J. Mol. Biol. (1991) 219, 123-132

Crystal Structure of a, Barnas+d(GpC) l-9 A Resolution

Complex at

Sylvie Baudet and Jog1 Janin Laboratoire

de Biologic Physicochimique - CNRS U.A. Universite’ Paris-Sud 91405Orsay, France

(Received 19 September 1990; accepted 21 January

1131

1991)

The ribonuclease excreted by Bacillus amyloliquefaciens, Barnase, was co-crystallized with the deoxy-dinucleotide d(GpC). The crystal structure was determined by molecular replacement from a model of free Barnase previously derived by Mauguen et al. Refinement was carried out using data to 1.9 A resolution. The final model, which has a crystallographic R factor of 22%, includes 869 protein atoms, 38 atoms from d(GpC), a sulfate ion and 73 water molecules. Only minor differences from free Barnase are seen in the protein moiety, the root-mean-square C” movement being 045 A. The dinucleotide has a folded conformation. It is located near the active site of the enzyme, but outside the protein molecule and making crystal packing contacts with neighboring molecules. The guanine base is stacked on the imidazole ring of active site HislOB, rather than binding to the socalled recognition loop as it does in other complexes of guanine nucleotides with microbial nucleases. The deoxyguanosine is syn, with the sugar ring in C-2’-endo conformation; the deoxycytidine is anti and C-4’-exo. In addition to the stacking interaction, His102 hydrogen bonds to the free 5’ hydroxyl, which is located near the position where the 3’ phosphate group is found in other inhibitors of microbial ribonucleases. While the mode of binding observed with d(GpC) and Barnase would be non-productive for a dinucleotide substrate, it may define a site for the nucleotide product on the 3’ side of the hydrolyzed bond. Keywords: ribonuclease; X-ray structure;

1. Introduction Barnase is an endonuclease produced and excreted by Bacillus amyloliquefaciens (Nishimura & Nomura, 1958; Hartley, 1989). It belongs to a family of small microbial ribonucleases with similar structures and properties, which hydrolyze phosphodiester bonds on the 3’ side of guanosine nucleotides in RNA (Hill et al., 1983). Microbial ribonucleases are believed to use the same chemical mechanism as pancreatic RNase A: a 2’,3’ cyclic intermediate is formed by transesterification, and then it is hydrolyzed. Barnase has 110 amino acid residues and a molecular weight of 12,382. Its primary structure has been determined (Hartley & Barker, 1972). It is highly homologous to other bacterial ribonucleases, but the homology to fungal enzymes is low. The crystal structure of Barnase has been determined as a complex of three protein molecules with one Zn ion (Mauguen et al., 1982). Its active site has been investigated by site-directed mutagenesis (Mossakowska et al., 1989). On the basis of these results and of a comparison with other microbial ribonucleases (Hill, 1986, 1989), two

protein-nucleic

acid interactions

residues are implicated as general acid-base catalysts: Glu73 for the transesterification step and His102 for hydrolysis. Residues 56 to 62 form a “recognition” loop of polypeptide chain stabilized by a peculiar reverse-turn structure conserved throughout the microbial ribonuclease family. A specific pattern of hydrogen bonds from the base to the loop explains the G specificity. X-ray structures of fungal ribonucleases with bound guanine nucleotides have been available for some time (Heinemann & Saenger, 1982; Sugio et al., 1985; Hakoshima et al., 1988; Koepke et al., 1989), but not for Barnase or any bacterial enzyme homologous to Barnase until Pavlovsky 8: Karpeisky (1989) solved the complex of the ribonuclease from Bacillus &termed&s and 3’-GMP. Therefore, models of Barnase catalysis and specificity were based on indirect structural evidence. To test these models we tried and prepared crystals of the enzyme in the presence of small nucleotide inhibitors. We present here the crystal structure of a complex of Barnase with deoxyguanylyl-3’,5’-cytosine, d(GpC). The model, refined to a R factor of 22% to 1.9 A resolution (1 a = @l nm), shows that the dinucleotide

S. Baudet and J. Janin

124

Table 1 Characteristics

Crystal form

Space group

Barnase-Zn Barnase-d(GpC)

