J. Mol. Biol. (1992) 223, 1177-1182
Crystallization
of B-Galactosidase from Escherichia coli
Raymond H. Jacobson and Brian W. Matthews Howard
Institute of Molecular Biology Hughes Medical Institute and Department University
of Physics
of Oregon
Eugene, Oregon 97403, U.S.A. (Received 17 July
1991; accepted 29 October 1991)
Two crystal forms of P-galactosidase have been obtained from Escherichia coli. One crystal form is hexagonal space group P6,22 or enantiomorph, with cell dimensions a = b = 154 A, e = 750 A. The second form is monoclinic, space group P2,, with cell dimensions a = 107.9 A, b = 207.5 A, c = 5099 A, /3 = 947”. The monoclinic form seems better suited to detailed structural analysis. The crystals are radiation-sensitive, but by using synchrotron radiation in conjunction with a long (400 mm) crystal-to-film distance it was possible to resolve the individual reflections. On the basis of crystal density measurements, there are four tetramers each of molecular weight 465,000 per asymmetric unit. The Patterson function strongly suggests that two of the tetramers are related to the other two by translation. The data are consistent with the tetramers having 222 point symmetry, but this is not proven. Keywords: fl-galactosidase;
crystals; synchrotron
b-Galactosidase (EC3.2.1.23) from Escherichia coli is a disaccharidase that catalyzes the hydrolysis and transgalactosylis of fl-galactopyranosides. In E. co& this enzyme is responsible for the hydrolysis of the pl-4 linkage of lactose, producing galactose and glucose, which the bacteria can utilize as carbon sources, as well as the transgalactosylic formation of allolactose (galactosyl-P-n-(I-6)-glucopyranose), the natural inducer of the lactose operon (Huber et al., 1976, and references therein). The purification of the enzyme was first carried out in the early 1950s by Cohn and Monod, and study of the regulation of P-galactosidase production in E. coli led Jacob and Monod to propose the first model of operon regulation (Jacob & Monod, 1961; Wallenfels & Weil, 1972, and references therein). Many uses have been found for the enzyme because of the convenience of the calorimetric assay for determining its activity (Lederberg, 1950). Hybrid /l-galactosidase molecules that retain glycosidase activity are routinely made by fusing the polypeptide chain of a protein of interest to the /I-galactosidase sequence. These fusion proteins can then be used to detect new gene products with activities that are difficult to assay, to follow protein translocation across membranes, to produce mammalian hormones in E. co&, and in many other settings (Fowler & Zabin, 1983, and references therein). 1177 0022-2836/92/041177-06
$03.00/O
radiation
The functional enzyme is most probably a tetramer but might also be a higher aggregate of identical subunits (Appel et al., 1965; Fowler & Zabin, 1970, and references therein). Each monomer consists of 1023 amino acids with M, = 116,353 (Kalnins et al., 1983; Fowler & Zabin, 1978). The catalytic mechanism of B-galactosidase is thought to be a general acid-catalyzed hydrolysis somewhat analogous to that of hen egg-white lysozyme, with Tyr503 probably acting as the proton donor (Sinnot, 1978; Ring & Huber, 1990). Each tetramer contains four active sites (Cohn, 1957), and Mg*+ plus K+ or Na+ is required for maximal activity (Wallenfels & Weil, 1972, and references therein). Initial crystals of /?-galactosidase were obtained using purified protein supplied by Drs J. Harris and P. L. Khanna of Microgenics Corp., Concord, CA. Subsequently, fl-galactosidase was purified from strain A324-5 of E. coli using the method developed by Fowler & Zabin (1983). The purification follows the protocol described; however, a 60 ml p-aminobenzyl-1-thio-fi-n-galactopyranoside agarose (Sigma no. A-0414) column was also used between the DEAE and Sephacryl 5300 columns, as suggested by I. Zabin (personal communication). This additional column is composed of a non-hydrolyzable substrate analogue covalently attached to agarose beads and has high affinity for P-galactosidase. The 0
1992 Academic
Press Limited
1178
R. H. Jacobson and B. W. Matthews
use of a salt gradient to elute the bound protein from this column material serves to increase protein purity by separating lower molecular weight contaminants from the intact j?-galactosidase molecules. The pooled fraction containing the j?galactosidase from the DEAE column was loaded onto the substrate affinity column, which had been equilibrated with 10 column volumes of 40 mw-Tris * HCl(pH 75), 1 m&r-MgCl,, O-5 mM EDTA, and 10 mM-fl-mercaptoethanol #ME). The column was washed with 250 ml of the equilibration buffer and then with 400 ml of the equilibration buffer containing 300 mM-KCl. At this point the fi-galactosidase was eluted from the column using an 800 ml gradient containing 200 mM Tris. HCI (pH 8+6), 1 mM-MgCl,, 95 mM-EDTA, 10 mM$ME and 0.5 or I.5 M-KC]. The contaminants elute first, followed by the intact B-galactosidase tetramers. The pooled fraction from the affinity column was precipitated with ammonium sulfate, resuspended at 25 mg/ml and loaded onto the Sephacryl 300 column. Purification then proceeded as described by Fowler & Zabin (1983). The purified enzyme was divided into 1 ml aliquots at approximately 10 mg/ml and flash frozen in liquid nitrogen for storage at -80°C. To prepare the j?-galactosidase for crystallization, a 1 ml aliquot was thawed and dialyzed against 10 miv-Tris.HCl (pH 7.5), 50 m&i-NaCl (without /?ME). The dialyzed protein was then concentrated to 20 mg/ml in a Centricon microconcentrator with a molecular weight cut-off of 30,000 M,. The protein solution was left at 4°C for approximately one week prior to use. Crystals were grown by the method of hanging-drop vapor diffusion (McPherson, 1982) using a mixture of polyethyleneglycol (PEG) 8000, 40 mM-cacodylic acid (pH 59), 80 mM-MgSG, and BME. A typical crystallization experiment screened well solutions containing PEG concentrations between 7 and 11 y0 (w/w) and BME concentrations between 70 and 140 mM. A 5 ~1 sample of protein was mixed with an equal volume of well solution, suspended over 1 ml of well buffer, and kept at room temperature. Two different crystal forms appeared after approximately three to four weeks. In some cases both crystal forms appeared in the same drop. The first crystal form was more likely to appear under conditions containing 70 to 105 mM-flME. These crystals grew as hexagonal rods and reached dimensions of 1.0 mm x 64 mm x 94 mm. Diffraction patterns taken on a Rigaku rotatinganode generator operating at 40 kV and 150 mA showed diffraction spectra to 3-l A (1 A = 01 nm). These crystals were determined as belonging to space group P6,22 or its enantiomorph, with cell dimensions of a = b = 154 A, c = 750 A. The 750 A cell edge was not resolvable using in-house facilities and was estimated from the radius of the first level reflections on a still photograph with the crystal oriented to the hk0 zone. The initial attempts to characterize the hexagonal crystals included the use of precession photography. A 6” precession was
