Eur. J. Biochem. 208,411-417 (1992) 0FERS 1992

Three-dimensional structure of phenylalanyl-transfer RNA synthetase from Thermus thevmophilrcs HB8 at 0.6-nm resolution Ludmila RESHETNIKOVA’, Maia CHERNAYA ’,Valentina ANKILOVA2, Olga LAVRIK *, Marc DELARUE ’,

Jean-Cloud THIERRY ’, Dino MORAS and Mark S A F R 0 4

’ Institute of Molecular Biology, Academy of Sciences of the USSR, Moscow, USSR Institute of Bioorganic Chemistry, Academy of Sciences of the USSR, Novosibirsk, USSR ’ Institut de Biologie Moleculaire et Cellulaire du Centre National de la Recherche Scientifique, Strasbourg, France Department of Structural Biology, Weizmann Institute of Science, Rehovot, Israel (Received January 6/April6, 1992) - EJB 92 0006

The three-dimensional structure of the heterodimeric a2D2enzyme phenylalanyl-tRNA synthetase from Thermus thermophilus HB8 has been determined by X-ray crystallography, using the multipleisomorphous-replacement method at 0.6 nm resolution. Trigonal crystals of space group P3,21 have cell dimensions a = b = 17.6 nm and c = 14.2 nm. Assuming one heterodimeric molecule/ asymmetric unit, the ratio of unit cell volume/molecular mass was V = 0.00244 nm3/Da, which is in the middle of the range normally observed. However, after a rotation-function calculation and measurement of the density of the native crystals, we postulate the existence of only the tlp dimer in the asymmetric units. This implies 73% solvent content in the unit cell. Three heavy-atom derivatives [K,PtCl,, KAu(CN)~and Hg(CH3C00),] and the solvent-flattening procedure were used for electron-density-map calculations. This map confirmed our hypothesis and revealed a remarkably large space filled by solvent, with aj?dimer only in the asymmetric unit. The phenylalanyl-tRNA synthetase from T. thermophilus molecule has a ‘quasi-linear’ subunit organization. As can be concluded at this level of resolution, there is no contact between small tl subunits in the functional heterodimer. The aminoacyl-tRNA synthetases are a family of enzymes synthetase isolated from Saccharomyces cerevisiae. In this which catalyze the attachment of an amino acid to its cognate situation, only information on the three-dimensional structures of synthetase from different sources will provide a cleartRNA molecule at the 3’4erminus. Among the 20 aminoacyl-tRNA synthetases, only two of cut answer regarding the role of the quaternary structure in them, phenylalanyl-tRNA synthetase and glycyl-tRNA functional activity. tRNA synthetases are partitioned, as was recently shown synthetase, have a quaternary structure of the a2pZtype [i]. The other tRNA synthetases are oligomers having quaternary by Eriani et al. [8], into two classes. Phenylalanyl-tRNA structures of a, t12 and a4 types. The functional necessity and synthetase belongs to class-I1 synthetases because the smaller evolutional origin for the a2P2structure is not yet clear. Studies subunit (a subunit) contains the definite amino acid sequence [2 - 51 of isolated glycine, as well as of phenylalanine-tRNA motifs l(XaaProXaa), 2(XaaArgXaa) and 3(XaaArgXaa) (see synthetase subunits from Escherichia coli (FRSEC) and definition in [S]). This class also involves the Ala(a4)-, Thermus thermophilus (FRSTT), showed that neither a and fl Gly(dL)-, pro(^)-, Ser(ad-, T W a d - , Asp(a2)-, Asn(a2)-, monomers nor az and f12 dimers manifested a catalytic activity His(t12)- and Lys(a,)-tRNA synthetases. For all enzymes of in the reaction ofaminoacylation. It was also shown by affinity this class, except FRSTT, aminoacylation occurs on the 3‘labelling [6] of FRSEC, that the active site of tRNA OH of the ribose of the last nucleotide of tRNA. For FRSTT, this process occurs on the 2’-OH of the ribose. aminoacylation is formed in the contact region between subThis type of tRNA aminoacylation is characteristic for the units. Recently F. Fasiolo et al. [7] have found functional activ- synthetases belonging to class I, and involves Glu(a)-, Arg(a)-, ity for one subunit of mitochondria1 phenylalanyl-tRNA Cys(a)-, Val(&, Ile(a)-, Leu(a)-, Met(a2)-, Gln(a)-, Trp(a2)-, Tyr(a,)-tRNA synthetases. So, from the structural point of view, these intriguing properties allow us to identify phenylalaCorrespondence to M . Safro, Department of Structural Biology, nyl-tRNA synthetase as of the class-I1 type, and from the Weimann Institute of Sciences, Rehovot, Israel IL-76100 functional point of view as of the class-I type. Abbreviations. FRSTT, phenylalanyl-tRNA synthetase from The structures of E. coli [9] glutaminyl-tRNA synthetase Thermus thertnophilus HB8 ; FRSEC, phenylalanyl-tRNA synthetase and yeast aspartyl-tRNA synthetase [lo], complexed with the from Escherichia coli. Enzymes. Phenylalanyl-tRNA synthetase (EC 6.1 .1.20); glycyl- correspondent tRNA molecules, were solved at high resoltRNA synthetase (EC 6.1.1.14); glutaminyl-tRNA synthetase ution. Specific structural recognition elements were identified (EC 6.1.1.17); methionyl-tRNA synthetase (EC 6.1.1.10); seryl- for these synthetases in accordance, as it is now clear, with tRNA synthetase (EC 6.1.1.11); tyrosyl-tRNA synthetase (EC 6.1.1.1); their resemblance to the two different classes, based on their aspartyl-tRNA synthetase (EC 6.1.1.12). primary structure. The association between tRNA and en-