P3, P3,21

590, 81.6 58.0, 854

and Data Collection

cloned from has been The Barnase gene B. amyloliquefaciens into Escherichia coli by Hartley (1988), together with a gene for its small natural protein inhibitor Barstar, into a plasmid downstream from a promoter and signal sequence for the outer membrane protein PhoA. The enzyme was overexpressed and excreted into the supernatant, from which it was isolated following a protocol established by Hartley et al. (1972), and Hartley (1988). Barnase activity was tested on 20 mM-GpA at pH 5.0 using the coupled assay with adenosine deaminase (Ipata & Felicioli, 1968). All dinucleotides were from Pharmacia. Crystals of Barnase were grown in the presence of ZnSO, according to a procedure of Dr R. Hartley (Bethesda, U.S.A.). These crystals, which are isomorphous to those used by Mauguen et aZ. (1982), did not bind small nucleotides when added by infiltration, presumably because crystal packing blocks the active site. We found that the deoxy-dinucleotides d(GpA) and d(GpC), at concentrations of 75 mM, inhibit hydrolysis of GpA. Hence. we tried co-crystallization and obtained crystals by vapor diffusion in hanging drops containing @7 mm-Barnase and 3.5 mM-d(GpC) in 1.5 M-ammonium sulfate over pits containing 3.0 M-ammonium sulfate, at pH 8 and 20°C. The procedure produced no crystals when d(GpC) was omitted or when ZnSOL was added. The crystals were hexagonal sticks and grew in size up to @3 mm x @3 mm x 1 mm. The space group was identified as either P3121 or P3,21 on precession photographs, the true space group found later to be P3,21. The cell dimensions are compatible with either 1 or 2 molecules

Data collection Measurements Total Independent

Data set

(4

I. 2. 3. 4.

Diffractometer Film FAST FAST

49 2.5 >2 1.9

R,,, = E[I,. i - (I,,)]pI,,,

i, where I,. i are individual

7728 12,766 35,224 26,792

829 6200 6692 12,204

& (‘$&) 91 6.9 8.1 5.8

values and

(Zh is the mean value of the intensity of reflection h. The wavelength was 140 A for the film data set 2 collected with synchrotron

radiation

and 1.54 A otherwise.

90, 120 90, 120

9 6

2.2 3.3

44 51

per asymmetric unit. Density measurements in a Ficoll gradient (Westbrook. 1982) indicated t)hat it contains only one. Table 1 details the chararteristics of the 2 crystal forms of Barnase. It can be seen that they have different symmetries even though both forms are trigonal and have similar cell parameters. Free Barnase crystals have 3 protein molecules (and 1 Zn ion) per asymmetric unit, crystals grown with d(GpC) have only 1. Their packings are unrelated. X-ray diffraction data were first taken at low resolution on a 4-circle diffractometer to test the quality of thta crystals. High-resolution data were then collected on films using a Arndt-Wonacott rotation camera at the synchrotron radiation source LURE-DC1 in Orsay. The wavelength was set to 1.4 A. A complete data set was collected to 2d A resolution with 1 single crystal. The rotation angle was 1.5” per exposure, with a total angular range of 30”. Films were digitized on a Optronix densitometer PlOOO with a 50 pm step and diffracted int.ensities were integrated with the program DEPU‘ZO (Dr Z. Otwinowski. Chicago, U.S.A.). Further data processing used t,he Program Suit for Protein Crystallography distributed by CCP4, Daresbury, England. Other high-resolution data sets were collected on an ENRAF-Nonius FAST area, detector at EMBL, Heidelberg, F.R.G. Intensities were recorded over an angular range of 120”. first with 2 crystals to 2.2 A resolution and. later. to l-9 A resolut,ion on 2 orientations of a single crystal. These data were processed with the Groningen package. Details of the collected data sets are given in Table 2. The film data set 2 and the FAST data sets 3 and 4 were of good quality on the criterion of internal agreement. merging R factors being 6 to So/b. Yet. they agreed poorly. with r.m.s.t discrepancies of 13 to 16% between pairs. In addition, the cell parameters were found to change by 0.1 to @3 A from one crystal batch to another. Parameters quoted in Table 1 are from data set 4 for crystals soaked in 2 m-ammonium sulfate. For these reasons, the highresolution data set were not merged and we used them at separate stages of structure determination and refinement.

3. Crystal Structure Determination

Table 2 Resolution

crystals

Cell parameters m= 8, Y a = b, c (A) (deg.)

binds in a non-productive mode at a surface site implicated in crystal contacts. We rationalize this result and propose that the deoxyguanosine moiety of d(GpC) occupies the site for the nucleoside 3’ to the hydrolyzed bond, rather than the expected G-specific site in the 5’ position.

