636 Nuclear Instruments and Methods in Physics Research A266 (1988) 636-644 North-Holland, Amsterdam

P R O T E I N C R Y S T A L L O G R A P H I C D A T A A C Q U I S I T I O N A N D P R E L I M I N A R Y A N A L Y S I S

U S I N G K O D A K S T O R A G E P H O S P H O R P L A T E S

D o n a l d B I L D E R B A C K 1), K e i t h M O F F A T 2) J a m e s O W E N 3) B y r o n R U B I N 3)

Wi l f r i ed S C H I L D K A M P 2), D o l e t h a S Z E B E N Y I 2}, B r e n d a S M I T H T E M P L E 4), K a r l V O L Z 2)

a n d Bruce W H I T I N G 3)

0 Cornell High Energy Synchrotron Source, Wilson Laboratory, Cornell University, Ithaca, NY 14853, USA 2) Section of Biochemistry, Molecular and Cell Biology, Clark Hall, Cornell University, Ithaca, NY 14853, USA 3) Research Laboratories, Eastman Kodak Company, Rochester, NY 14650, USA 4) School of Applied and Engineering Physics, Clark Hall, Cornell University, Ithaca, N Y 14853, USA

X-ray diffraction data from single crystals of typical proteins are very weak, numerous, and subject to systematic errors arising from radiation damage at long exposure times. Compared with films, the Kodak storage phosphor technology described in the accompanying paper [1] offers the prospect of greatly improved signal-to-noise, increased sensitivity particularly at shorter wavelengths, and wide dynamic range, though with more modest spatial resolution. To assess the suitability of this technology for protein crystallographic data collection, we have collected both monochromatic oscillation and wide bandpass Laue data at CHESS on crystals ranging in unit cell size from - 5 0 ,~ (lysozyme) to -300 ,~ (viruses). A direct comparison of the Kodak storage phosphor with conventional Kodak Direct Exposure Film (DEF-5) was obtained by making immediately sequential exposures on the same crystal with the two detector systems. Even with an exposure time one order of magnitude less than with the corresponding film, the storage phosphor yielded data with improved signal-to-noise. Thus, storage phosphors enable more data to be acquired per crystal, with less radiation damage, and with better precision. Such detectors appear extremely well suited to protein crystallographic applications, both static and time-resolved, with both monochromatic and polychromatic X-ray sources.

1. Introduction

Complete X-ray diffraction data from single crystals of macromolecules typically consist of 10 4 to 10 7 ob- servations of individual diffraction maxima, depending on the size of the unit cell, the limiting resolution of the crystal, and the extent of redundancy of the data. For all but the smallest unit cells, area detectors that simul- taneously record a substantial fraction of the diffraction data provide the most efficient means of data collection. Unti l the recent introduction of two-dimensional multi- wire proportional counters (MWPCs) and TV-based detectors [2], photographic film has been the detector of choice. While MWPCs offer advantages such as high precision when used with laboratory sources, their severe limitation on total count rate does not fit them well for use with synchrotron sources. Further, they have limited spatial resolution, limited active area, and limited sensi- tivity to hard X-rays, < 1 A. TV-based systems do not suffer from all these limitations, but are as yet relatively untried detectors. There is therefore demand for a sim- ple area detector, suitable for use at synchrotron sources.

Macromolecular diffraction data cover a wide dy- namic range though the majority are very weak. Subse- quent crystallographic calculations based on these data, such as phase determination by multiple isomorphous

0168-9002/88/$03.50 © Elsevier Science Publishers B.V. (North-Holland Physics Publishing Division)

replacement combined with anomalous scattering, de- pend critically on precise comparison of diffraction intensities that typically differ by less (often, much less) than 15%. Precise measurement of the numerous weak data is, therefore, essential. However, film suffers from a high noise level due to intrinsic chemical fog. At 1.54 A, 106 photons mm -2 will produce an optical density of 1, but will yield a precision of roughly 10% in intensity, rather than the 0.1% that would be attained for a perfect, noise-free detector. It is not effective to increase the exposure time in an attempt to increase the signal, as sample radiation damage at longer exposure times lowers the accuracy substantially. The key to accurate data measurement in macromolecular crys- tallography lies more in noise reduction than in signal enhancement.