collected of the h01 zone at a crystal-to-film distance of 150 mm using a 92 mm x 0.2 mm collimator with a pin-hole focus. Although in general the reflections along the 1 direction were not resolved, the photograph displayed mm symmetry and a clear pattern of systematic absences was observed for the 002 reflections. On the basis of the estimate of the c cell edge and the spacing of the 001 reflections, these were determined as corresponding to reflections obeying the relationship 1 = 3n; hence the space group was P6,22 or its enantiomorph. The crystals were very radiation-sensitive, with essentially no diffraction remaining after 24 hours under the beam conditions described above. The second crystal form usually appeared under conditions containing greater than 105 mM-pME and grew as rhombic prisms. These were often semihollow at one end and could attain dimensions of I.0 mm x 0.3 mm x 93 mm. The crystals were found to belong to space group P2, with cell dimena = 107.9 A, b = 207.5 ii, c = 509.9 A, sions fi = 947”. Using a xylene/carbon tetrachloride gradient, the density of the monoclinic crystals was determined to be I.18 g/cm3 and that of the solvent I.04 g/cm3. From these values it was determined (Matthews, 1985) that each asymmetric unit (a.s.u.) contained 2,100,OOO M, of protein. This value corresponds to 4.4 tetramers/a.s.u. and yields a v, of 2.6 A3/dalton, which is within the normal range (Matthews, 1968). Diffraction spectra on still photographs were observable to at least 2.7 A. The intensity of diffraction was also stronger, by about a factor of 2, relative to that observed from t,he hexagonal crystals. However, the monoclinic crystals showed no improvement over the hexagonal crystals in radiation sensitivity. The cell dimensions listed above were measured from oscillation photographs taken at the CHESS synchrotron facilit’y at Cornell University. Native data sets have been collected for both crystal forms at CHESS. The data sets were recorded using 6908 A radiation on 8 inch x 10 inch pieces of Kodak DEF-5 X-ray film. The hexagonal data set was collected at a crystal-to-film distance of 500 mm with a 200 pm collimated beam. The maximum resolution that could be recorded in this configuration was approximately 4.0 A at the top and bottom edges of the film. Oscillations of 2” were used and diffraction extended to approximately 5.0 A. The limited resolution was partly due to the size of the crystals that were available at the time of typically 63mmx visit to CHESS, the 93 mm x 0.15 mm. The monoclinic data set was collected at a crystal-to-film distance of 400 mm with a 300 pm collimated beam. Oscillations of 2” were also used for the monoclinic crystals and diffraction extended to the top and bottom edges of the films. At the crystal-to-film distance of 400 mm this corresponded to better than 3.0 A resolution. A typical film from the monoclinic data set is shown in Figure 1. Recorded on this film are approximately 18,000 reflections. To maximize data completeness the c* axis was aligned along the
Crystallization
Notes
1179
Figure 1. Diffraction pattern of the monoclinic crystals recorded at CHESS. The inset shows an enlarged view of the diffraction pattern above the beam stop. The scale is marked in millimeters. The blackening of the film to the left of the beam stop is not due to misalignment but to the very high intensity of the (0,0,6) reflection. 2” oscillation photograph; exposure time 2.5 min; crystal-to-film distance 400 mm; I = 0.908” 8; film size 8 inches x 10 inches; crystal size 0% mm x 02 mm x @2 mm. The diffraction limit at the top and bottom edge of the film is 30 d. rotation axis using still photographs on Polaroid film and each crystal was then rotated to its appropriate oscillation range. Alignment and two oscillation photographs 2.5 min exposures) could be obtained from each crystal before it had decayed. A total of 19 crystals with dimensions of approximately @8 mm x @2 mm x @2 mm were used to collect 38 oscillation photographs. These have been digitized using an Optronics P-1000 drum scanner at a 50 pm raster step size. The digitized images were processed using the program of Rossmann as modified by Schmid (Rossmann, 1979; Schmid et al., 1981). The resultant data set contained 466,195 measurements, which reduced to 199,493 unique reflections. The data set is 70% complete to 3.5 d
resolution with an agreement between independently measured intensities (Emerge) of 9.7%. The missing data correspond roughly to two oscillation wedges, each of about lo”, near the principal zones. Inspection of the films making up the data set from the monoclinic crystals led to the identification of 22 low-resolution reflections (Bragg spacings, 25 A or greater) that had intensities several orders of magnitude greater than the mean intensity of the diffraction pattern. One such reflection, with Miller indices (0,0,6), is visible just to the left of the beamstop shadow in Figure 1. Owing to the extreme saturation of these reflections it was not possible to measure their intensities accurately. Their indices obey the relation h+E = 2n, which corresponds to a
R. H. Jacobson and B. W. Matthews
1180
--i---*------------------------------__..--_-..-____--_______________ I ,
C
0
