Crystal structure of a deletion mutant of a tyrosyl-tRNA synthetase complexed with tyrosine

J. Mol. Bid. (1987) 194. 287-297

Crystal Structure of a Deletion Mutant of a Tyrosyl-tRNA Synthetase Complexed with Tyrosine Peter Brick and David M. Blow? Blackett Laboratory, Imperial College London SW7 2BZ, England (Received 18 September 1986) The crystal structure of a deletion mutant of tyrosyl-tRNA synthetase from Bacillus stearothermophilus has been determined at 2.5 L%resolution using molecular replacement techniques. The genetically engineered molecule catalyses the activation of tyrosine with kinetic properties similar to those of the wild-type enzyme but no longer binds tRNATy’. It contains 319 residues corresponding to the region of the polypeptide chain for which interpretable electron density is present in crystals of the wild-type enzyme. The partly refined model of the wild-type enzyme was used as a starting point in determining the structure of the truncated mutant. The new crystals are of space group P2, and contain the molecular dimer within the asymmetric unit. The refined model has a crystallographic R-factor of 18.746 for all reflections between 8 and 2.5 A. Each subunit contains two structural domains: the a//I domain (residues 1 to 220) containing a six-stranded P-sheet and the a-helical domain (residues 248 to 319) containing five helices. The a//I domains are related bv a non-crystallographic dyad while the a-helical domains are in slightly different oiientations in the two subunits. The tyrosine substrate binds in a slot at the bottom of a deep active site cleft in the middle of the a//II domain. It is surrounded by polar side-chains and water molecules that are involved in an intricate hydrogen bonding network. Both the a-amino and hydroxyl groups of the substrate make good hydrogen bonds with the protein. The amino group forms hydrogen bonds with Tyrl69-OH, Asp78-ODl and Gln173-OEl. The phenolic hydroxyl group forms hydrogen bonds with Asp76-ODl and Tyr34-OH. In contrast, the substrate cacboxyi group makes no direct interactions with the enzyme. The results of both substrate inhibition studies and site-directed mutagenesis experiments have been examined in the light of the refined structure.

1. Introduction

refined (Bhat & Blow, 1983) but the model included no solvent molecules, and it had not been possible to construct a molecular model for 100 amino acid residues at the C terminus of the molecule that are disordered in the crystals. This made the refinement more difficult, but, by calculating structure factors for the uninterpreted electron density, and adding these to the structure factors calculated from the atomic model, it had been possible to lower the crystallographic R-factor to 32% using all data to 2.1 A. The R-factor increased to 34% when the additional density was omitted from the calculation. At this level of agreement between the observed and calculated structure factors, the conformation of many of the side-chains and of the more poorly ordered parts of the peptide chain is uncertain. By taking advantage of t,wo conveniently situated Hinf restriction sites at codons 317 and 417, Waye and co-workers (Waye et al., 1983) removed the DNA coding for the disordered

Tyrosyl-tRNA synthetase from Bacillus stearothermophilus is a dimeric enzyme containing 419 amino acid residues in each subunit, which catalyses the aminoacylation of tRNATY’ in a twostep reaction. The enzyme-bound tyrosyl-adenylate complex is first formed from the amino acid and ATP. The aminoacyl group is then transferred to a molecule of tRNATY’ to form tyrosyl-tRNATy’. The enzyme has been cloned and sequenced by Winter et al. (1983) and is currently being used as a test system to study the effect of mutating specific amino acid side-chains on catalytic activity (Fersht et al., 1984). The choice of mutation sites was based initially on the three-dimensional crystal structure of the enzyme reported by Bhat et al. (1982). This structure had subsequently been partly 7 Author addressed

to whom

correspondence

should

be

287

0 1%37 Academic Press [nc. (London)

Ltd.

P. Brick and D. &I. Blow

288

C-terminal region from the structural gene and expressed the deletion mutant in Escherichia coli. The truncated enzyme, corresponding to residues 1 to 317, 418, 419 of the wild-type enzyme, catalyses the formation of tyrosyl adenylate with unchanged k,,, and K, values, but no longer binds tRNATy’ or transfers tyrosine to tRNATy’. This paper describes the crystallization and structure determination of the truncated enzyme complexed with the tyrosine substrate. The improved molecular model has been used in the analysis of inhibition studies and site-directed mutagenesis experiments.

2. Experimental Procedures (a) Crystalliz4xtion The protein was prepared by infecting an E. coli host with recombinant Ml3 phage carrying the gene for the truncated enzyme and then purified as described (Waye et al., 1983). Two crystal forms of the truncated enzyme were obtained in the presence of tyrosine using the vapour diffusion technique (see Table 1). The first. form (type I)

was grown using 2 M-ammonium sulphate as a precipitant at pH 7.0. These crystals diffracted well but invariably grew as aggregates and could not be used for high resolution data collection. The second form (type II) was obtained by allowing drops containing 15 mg protein/ml, 1 mw-tyrosine. 10 mM-MgCl, 50 mM-Tris-acetate buffer (pH 7.5) and 6% (w/v) polyethylene glycol 4000 to equilibrate at 18°C against a reservoir solution containing 12% polyethylene glycol. The flat lozenge-shaped crystals could be grown to over 1 mm in length in 2 or 3 days and were suitable for high resolution data collection. These crystals contain the dimeric molecule within the crystallographic asymmetric unit. Table 1 includes the cell parameters of the crystal form of the wild-type enzyme obtained using ammonium sulphate as a precipitant that was used to determine the original wildtype structure (Irwin et al., 1976). (b) Data collection, and processing