412 zyme were dctcrmined recently for tryptic fragments of the E. coli Met-tRNA synthetase [ll]. Probably this procedure can bc performed for dimeric Ser-tRNA synthetasc isolated from E. coli and Tyr-tRNA synthetase from Bacillus Stearothermophilus, whose structures are also known at high resolution [12, 131. The structure of heterodimeric FRSTT can provide a new additional understanding of the molecular basis of the specificity of the aminoacyl-tRNA synthetases. In this paper, we describe the determination of the threedimensional structure of phenylalanyl-tRNA synthetase from T. tkrrmophilus HB8 by X-ray crystallography using the multiple isomorphous-replacement method, at 0.6 nni resolution.

saturated ammonium sulfate. The symmetry of the crystals is P3221 (or its enantiomorph) and unit cell parameters are a = b = 17.6 nm and c = 14.2 nm, as determined from 8' precession photographs. Heavy-atom derivatives wcrc prepared by soaking crystals at concentration of 1 - 10 mM in mother liquor for periods of several hours up to 18 days. These results are summarized in Table 1. Data collection was carried out on a Xentronics area detector (Strasbourg, IBMC) and processed using the software XENGEN. Data collection statistics are given in Table 2. The maximal resolution of the native crystals for which the diffraction intensities have been collected is 0.36 nm.

MATERIALS AND METHODS FRSTT was isolated, characterized and crystallized as described [14]. The molecular mass of the native enzymc was determined by gel filtration and proved to be 260 kDa. The existence of two polypeptide chains with molecular masses of 40 kDa (N) and 92 kDa (8) was determined by denaturating polyacrylamide gel electrophoresis [15]. These results provide evidence for an oligomeric structure of the azP2 type. Crystals were grown at 4°C by vapor diffusion of 30% saturated ammonium sulfate in 20 mM imidazol/HCl, pH 7.8 in the presence of 1 mM MgC1,. Crystals appeared after 12 weeks and grew to maximal dimensions of 0.4x0.4 x 0.3 mm during a few weeks. They were then kept in 35%

Table 1. Summary of space group and heavy-atom derivatives. Phenylalanyl-tRNA synthetase from T. thermophilus HB8. Spacc group, P3,21; unit cell parameters, a = h = 17.6 nm, c = 14.2 nm; molecular mass, 260kDa. FNAT and FPH are the observed structure factors for the native protein and of the heavy-atom derivative respectively.