Figure 2. Section r = 0.03 of the Patterson function for the monoclinic crystal form with 22 very intense low-order reflections omitted (see text). Resolution 5 A. The peak at the center of the section is of height approximately 20 cr, where o is the average value of the function throughout the unit cell. This peak is about 10% of the height of the origin peak. Contours are drawn at increments of 2 u.
pseudo
B-face-centering
operation,
i.e. orthogonal
to the crystallographic 2, axis. The Patterson function at 5 a resolution, computed with F values from the processed data (i.e. not including the super-intense reflections)
contains a peak at (u,v,w) = (05,@03,0.5) (Fig. 2), which is approximately 20 standard deviations above the noise level of the map or about 10% of the height of the origin. Apart from this peak the map is essentially featureless. (The absence of the
Figure 3. Section K = 180” of the locked rotation function (Tong & Rossmann, 1990). Resolution 190 to 5.5 8. This Figure displays the result of the calculation as a stereographic projection. The directions in space are plotted in spherical polar co-ordinates where the axial tilt of the rotation vector away from the c axis of the crystal co-ordinate system ($) is plotted latitudinally and the rotation within the a*b* plane (4) longitudinally. The 3 crosses indicate 1 choice of a set of 3 orthogonal axes of Z-fold symmetry (see the text). Contours are drawn at increments of 1 0 with the lowest contour drawn at 1.3 0 (see legend to Fig. 2).
Crystallization
intense reflections may perturb the actual peak height but’ should not otherwise materially effect the result.) Tt was of particular interest that the Harker section (V = @5) was featureless, since if a molecular f-fold was oriented parallel to the crystallographic 2, screw axis a peak in the Harker section should be present. The magnitude of the peak at (0.5,@03,0.5) suggests that the contents of the asymmetric unit can be considered as being made up of two pairs of tetramers (physically, this might correspond to two octamers. but it is not required). One pair of tetramers would be related to the second pair by an approximate translation, z = 05, y = @03, z = 05. giving rise to the strong peak in the Patterson function. Within a given pair, the two tetramers would have to have different orientations t)o avoid the presence of additional translation vectors. which would be apparent in the Patterson function. The approximate dimensions of tetrametic /&galactosidase have been estimated from electron microgra,phs (Karlsson et al., 1964) to be This suggests that t.he 120 A x 120 LAx 70 8. tet)ramers must be packing essentially side-by-side to be consistent with the 509 A cell edge and the narrow 108 A cell edge. The self-rotation function has been calculated using several resolution ranges of the data. The different, choices of resolution shells for the calculation all resulted in similar maps. The IC = 180” section, which shows the orientations of potential non-crystallographic 2-fold axes, always contained the most significant features. The K = 90” section, which is expected to show the orientations of potential molecular 4-fold axes, never contained significant peaks. Figure 3 shows the JC= 180” section of the self-rot)ation function, produced using the General Locked Rotation Function program (Tong & Rossmann, 1990). The large peak at the center of the Figure corresponds to the crystallographic 2, screw axis along b. This axis relates the right and left halves. so that only half of the Figure is unique. However. the pseudo mirror symmetry seen in the Figure is not required by space group symmetry. Several potential 2-fold axes are apparent. Since many tetrameric proteins exhibit 222 point symmet,ry (Matthews & Bernhard, 1973), it was of interest t.o examine the spatial relationships between the candidates for orthogonality. One possible choice of three orthogonal vectors might include the peaks at ($,@) = (0”,48”),(0”,138”), with the third lying in the same direction as the crystallographic 2, axis (90”.90”). However, as previously mentioned. a molecular 2-fold axis that is oriented parallel to the crystallographic 2, axis generates a t.ranslation vector that would generate an obvious peak in thr Harker section of the Patterson function. The absence of any such feature rules out the possibility that these three rotation axes could form the 222 point symmetry elements of a tetramer. An alternative set of orthogonal 2-fold rotation axes, shown in Figure 3. is formed by the peaks at (0”,48”),(146”,47”) and (72”. 108”). This suggests the presence of a set of subunits, possibly a single
1181
Notes
tetramer, having 222 point symmetry. Because there is a peak midway between the second and third peaks it is possible that a second tetramer is related to the first by rotation of 180” about an axis that bisects a pair of the 222 tetrameric symmetry axes. In summary, two different crystal forms of bgalactosidase have been obtained. Both have one long unit cell edge but the use of synchrotron radiation in conjunction with large crystal-to-film distances (and large film sizes) makes it feasible to resolve the diffraction spectra and to measure intensities of the monoclinic crystals to Bragg spacings of at least 3.5 W resolution. The asymmetric unit of the monoclinic crystals contains four tetramers that seem to be related, by translation; in pairs. The data are consistent with the postulate that each tetramer has 222 point symmetry. The first. crystals were obtained using protein kindly provided by Drs *Jeffrey Harris and Pyare Khanna of Microgenics Corporation, Concord, CA. Drs Scott Eisenbeis and Mark Krevolin, also of Microgenics. provided the E. coli cells used for subsequent protein purification. We also thank Dr Don Bilderback at CHESS for invaluable assistance configuring the beam line to make the data collection feasible, Drs Irving Zabin and R. E. Huber for helpful discussions and advice on the biochemistry and purification of fl-galactosidase, Dr M. G. Rossmann for advice concerning film processing and R. Albright and Dr S. Roderick for much needed help with the data collection. This work was supported in part by grants from the National Institut’es of Health (GM20066) and the Lucille P. Markey Charitable Trust.