oscillation camera (Enraf-Nonius, Delft) and graphitemonochromatized CuKa radiation from an Elliot GX20 rotating anode generator. The crystal-to-film distance was 80 mm. The crystals were cooled t,o +5”C using a wide stream of dry air. The diffraction patterns were recorded on flat cassettes containing 3 sheets of Kodak DEF-59T film. A helium-filled cone placed between the crystal and the film cassette reduced air-scatter of the direct X-ray beam. Oscillation photographs were digitized on a Joyce-Loebl Scandig 3 rotating-drum microdensitometer controlled by a Data General Nova 3/12 computer. All scanning was performed using a 50 pm raster step size and an optical density range of 0.0 t,o 2.0 units. The digitized data were processed using a suite of programs originally described by Nyborg & Wonacot,t (1977) and subsequently extensively rewritten at Imperial College. Film-pack sca,ling was performed using the Fox & Holmes (1966) algorithm. Details of the data collection and processing are given in Table 2. (c) Rotation functio?,. The fast rotation function (Crowther. 1972) was used to determine the orientation of the truncated molecule in the unit cell. Since the crystals contain a dimer within the crystallographic asymmetric unit, a self-rotation function was calculated to determine the orientation of the noncrystallographic a-fold axis using data between IO and 4 A and radial limits of 10 to 22 A. Fig. 1 shows the results plotted as a stereographic projection. A cross-rotation function was calculated using a search model based on the partly refined wild-type enzyme dimer. Residues in the wild-type model with average atomic temperature factors above 25 8’ were omitted as the position of many of these residues was uncertain. Atomic co-ordinates for a perfect dimer were generated from the remaining 270 residues. Triclinic structure factors were calculated for the dimer placed in a large orthogonal cell of dimensions 64 A x 140 a x 84 -4 such that no intermolecular vect,ors shorter than 23 a occurred. The cross-rotation function was calculated as a function of Crowther’s Eulerian angles on a 5’ grid using data between 6.0 and 4.5 d and radial limits in the Patterson map of 8 and 20 a. The section of the croxsrotation function that cont,ains the peak corresponding to the correct solution is shown in Fig. 2.

X-ray data to 2.5 A were collected from 7 crystals of the bruncated enzyme using an Arndt-Wonacott

Table 2 Details of th,e data collectio?l

Table 1 Cell parameters of crystals obtained for the truncated and wild-type enzymes Truncated Precipitant

2 M-(NH,),SO,

space group n (4 b (& c (4 Alpha (deg.) Beta (deg.) Gamma (deg.) Contents of the asymmetric unit F V,.,,(A3/dalton) t PEG, polyethylene

I

Truncated II Wild-type 12% (w/v) 2 M-(NH,),So, PEG?

c2 107.4 64.5 71-9 90.0 129.67 90.0

p2, 95.5 67.1 61.4 90.0 90.78 90.0

P3,21 64.5 64.5 238.1 90.0 90.0 120.0

Monomer 2.72

Dimer 2.13

Monomer 3.04

glycol.

Crystal-to-film distance Exposure time

80 mm 1.5 x IO4 sidrg.

Temperature

+54”(’

Nominal resolution Oscillation angle Xumber of crystals Number of film packs Number of measurements Number of independent reflections Merging R-factort Overall From 2.63 to 2.5 A Percentage of independent reflect,ions in data set From 10 to 2.5 x From 2.66 t,o 2.5 A

2.5 A 24 deg. 7 60 64.683

“6.176 ,5.X0,,

l6.7’! 0 94-30 93.7 (;

t The merging R-factor is defined as X(1(i)-(/))jXr(i). whew I(i) are the intensitv values of the individual measurements. and (I> the correspond& mean values.

289

Structure of Tyrosyl-tRNA Synthetase TR-Tyr-RS

*

Cross Rob

Beto = 35

c

Figure 1. A stereographic projection down the y axis of the truncated tyrosyl-tRNA synthetase self-rotation function. The rotation function was calculated using reflection data between 10 and 4 A. The Patterson shell was from 10 to 20 A. The crystallographic dyad peak height at the centre was first scaled to 100 and then the map was contoured in steps of 6 with the first level at 14.

(d) R-f&or

search

An R-factor search procedure (Nixon & North, 1976; Derewenda et al., 1981) was used to position the molecule within the unit cell. The correctly oriented search object (a molecular dimer) was placed in an arbitrary position in the unit cell and partial structure factors computed for a single search object in space group Pl using reflections between 5 and 4 A. The structure factors for a symmetryrelated search object were generated by application of the appropriate symmetry operator. For each grid point on the search lattice these partial structure factors were added using appropriate relative phase angles. For space group P2, the y co-ordinate of the molecular origin is arbitrary and there are 4 symmetry equivalent origin positions, so it was only necessary to search over one quarter of the y =0 plane. The lowest 3 values of the R-factor obtained using a 1.2 A grid were 45.1%, 48.0% and 48.1% compared with a mean value of 49.2%. The search was repeated around the minimum value using a O-2 A grid and gave a best R-factor of 43.1%.

