Laboratory of Molecular Biophysics
Laboratory Journal 2003
Elspeth Garman

Return to: Contents.

Elspeth Garman

Methods development for protein crystallography

Introduction

The work of the group this year has been largely in the four areas detailed below, with the main emphasis being on the radiation damage studies. The Long Term Project allocation of beamtime at the ESRF for Radiation Damage studies ended in September 2003 and was replaced by a Block Allocation Group for continuation of the work, in collaboration with Drs. Raimond Ravelli (ESRF), Sean McSweeney (ESRF) and Martin Weik (IBS, Grenoble), and Professor Martin Caffrey (Ohio State and Limerick University, Ireland).
James Murray spent a month at the ESRF in the spring of 2003 working both on the installation of an on-line microspectrophotometer on ID14-4 and on the computer program RADDOSE, briefly described below.
The Third International Workshop on Radiation Damage to Crystalline Biological Samples will take place in November 2003 at the ESRF, and provide a valuable forum for discussions on the future direction of our work. This follows on from the second workshop held at the APS in December 2001, our papers from which were published in 2002 (Garman and Nave 2002, Murray and Garman 2002, O'Neill, Stevens and Garman 2002).
As well as the research reported below, we have been active on a number of smaller technical projects (e.g. testing of new plastic cryo-loops and development of an improved backstop). With a view to improving the success rate of transporting cryocooled crystals to synchrotrons, we have performed temperature measurements on the ACTOR robot pucks and on commonly used cryovial arrangements. Four calibrated platinum resistance thermometers were mounted in top-hats to simulate crystals for the investigation. The results showed a) a temperature gradient existed within a dry transport dewar, but after 2 weeks of monitoring, was sufficiently small (7K: 78K at the bottom of a cane and 85K at the top position) not to affect the crystals, and b) and that when lifted out of the dry dewar, the pucks warmed up more slowly than the cryovials stored on canes, providing more time for crystal manipulation.

References:

Garman, E. and Nave, C. (2002) Radiation damage to crystalline biological molecules: current view. Journal of Synchrotron Radiation 9, 327-328.
Murray, J.M. and Garman, E.F. (2002) Investigation of Possible Free Radical Scavengers and Metrics for Radiation Damage in Protein Crystallography. Journal of Synchrotron Radiation 9, 347-354.
Neill, P., Stevens, D.L. and Garman, E.F. (2002) Physical and chemical considerations of damage induced in protein crystals by synchrotron radiation: a radiation chemical perspective. Journal of Synchrotron Radiation 9, 329-332.

1. Radiation Damage in Cryo-cooled crystals

James Murray, James Tichler, Enrique Rudino Pinera and Elspeth Garman

We have made progress of three fronts: investigating radical scavengers to slow down the rate of radiation damage in cryo-cooled crystals, theoretical

Figure 1. Photo of microspectrophotometer installed on ID14-4 at the ESRF, Grenoble.

calculations on the X-ray absorption coefficients of protein crystals to look at the effect of X-ray wavelength and the presence of heavy atoms (including selenium) on the expected crystal lifetime in the beam, and lastly a study of unit cell and structural changes in crystals of holo- and apo-ferritin as a function of cryogen temperature and radiation load.
The results reported in Murray and Garman 2002 on ascorbate as a radical scavenger were obtained by looking at sequential datasets taken from HEWL crystals with and without ascorbate as a co-crystallistion additive. The rate of damage to disulphide bonds was monitored in electron density maps of refined structures, and the absorption spectrum of the crystal before and after irradiation was observed, as detected using an off-line microspectrophotometer at the ESRF: in particular the 400nm peak from a disulphide radical species. The spectral results were very difficult to reproduce due to problems with ice formation, both during transport round the storage ring to the off-line microspectrophotometer and with the multiple crystal handling required for an experiment. In addition, the alignment of the spectrometer light beam onto the part of the crystal which had been
irradiated with the X-ray beam was time consuming and introduced an unecessary uncertainty into the measurements. We thus installed a microspectrophotometer on ID14-4 at the ESRF, Grenoble, funded by the Royal Society Equipment Fund. A convenient robust adjustable stand was designed and built at the ESRF. After initial commissioning and testing, this device (Figure 1) is in routine use during our radiation damage experiments. We are now in a position to conduct a broad survey of putative scavengers and the effects of various cryoprotective agents on the rate of damage. Fig 2 shows the absorption spectrum of cryoprotected cystine solution without (Fig 2a) and with (Fig2b) 0.5M ascorbate. It is clear that the 400nm disulphide species radical is suppressed in the presence of ascorbate.