References Appel, S. H., Alpers, D. H. & Tomkins, G. M. (1965). Multiple molecular forms of fl-galactosidase. J. Mol. Biot. 11, 12-22. Cohn, M. (1957). Contributions of studies on the pgalactosidase of Escherichia coli to our understanding of enzyme synthesis. Bacterial. Rev. 21. 140-168. Fowler. A. & Zabin, I. (1970). The amino acid sequence of fi-galactosidase. I. Isolation and composition of tryptic peptides. J. Biol. Chem. 245. 5032-5041. Fowler, A. & Zabin, I. (1978). Amino acid sequence of pgalactosidase. XI. Peptide ordering procedures and Chem. 253. the comp1et.e sequence. d. Biol. 5521-5525. Fowler, A. & Zabin, I. (1983). Purification. structure, and properties of hybrid fi-galartosidase proteins. J. Nol. Biol. 104, 541-555. Huber, R. E. Kurz, G. & Wallenfels. K. (1976). A quantitation of the factors which affect the hydrolase and transgalactosylase activities of fl-galactosidase (E. coli) on lactose. Biochemistry. 15. 1994-2001. Jacob, F. & Monod. tJ. (1961). Genetic regulatory mechanisms in the synthesis of prot,eins. J. Mol. Biol. 3, 318-356. Kalnins, A.: Otto, K., Ruther, U. & Miiller-Hill. B. (1983). Sequence of the 1acZ gene of Ir:scherichia coli. EMBO
J. 2, 593-597.
Karlsson, G., Koorajian, S., Zabin, I., Sjiistrand, F. S. & Miller. A. (1964). High resolution electron microscopy on highly purified fl-galactosidase from Escherichia coli. J. Ultrastruct.
Res. 10, 457-469.
1182
R. H. Jacobson and B. W. Matthews
Lederberg, J. (1950). The /I-D-galactosidase of Escherichia coli strain K-12. J. Bucteriol. 60, 381-399. Matthews, B. W. (1968). Solvent content of protein crystals. J. Mol. Biol. 33, 491497. Matthews, B. W. (1985). Determination of protein molecular weight, hydration, and packing from crystal density. Methods Enzymol. 114, 176-186. Matthews, B. W. & Bernhard, S. A. (1973). Structure and symmetry of oligomeric enzymes. Annu. Rev. Biophys. Bioeng. 2, 257-317. and Analysis of McPherson, A. J. (1982). Preparation Protein Crystals, pp. 82-160, John Wiley and Sons, New York. Ring, M. & Huber, R. E. (1990). Multiple replacements establish the importance of tyrosine 503 in /Igalactosidase (Escherichia co&). Arch. Biochem. Biophys.
283, 342-350.
data for very large unit cells with an automatic convolution technique and profile fitting. ,I. Appl. Crystallogr. 12, 225-239. Schmid, M. F., Weaver, L. H., Holmes, M. A., &utter. M. G., Ohlendorf, D. H., Reynolds. R. A.. Remington, S. J. & Matthews, B. W. (1981). An oscillation data collection system for high-resolution Acta Crystallogr. sect. A, 37. protein crystallography. 701-710. Sinnott, M. L. (1978). Ions, ion-pairs and catalysis by the 1acZ /I-galactosidase of Escherichia coli. FEBR Letters,
Rossmann, M. G. (1979). Processing oscillation diffraction
Edited
94, 1-9.
