The highly refined solution structure of the cytotoxic ribonuclease α-sarcin reveals the structural requirements for substrate recognition and ribonucleolytic activity1

doi:10.1006/jmbi.2000.3813 available online at http://www.idealibrary.com on

J. Mol. Biol. (2000) 299, 1061±1073

The Highly Refined Solution Structure of the Cytotoxic Ribonuclease a -Sarcin Reveals the Structural Requirements for Substrate Recognition and Ribonucleolytic Activity Jose Manuel PeÂrez-CanÄadillas1, Jorge Santoro1 RamoÂn Campos-Olivas1, Javier Lacadena2, Alvaro MartõÂnez del Pozo2 Jose G. Gavilanes2, Manuel Rico1 and Marta Bruix1* 1

Instituto de Estructura de la Materia, Consejo Superior de Investigaciones Cientõ®cas Serrano 119, 28006 Madrid Spain 2

Departamento de BioquõÂmica y BiologõÂa Molecular I, Facultad de QuõÂmica, Universidad Complutense, 28040 Madrid Spain

a-Sarcin selectively cleaves a single phosphodiester bond in a universally conserved sequence of the major rRNA, that inactivates the ribosome. The elucidation of the three-dimensional solution structure of this 150 residue enzyme is a crucial step towards understanding a-sarcin's conformational stability, ribonucleolytic activity, and its exceptionally high level of speci®city. Here, the solution structure has been determined on the basis of 2658 conformationally relevant distances restraints (including stereoespeci®c assignments) and 119 torsional angular restraints, by nuclear magnetic resonance spectroscopy methods. A total of 60 converged structures have been computed using the program DYANA. The 47 best DYANA structures, following restrained energy minimization by GROMOS, represent the solution structure of a-sarcin. The resulting average pairwise rootÊ for backbone atoms and 1.47 A Ê for all mean-square-deviation is 0.86 A heavy atoms. When the more variable regions are excluded from the analyÊ and 1.00 A Ê, sis, the pairwise root-mean-square deviation drops to 0.50 A for backbone and heavy atoms, respectively. The a-sarcin structure is similar to that reported for restrictocin, although some differences are clearly evident, especially in the loop regions. The average rmsd between the structurally aligned backbones of the 47 ®nal a-sarcin structures and the Ê . On the basis of a docking model crystal structure of restrictocin is 1.46 A constructed with a-sarcin solution structure and the crystal structure of a 29-nt RNA containing the sarcin/ricin domain, the regions in the protein that could interact speci®cally with the substrate have been identi®ed. The structural elements that account for the speci®city of RNA recognition are located in two separate regions of the protein. One is composed by residues Ê away in 51 to 55 and loop 5, and the other region, located more than 11 A the structure, is the positively charged segment formed by residues 110 to 114. # 2000 Academic Press

*Corresponding author

Keywords: a-sarcin; ribosome-inactivating protein; NMR solution structure; protein-RNA interaction; ribonucleolytic activity

Present addresses: R. Campos-Olivas, Laboratory of Chemical Physics, NIDDK, National Institute of Health, Bethesda, Maryland 20892-0505, USA; J. Lacadena, Facultad de BiologõÂa, Universidad SEK, 40003 Segovia, Spain. Abbreviations used: pH*, pH reading without correction for the deuterium isotope effect; NOESY, nuclear Overhauser enhancement spectroscopy; NOE, nuclear Overhauser effect; RNase, ribonuclease; SRD, sarcin/ricin domain; TAD, torsion angle dynamics. E-mail address of the corresponding author: [email protected] 0022-2836/00/041061±13 $35.00/0

Introduction a-Sarcin is an extracellular cytotoxin produced by the mold Aspergillus giganteus (Olson & Goerner, 1965; Olson et al., 1965). It is a small ribosome-inactivating protein than inhibits protein biosynthesis by cleaving a single phosphodiester bond in a strictly conserved RNA sequence located in the ``sarcin/ricin'' loop of the largest ribosome # 2000 Academic Press