(e) Crystallographic

rejinement and model building

The values of the positional and orientational parameters obtained from the molecular replacement technique were improved by using the constrained leastsquares reciprocal space refinement program CORELS (Sussman et aZ., 1977). The molecule was treated initially as a single rigid body. Three cycles of refinement using reflections between 10.0 and 4.0 A reduced the R-factor from 45.2% to 43.3%. The molecule was then broken into 4 separa.te rigid groups by dividing each monomer at the doma.in boundary so that the first rigid group contained residues 2 to 228, while the second rigid group contained residues 238 to 319. (Residues 229 to 237 were poorly ordered in the starting structure and have been omitted.) Three further cycles of refinement using the same resolution limits lowered the R-factor to 40.4%. The molecular model obtained from the rigid body refinement was used as a starting model for restrained least-squares structure factor refinement using the

Stole Alpho

= horizontal

Gommo vertical

Section Bet0 sections

Figure 2. Section of the cross-rotation function map at /I = 35” containing the peak corresponding to the correct solution. The rotation function was calculated using reflection data between 6 and 4.5 A. The Patterson shell was between 8 and 20 A. The map has been contoured at intervals of the root-mean-square (r.m.s.) level of the whole map with the first level omitted. The peak has a value of 8.8 times the r.m.s. level.

program PROLSQ (Hendrickson & Konnert, 1980). As originally coded, the program used a conventional structure factor calculation procedure to obtain both the structure factors and their derivatives. For large molecules this method requires a large amount of central processor time (Honzatko et aE., 1985) and is not practicable on currently available mini-computers without the addition of an array processor. In order to reduce the computation, a space-group-specific fast Fourier procedure was employed based on the algorithm proposed by Agarwal (1978) with modifications described by Isaacs (1982). The structure factor contributions to the normal equations were calculated as a separate step in the refinement cycle and read in by the program PROLSQ. One complete cycle of refinement on a VAX-11/750 computer (Digital Equipment Corporation) of both the atomic positions and their temperature factors at 2.5 A resolution required 45 min of central processor time.

i? Brick and D. M. Blow When the refinement had converged the model coordinates were used to generate electron density maps, and model building was carried out, on an Evans and Sutherland PS300 interactive graphics system using the program FRODO (Jones, 1978). Electron density maps were calculated using Fourier coefficients (3F,,, - 2B,,,,), ~~~~~ and difference maps using (Fobs- Fcalc), clCalC. For the first 4 rounds of model building the electron density for the 2 subunits was averaged (Bricogne, 1976) and the model for only one monomer was rebuilt. The electron densities for the 2 structural domains of the monomer were averaged separately as the relative orientations of the domains were significantly different in the 2 monomers. The rotation and translation parameters required to minimize the deviation between equivalent a-carbon atoms were determined using a least-squares procedure (Hendrickson, 1979). After model building the co-ordinates of the complete molecule were regenerated using the same rotation and translation parameters. The very weak electron density for both the loop between residues 110 and 115 and for the connecting regions between the two domains (220 and 240) did not superimpose well. These regions were therefore omitted from the model at this stage in t,he refinement. Noncrystallographic symmetry restraints were not applied at any stage during the course of the refinement. Initially only data to 3.0 A were included in the refinement and the atomic temperature factors were not refined. The first round of model building was undertaken when the R-factor had been reduced to 35.1%. Density for the substrate tyrosine was clearly visible at this stage in the difference map and a model for this amino acid was built into the density and included in further refinement. The resolution was then extended to 2.5 A and refinement of atomic temperature factors introduced. This lowered the R-factor to 28%, at which point the molecular model was again examined using the computer graphics system. Water molecules were first introduced into the atomic model during the third round of model building when the R-factor was 24.9%. Well-defined peaks in difference maps were identified as water molecules if they were within 3.5 A of either a polar group on the protein or another water molecule. Water molecules that refined to temperature factors were eliminated. The high occupancies of the water molecules were set to unity and not refined. Where difference density associated with the protein could not easily be interpreted, the neighbouring protein residues were eliminated from the atomic model. A satisfactory model could then usually be built into the density in the following round of model building. The final model was obtained after 5 rounds of model building and refinement. It was not possible to position residues Glu212 and Ala213 in either subunit and in the B subunit there was no density above the noise level for the final 6 residues. In addition, there are a total of 20 hydrophilic side-chains for which in neither subunit is there sufficient electron density.

3. Results and Discussion (a) Overall structure has resulted in a structure with good geometry and a low crystallographic R-factor. A summary of the deviations of the structure from ideal geometry is given in Table 3. At this resolution details of the molecular model are very dependent on the The

least-squares

refinement

procedure

geometric restraints used in the refinement. The final model contains 4830 protein atoms and 143 solvent atoms. The overall R-factor is 18.70/, for the 25,459 reflections between 8 and 2.5 8. If the solvent molecules are omitted from the structure factor calculation, the R-factor increases to 21 .S%. The general folding of the polypeptide chain is similar to that previously reported for the wild-type structure (Bhat & Blow, 1983; Blow & Brick, 1985). The molecular model comprises the two crystallographically independent subunits (denoted as A and B) of the molecular dimer. Figure 3 provides a schematic view of the structure of the A subunit. Most, of the major changes to the model that have occurred during the course of the refinement procedure have been in stretches of the chain that have relatively high atomic temperature factors. All non-glycyl residues have main-chain dihedral angles that lie within or very near the acceptable regions of a &Ic/ plot (Ramakrishnan & Ramachandran, 1965). Of the non-glycyl residues only Arg292 has a positive value of (b. At this resolution (2.5 il) it’ is sometimes difficult, to determine the side-chain torsion angles. especially for the less well-ordered sections of t,he molecule. In particular when examining the electron density map there is often an ambiguity in x1 of 180” for valine and threonine side-chains, and a similar ambiguity in xz for leucine side-chains. In

such cases the side-chain conformation with the energetically more favourable torsion angles has been selected. The refined temperature factors for the t,wo subunits

are presented

in Figure

4. The temperat,ure

factors exceeding about 40 A2 represent regions of in the the structure for which precise interpretation form of at,omic co-ordinates becomes impossible, and there is no prospect of improvement, by