Tong, L. & Rossmann, M. G. (1990). The locked rotation function. Acta Crystallogr. sect. A, 46, 783-792. Wallenfels, K. & Weil, R. (1972). /I-Galactosidase. In The Enzymes (Boyer, P. D., ed.), 3rd edit.. vol. 7, pp. 617-663, Academic Press, London.
by W. Hendrickson
Crystallization
of B-Galactosidase from Escherichia coli
Raymond H. Jacobson and Brian W. Matthews Howard
Institute of Molecular Biology Hughes Medical Institute and Department University
of Physics
of Oregon
Eugene, Oregon 97403, U.S.A. (Received 17 July
1991; accepted 29 October 1991)
Two crystal forms of P-galactosidase have been obtained from Escherichia coli. One crystal form is hexagonal space group P6,22 or enantiomorph, with cell dimensions a = b = 154 A, e = 750 A. The second form is monoclinic, space group P2,, with cell dimensions a = 107.9 A, b = 207.5 A, c = 5099 A, /3 = 947”. The monoclinic form seems better suited to detailed structural analysis. The crystals are radiation-sensitive, but by using synchrotron radiation in conjunction with a long (400 mm) crystal-to-film distance it was possible to resolve the individual reflections. On the basis of crystal density measurements, there are four tetramers each of molecular weight 465,000 per asymmetric unit. The Patterson function strongly suggests that two of the tetramers are related to the other two by translation. The data are consistent with the tetramers having 222 point symmetry, but this is not proven. Keywords: fl-galactosidase;
crystals; synchrotron
b-Galactosidase (EC3.2.1.23) from Escherichia coli is a disaccharidase that catalyzes the hydrolysis and transgalactosylis of fl-galactopyranosides. In E. co& this enzyme is responsible for the hydrolysis of the pl-4 linkage of lactose, producing galactose and glucose, which the bacteria can utilize as carbon sources, as well as the transgalactosylic formation of allolactose (galactosyl-P-n-(I-6)-glucopyranose), the natural inducer of the lactose operon (Huber et al., 1976, and references therein). The purification of the enzyme was first carried out in the early 1950s by Cohn and Monod, and study of the regulation of P-galactosidase production in E. coli led Jacob and Monod to propose the first model of operon regulation (Jacob & Monod, 1961; Wallenfels & Weil, 1972, and references therein). Many uses have been found for the enzyme because of the convenience of the calorimetric assay for determining its activity (Lederberg, 1950). Hybrid /l-galactosidase molecules that retain glycosidase activity are routinely made by fusing the polypeptide chain of a protein of interest to the /I-galactosidase sequence. These fusion proteins can then be used to detect new gene products with activities that are difficult to assay, to follow protein translocation across membranes, to produce mammalian hormones in E. co&, and in many other settings (Fowler & Zabin, 1983, and references therein). 1177 0022-2836/92/041177-06
$03.00/O
radiation
The functional enzyme is most probably a tetramer but might also be a higher aggregate of identical subunits (Appel et al., 1965; Fowler & Zabin, 1970, and references therein). Each monomer consists of 1023 amino acids with M, = 116,353 (Kalnins et al., 1983; Fowler & Zabin, 1978). The catalytic mechanism of B-galactosidase is thought to be a general acid-catalyzed hydrolysis somewhat analogous to that of hen egg-white lysozyme, with Tyr503 probably acting as the proton donor (Sinnot, 1978; Ring & Huber, 1990). Each tetramer contains four active sites (Cohn, 1957), and Mg*+ plus K+ or Na+ is required for maximal activity (Wallenfels & Weil, 1972, and references therein). Initial crystals of /?-galactosidase were obtained using purified protein supplied by Drs J. Harris and P. L. Khanna of Microgenics Corp., Concord, CA. Subsequently, fl-galactosidase was purified from strain A324-5 of E. coli using the method developed by Fowler & Zabin (1983). The purification follows the protocol described; however, a 60 ml p-aminobenzyl-1-thio-fi-n-galactopyranoside agarose (Sigma no. A-0414) column was also used between the DEAE and Sephacryl 5300 columns, as suggested by I. Zabin (personal communication). This additional column is composed of a non-hydrolyzable substrate analogue covalently attached to agarose beads and has high affinity for P-galactosidase. The 0
1992 Academic
Press Limited
1178
R. H. Jacobson and B. W. Matthews
use of a salt gradient to elute the bound protein from this column material serves to increase protein purity by separating lower molecular weight contaminants from the intact j?-galactosidase molecules. The pooled fraction containing the j?galactosidase from the DEAE column was loaded onto the substrate affinity column, which had been equilibrated with 10 column volumes of 40 mw-Tris * HCl(pH 75), 1 m&r-MgCl,, O-5 mM EDTA, and 10 mM-fl-mercaptoethanol #ME). The column was washed with 250 ml of the equilibration buffer and then with 400 ml of the equilibration buffer containing 300 mM-KCl. At this point the fi-galactosidase was eluted from the column using an 800 ml gradient containing 200 mM Tris. HCI (pH 8+6), 1 mM-MgCl,, 95 mM-EDTA, 10 mM$ME and 0.5 or I.5 M-KC]. The contaminants elute first, followed by the intact B-galactosidase tetramers. The pooled fraction from the affinity column was precipitated with ammonium sulfate, resuspended at 25 mg/ml and loaded onto the Sephacryl 300 column. Purification then proceeded as described by Fowler & Zabin (1983). The purified enzyme was divided into 1 ml aliquots at approximately 10 mg/ml and flash frozen in liquid nitrogen for storage at -80°C. To prepare the j?