2. Crystallization

of Barnase

As crystals of Barnase-d(GpC) are not isomorphous to those of the free Barnase studied by Mauguen et al. (1982). we used the molecular replacement method to solve thei structure. Atomic co-ordinates were kindly provided by Dr Mauguen (Chatenay-Malabry, France). The Barnase model contains 3 molecules of Barnase labeled A, B and (’ with very similar conformations. We used Barnase molecule C as a starting point for locating the protein in the d(GpC) complex. Its orientation was determined with t,hr fast rotation function described by Crowther (1972) t Abbreviations used: r.m.s.. root-mean-square;, fast-Fourier transform.

FFT.

Crystal Structure of Barnase-d(GpC)

Cmplex

125

(b)

Figure 1. Molecular replacement. (a) Section fl = 30” of the cross-rotation function map calculated using reflections of data set 3 between 10 and 3 A. with R = 4 to 18 A in the Patterson shell. Contours every 1.3 cr from 1.3 u (dotted line). The peak at a = 20”, y = 130” is 5 Q high. (b) Section z = 911 of the translation function map calculated using the same reflections. Contours every I a from 1 u. The peak at z = 073. y = 0.24 is 13 Q high.

implemented in program ALMX of the CCP4 suite. We computed a cross-rotation function between our data and structure factors calculated for molecule C placed in a Pl orthogonal cell of dimensions (50 A)3. We tested several resolution shells and radial limits for the Patterson map. The highest peak gave an approximately correct solution. We then made a translation search with the T2 translation function described by Crowther & Blow (1967) implemented by Dr I. Tickle (London, U.K.) in the program TFSGEN. The correctly oriented model was placed in an arbitrary position of the trigonal unit cell. Structure factors for the FFT evaluation of the T2 function were computed using reflections between 10 and 3 A. Both enantiomeric space groups P3121 and P3,21 were tested. A peak 13~ high was obtained in P3,21 (Fig. 1). Translation parameters were derived from this peak. In P3,21. no value higher than 60 was observed. The initial model derived from molecular replacement had a R factor of 45% at 3 A resolution, which dropped to 36% after refinement of the positional and orientional parameters with the constrained least-squares refinement program CORELS (Sussmann et al., 1977), Barnase being treated as a single block. This rather low value of R, obtained before any atomic refinement, confirmed that the protein molecule undergoes no major conformation changes in the complex. After 24 cycles of CORELS, the molecule had moved by 97 A from the position determined by molecular replacement and rotated by 22”. This position and orientation of Barnase molecule C were then used for crystallographic restrained refinement. We first carried out structure refinement against data set 3 including 6589 reflections in the range 10 to 2.24 A. Early steps used the molecular dynamics technique imple-

mented by Fujinaga et aE. (1989) in GROMOS (Van Gunsteren, 1989) followed with restrained least-squares refinement using an FFT version of PROLSQ (Hendrikson, 1985). At various stages, electron density maps with 2F,, - Fcalscoefficients were calculated and displayed on the Evans t Sutherland PS390 color graphics system using PROD0 (Jones. 1985). Difference Fourier maps with Fobs- Fcalccoefficients were used to locate solvent molecules and a few ill-defined side-chains. These were excluded from the calculation of Fobsto reduce the bias towards previously determined atomic positions. At the stage where R reached 20% against all reflections of data set 3, data set 4 became available and further refinement was carried out with PROLSQ against these new data. This introduced small changes in the structure, the r.m.s. movement of protein and d(GpC) atoms being 949 A. Yet, a number of solvent molecules had to be removed or added. The final model has an R factor of 22.4% against 9679 (80% of expected, 92 yc of recorded) reflections between 5.5 and 1.9 A, and good stereochemistry. Parameters from the last step of leastsquares refinement are quoted in Table 3. The dependence of R on resolution (Fig. 2) is compatible with a r.m.s. error in atomic positions of about 62 A. Another estimate of the error was obtained by removing the restraints in least-squares refinement. The r.m.s. atomic shift after 12 cycles was 0.17 A. Obvious electron density was found for d(GpC) in maps calculated at early stages of refinement. It was located at the entrance of the active site, but outside the protein rather than within the active site and recognition loop as could have been expected. An ideal model of the dinucleotide was cut from B-DNA in the Brookhaven Protein

126

S. Baudet and J. Janin

Table 3 Refinement statistics Resolution (A) Reflections? Atomsf Solvent$ R factor Average B factor (AZ) Departure from standard stereochemistry (A) Bond lengths (I-2 distances) Rand angles (1-3 distances) Dihedral angles (14 distances) Planarity Chiral volumes (A’) Single-torsion contacts Multiple-torsion contacts

I

I

0-l

0.2

C

2 (x-‘)

Figure 2. R factor of the refined structure. R factors are from the last cycle of least-squares refinement with PROLSQ against data set 4.