The Kodak storage phosphor detector described in the accompanying paper [1] offers the advantage of very low noise and combines it with high sensitivity and wide dynamic range, though with more modest spatial resolu- tion than film. Similar features are exhibited by the Fuji image plate detector [3,4]. Do these advantages in fact lead to macromolecular diffraction data of higher qual- ity, when using an X-ray source such as CHESS, the Cornell High Energy Synchrotron Source?

a

638 D. Bilderback et al. / Protein crystallographic data acquisition

Fig. 1. Continued.

D. Bilderback et al. / Protein crystallographic data acquisition 639

Fig. 1. Continued.

IV(c). IMAGE PLATES

640 D. Bilderback et a L / Protein crystallographic data acquisition

2. Results

We describe here a direct comparison of Kodak storage phosphors with conventional X-ray film by col- lecting identical X-ray data on both detectors. We have used both monochromatic and polychromatic, Laue sources, and applied them to crystals with small and large cell dimensions, exemplified by hen egg white lysozyme (HEWL) and viruses. In each case, a brief exposure on a given sample was recorded on a Kodak storage phosphor, then immediately followed with a usually longer but otherwise identical exposure recorded on Kodak Direct Exposure Film. The storage phosphors were scanned at Kodak, using the scanner described elsewhere [1], and the films were scanned either at Kodak or at Cornell using a conventional micro- densitometer.

Qualitatively, it was immediately obvious that the storage phosphors were more sensitive than film, in the sense of needing a substantially briefer exposure to produce an image of comparable signal-to-noise. In fig. 1, we illustrate Laue diffraction patterns [5] for a single crystal of HEWL, stimulated by a polychromatic beam derived from unfocussed, bending magnet radiation from CHESS. Only a few reflections are visible on the 1 s exposure on film (fig. la), but many more are re-

corded with excellent signal-to-noise in the correspond- ing 1 s exposure on the storage phosphor (fig. lc). Even a 100 ms exposure on storage phosphor had good signal-to-noise (fig. lb). These observations are quanti- tated in fig. 2, where the cumulative distribution of reflections as a function of the ratio of the "integrated optical density", iod (more accurately, the integrated intensity), to its standard deviation, o, is given. Many more reflections have large values of the ratio of iod to o, and fewer have physically unrealistic negative values for the storage phosphor than for film. Indeed, this ratio exceeds 14 for half of all storage phosphor data (inset, fig. 2). That is, for identical exposures recorded on storage phosphor and film, the storage phosphor data are of much higher precision. These and other data are quantitated in another form in table 1. Even when the storage phosphor exposure time was reduced to 100 ms (phosphor denoted A/final, fig. lb), 1.5 times more single reflections could be measured on the storage phosphor than on film (1134 reflections versus 673). Furthermore, the ability to refine the crystal orienta- tion, cell and camera parameters [6], as measured by the RMS difference in the observed and predicted reflection centers, was slightly superior for the storage phosphors: 19-28 /~m, against 31 /~m for the corresponding film (table 1). This result was unexpected in view of both the

g_

film

~ ~ 1//2 max % error .~ / / iod/sigma in iod

~o / / 1//2max [film :~8 35.7 .~ o [ phosphor 14.0 7.1

/ / I film 18.8 5.3

£ / / 9/10 max I phosphor 372 22

-10.0 4.0 lO.O 20.0 30.0 40.0 50.0 60.0 iod/sigma

Fig. 2. Cumulative frequency distribution of reflections, as a function of the ratio of integrated optical density (iod) to the standard error in that measurement, o, for a Laue exposure. Curves are shown for film and for the corresponding storage phosphor.

D. Bilderback et al. / Protein crystallographic data acquisition 641

Table 1 Quantitation of Laue data for Kodak storage phosphors and film. "Distance" denotes the crystal-to-detector distance; "exposure" denotes the exposure time; "range" denotes wavelength range of the incident X-ray beam; "raster" denotes the step size used in scanning the data. Under "refinement", "rms d x " denotes the root mean square difference between the observed and calculated reflection eentroids, in either raster units (ru) or #m; "rms bk" denotes the root mean square deviation of background from zero, on a scale where the maximum peak height is 100. Under "integration", entries are the numbers of reflections that contain only one (single) or more than one (multiple) structure factors.