Table 3 Deviations from ideal geometry at th,e end of rejinement Input rs I‘.~.Y. deviat8ion Distanrr
Bonds (l-2 neighbour) Angles (l-3 neighbour)

0.01 0.03

Intrapianar (14 neighbour)

0.05

0~044 0.057

Planar groups (A) Chiral centres (A3) Torsion angles (deg.) Staggered (e.g. aliphatir 1,) Transverse (e.g, aromatic x2)

032 0.15

0.016 0.17

15 46

30.1

0.5 0.5

0.30

1.50

3.20

1.25 I.50 5.25

4.80 3~87

Contacts

(A)

Single torsion Multiple torsion Thermal

17.9

factors

0.20

(A’)

Main-chain bond (l--2 neighbour) Main-chain angle (l-3 neighbour) Side-chain bond Side-chain angle

5.79

The values of (r given in the Table are those input, t,o t,he refinement program and determine the relative weights of the corresponding geometric restraints (Hendickson & Konnert, 1980). r.m.s.. root-mean-square.

Structure of Tyrosyl-tRNA

291

Synthetase

,

Figure 3. Stereo view showing the u-carbon positions of the A subunit of the truncated tyrosyl-tRNA synthetase model. The substrate tyrosine has been drawn using a ball and stick representation. The molecular dyad axis is shown as a broken line on the left. The unplaced residues Glu212 and Ala213 have been omitted.

extending the resolution. There is generally close correspondence between the temperature factors of equivalent residues in the two subunits. In both subunits, the highest degree of order is in the parallel &strands of the a//? domain (residues 1 to 220), and in adjacent parts of the structure. The a-helical domain (residues 248 to 319) and the chain linking the two domains (residues 221 to 247) are less well-ordered. The most significant difference

between the two E/P domains residue 112 that is involved contacts (see below).

is in a loop near in intermolecular

(b) Molecular symmetry The two subunits of the truncated enzyme form a dimer that closely resembles the wild-type dimer. The dimer is formed by the association of a

80 70 60 50 8

40 30 20 IO

O’~““““““““““““““lII 50

100

150 Residue number (a)

200

250

300

60

Figure 4. Average main-chain atomic temperature factors plotted against residue number for the 2 crystallographically independent subunits of the dimeric molecule. The unplaced residues GM?12 and Ala213 have been omitted.

P. Brick and D. M. Blow

Residue number (0)

Residue number (b)

Figure 5. The r.m.s. separation between equivalent main-chain atoms in the A and B subunits of the truncated enzyme after superposition as a function of residue number. In (a) the a/B domains (residues 1 to 220) of the A and B subunits have been optimally superposed while in (b) residues 248 to 313 of the a-helical domains have been superposed.

to A

chains (Arg137 and Lysl41) involved in these interactions are partly buried when the dimer is formed. A salt-bridge is formed between Lysl41 in one subunit and Glu85 in the other. Mutant enzymes containing either of the single amino-acid substitutions Arg137 to Gln137 or Lys141 to Asn141 are able to complement a mutant strain of B. stearothermophilus containing a thermosensitive mutation in the genomic tyrS gene (Bedouelle & Winter, 1986). As the dimer is required for the amino-acylation reaction (Bedouelle & Winter, 1986; Jones et al., 1985) this suggests that these substitutions do not disrupt the association of the two subunits to form a dimer. The two subunits can be compared in detail by applying a transformation that gives the best fit between the equivalent a-carbon atoms, and studying the residual co-ordinate differences between them. The transformation t.o be applied t)o the B subunit co-ordinates xs, which gives the best fit to residues 1 to 220 of subunit A is:

3.1 2.8 2.9

x=

hydrophobic patch on each subunit surface which is accompanied by the loss of accessible surface area (Lee & Richards, 1971) of 1520 A2 on each subunit, of which 1030 A2 corresponds to hydrophobic surface. The greatest contribution is made by Phe164, which loses 142 A2 of solvent-accessible hydrophobic surface. Jones et al. (1985) used sitedirected mutagenesis to change Phe164 to Asp164, and have shown that the mutant enzyme readily dissociates at high pH. At the periphery of this hydrophobic patch there are a number of surface polar groups, which in the dimer form polar interactions between the subunits. These polar interactions are listed in Table 4. Two charged sideTable 4 Hydrogen bonds between subunits of the dimer (distances in A) A. Hydrogen bonds between main-chain atoms A to B

Va1145-N Ile162-0 Phe143-0 Phe164N Va1132-N Leul30-0 B. Hydrogen bonds involving side-chains

3.2 2.8 3.0

Thr165-OGl Glu85-OE 1 Arg86-0 Arg137-NH2 Arg137-NH1

‘ildE~~~~~~~Iiligf)

x,+

(“T:%)

( A

Leu136-0 LyslQl-NZ

B

to B 2.9 2.5 2.6 2.5 2.8

Leu75-0 Asn89-0 C. Hydrogen bonds involving bridging water ~molecules A to B 2.6, 2.9 Ile133-N Ile127-0 Wat362 2.4, 2.6 Ile76-0 Argl41-NZ Wat356

B to A 2.3 2.7 2.9 2.5 2.7 B to A 2.8, 2.9 2.3, 2.6

(The co-ordinates are in the standard Protein Data Bank orthogonal system with Cartesian co-ordinates parallel to the crystallographic a. h. c*.) This transformation represents a pure rotation of 179.2”. Figure 5(a) shows the r.m.s.t separation of the main-chain atoms for each residue after applying t Abbreviation used: r.m s.. root-mean-scluarr.