-galactosidase for crystallization, a 1 ml aliquot was thawed and dialyzed against 10 miv-Tris.HCl (pH 7.5), 50 m&i-NaCl (without /?ME). The dialyzed protein was then concentrated to 20 mg/ml in a Centricon microconcentrator with a molecular weight cut-off of 30,000 M,. The protein solution was left at 4°C for approximately one week prior to use. Crystals were grown by the method of hanging-drop vapor diffusion (McPherson, 1982) using a mixture of polyethyleneglycol (PEG) 8000, 40 mM-cacodylic acid (pH 59), 80 mM-MgSG, and BME. A typical crystallization experiment screened well solutions containing PEG concentrations between 7 and 11 y0 (w/w) and BME concentrations between 70 and 140 mM. A 5 ~1 sample of protein was mixed with an equal volume of well solution, suspended over 1 ml of well buffer, and kept at room temperature. Two different crystal forms appeared after approximately three to four weeks. In some cases both crystal forms appeared in the same drop. The first crystal form was more likely to appear under conditions containing 70 to 105 mM-flME. These crystals grew as hexagonal rods and reached dimensions of 1.0 mm x 64 mm x 94 mm. Diffraction patterns taken on a Rigaku rotatinganode generator operating at 40 kV and 150 mA showed diffraction spectra to 3-l A (1 A = 01 nm). These crystals were determined as belonging to space group P6,22 or its enantiomorph, with cell dimensions of a = b = 154 A, c = 750 A. The 750 A cell edge was not resolvable using in-house facilities and was estimated from the radius of the first level reflections on a still photograph with the crystal oriented to the hk0 zone. The initial attempts to characterize the hexagonal crystals included the use of precession photography. A 6” precession was
collected of the h01 zone at a crystal-to-film distance of 150 mm using a 92 mm x 0.2 mm collimator with a pin-hole focus. Although in general the reflections along the 1 direction were not resolved, the photograph displayed mm symmetry and a clear pattern of systematic absences was observed for the 002 reflections. On the basis of the estimate of the c cell edge and the spacing of the 001 reflections, these were determined as corresponding to reflections obeying the relationship 1 = 3n; hence the space group was P6,22 or its enantiomorph. The crystals were very radiation-sensitive, with essentially no diffraction remaining after 24 hours under the beam conditions described above. The second crystal form usually appeared under conditions containing greater than 105 mM-pME and grew as rhombic prisms. These were often semihollow at one end and could attain dimensions of I.0 mm x 0.3 mm x 93 mm. The crystals were found to belong to space group P2, with cell dimena = 107.9 A, b = 207.5 ii, c = 509.9 A, sions fi = 947”. Using a xylene/carbon tetrachloride gradient, the density of the monoclinic crystals was determined to be I.18 g/cm3 and that of the solvent I.04 g/cm3. From these values it was determined (Matthews, 1985) that each asymmetric unit (a.s.u.) contained 2,100,OOO M, of protein. This value corresponds to 4.4 tetramers/a.s.u. and yields a v, of 2.6 A3/dalton, which is within the normal range (Matthews, 1968). Diffraction spectra on still photographs were observable to at least 2.7 A. The intensity of diffraction was also stronger, by about a factor of 2, relative to that observed from t,he hexagonal crystals. However, the monoclinic crystals showed no improvement over the hexagonal crystals in radiation sensitivity. The cell dimensions listed above were measured from oscillation photographs taken at the CHESS synchrotron facilit’y at Cornell University. Native data sets have been collected for both crystal forms at CHESS. The data sets were recorded using 6908 A radiation on 8 inch x 10 inch pieces of Kodak DEF-5 X-ray film. The hexagonal data set was collected at a crystal-to-film distance of 500 mm with a 200 pm collimated beam. The maximum resolution that could be recorded in this configuration was approximately 4.0 A at the top and bottom edges of the film. Oscillations of 2” were used and diffraction extended to approximately 5.0 A. The limited resolution was partly due to the size of the crystals that were available at the time of typically 63mmx visit to CHESS, the 93 mm x 0.15 mm. The monoclinic data set was collected at a crystal-to-film distance of 400 mm with a 300 pm collimated beam. Oscillations of 2” were also used for the monoclinic crystals and diffraction extended to the top and bottom edges of the films. At the crystal-to-film distance of 400 mm this corresponded to better than 3.0 A resolution. A typical film from the monoclinic data set is shown in Figure 1. Recorded on this film are approximately 18,000 reflections. To maximize data completeness the c* axis was aligned along the
Crystallization
Notes
1179
Figure 1. Diffraction pattern of the monoclinic crystals recorded at CHESS. The inset shows an enlarged view of the diffraction pattern above the beam stop. The scale is marked in millimeters. The blackening of the film to the left of the beam stop is not due to misalignment but to the very high intensity of the (0,0,6) reflection. 2” oscillation photograph; exposure time 2.5 min; crystal-to-film distance 400 mm; I = 0.908” 8; film size 8 inches x 10 inches; crystal size 0% mm x 02 mm x @2 mm. The diffraction limit at the top and bottom edge of the film is 30 d. rotation axis using still photographs on Polaroid film and each crystal was then rotated to its appropriate oscillation range. Alignment and two oscillation photographs 2.5 min exposures) could be obtained from each crystal before it had decayed. A total of 19 crystals with dimensions of approximately @8 mm x @2 mm x @2 mm were used to collect 38 oscillation photographs. These have been digitized using an Optronics P-1000 drum scanner at a 50 pm raster step size. The digitized images were processed using the program of Rossmann as modified by Schmid (Rossmann, 1979; Schmid et al., 1981). The resultant data set contained 466,195 measurements, which reduced to 199,493 unique reflections. The data set is 70% complete to 3.5 d
resolution with an agreement between independently measured intensities (Emerge) of 9.7%. The missing data correspond roughly to two oscillation wedges, each of about lo”, near the principal zones. Inspection of the films making up the data set from the monoclinic crystals led to the identification of 22 low-resolution reflections (Bragg spacings, 25 A or greater) that had intensities several orders of magnitude greater than the mean intensity of the diffraction pattern. One such reflection, with Miller indices (0,0,6), is visible just to the left of the beamstop shadow in Figure 1. Owing to the extreme saturation of these reflections it was not possible to measure their intensities accurately. Their indices obey the relation h+E = 2n, which corresponds to a
R. H. Jacobson and B. W. Matthews
1180
--i---*------------------------------__..--_-..-____--_______________ I ,
C
0