Heavy atom compounds

Resolution

C (I FPH-

Soaking time

Concentration

nm

days

mM

Yo

0.4

4

0.5 0.4

7 15

3 1 5

21 15

0.36

8

1

22

FNATi)/ C(FNAT) ~

KZPtC14 Hg(CH,COO)Z KAu(CN), p-chloromercurbenx n e sulfonate

19

RESULTS AND DISCUSSION Crystal density and molecular mass in asymmetric units The volume of the crystallographic asymmetric unit is about 634.9 nm3. Assuming one molecule/asymmetric unit, the ratio of unit cell volume/molecular mass (V,) is 0.00244 nm3/Da [16]. If this is the case, a non-crystallographic twofold axis has to be present in the asymmetric unit, relating $3 dimers of the molecule. We checked this possibility using the self-rotation function. The program POLARRFN from the CCP4 programming package was used. All possible orientations of non-crystallographic twofold axes should be found in the section K = 180". We varied the limits of resolution and the radius of integration, but could not detect any significant peaks on the rotation function maps. Only peaks corresponding to crystallographic twofold axes in the (001) plane were found. The ratio of the peak for the twofold axis [loo] along to the next highest value was 6. We also checked different sections of K , beginning at 5" and proceeding up to 17Y, with steps or K = 5". Only IC = 120" gave a significant peak along the threefold axis in the [OOl] direction. So, if the twofold noncrystallographic axis is real. it has to be parallel or coincide with the crystallographic twofold axes. The Patterson function with coefficients (FNAT) was calculated to check for any possible translational shift. N o significant peaks were detected. The density of the crystal was then measured. In 35% saturated ammonium sulfate, the dcnsity was 1.16 gjcm. Using formula 7 of [17], we found that the molecular mass of the protein in the asymmetric unit is 142000 Da. Since the molecular mass of a dimer measured in polyacrylamide gel electrophoresis in the presence of sodium dodecyl sulfate is

Table 2. Data processing information. NOBS, total number of observations; NUR, number of unique reflections; KSYM n.d., not determined.

Compound

Low rcsolution data Resolution limit

NOBS

0.64 0.72 0.50 0.72 0.64

C (lI-q)/Z

Medium resolution data NUR

RSYM

RCSO-

NOBS

NUR

KSY M

lution limit

nm

Native KZPtC1, Hg(CH,COO), KAu(CN), ClHgBzOH

=

22733 8069 20106 9120 19033

5288 3682 8954 3750 5283

Yo

nm

9.18 6.53 9.20 8.17 9.50

0.357 0.40 n.d. 0.40 0.36

010

91638 3721 5 n.d. 43181 80395

27198 17018 n.d. 18345 28180

16.0 11.0 n.d. 14.5 13.0

(0.

42 3 Table 3. Heavy-atom parameters. All temperature factors at this resolution were isotropic, not varied and B = 0.4 nm’.