Structure of Tyrosyl-tRNA this transformation. The r.m.s. residual distance for the main-chain atoms of residues 1 to 220 is 0.32 A. Significantly higher residual distances are observed in the a-helical domain. All the residues of a/j domain (residues 1 to 220) fit better than 1-O A except for residues 111 to 113, which have temperature factors > 30 A2 in subunit A, and may have been poorly placed. They form part of a loop on the surface of the molecule which is involved in inter-molecular interactions that are of different types in the A and B subunits. When the same procedure is used to fit the a-helical domain (residues 248 to 319) of the two subunits, the transformation is: f 19*78\ /- O-0936 - 0.7155 0.6923 \ X= The rotational part of the transformation represents a rotation of 176.7”, showing that the a-helical domains are not related by an accurate 2-fold rotation. After the rotation for optimum fitting of residues 1 to 220, a further rotation of 4.7” is required to overlap the a-helical domains. Figure 5(b) shows the r.m.s. separation of the main-chain atoms when this transformation is applied. The r.m.s. residual distance for the main-chain atoms in the a-helical domain is 0.27 A. The two domains of each subunit of the truncated enzyme behave, to the accuracy of the refined co-ordinates, as rigid bodies. In the truncated dimer, the a//? domains are related by an exact 2-fold local symmetry; however, one of the a-helical domains is rotated 4.7” away from an orientation of exact 2-fold symmetry. The dimensions of the molecule are 34 A in a direction parallel to the molecular dyad and 40 A x 117 A in the other two directions. The dimer is thus a rather elongated molecule so that any forces generated by intermolecular contacts can easily produce distortions. It is unlikely that domain rotation is related to t’he anti-co-operative binding of tyrosine to the dimer observed in studies of the enzyme in solution (Jakes 85 Fersht, 1975), since the tyrosine binding site is situated well away from the interface between the two structural domains (Fig. 3). The two tyrosine binding sites are separated by 33 A. The main-chain temperature factors for the residues in the helical C-terminal domain (Fig. 4), correlate with the residual separation of the atoms after superposition of the a/b domains shown in Figure 5(a). These residuals are also much larger than those obtained when the helical domains themselves are superimposed (Fig. 5(b)). This suggests that the motional or positional disorder bringing about the observed atomic temperature factors might be due t.o the rotation of t,he a-helical domain. (c) Solvent structure The molecular model includes 143 water molecules within the crystallographic asymmetric unit of which an approximately equal number is

Synthetase

associated with each subunit. There are 45 pairs of water molecules that occupy equivalent positions in each of the two subunits. These 90 molecules are within 1.0 A of the non-crystallographic symmetryof ordered related position. A large proportion solvent is situated in and around the two active site clefts. There are 36 molecules within 15 A of the a-carbon of each of the substrate tyrosines. The solvent molecules hydrogen bond to the wellordered residues in this region of the enzyme (Table 5). In contrast there is much less electron density for solvent molecules around the relatively mobile external loops on the enzyme surface and around the more mobile a-helical domains. There are five buried water molecules within each subunit. One of these (Wat349) is within the a-helical domain, and another (Wat351) is at the domain interface. The remaining three are near the tyrosine binding pocket. Wat345 is 5.9 A from the substrate hydroxyl and hydrogen bonds to Gln173-0, Asn198-0 and Gln189-NE2. Wat344 and Wat365 are adjacent to one another with Wat365 forming a hydrogen bond with Asn198-ODl (shown in Fig. 8). These two water molecules are blocked from contact with the solvent by the side-chain of Ser80. There is a cluster of water molecules (Wat348, Wat359, Wat360, Wat368 and Wat379) in a crevice on the surface of the molecule that terminates with this serine residue. In addition to the buried solvent molecules there are tightly bound molecules close to the tyrosine Table 5 Hydrogen bonds around the tyrosine binding pocket (distances in ii) A subunit

R subunit

A. Hydrogen bombs between the substrate and thr enzyme Tvr-NH3 3.2 Gln173-OEl 3.1 2.6 Tyr169-OH 2.8 Asp78-ODl 2.6 2.9 Tyr-OH Asp176-ODl 2.3 2.3 Tyr34-OH 3.0 2.9 H. Hydrogen bonds between side-chains Ssp78-ODl Gln195-NE2 Asp78-OD2 S&O-N Tyr169-OH Asp38-ODl Gln 173-NE2 Gln 195.OEl Gln189-OEl GIn173-OEl Asn198-ND2 Asp176.ODl Asn 123-ND2 $sp176-OD2 Trpl26-NE1 Thr73-OGI Asp38-OD2 .4sn123-ND2

2.8 2.9 2.5 3.2 3.1 3.0 2.9 2.7

3.2

3.2 3.0 2.5 3.0 3.2 2.9 2.8 2.8 2.7 3.3

C. Hydrogen bonds involving water moleculas LV&352 Wat326 3.1 Gln189-OEI 2.2 IIe190-0 2.6 Gly36-N 3.1 Wat326 Wat352 3.1 Gly36-0 6.8 m-at344 Glu166-0 2.8 Phe167-0 2.9 LM&l70.N 3.0 2.5 Wat365 W&359 Se&O-OG 2.9 Gln194-0 2.6