1062 subunit (Endo & Wool, 1982). Due to this very speci®c activity, a-sarcin represents an important tool for analyzing the structure and function of ribosomes (Wool, 1984). This protein is a representative member of a distinctive family of fungal ribonucleases, called ribotoxins, with very similar amino acid sequences (about 86 % identity) and produced by different Aspergillus species: gigantin (Wirth et al., 1997) also from A. giganteus, mitogillin (FernaÂndez-Luna et al., 1985) and restrictocin (LoÂpez-Otin et al., 1984), both from A. restrictus, allergen I (Arruda et al., 1992) from A. fumigatus and clavin (Parente et al., 1996) from A. clavatus. Ribotoxins are more distantly related to RNase T1 subfamily of microbial ribonucleases (ManchenÄo et al., 1995), but have additional abilities, including exquisite substrate speci®city and the ability to interact with and translocate across phospholipid bilayers (Gasset et al., 1994), which RNase T1 subfamily members lack. Extensive studies on the a-sarcin ribonucleolytic activity (Endo et al., 1983; Lacadena et al., 1999) and the protein lipid interaction of a-sarcin (Gasset et al., 1990, 1991) have been reported. Moreover, a-sarcin's spectroscopic characterization (MartõÂnez del Pozo et al., 1988) and the relationship between electrostatic interactions, structure and catalytic function have been recently reported (PeÂrez-CanÄadillas et al., 1998). a-Sarcin is cytotoxic for normal human cells (Turnay et al., 1993) and for many tumor cell lines, particularly sarcoma and carcinoma. Since a-sarcin shows a selective, high level of toxicity against cells with altered or damaged cell membranes (Otero & Carrasco, 1988), it is a promising candidate for the treatment of cancer and viral infections such as AIDs (FernaÂndez-Puentes & Carrasco, 1980; Wawrzynczak et al., 1991). How this small protein recognizes and cleaves a single phosphodiester bond out of the more than 7000 present in the rRNA is a very important puzzle to solve. a-Sarcin's speci®city undoubtedly arises from its structure, the structure of its target rRNA and their interactions. Thus, knowledge of the three-dimensional structure is the ®rst step towards the ®nal goal of understanding the cytotoxic action of a-sarcin at a molecular level. Here, we report the high-resolution solution structure of a-sarcin obtained by nuclear magnetic resonance (NMR) methods.

Results Structure calculations Analysis of the nuclear Overhauser enhancement spectroscopy (NOESY) spectra provided 7341 unambiguously assigned NOEs which were translated into 3359 upper-limit distance constraints leading to a set of 2658 conformationally relevant distance constraints (Table 1). These correspond to an average of about 18 distance constraints per residue and are classi®ed as follows: 564 intraresi-

Solution Structure of -Sarcin

dual, 480 sequential (ji ÿ jj ˆ 1), 285 medium range (2 4 ji ÿ jj 4 5) and 1335 long range (ji ÿ jj > 5). The distribution of NOE-derived distance constraints is not homogeneous throughout the sequence and ranges from a low of three constraints for Gly143 to a high of 110 constraints for Trp4. Stereospeci®c assignments were made by analyzing preliminary structures with the program GLOMSA and using the iterative procedure described in Materials and Methods. A total of 12 pairs of bb0 protons (Pro13, Gln27, Phe52, Asn54, Asp75, Asp77, Tyr93, Phe97, His104, Tyr124, Tyr126 and Cys148), one pair of aa0 protons (Gly58) and three pairs of dd0 protons of Pro residues (Pro38, Pro101 and Pro127) were stereospeci®cally assigned. In addition, 61 restrictions involving Tyr and Phe residues symmetrically related ring protons could be uniquely assigned. Moreover, the methyl group pairs of all Leu and Val residues (except for Val130, whose methyl groups are degenerated) were stereospeci®cally assigned by analyzing the 13C-HSQC spectrum of a [U-20 % 13C] sample. In this spectrum, the pro-R methyl groups (g1 Val and d1 Leu) appear as doublets, showing the 35 Hz separation corresponding to the 13C-13C coupling constant, while the pro-S groups (g2 Val and d2 Leu) appear as singlets (Neri et al., 1989). A total of 119 f angle restraints, derived from the experimentally determined 3JHN,aH coupling constants, were introduced in the last step of the calculation process. These restrictions correspond to all residues except proline residues, glycine residues, Val2, Phe71, Asp102 and Asn116. The ranges of these restrictions were determined as described in Materials and Methods. The ®nal DYANA (GuÈntert et al., 1997) calculation produced a set of 60 structures that satis®ed the distance constraints, with reasonable van der Waals packing. The best 47 structures, with target functions within three units of the minimal target function (0.52-3.52), were selected for restrained energy minimization by GROMOS (van Gunsteren & Berendsen, 1987). The 47 ®nal structures of a-sarcin are remarkably well de®ned, showing a Ê for the backbone atoms and RMSD of 0.86 A Ê for all heavy atoms. These values indicate, 1.47 A despite the fact that two-thirds of the residues are found outside regular secondary structural elements in long loops, that the de®nition of this family of structures is as good as many others with a higher percentage of secondary structure. The structural statistics of the ®nal structures are summarized in Table 1, and the backbone superposition of the 47 a-sarcin structures together with a ribbon representation are shown in Figure 1. Not all regions in the protein superimpose equally well; the regions composed of the N-terminal residues (Ala1 and Val2) and residues 11 to 18, 55 to 67, 70 to 75, 84 to 91, 112 to 117 and 143 to 144 show values of global RMSD higher than the mean value. When these variable regions are excluded from the comparison, the average pairwaise back-