Derivatives

Coordinates Y

Z

Relative Heavy occuatom PancY

-0.556

-0.155

0.78

Hg

- 0.570 -0.556 -0.574 -0.614 -0.585 -0.031

-0.453 -0.288 -0.281 -0.780 -0.694 -0.885

-0.152 -0.111 -0.047 -0.065 -0.148 -0.152

0.70 0.71 0.55 0.59 0.16 0.50

PtA PtB PtC PtD

-0.583 -0.560 -0.430 -0.760 -0.750

-0.607 -0.262 -0.309 -0.724 - 0.750

-0.017 -0.032 -0.162 - 0.067 0.000

0.57 0.37 0.43 0.21 0.17

Aul

X

Hg(CH3COO)Z -0.406 KlPtC14

KAu(CN)Z

PtE PtF Au2 Au3 Au4 Au5

132000 Da, this implies the existence of the ctp dimer in the asymmetric unit of the crystal. The solvent content of the unit cell will then be 73%. This high percentage of liquid in the crystal may be the reason for the relatively weak diffraction of the perfectly shaped crystals of FRSTT. Locations of heavy-atom positions and their refinement Experimental crystallographic phases have been determined by the multiple heavy-atom isomorphous replacement method and were based primarily on three derivatives. The difference Patterson synthesis for the Hg(CH3C00)2 derivative, calculated at 0.6 nm resolution, was first interpreted in terms of one site. This solution has been confirmed by results obtained with program HASSP (Terwilliger, T.) which performs automatic searches of heavy-atom positions in Patterson maps. The height of the peak corresponding to the heavy-atom site, found by visual inspection of Patterson synthesis of this derivative, was five-times more than the height of the second possible solution. At this stage, difference Fourier maps, computed with phases derived from mercury derivative, were used to determine heavy-atom positions in K2PtC14,KAu(CN)~ andp-chlormercurbenzene sulfonate derivatives and to refer all heavy-atom sites to a common origin. All heavy-atom positions in derivatives were confirmed by direct and cross-difference Fourier maps. The heavy-atom sites in KAU(CN)~ were identified only after two other derivatives were available for phase cross-difference Fourier maps. The interpretation of the Patterson function of K A U ( C N )by ~ visual inspection was impossible. As it appeared later, the second mercury derivative, p-chlormercurbenzene sulfonate had the same main heavy-atom-binding site as the Hg(CH3C00)2 derivative, so we did not use this derivative at low resolution for the phasing procedure. The coordinates of the heavy-atom positions are listed in Table 3. At this stage, using a combination of isomorphous-scattering and anomalous-scattering information for the Pt derivative and isomorphous information for mercury derivative, we found the correct enantiomorph which corresponds to space group P3221. Refinement of the heavy-atom parameters was performed using the CCP4 program PHARE. Refinement was carried

Table 4. Statistics of refinement. RCullis = C (I FPHCaI,- FPHobsl)/2 ( 1 FPH,b, - FNATI), where FPHcalc and FPHobs are the calculated and observed structure factors for derivative, FNAT are observed structure factors for the native protein. Summing is over the centric reflections only. Phasing power, f/s, where f is the rms heavy atom structurefactor and E is the rms lack of closure error. The % complete reflection is shown in parenthesis.

Derivative

Unique reflection

Native

5729(94) 5410(89) Hg(CH3CO0)2 4523(74) 5604(92) KAu(CN)Z KZPtC1,

Resolution Rlimit

Cullis

nm

%Q

1.2-0.6 1.2-0.6 1.2 -0.6 1.2-0.6

60 75 79

Phasing F.O.M. power

76 2.97 2.02 1.39

out over 1.2- 0.6 nm resolution. The temperature factors were not refined. As can be seen from Table3, the KAu(CN)~ derivative has low phasing power, but addition of this derivative resulted in a 3% rise in the overall figure of merit, suggesting that these data are useful for resolve at least some phase ambiguities. To check the phasing power of each derivative, we computed three electron density maps for FRSTT using single isomorphous-replacement phases from each derivative, then compared all three maps. All these maps have different levels of noise according to their rcspective phasing power, but all maps exhibit common region of electron density. Taking into account that there are no common binding sites in these derivatives, and knowing that maps are mainly determined by phases, it is therefore possible to conclude that phase angles determined for derivatives are in agreement with one another. Statistical data for phasing are presented in Table4 and are as given by the final run of the PHARE program.

Electron density map The 0.6-nm electron density map was calculated using ‘best’ phases and plotted in sections perpendicular to the crystallographic c axis. The information regarding the quarternary organization of FRSTT is of special interest because it is the first representative of the structure of synthetases with a heterodimeric organization. Part of the multiple isomorphousreplacement map is shown in Fig. 1. The regions of moredcnse protein and less-dense solvent can easily be separated one from another. In the plane of the figure, as well as in the direction perpendicular to this plane (Fig. 2), regions of the density map which appear like dense continuous rods of 0.6nm diameter can be seen. Some of them have a length of about 2.5 nm or longer and probably can be identified as E helixes. Typical spacing between some of them varies over 1.01.3 nm. The delineation of the a2 82 molecular boundaries could be performed using the multiple isomorphous-replacement map. Taking into account the unusually high percentage of solvent in the FRSTT crystals, we used the ‘solvent-flattening’ procedure to improve the electron density maps. A density-modification procedure using solvent flattening includes a stage of automatic mask calculations. We used this step only once, taking the solvent content of the crystal volume equal to 50% and the averaging radius as 1.2 nm. This stage

414

Fig. 1. Superposition of the sections over 2 = 0.04- 0.09 of the 0.6-nm resolution multiple-isomorphous-replacement electron density map of the FRSTT. There are six sections in the stack (0.74 nm) along the c axis. The protein density is outlined by a solid line. 0.9800 mm represent 0.1 nm.