Data Bank (file 1BKA: Dickerson & Drew. 1981: Bernstein et al., 1977). All 38 atoms of d(GpC) could be positioned unambiguously in density simply by rotating about single bonds. Fig. 3 shows portions of the 2Fobs - Fca,c electron density map around the dinucleotide. Full occupancy was asssumed in refinement and confirmed by the lower than average B factor (22 8’) of nucleotide atoms. Solvent molecules were located in difference maps and included when they made sensible interactions with protein atoms; 73 water molecules were thus located, their position and temperature factor refined. A large peak located near the guanine moiety of d(GpC) was attributed to a sulfate ion, although the sulfur and oxygen atoms were not resolved in the density. As refinement assuming full occupancy led to a high temperature factor, the sulfate occupancy was fixed to 07.5. The occupancy factor of water molecules was assumed to be 1 and was not refined.

551.9 9679 985 78 0,224 396

r.m.s. 0020 0.059 0.06 1 CO14 0221 0.222 0.308

Input (CT) 6020 0.050 0050 0.020 Wl56 0500 (k500

Statistics from PROLSQ output. t Reflections with F larger than 3a, or 924,) of 10,578 observed reflections in data set 4 in the resolution range 5.5 to 1.9 A. $ 868 protein atoms, 38 dinucleotide atoms. 0 Include 5 atoms from a sulfate ion.

The final electron density remained poor at 2 locations. One was at the N terminal which is disordered in this crystal form as it is in free Barnase. Ala1 and the sidechain of Gln2 could not be built with confidence and were omitted from the final model. The other disordered region was the recognition loop (residues 56 to 62), which is discussed below. Fig. 4 shows that temperature factors in these regions refine to values higher than 40 A2, while the average value for C” atoms is 29 A’. The K’ terminus and C-terminal residues 108 to 110 also have high temperature factors.

4. The Structure of Barnase The structure of Barnase in the d(CrpC) complex is essentially the same as that of the free enzyme described by Mauguen et a2. (1982) and Hill (1986). Superposition of the main-chain onto that of mole-

cule C of the free Barnase crystal yields an r.m.s.

Figure 3. Electron density for d(GpC). Part of the refined model is superimposed on a 2Fot,, - Fcalc electron density map contoured at 2 o and drawn with FRODO (Jones, 1985). The phosphorus atom of d(GpC) and the sulfur atom of t,he sulfate ion are labeled.

Crystal Structure of BarnaseA(GpC)

20

40

60

Residue

number

80

100

Figure 4. Temperature factors and main-chain moveRefined B factors (0) of C” atoms in the complex and C” displacements from positions observed in free Rarnase molecule C (continuous line) are plotted for residues 3 to 110. ments.

distance between all c” atoms of @49 A; with molecule A, it is 045 A; with molecule B, @51 A. These discrepancies are slightly larger than between pairs of molecules A,B,C of the free Barnase crystals (about 033 A in the co-ordinate set we used). C” discrepancies larger than 0.8 A are seen at the N and C termini, at residues 5%60 in the recognition loop and at residues 67-68 (Fig. 4). The fit is also good for side-chains. Excluding those of residues 2-3 and 59-60, which have high temperature factors, 841 protein atoms can be superimposed to within 09 A r.m.s. While differences observed at the N terminus and in the recognition loop are probably not significant, those seen around residue 67 can be attributed to the molecular packing. As we shall see, Lys66 and Ser67 interact with the nucleotide of a neighboring mole-