2.7 2.3 2.7 3.3 2.7 2.8 2.9 3.1 3.1 2% 2.7 2.7

2.6

294

I? Brick

and D. M. Blow

Figure 6. Stereo view showing the tyrosine substrate in the binding pocket. The substrate is shown with lighter bonds while hydrogen bonds are represented by broken lines. The Figure includes water molecules Wat352. Wat326 and Wat347-drawn as double circles. pocket. The three water molecules (Wat352, Wat326 and Wat347), which occupy one end of the tyrosine binding slot, are shown in Figures 6 and 7(b). The side-chain of Lys82, which forms a salt bridge with the substrate, partly blocks the entrance to a pocket on the surface of the molecule containing seven water molecules (Wat328, Wat333, Wat334, Wat335, Wat354, Wat355 and Wat356), of which several are very well ordered.

(d) Binding

of the tyrosine

substrate

Tyrosine was present at high concentration of the enzyme, and (1 IIIM) during crystallization there is good electron density for the substiate in both subunits of the dimer. The atomic temperature factors for the amino acid are some of the lowest in the molecule and are similar to those of the adjacent protein side-chains. This suggests that both tyrosine binding sites are fully occupied. The binding of tyrosine to the truncated mutant appears to be identical with its binding to the wildtype enzyme (Monteilhet & Blow, 1978), and provides two independently refined models of the interactions. The tyrosine occupies a slot at the bottom of the deep active site cleft close to the carboxyl-terminal end of t.he middle strands of the parallel p-sheet (Fig. 3). The depth of the active site cleft can be clearly seen in Figure 7(a). The binding slot is surrounded by bot,h polar side-chains and tightly bound water molecules that form part of an intricate

hydrogen-bonding

network

(Figs 6 and 8,

and Table 5). One end of the slot is occupied by three tightly bound water molecules (Wat352, Wat326 and Wat347; Fig. 7(b)). The slot is sufficiently

narrow

for there

to be van der Waals’

contacts with both sides of the aromatic ring of the t,vrosine

(Fig. 7(c)). One face of the slot is formed

by

61~36 and the side-chain of Leu68. The other face is formed by the side-chain of Gln173. There are two polar groups (Asp176 and Tyr34) at the bottom of the slot that hydrogen bond with the hydroxyl group

of

Asp176-ODl strong

the

tyrosine.

and Tyr-OH

interaction

The

distance

is short

between

suggesting

(Table 5). The a-amino

group

a of

the tyrosine substrate makes three hydrogen bonds with polar groups of the protein. The observed distances suggest that the interactions with Tyrl69 and Asp78 are stronger than that with Gln173. The carboxyl group of the substrate makes no direct hydrogen bond with the enzyme, but there is sufficient space for water molecules that, form hydrogen bridges between the carboxyl oxygens and polar groups on the enzyme. The side-chain of Lys82 makes a charge interaction with one of the carboxyl oxygens through a bridging water molecule (Wat353). If the lysine at position 82 is changed to an asparagine, the resulting mutant has an impaired ability to activate tyrosine (Bedouelle & Winter, 1986). Calendar & Berg (1966), in a study of the substrate specificitv of the closely homologous enzyme coli, showed that substitution of from Escherichia chloro, iodo, amino or nitro groups at the 3 (meta) position of the phenol ring eliminated the substrate activity (as measured by the pyrophosphate exchange reaction), but when either a tluoro or hydroxy group was introduced into the same position the derivative was still active, but had a lower V,,, and a higher K, value. It can be seen from Figure 7(b) that there is very little space to accommodate an extra atom at, the meta position. The substituted tyrosine must either bind in a slightly different, position or there must be a shift in the position of the Asn123 or Gln189 side-chains. Bot$h Calendar & Berg (1966) and Santi & Pefia (1973) have shown t,hat the activation of tyrosine can be strongly inhibited by a number of carboxyl-substituted derivatives of tyrosine including tyrosinol and tyramine in which the carboxyl group in absent. Monteilhet et al. (1984) showed that tyrosinol binds to the wild-type enzyme in exactly the same position as tyrosine. These results are consistent, with

the observed

lack

of interaction

between

the

substrate carboxyl group and the protein. (e) Mutations

that afect tyrovine

binding

Both Tyr34 and Tyrl69 have been t,he subject of site-directed mutagenesis experiments in which

Structure of Tyrosyl-tRNA

(b)

Figure 7. Sections through a space-filling van der Waals’ model of the molecule showing the contacts between the tyrosine substrate and the surrounding (a) A cross-section