I

I I I V

N

was followed by four cycles of the phase combination in reciprocal space as implemented in the CCP4 package by A. Leslie (181. The final figure of merit for the 5729 reflections was 87%. The density-modified map (Figs 3 - 5) showcd better contrast than the multiple isoinorphous-replacement map and, in somc parts, better connectivity. The rms of the electron dcnsity map fell from 11.5 (multiple isomorphous-replacement map) to 7.6 (density-modified map) at the same value of RHOMAX (50.0 in arbitrary units). The resulting map was clcarly interpretable in terms of molecular boundaries and subunit organization. This map confirmed our hypothesis that, in the asymmetric unit of the crystal unit cell, there is only one CIP dimer. As can be seen from the three stacks of electron-density sections comprising almost thc whole C I ~ dimer (Figs 3 5 ) , there is a remarkably large space filled by solvent. The intramolecular twofold axis connecting a and P dimers in the a J z heterodimer coincides with the crystallographic tworold axis directed along the diagonal in the (001) plane. The separation of subunits is also clearly visible in Figs 3 and 5. The large differences in the masses ( a = 40 kDa and fi = 92 kDa), it is possible to conclude that thc small CI subunit is placed in the distal (from the diagonal twofold axis) part of the electron-density map, and the large subunit, called 11, placed around the crystallographic twofold axis which crcatcs heterodimer rx2fi2 from the two x / j dimers (Fig. 6). The contacts between the fi subunits in the molecule are very tight and extensive. There is no contact between small ct subunits in the functional hetcrodimer. This ‘quasi-linear’ subunit organization of FKSTT molecules differs froni the oligomeric structure of phenylalanyltRNA synthetase isolated from E . coli, as dcscribed from ~

Fig. 2. Superposition of the section over y = 0.59 - 0.63 of the 0.6-nm resolution multiple isomorphous-replacement electron density map of the FRS’TT. Thcrc are six sections in the stack (0.8 nm) along the b axis. 0.9800 mm represent 0.1 nm.

415

Fig.3. The electron density map at 0.6 nm resolution modified by the solvent-flattening procedure, viewed along the c axis. Six sections (4- 9; comprising 0.74 nm) in this directions. Solvent content-50%, averaging radius-I .2nm. Comparing with the same stack shown on the Fig. 1, it can bc concluded that density modification of the MIK map mainly kept its contours. as, small subunit of symmetry related molecule. 0.9800 mm represent 0.1 nm.

Fig. 4. Part of the ‘modified’ electron density map at 0.6 nm resolution, mimicing in this slice the ‘L-shape’ of the tRNA molecule. Elevcn sections from z = 0.1 -0.21 (sections 10-20; comprising 1.47 nm) stacked onto each other. (+) Position ofone heavy-atom site (Pt €3) from platinum derivative. located in the interface area between z and fl subunits. 0.9800 mm represent 0.1 nm.

neutron small-angle-scattering experiments [18]. The type of ‘linear’ subunit organization was also suggested for yeast cytoplasmic phenylalanyl-1RNA synthetase [19]. It gives a clearcut answer to the results of denaturation experiments [4],

where it was discovered that j subunits exists mainly in the lorm of l j 2 dimers. The absence of a/]dimers can also be explained by the relatively weak contacts between I and p subunits in the three-dimensional structure.

416

Fig. 5. Superposition of the twelve sections of the ‘modified’ electron density map at 0.6 nm resolution along e axis from z = 0.22 - 0.33 (sections 21 -32; comprising 1.56 nm). The small CI subunit of a referenced molecule outlined by solid line. 0.9800 mm represent 0.1 nm.

Fig. 6. Model of a& FRSTT molecule. The intramolecular twofold axis connecting CIBdimers in the heterodinier is depicted by a solid line

parallel to the plane of the picture. The molecular cnvelopc was constructed from thc ‘averaged’ electron density map. The multipleisomorphous-replacement map at each grid point is replaced by the weighted average of the electron density at all surrounding grid points within a sphere ol‘ radius 1.2 nm. The area of intersubunit contacts. smoothcd out by this procedure, is shown by a dashed line.

