaxation properties of the backbone will be somewhat different to those for similar proteins in which deuteration has not been done but unlike the latter proteins, which are routinely studied in many laboratories by these approaches, deuterated proteins are still in the hands of relatively few groups. Moving onto side chain dynamics, deuterated proteins will require specific labelling protocols; for example, introducing protonated methyl groups of particular amino acids into an otherwise deuterated protein. So, the methodology may all be in place for you to characterise the global dynamics of a 50 kDa enzyme over a wide range of temperatures but note that although various groups have described methods that may lead to the fold of a large protein being determined by NMR,4a high-resolution NMR structure of a monomeric protein greater than 30 kDa has not been reported.1 A. G. Palmer III,Curr. Opin. Struct. Biol., 1997,7, 732.2 J. Engelke and H. Rüterjans, inBiological Magnetic Resonance: Structure Computation and Dynamicsin Protein NMR, Kluwer Academic/Kluwer Publishers, New York, 1999, vol. 17, p. 357–417.3 K. Pervushin,Quart. Rev. Biophys., 2000,33, 161.4 B. T. Farmer III and R. A. Venters,J. Biol. NMR, 1996,7, 59; B. O. Smith, Y. Ito, A. Raine, S. Teichmann, L. Ben-Tovim, D. Nietlispach, R. W. Broadhurst, T. Terada, M. Kelly, H. Oschkinat, T. Shibata, S. Yokoyama and E. D. Laue,J. Biol. NMR, 1996,8, 360; W.-Y. Choy, M. Tollinger, G. A. Mueller and L. E. Kay,J. Biol. NMR, 2001,21, 31.