On the theoretical side, the computer program RADDOSE, originally written by Raimond Ravelli (EMBL, Grenoble) was developed and extended to allow experimenters to enter the amino acid and heavier atom content of their crystals, the mM composition of the buffer, and the incident beam conditions. The program then calculates the estimated time to reach an absorbed dose of 2 x 10⁷ Gy (J kg ^-1), the so called `Henderson limit' after which half the diffracting power of the crystal is predicted to be lost. This program is now available through: http://www.esrf.fr/exp_facilities/ID14-4/ID14-4.html and a paper describing it and the implications of the predicted lifetimes has been submitted.


Figure 2a. Absorption spectrum of cystine following X-ray beam irradiation, showing 400nm di-sulphide radical species peak.	Figure 2b. Same but with 0.5M sodium ascorbate solution added. The peak at 350nm is from ascorbate. There is no 400nm peak visible, showing suppression

Reference:

Murry J.W. and Garman E.F. (First one from PI)

2. Elemental analysis of proteins using microPIXE.

David Yates and Elspeth Garman, with Geoff Grime, University of Surrey.

The Oxford Scanning proton microprobe moved to the National Ion Beam Centre at the University of Surrey, Guildford, in October 2002. The two year old accelerator there had never been used to produce proton beams and some extra source equipment was especially installed for this purpose in November 2002. While this modification was underway, we tried a new experiment on proteins: inducing X-ray emission using 1 MeV alpha particles ( alpha

IXE) instead of protons. The ionisation cross sections for alphas are lower than for protons, and although we were able to assess the suitability of alpha

IXE for our purpose and check the calibration of the X-ray Si(Li) detector etc, it was clear that the technique does not hold long term promise for analysing proteins.
   Following the production of a proton microbeam at Guildford, we undertook a thorough search for a better backing material for use when analysing proteins (Yates 2003), since the mylar `spectrofilm' we have always used has trace contamination of calcium and phosphorus. We tested 0.9µm and 1.5µm thick PET (polyethylene terephthalate), 0.1µm silicon nitride membrane windows, sub-micron carbon film supported on a copper grid support, pioloform film (polyvinyl butyral) and formvar film (polyvinyl formal), as well as 2.5µm, 3.6µm and 6.3µm spectrofilm. Of these, the original 2.5µm or samples where the calcium concentration is required. Interestingly, the contaminant P/Ca ratio in spectrofilm was found to be constant over large areas of the film, and thus it is possible to calculate the calcium background as long as the sample contains no phosphorous, and vice versa.

     Elemental analysis of a variety of liquid and crystalline protein samples (Garman 1999) was carried out on:
1) Bs1I solution for Eva Vanamee (Mount Sinai Hospital, New York) [zinc] (Vanamee et al., 2003),
2) NHIS and TFB for Ben Luisi and Anastasia Callaghan (Biochemistry, Cambridge) [zinc],
3) FIH and AlkB solution for Jon Elkins (OCMS, Oxford) [iron],
4) anastellin solution for David Staunton and Luke Rooney (Biochemistry, Oxford) [paramagnetic metal],
5) NapC solution for Michael Cartron and Stuart Ferguson (Biochemistry, Oxford) [iron],
6) an EMR2 crystal for Susan Lea and Rachel Abbott (LMB, Oxford) [barium], (Garman and Murray 2003)
7) DAF1234 crystals for Susan Lea and Petra Lukacik [gold, selenium, mercury] (Garman and Murray 2003),
8) NagA crystals for Gideon Davies and Florence Vincent (Structural Biology, York) [iron] (Vincent et al, in press),
9) two proteins for Chris Phillips [Pfizer],
10) PNPase + RNA crystals for Ben Luisi (Cambridge) [phosphorus],
11) a putative 10mm protein crystal for Jochen Zimmer and Declan Doyle (LMB, Oxford) [sulphur],
12) HHARI RING2 liquid for Allan Capili (Mount Sinai Hospital, New York) [zinc],
13) and crystals of formate dehydrogenase for Teresa Santos Silva and Maria Romao (Lisbon, Portugal) [iron].

Calibration standards of solutions of pseudoazurin (1 copper per protein molecule) and cytochrome c⁵⁵⁰ (1 iron atom per protein molecule) were kindly supplied to us by James Allen (Biochemistry, Oxford) to allow us to check the new experimental arrangement at Guildford.

The large number of liquid samples measured this year has allowed us to refine our graph of the minimum protein concentration in mg/ml against the proportion of sulphur containing residues. The new graph is shown in Figure 3. We also exhaustively identified buffers which made protein samples unsuitable for PIXE measurements due to their sulphur content, including BES, DTT and HEPES.

Figure 3. Graph to determine the minimum useful liquid protein sample concentrations for microPIXE analysis. Liquid protein samples must be above the black dotted line. The minimum ratio of sulphur atoms/amino acid per protein is 5.8 x 10^-5 . Note that crystal samples are usually around 600mg/ml, so will always be of sufficient concentration provided there are some suphur containing amino acids (cysteines and methionines).

References:

Garman, E. (1999) Leaving no element of doubt: elemental analysis of proteins by microPIXE. Structure 7, R291-299.