Complex

127

cule in the crystalline complex, and their side-chains change conformation compared to the free enzyme. Conversely, His18 of molecule C is engaged in intermolecular contacts in the free Barnase crystal where it is co-ordinated to the Zn ion. His18 is free in the complex, and its x1 angle changes from -45” to a more favorable - 68”. Free and liganded Barnase have the same secondary structure. There are two a-helices 6 to 18 and 26 to 34 and a five-stranded /?-sheet. Hydrophobic residues involved in helix-sheet packing are important for stability (Kellis et al., 1988) and are not perturbed by dinucleotide binding. The first residues or “N-caps” of each helix (Richardson & Richardson, 1988) are threonine residues. Site-directed mutagenesis has shown that they stabilize the protein by forming a hydrogen bond with peptide groups in the first helix turn (Serrano & Fersht, 1989). The side-chain conformation of Thr6 and Thr26 is the same in the complex as in free Barnase. Their Oy oxygen atoms can accept a hydrogen bond from the NH of residues 9 and 29. The distance from Thr6 Oy to NH Gly9 is 3.2 b in the d(GpC) complex, that from Thr26 Oy to NH Glu29 is 3.3 8. These distances indicate weak hydrogen bonds; those to the NH of other residues in the helices are all too large for hydrogen bonding. Helix 6 to 18 of Barnase also has a C-cap, which stabilizes the last helix turn by hydrogen bonding to a main-chain carbonyl (Sali et al., 1988); it is Hisl8, the N” atom of which is within bonding distance of 0 Gln15. In the complex, N” cannot make the bond, but the Nd atom can with a distance of 3.3 A.

5. The Dinucleotide The d(GpC) dinucleotide is located at the mouth of a cleft forming the active site of Barnase (Fig. 5), rather than inside the cleft as we expected of a

Figure 5. Comparison of free Barnase and of the d(GpC) (h eavy line) complex after least-squares superposition (McLachan. 1979). The arrow points to the recognition loop (residues 56 to 62) which, in other microbial nuclease crystals structures. binds the G base. The difference between the 2 Barnase structures is largest, at residue 59.

S. Baudet and J. Janin

128

Table 4 Conformation of d(GpC) in complex Dihedral angle

Bond

-P-05’-o-5’-c-5’. .C-5’4.4’. .c4q-3’. J--g-o-3’. -0.3’-p-

i

Sugar puckering P Glycosidic bond x

-C-l’-N-

G

c -45 130 52 73

160 145 -85 -85 C-2’-end0 155

C-4’.exe 55

90 syn

- 107 anli

For the definition of the dihedral angles and of the pseudorotation phase angle P governing sugar puckering, see Saenger (1984), pp. 17-19.

substrate analog. It is also at a molecular interface: d(GpC) at the active site of Barnase molecule # 1 makes contacts with two other molecules related by screw 3-fold and 2-fold symmetries and labeled #2 and #3. The nucleotide has a closed quadrangular shape, with the bases on two adjacent sides and the phosphate group on the opposite corner (Fig. 5). A contact between the N-3 imino group of C and the N-2 amino group of G closes the quadrangle. The

distance is 3-O ,k, which implies a hydrogen bond even though the angular geometry is poor. the planes of the two bases being almost orthogonal. The conformational parameters of d(GpC) are summarized in Table 4, its interactions in Table 5. The main interaction with Barnase molecule # 1 is the stacking of the guanine base on the imidazole ring of HislOB. The planes of the two aromatic groups are parallel and their distance is 3.5 A. The G base is sandwiched between the imidazole and alipathic part of a well-defined lysine side-chain from neighboring molecule # 2, Lys66 # 2 (Fig. 6). The side-chain lies flat on the base and is in direct contact with it. In addition to these non-polar contacts, polar groups of the G base a.re extensively involved in hydrogen bonds. The N-2 amino group donates one to the cytosine base as already noted. and another to a sulfate ion. The sulfate ion probably also hydrogen bonds t,o the N-l imino group. though the distance is long. N-7 accepts a bond from the side-chain of Ser85. The 0-6 carbonyl group hydrogen bonds across the molecular interface to a main-chain NH of molecule #2. and also to a water molecule. The glycosyl bond of t,he deoxyguanosine is syn, and the sugar puckering is clearly C-2’-endo. The free 5’ hydroxyl does not interact with the base, as often observed with sy~ nucleosides (Saenger, 1984). Rather, it> points away

Table 5 Interactions of d (GpC) WW) A. Guanosine Polar N-l G N-2 G O-6 G N-7 G 0-‘5 G

O-4’ G Non-polar G base G base U. Phosphate O-l P o-2 P c.

Distance (A)

(3.6) 34 30 25 2.9 26 2.6 3.3 2.5 31

Wp’J

Barnase

Solvent

O-4 sulfate 0-l sulfate N-3