This type of packing makes the heterodimer a relatively compact molecule. The maximal length of the o!fl dimer is about 13 nm. If these dimers were connected in the same way as Tyr-tKNA synthetase, the length of the molecule constructed in a such manner would be around 30 nm. It is interesting to note that the dimer (Fig. 6) FRSTT looks like the tRNA

with an ‘L-shape’ structure. The length of the large shoulder, presumably created by the p subunit, is about 10 nm and the length of the small shoulder, which could be constructed both from p and o! subunits, is about 6 nm. The separation of o! and p monomers can probably be marked by some of the binding sites in the K2PtC1, derivative.

417

In the presence of ammonium sulfate this derivative can be bound to anionic site in the FRSTT. At the anionic sites, negatively charged substrates like ATP and tRNA can be bound. In this case, K,PtCl, should be an inhibitor for FRSTI’ in the process of aminoacyl-tRNA formation. This was recently confirmed by one of us (unpublished results). However, the active sites have to be located in the area of afl contacts [ 6 ] .Only one of the binding sites in this derivative, namely PtB (Fig. 4), with maximum relative occupancy, is placed in the area of the c$ interface. The above mentioned ‘quasi-linear’ molecular structure and the intersubunit contacts in the a/] dimer and in the heterodimer a2p2 allows us to assume the following mechanism of oligomerization of the FRSTT: P+B-P2

One of the authors, M. S. is very grateful to Prof. A. Yonath for her interest in this project and to the Kimmelman Center for Assemblies of Biomolecdes for continuous support to this project.

REFERENCES 1. Schimrncl, P. (1987) Ann. Rev. Biochenz. 56, 125-128. 2. Ducruix, A.. Hounwanou. N., Reinbolt, J., Boulanger, I. & Blanquet, S. (1983) Biochim. et Biophys. Acra 771, 244-250. 3. Zykova, N., Nevinskii, G. & Lavrik, 0. (1982) Mo/. Biol. 16, 928 - 935.

4. Bobkova, ti., Volfson, A.. Ankilova, V. & Lavrik, 0. (1990) Biochemisfry 55, 389-395. 5. McDonald. J., Brute, L., Pangburn, K . L. W., Mom, S., Manser, J . & Nagel, G. (1980) Biochemistry 19, 1402- 1409. 6. Khodyreva, S . , Moor. N.. Ankilova, V. & Lavrik, 0. (1985) Biochim. Biophys. Actu 830, 206-216. 7. Sanni, A ,, Walter, P., Boulanger. Y . , Ebel, J.-P. & Fasiolo, F. (1991) Proc. N u f / A c u d . Sci. I U S A ) 88, 8387-8391. 8. Eriani, G., Delarue, M., Poch. O., Gangloff, J . & Mords, D. (1990) Narure 347, 203-206. 9. Rould. M., Perona, J., Soll, D. & Steitz, T. (1989) Science 246, 1135- 1142. 10. Ruff, M., Krishnaswamy, S.. Boeglin, M., Roterszman, A., Mitschler, A., Podjarny, A., Rees, B., Thierry. J.-C. & Moras, 1).(19Y1) Science 252, 1682-1689. 11. Perona, J., Rould, M., Steitz, T.. Rislcr. J.-L., Zelwer, C. & Brunie, S. (1991) Proc. Natl Acnd. Sci. 88, 2903-2907. 12. Brick. P., Bhat, T. & Blow, D. (1989) J . Mol. Biol. 208, 83-98. 13. Cusack, S., Berthet-Colominas, C., Hartlein, M., Nassar, N., Lcbcrrnan, R. (1990) Nature 347, 249-255. 14. Chernaya, M., Reshetnikova, L., Korolev, S. & Safro, M. (1987) J . Mol. B i d . 198, 555- 556. 15. Ankilova, V., Rcshetnikova, M., Chernaya, M. & Lavrik, 0. (1988) FEBS Lett. 227. 9-13. 16. Matthews, B. (1968) J . Mol. Biol. 33,491 -497. 17. Matthew, B. (1974) J . Mol. B i d . 82, 513-526. 18. Leslie, A. (1987) Actu Crysf.A43, 134 136. 19. Dessen, P., Ducruix, A,. May, R. & Blanquet. S . (1990) BiochemistrJ’ 29, 3039 - 3046. 20. Fasiolo. F., Sanni, A., I’otier, S., Ebel, J.-P. & Boulanger, Y. (1089) FEBS I x t t . 242, 351 -356 ~