Garman, E. and Murray, J. (2003) Heavy atom derivatisation. Acta Cryst. D 59, 1903-1913.

Vanamee, E.S., Hseih, P-C., Zhu, Z., Yates, D., Garman, E., Xu S-Y. and Aggarwal, A. (2003) Glucocorticoid-like Zn(Cys)4 Zn motifs in BslI restriction endonuclease J. Mol. Biol. 334, 595-603.

Vincent, F., Yates, D., Garman, E, Davies, G. and Brannigan, J. (2004) The 3-D structure of the N-acteylglucosamine-6-phosphate deacetylase, NagA, from Bacillus subtilis: a member if the urease superfamily. JBC (in press)

Yates, David, (2003) MicroPIXE analysis of protein samples - improving and extending the technique. University of Oxford M.Biochem., Part II undergraduate project report.

3. Predetermination of Protein Crystal Diffraction Quality

Robin Owen and Elspeth Garman

Using an Oxford Cryosystems Metripol birefringence microscope, we have found a correlation between the width of the slow optical axis of Hen Egg White Lysozyme (HEWL) crystals and their diffraction quality (Owen et al. (2003)). Diffraction quality was assessed from Rmeas ( Fig4) and from the I/ sigma

(I) of the whole image as well as that of the highest resolution shell. We observe that the smaller the variation in the slow optical axis position (SOAP), the lower the Rmeas and the higher the I// sigma

(I). However there appears to be no correlation between the optical axis width and the mosaicity of the crystals or their volume. This is an exciting finding which opens up the possibility of pre-screening crystals in a crystallisation drop to select those which should diffract best. Clearly the method cannot be used on cubic crystals. We are now extending the measurements to glucose isomerase crystals, both at room temperature and cryocooled to 100K.

Figure 4. Graph showing width of the slow optical axis of ten different crystals of HEWL against their diffraction quality as measured by the Rmeas, PVC and Rsymm of a complete dataset to 2.1 Å. The correlation coefficient between Rmeas and SOAP (fit shown as a black line) is 0.96 with a standard deviation of 0.008.

Reference:

R.L.Owen, R.L., Geday, M.A., Glazer, R.L. and Garman, E.F. A new method for predetermining the diffraction quality of protein crystals. ECM Durban 2003 Poster f1.m2.p2

4 Structure determination of Fibronectin domains.

Enrique Rudino Pinera and Elspeth Garman

We are collaborating with Drs. Jennifer Potts and Ulrich Schwarz-Linek with the aim of determining the structure of the ²F1 and ³F1 human fibronectin domains, with and without a synthetic peptide (STATT1) from Staphylococcus aureus bound. We have used the in-house Tecan crystallisation robot to great effect to obtain crystals diffracting to at least 2.0Å in-house and better than 1.7 Å at the ESRF of both native and peptide-bound protein.
The crystals appeared to be twinned as judged by some but not all of the standard twinning tests. Crystallisation conditions have been modified to try to minimise the chance of twinned crystals (e.g. the addition of MPD, glycerol and dioxane), but these strategies are rather anecdotal in the literature and their success often not open to rational interpretation.
X-ray diffraction datasets to high resolution limits of between 2.2Å and 1.7Å were collected from 7 different crystals: 4 in-house on the LMB rotating anode (3 native and 1 with STATT1), and 3 at synchrotron radiation sources: 1 native at the ESRF, Grenoble, France and 1 platinum derivative and 1 native at SRS, Daresbury, Cheshire. Following analysis of these data, some of which have very high multiplicity (up to 40), we now believe the crystals to be single and of space group P4₃2₁2 or P4₁2₁2.
Attempts to solve the structure by molecular replacement using an average of the 15 NMR structures of ²F1 (Schwarz-Linek et al. (2003)) have failed, probably due to a combination of factors including uncertainty in which model might be the most appropriate. All available molecular replacement software was sought and tried: CPP4 (Molrep), CNS, Phoenix, Beast, Phasor and combinations of these. Efforts are now being concentrated on using the platinum soaked dataset to solve the structure by the SAD method. We have also collected 540º of in-house native data for sulphur SAD. There are 8 cysteines and 2 methionines in 90 residues, so this should be feasible.
We have also started to search for crystallisation conditions for a construct containing all five F1 human fibronectin modules.

Figure 5. Crystals of fibronectin domains ²F1 and ³F1. The size is approximately 0.2 mm x 0.2mm x 0.2mm and diffract, the crystals to 2.0Å on our in-house generator

Reference:

Schwarz-Linek, U., Werner, J. M., Pickford, A. R., Kim, S. Gurusidda J. H., Pilka, E. S., Briggs, J. A. G., Gough, T. S., Hook, M., Campbell, I. D., Potts, J. R. (2003): Pathogenic Bacteria Attach to Human Fibronectin Through a Tandem Beta-Zipper. Nature 423 177