Storage phosphors Film 1

B/final B/prescan A/final

Optics platinum mirror/aluminum filter combination Distance (ram) 180 180 180 180 Exposure 1 s - 100 ms 1 s Range (,~) 0.5-1.9 0.5-1.9 0.5-1.9 0.5-1.9 Raster (#m) 100 100 100 60 Refinement

Rms dx (ru) 0.19 0.23 0.28 0.52 Rms dx (#m) 19 23 28 31 Rms bk (pu) 0.16 0.13 0.11 0.11

Integration Singles 1775 1154 1134 673 Multiples 847 785 785 607

E/final Film 3

Optics aluminum filter/no mirror Distance (mm) 250 250 Exposure (ms) 100 500 Range (,~) 0.35-1.40 0.35-1.40 Raster (/~m) 100 60 Refinement

Rms dx (ru) 0.39 0.80 Rms dx (/~m) 39 48 Rms bk (pu) 0.22 0.36

Integration Singles 734 337 Multiples 427 200

larger raster size and the reduced spatial resolution of the phosphors, but is accounted for when the sources of noise in the phosphors and film are considered [1].

The spatial resolution of a detector is important for resolving closely adjacent reflections. Fig. 3 shows a port ion of the Lane diffraction pattern of a single crystal of bean pod mott le virus, recorded on a storage phosphor with a 10 s exposure.

Of most importance for crystallographic purposes, the crystallographic R-factors, here equivalent to a film- to-film scaling within a film pack, are lower for the storage phosphor than for film (table 2), and lie in the range of 4.2-5.2%. These values may be regarded as an upper limit, as certain small systematic errors are yet to be removed.

Preliminary information on the sensitivity of the storage phosphor as a function of wavelength was ob- tained by comparing the ratio of the intensities recorded on a storage phosphor to those on film. This ratio increased as the wavelength decreased, except for a

sharp drop at the bromine K absorption edge (data not shown; see also fig. 5 of ref. [1]).

The restricted dynamic range of film is a drawback that necessitates the use of several films in a pack, sometimes with interleaved thin metal foils. The dy- namic range of a single storage phosphor is substan- tially larger. For example, comparison of the strongest reflections on a 5% prescan [1] of a storage phosphor with the weakest reflection recorded on the correspond- ing 95% full scan yielded an integrated optical density ratio of 1200, whereas a more typical value for a 3-film pack would be 150. The value of 1200 is a lower limit as it is affected also by the parameters of the phosphor scanner, which here used only a 12°bit analog to digital converter.

A comparison of storage phosphor to film was also carried out for monochromatic oscillation data collected using.the A1 focussing wiggler station at a wavelength of 1.55 A. In fig. 4 we compare the distribution of reflec- tions as a function of the ratio of integrated optical

IV(c). IMAGE PLATES

642 D. Bilderback et al. / Protein crystallographic data acquisition

Fig. 3. Lane exposure of a single crystal of bean pod mottle virus, which belongs to space group P21212 with cell dimensions of 312 × 280 × 345 ,~,. Exposure time 10 s at 4 ° C, on Kodak storage phosphor. Synchrotron running at 5.17 GeV, 18.9 mA. Float glass

mirror set at 2.1 mrad, plus 125/.tin A1 foil. Crystal-to-detector distance of 500 mm.

D. Bilderback et a L / Protein crystallographic data acquisition 643

50

40

3O No.

Refls. 20

10

0 0

r-- 4 6 8 10 12 14 16 18

I/a I

77

Storage phosphor R, lysozyme 12 sec, 3 ° osc., 100/~, old scanner Resolution: 2.7J[ R=ym: 0.055 (142 overlaps) Unobserved whole refls.: 2.4~;

19

50

4O

30 No.

Refls. 20

10

0 0

7-

2 4 6 8 10 12 14 I/a,

178

16 18 >19

Storage phosphor L15, lysozyme 16 sac, 4 ° osc., 50/~, new scanner Resolution: 2.0~ (max) R=ym: 0.039 (127 overlaps) Unobserved whole refls.: 9.2%

3O

No. 20 I Refls. 10

0 0

1 2 4 6 8 10 12 14 16 18 >19

I/~,

Film 18-2A, placental Ioctogen 2 rain, 6 ° osc., 50/z Resolution: 3.2~, Riym: 0.059 (55 overlaps) Unobserved whole refls.: 24.5~

20 //----] Film 5A (partial), lysozyme

I ~ 48 sec, 3* OSC., 100# No. 10 H= J Resolution: 2.7~ (max)