Figure 2. Section r = 0.03 of the Patterson function for the monoclinic crystal form with 22 very intense low-order reflections omitted (see text). Resolution 5 A. The peak at the center of the section is of height approximately 20 cr, where o is the average value of the function throughout the unit cell. This peak is about 10% of the height of the origin peak. Contours are drawn at increments of 2 u.
pseudo
B-face-centering
operation,
i.e. orthogonal
to the crystallographic 2, axis. The Patterson function at 5 a resolution, computed with F values from the processed data (i.e. not including the super-intense reflections)
contains a peak at (u,v,w) = (05,@03,0.5) (Fig. 2), which is approximately 20 standard deviations above the noise level of the map or about 10% of the height of the origin. Apart from this peak the map is essentially featureless. (The absence of the
Figure 3. Section K = 180” of the locked rotation function (Tong & Rossmann, 1990). Resolution 190 to 5.5 8. This Figure displays the result of the calculation as a stereographic projection. The directions in space are plotted in spherical polar co-ordinates where the axial tilt of the rotation vector away from the c axis of the crystal co-ordinate system ($) is plotted latitudinally and the rotation within the a*b* plane (4) longitudinally. The 3 crosses indicate 1 choice of a set of 3 orthogonal axes of Z-fold symmetry (see the text). Contours are drawn at increments of 1 0 with the lowest contour drawn at 1.3 0 (see legend to Fig. 2).
Crystallization
intense reflections may perturb the actual peak height but’ should not otherwise materially effect the result.) Tt was of particular interest that the Harker section (V = @5) was featureless, since if a molecular f-fold was oriented parallel to the crystallographic 2, screw axis a peak in the Harker section should be present. The magnitude of the peak at (0.5,@03,0.5) suggests that the contents of the asymmetric unit can be considered as being made up of two pairs of tetramers (physically, this might correspond to two octamers. but it is not required). One pair of tetramers would be related to the second pair by an approximate translation, z = 05, y = @03, z = 05. giving rise to the strong peak in the Patterson function. Within a given pair, the two tetramers would have to have different orientations t)o avoid the presence of additional translation vectors. which would be apparent in the Patterson function. The approximate dimensions of tetrametic /&galactosidase have been estimated from electron microgra,phs (Karlsson et al., 1964) to be This suggests that t.he 120 A x 120 LAx 70 8. tet)ramers must be packing essentially side-by-side to be consistent with the 509 A cell edge and the narrow 108 A cell edge. The self-rotation function has been calculated using several resolution ranges of the data. The different, choices of resolution shells for the calculation all resulted in similar maps. The IC = 180” section, which shows the orientations of potential non-crystallographic 2-fold axes, always contained the most significant features. The K = 90” section, which is expected to show the orientations of potential molecular 4-fold axes, never contained significant peaks. Figure 3 shows the JC= 180” section of the self-rot)ation function, produced using the General Locked Rotation Function program (Tong & Rossmann, 1990). The large peak at the center of the Figure corresponds to the crystallographic 2, screw axis along b. This axis relates the right and left halves. so that only half of the Figure is unique. However. the pseudo mirror symmetry seen in the Figure is not required by space group symmetry. Several potential 2-fold axes are apparent. Since many tetrameric proteins exhibit 222 point symmet,ry (Matthews & Bernhard, 1973), it was of interest t.o examine the spatial relationships between the candidates for orthogonality. One possible choice of three orthogonal vectors might include the peaks at ($,@) = (0”,48”),(0”,138”), with the third lying in the same direction as the crystallographic 2, axis (90”.90”). However, as previously mentioned. a molecular 2-fold axis that is oriented parallel to the crystallographic 2, axis generates a t.ranslation vector that would generate an obvious peak in thr Harker section of the Patterson function. The absence of any such feature rules out the possibility that these three rotation axes could form the 222 point symmetry elements of a tetramer. An alternative set of orthogonal 2-fold rotation axes, shown in Figure 3. is formed by the peaks at (0”,48”),(146”,47”) and (72”. 108”). This suggests the presence of a set of subunits, possibly a single
1181
Notes
tetramer, having 222 point symmetry. Because there is a peak midway between the second and third peaks it is possible that a second tetramer is related to the first by rotation of 180” about an axis that bisects a pair of the 222 tetrameric symmetry axes. In summary, two different crystal forms of bgalactosidase have been obtained. Both have one long unit cell edge but the use of synchrotron radiation in conjunction with large crystal-to-film distances (and large film sizes) makes it feasible to resolve the diffraction spectra and to measure intensities of the monoclinic crystals to Bragg spacings of at least 3.5 W resolution. The asymmetric unit of the monoclinic crystals contains four tetramers that seem to be related, by translation; in pairs. The data are consistent with the postulate that each tetramer has 222 point symmetry. The first. crystals were obtained using protein kindly provided by Drs *Jeffrey Harris and Pyare Khanna of Microgenics Corporation, Concord, CA. Drs Scott Eisenbeis and Mark Krevolin, also of Microgenics. provided the E. coli cells used for subsequent protein purification. We also thank Dr Don Bilderback at CHESS for invaluable assistance configuring the beam line to make the data collection feasible, Drs Irving Zabin and R. E. Huber for helpful discussions and advice on the biochemistry and purification of fl-galactosidase, Dr M. G. Rossmann for advice concerning film processing and R. Albright and Dr S. Roderick for much needed help with the data collection. This work was supported in part by grants from the National Institut’es of Health (GM20066) and the Lucille P. Markey Charitable Trust.