295

side-chains have been replaced by tyrosine phenylalanine (Fersht et aE., 1985). The properties of the mutant enzymes have been measured using equilibrium dialysis and rapid reaction kinetics (Wells & Fersht, 1985). The results show that although both mutants (Tyr-Phe34 and TyrPhe169) have an increased dissociation constant of tyrosine from the enzyme, for neither mutant is there a significant change in the first-order rate constant for the formation of tyrosyl adenylate from the bound reagents. The hydroxyl moieties of Tyr34 and Tyr169, therefore, appear to interact equally well with the tyrosine substrate in its unreacted form and in the transition state of the reaction. The apparent binding energies of the substrate to the side-chains obtained from the ratio of the mutant and wild-type dissociat,ion constants (Wells & Fersht, 1985) are 0.52 kcal/mol for Tyr34OH and the significantly larger value of 2.58 kcal/mol for Tyr169-OH, which interacts with charged NH: group of the ligand the (1 kcal = 4.184 kJ). The polar interactions made by the enzyme with the phenolic hydroxyl group of the tyrosine lead to the preferential binding of the cognate substrate compared to phenylalanine, and are fundamental for the biological specificity of the enzyme. Fersht et al. (1985) have shown that the mutation Tyr-Phe34 reduces the relative specificity of the enzyme (as measured by the ratios of k,,,/K,) for tyrosine compared with phenylalanine by a factor of 15. When Gln195 is changed to glycine, the measured K, value for tyrosine increases from 2,23 PM for the wild-type enzyme to 166 FM for the mutant, while the turnover number decreases from 8.35 s-l to 0.19 s- ‘. This produces an apparent change in the transition state binding energy of 4.49 kcal/mol. The substitution also has a deleterious effect on the binding of ATP. Gln195 does not form any hydrogen bonds directly with the substrate tyrosine, but instead interacts with Asp78 and Gln173 (Figs 6 and 8, and Table 5). The side-chain of Gln195 is in van der Waals’ contact with the a-carbon of the substrate. The effects of mutations of His48 on the dissociation constant of tyrosine from the enzyme have been studied by equilibrium dialysis (Lowe et al., 1985). The dissociation constant is raised from 11.6 pM found for the wild-type enzyme, to 23 PM for the Gly48 mutant and 22 PM for the Asn48 mutant. Replacement of histidine by lysine increases the dissociation constant by a factor of seven and abolishes all measurable catalytic activity. A possible explanation is that the imadazole group of the wild-type enzyme makes a favourable electrostatic interaction with the

(0)

protein.

Synthetase

of the molecule shows the

depth of the active-site cleft. In (b) the molecule has been sectioned through the plane of the phenol ring of the substrate, while in (c) the molecule has been sectioned perpendicular to this plane. A stick representation of the substrate has been superimposed on the sections. Water molecules have been drawn using double circles.

296

and Il. M. Blow

P. Brick

Ser80 -NH :

Asp78 NE2 a’

Glnl95

-4OEI

,’

’

\

I \

\

\

Asp38 \

I’ ,

,’

Glnl89

Thr73 I

Figure 8. Schematic diagram substrate binding pocket.

showing part of the intricate

carboxyl group of the substrate. However, the two moieties are separated by over 6 A, and although both make hydrogen bonds with solvent molecules, there is no electron density for a bridging water molecule. (f) Speci$city

jor tyrosine

One of the primary reasons for the crystallographic study of an aminoacyl-tRNA synthetase was the desire to understand the specificity of these enzymes in relation to the fidelity of protein synthesis. The error rate in the total process of translating proteins from mRNA is a few in lo4 (Loftfield & Vanderjagt, 1972; Edelmann & Gallant, 1977). The error rate in aminoacylation of tRNA by the cognate amino acid in viva must therefore be less than this. In critical cases, such as the rejection of threonine for valine, this is achieved by an “editing” mechanism in which an incorrectly charged tRNA is hydrolysed (Fersht & Kaethner, 1976). Such a mechanism requires a second binding site for t.he hydrolytic step. There is no evidence that tyrosyl-tRNA synthetase has an analogous hydrolytic binding site. In vivo it is probable that the most important competitor,

which

must

be rejected

in the amino-

acylation reaction, is phenylalanine. Fersht et al. (1985) showed that the selectivity of the wild-type enzyme against phenylalanine is 1.5 x 106, well beyond the general error rate. The overall picture of the specificity of the first step of the reaction (formation of the aminoacyl

hydrogen

bonding network

(broken lines] around hhe

adenylate) can be understood quite clearly from the refined structure of the truncated mutant. The enzyme contains a deep cleft, capable of binding ATP, though there is not yet any crystallographic evidence how ATP is bound prior to formation of the aminoacyl-adenylate (Monteilhet et al., 1984). On t.he basis of strong evidence for the involvement of Thr40 in this step, and that’ His45 is certainly essential, Leatherbarrow et al. (1985) have postulated a mode for the pre-transition state binding of ATP, but this has not been realized in a diffraction experiment’. In the botton of the cleft is the deep slot which binds tyrosine (Fig. 7(a)). The width of the slot is well-designed for an aromatic residue. but it was a surprise to find that it is not precisely shaped to accept tyrosine; and in particular to discover that in the absence of tyrosine, adenine binds in this site (Monteilhat & Blow, 1978). One requirement for acylation of ATP will be that a carboxyl group is appropriately orientated to attack the a-phosphoryl group of ATP. In practice this will require a carboxyl group on a carbon atom adjacent to a carbon atom carrying the aromatic ring that, is bound in the slot). The three hydrogen bonds made by the or-amino group of tyrosine help to orient the carboxyl group, if the acylating group is an a-amino acid. (But it is notable that tyrosyltRNA synthetase is unusual in having no strong requirement 1966).)

for an [,-amino

acid (Calender

&, Berg,

Two relevant features of the slot are. first, two strongly polar groups at t,he very bottom (Tyr34 and Asp176); second. a polar environment at one

Structure

of Tyrosyl-tRNA

side of the slot, including two water molecules (Wats52 and Wat326) adjacent to the side-chains of Thr73, Tyr169 and Gln173. The polar groups at the bottom clearly favour an aromatic ring with a at the hydroxyl group para to the substituent entrance to the slot. The polar environment generally favours a more polar aromatic ring. These are the main features which provide the strong specificity of 1.5 x lo6 for tyrosine rather than phenylalanine. As noted by Calendar 6 Berg (1966) and Santi & Pefia (1973), the binding of other aromatic groups, particularly those with para-hydroxyl groups and also ortho or meta substitution, is entirely possible. Apparently in viva such compounds, which include L-dopamine, have not been so abundant that they must be discriminated against. Formation of an aminoacyl-adenylate is only possible in cases like 3-fluoro- or 3hydroxy-L-tyrosine (Calendar & Berg, 1966), where the a-aminoacyl function is intact. We thank Philip Evans and Eleanor Dodson for providing computer programs. This work was funded by the Medical Research Council of the United Kingdom.