Refls. R=ym: 0.0,36 (13 overlaps) 0

0 2 4 6 8 10 12 14 16 18 > 19 Unobserved whole refls.: 4 .0~ i/a,

Fig. 4. Frequency distribution of reflections as a function of the ratio of iod to a (see legend to fig. 2), for monochromatic oscillation data. Upper two panels: data recorded on Kodak storage phosphors. Lower two panels: data recorded on film (third panel, CEA film; fourth panel, Kodak Direct Exposure Film). Note that for storage phosphor L15, the ratio of iod to a exceeds 19 for the vast

majority of the data.

densi ty to o for two storage phosphors and two films. (Fig. 4 shows the d is t r ibut ions themselves, whereas fig. 2 shows cumulat ive distr ibutions.) Fi lm 18-2A was a film of average quali ty taken on a crystal of placental lactogen using CEA film. Fi lm 5A was a high quali ty fi lm taken on a crystal of H E W L using Kodak Direct Exposure F i lm ( temporary l imitat ions on compute r

capacity required scanning only par t of this film, with a 100 /~m raster; 50 /~m would have been more ap- propriate) . Phosphor R was taken on the same lysozyme crystal with one-quar ter the exposure time. Phosphor L15 was taken on ano ther lysozyme crystal and processed on an improved version of the scanner used

for phosphor R.

IV(c). IMAGE PLATES

644 D. Bilderback et al. / Protein crystallographic data acquisition

Table 2 Image-to-image scaling of Laue data for storage phosphors. This process is analogous to film-to-film scaling within a film pack, prior to making wavelength-dependent corrections. "Psi" denotes the so-called function of merit [6], and is lower for superior data; "R-factor" is the standard crystallographic R- factor on intensities; "# refs" and "~ obs" denote the final number of merged reflections, and the number of initial ob- servations that were merged.

Storage phosphors Psi R-factor # refs/# obs

B/final 3.16 5 . 2 3 1261/3492 B/prescan A/final

B/final 1.30 4 . 3 4 1115/2230 B/prescan

B/final 1.78 4 . 1 6 1092/2184 A/final

A/final 1.21 4.91 994/1988 B/prescan

Films 5.5-8.0 (typical)

As for the Laue data, the storage phosphors again have many more reflections with a high signal-to-noise, particularly in the case of phosphor L15, i.e. when the improved scanner was used. Symmetry R-factors for the storage phosphors are comparable to those found on good quality films, and the percentage of unobserved whole reflections is lower for a particular resolution hmit.

3. Discussion

Although these are preliminary results based on the quantitation of a small number of phosphors and films, they point to several substantial advantages of the Kodak storage phosphor detector over film: increased sensitivity permitting shorter exposures and hence re-

duced radiation damage; superior signal-to-noise result- ing in lower crystallographic R-factors; sensitivity to a wide range of X-ray wavelengths; and spatial resolution adequate for almost all crystallographic applications. In a realistical experimental situation, a tradeoff will result between exposure time and signal-to-noise: a short ex- posure has higher random error, but lower systematic error due to radiation damage. From the data presented here, we estimate that exposure times at least one order of magnitude less than for film are possible for Kodak storage phosphors, while retaining superior signal-to- noise and crystallographic R-factors. It appears that the technical advantages of Kodak storage phosphors, de- scribed in detail in the accompanying paper [1], are retained in a realistic experimental situation. A more extensive quantitative analysis of Lane and monochro- matic data, both static and dynamic, is in progress.

Acknowledgements

Supported by NIH grant RR-01646 to K.M. We thank Mark Bommarito, Ying Chen, Sheri Wischusen, Armin Spirgatis and the CHESS staff for experimental assistance, and Professor Jack Johnson of Purdue Uni- versity for the bean pod mottle virus crystals.

References

[1] B.R. Whiting, J.F. Owen and B.H. Rubin, these Proceed- ings (5th Nat.Conf. on Synchrotron Radiation Instrumen- tation, Univ. of Wisconsin-Madison, 1987) Nucl. Instr. and Meth. A266 (1988) 628.

[2] U.W. Arndt, J. AppL Cryst. 19 (1986) 145. [3] J. Miyahara, K. Takahashi, Y. Amemiya, N. Kamiya and

Y. Satow, Nucl. Instr. and Meth. A246 (1986) 572. [4] Y. Amemiya, T. Matsushita, A. Nakagawa, Y. Satow, J.

Miyahara and J.-I. Chikawa, ref. [1], p. 645. [5] K. Moffat, D. Szebenyi and D. Bilderback, Science 223

(1984) 1423. [6] B.S. Temple and K. Moffat, Proceedings of the Daresbury

Study Weekend (1987) to be published.