References Appel, S. H., Alpers, D. H. & Tomkins, G. M. (1965). Multiple molecular forms of fl-galactosidase. J. Mol. Biot. 11, 12-22. Cohn, M. (1957). Contributions of studies on the pgalactosidase of Escherichia coli to our understanding of enzyme synthesis. Bacterial. Rev. 21. 140-168. Fowler. A. & Zabin, I. (1970). The amino acid sequence of fi-galactosidase. I. Isolation and composition of tryptic peptides. J. Biol. Chem. 245. 5032-5041. Fowler, A. & Zabin, I. (1978). Amino acid sequence of pgalactosidase. XI. Peptide ordering procedures and Chem. 253. the comp1et.e sequence. d. Biol. 5521-5525. Fowler, A. & Zabin, I. (1983). Purification. structure, and properties of hybrid fi-galartosidase proteins. J. Nol. Biol. 104, 541-555. Huber, R. E. Kurz, G. & Wallenfels. K. (1976). A quantitation of the factors which affect the hydrolase and transgalactosylase activities of fl-galactosidase (E. coli) on lactose. Biochemistry. 15. 1994-2001. Jacob, F. & Monod. tJ. (1961). Genetic regulatory mechanisms in the synthesis of prot,eins. J. Mol. Biol. 3, 318-356. Kalnins, A.: Otto, K., Ruther, U. & Miiller-Hill. B. (1983). Sequence of the 1acZ gene of Ir:scherichia coli. EMBO
J. 2, 593-597.
Karlsson, G., Koorajian, S., Zabin, I., Sjiistrand, F. S. & Miller. A. (1964). High resolution electron microscopy on highly purified fl-galactosidase from Escherichia coli. J. Ultrastruct.
Res. 10, 457-469.
1182
R. H. Jacobson and B. W. Matthews
Lederberg, J. (1950). The /I-D-galactosidase of Escherichia coli strain K-12. J. Bucteriol. 60, 381-399. Matthews, B. W. (1968). Solvent content of protein crystals. J. Mol. Biol. 33, 491497. Matthews, B. W. (1985). Determination of protein molecular weight, hydration, and packing from crystal density. Methods Enzymol. 114, 176-186. Matthews, B. W. & Bernhard, S. A. (1973). Structure and symmetry of oligomeric enzymes. Annu. Rev. Biophys. Bioeng. 2, 257-317. and Analysis of McPherson, A. J. (1982). Preparation Protein Crystals, pp. 82-160, John Wiley and Sons, New York. Ring, M. & Huber, R. E. (1990). Multiple replacements establish the importance of tyrosine 503 in /Igalactosidase (Escherichia co&). Arch. Biochem. Biophys.
283, 342-350.
data for very large unit cells with an automatic convolution technique and profile fitting. ,I. Appl. Crystallogr. 12, 225-239. Schmid, M. F., Weaver, L. H., Holmes, M. A., &utter. M. G., Ohlendorf, D. H., Reynolds. R. A.. Remington, S. J. & Matthews, B. W. (1981). An oscillation data collection system for high-resolution Acta Crystallogr. sect. A, 37. protein crystallography. 701-710. Sinnott, M. L. (1978). Ions, ion-pairs and catalysis by the 1acZ /I-galactosidase of Escherichia coli. FEBR Letters,
Rossmann, M. G. (1979). Processing oscillation diffraction
Edited
94, 1-9.
Tong, L. & Rossmann, M. G. (1990). The locked rotation function. Acta Crystallogr. sect. A, 46, 783-792. Wallenfels, K. & Weil, R. (1972). /I-Galactosidase. In The Enzymes (Boyer, P. D., ed.), 3rd edit.. vol. 7, pp. 617-663, Academic Press, London.
by W. Hendrickson