References

Agarwal, R. C. (1978). Acta Crystalbgr.

sect. A., 34,

791-809.

Bedouelle, H. & Winter, G. (1986). Nature (London), 320, 371-373. Bhat, T. N. & Blow, D. M. (1983). Acta Crystallogr. sect. A, 39. 166170.

Bhat, T. iV., Blow, D. M., Brick, P. & Nyborg, J. (1982). J. Mol. Biol. 158, 69!&709.

Blow, D. M. & Brick, P. (1985). In Biological Macromolecules and Assem.blies (Jurnak, F. A. & McPherson, A.. eds), vol. 2. pp. 442-469, Wiley, New York. Bricogne, G. (1976). Acta Crystallogr. sect. A, 32, 832-847. Calendar, R. Br Berg, P. (1966). Biochemistry, 5, 1690-1605. Crowther. R. A. (1972). In The Molecular Replacement Method (Rossmann, M. G., ed.), pp. 173-178, Gordon & Breach, New York. Derewenda: Z. S., Dodson. E. J., Dodson, G. C. & Brzozowski. A. M. (1981). Actu Crystallogr. sect. A, 37, 407-413. Edelmann. P. & Gallant, J. (1977). Cell, 10, 131-137. Fersht, A. R. & Kaethner, M. M. (1976). Biochemistry, 15: 3342-3346. Fersht. A. R.. Shi. J.-P.. Wilkinson. A. J., Blow, D. M.,

Edited

Synthetase

297

Carter, P., Waye, M. M. Y. & Winter, G. (1984). Angewandte Chemie, 23, 467473. Fersht, A. R., Shi, J.-P., Knill-Jones, J., Lowe, D. M., Wilkinson, A. J., Blow, D. M., Brick, P., Carter, P., Waye, MM. M. Y. & Winter, G. (1985). Nature (London), 314> 235-238. Fox. G. C. & Holmes, K. C. (1966). Acta CrystaElogr. sect. A, 20, 886-891. Hendrickson, W. A. (1979). Acta Crystallqr. sect. A, 35, 158-163. Hendrickson, W. A. & Konnert, J. H. (1980). In Crystallography (Diamond, R., Computing in Venkatesan, K., Ramasethan, S. & eds), Academy of Sciences, pp. 13.01-13.23. Indian Bangalore. Honzatko. R. B., Hendrickson, W. A. & Love, W. E. (1985). J. Mol. Biol. 184, 147-164. Irwin, M. J., Nyborg, J., Reid, B. R. & Blow, D. M. (1976). J. Mol. Biol. 105, 577-586. Isaacs, N. (1982). In Computational Crystallography (Sayre, D.. ed.). pp. 398408, Oxford University Press, New York. Jakes, R. & Fersht, A. R. (1975). Biochemistry, 14, 3344-3350. Jones, H. J., McMillan, A. J. & Fersht. A. R. (1985). Biochemistry, 24, 5852-5857. Jones, T. A. (1978). J. Appt. Crystallogr. 11, 268-272. Leat.herbarrow, R. J., Fersht, A. R. & Winter, G. (1985). Proc. Nat. Acad. Sci., U.S.A. 82, 7840-7844. Lee, B. & Richards, F. M. (1971). J. MoZ. Biol. 55, 379400. Loftfield, R. B. & Vanderjagt, D. (1972). Biochem. J. 128, 1353-1356. Lowe, D. M., Fersht, A. R. & Wilkinson. A. tJ. (1985). Biochemistry, 24, 5106-5109. Monteilhet, C. & Blow, D. M. (1978). J. Mol. BioE. 122, 407-417. Monteilhet, C., Brick, P. & Blow, D. M. (1984). J. Mol. Biot. 173, 477-485. Nixon. P. E. & North, A. C. T. (1976). Acta Crystullogr. sect. A. 32, 326325. Nyborg, J. & Wonacott, A. J. (1977). The Rotation Method in Crystallography, Xorth Holland, Amsterdam. Ramakrishnan, C. & Ramachandran. G. N. (1965). Riophys.

J. 5: 909-933.

Santi. D. V. & Peiia, V. A. (1973). J. Med. Chem. 16, 273-280. Sussman? J. L.. Holbrook, S. R., Church, G. M. & Kim, S. H. (1977). Actu Crystullogr. sect. A, 33. 806804. Waye. M. M. Y., Winter, G., Wilkinson. A. J. & Fersht, A. R. (1983). EMBO J. 2, 1827-1829. Wells, T. N’. C. & Fersht. A. F’. (1985). Xutnrv (London), 316, 65f.h-657. Winter, G.. Koch, G. L. E., Hartley. B. S. $ Barker, D. G. (1983). Eur. J. Biochem. 132. 3833387. by R. Huber