1063

Solution Structure of -Sarcin Table 1. NMR structure calculations summary Type of distance constraints

Total

Intraresidual (ji ÿ jj ˆ 0) Sequential (ji ÿ jj ˆ 1) Medium-range (2 4 ji ÿ jj 4 5) Long-range (ji ÿ jj > 5) All

564 480 285 1335 2658

Residual constraints violations in the final 47 structures Ê) Range (A 0.0-0.1 0.1-0.2 0.2-0.3 0.3-0.4 >0.4 Max. violation Average sum violations Total energy (KJ molÿ1) Lennard-Jones energy (KJ molÿ1) Electrostatic energy (KJ molÿ1) NOE term (KJ molÿ1) Backbone atoms All residues (150) Best defined residues (106)a Residues 2nd structure (39) Heavy atoms All residues (150) Best defined residues (106)a Residues 2nd structure (39) a

No. of constraints Ê Ê <3.5 A 3.5 < 4.5 A 283 151 3 60 497

Ê 54.5 A

186 100 72 160 518

95 229 210 1115 1649

Av. no. of distance constraints violations 91.4 42.4 18.5 3.9 1.3 0.43 17.3 Average ÿ4825 ÿ3904 ÿ3223 64

Range ÿ6651 to ÿ3423 ÿ4284 to ÿ3008 ÿ4861 to ÿ2122 40 to 101 Ê) Averaged pairwise RMSD (A 0.860.16 0.500.09 0.510.12 1.470.18 1.000.09 0.970.13

Residues: 3 to 10, 19 to 54, 68 to 69, 76 to 83, 92 to 111, 118 to 142, 145 to 150.

Figure 1. Stereoscopic representation of the solution structure of a-sarcin. (a) Superposition of the 47 solution structures. Residues in the N-terminal b-hairpin are depicted in gray, the central b-sheet is blue; the a-helix in red, loop 1 in yellow, loop 2 in green, loop 3 in cyan, loop 4 in pink and loop 5 in orange. Every tenth residue is numbered. (b) Ribbon diagram of a-sarcin structure. All structural elements (see the text) are indicated. Figures produced with the program MOLMOL.

1064 Ê and the RMSD for all bone RMSD drops to 0.50 A Ê heavy atoms falls to 1.00 A. All backbone torsion angles for the non-glycine residues fall within the allowed region of the f, c Ramachandran plot for all the structures, with the exception of Glu140 (Figure 2). Ten of the 13 glycine residues: 42, 55, 58, 60, 65, 72, 86, 103, 133 and 143, and Asn16, Ser47, Asn54 and Lys129 have positive f angles. The angular order parameters for the backbone angles f and c, and the global and local RMSD values for each residue were used to evaluate the local precision of the structure. The majority of the polypeptide residues show angular order parameters, S(f) and S(c), larger than 0.95. Only 26 residues, mostly located at the N terminus, segments 56 to 66, 86 to 92 and 139 to 144, have S(f) and S(c) lower than this value. Here, the high values of local and global RMSD and low values of S(f) and S(c) all indicate low-level structural de®nition and probably some degree of ¯exibility. In the case of the region comprised of residues 12 to 17, high values of S(f) and S(c) and low local RMSD values indicate that the local structure is well de®ned and relatively rigid; the global RMSD values for this region are, however, high, indicating the presence of alternative conformations, i.e. the poor global ®t in this region is not due to local disorder but to different orientations of this segment relative to the remainder of the protein. The multiple conformations observed may arise from a hinge movement. We are currently characterizing the local internal dynamics of the protein by 15N relaxation measurements, to clarify these issues.

Figure 2. Ramachandran plot for all non-glycine residues in the 47 ®nal structures of a-sarcin. Glu140, which has unusual c and f values, is labeled.

Solution Structure of -Sarcin

The coordinates of 20 representative conformers of the total ensemble have been deposited in the RCSB Protein Data Bank under the accession code 1DE3. Description of the a -sarcin structure Backbone conformation a-Sarcin, as well as other microbial ribonucleases, folds into an a ‡ b structure (Figure 1). The global fold of the protein consists of a central ®ve-stranded anti-parallel b-sheet and an a-helix of almost three turns (Gln27-Ala37) packed nearly orthogonally on one face of the b-sheet. This sheet is composed of strands b3 (His50-Phe52), b4 (Leu94-Phe97), b5 (Ala120-Tyr124), b6 (Gly133Thr138), and b7 (Glu144-Leu147) arranged in a ÿ1, ÿ1, ÿ1, ÿ1 topology. It is highly twisted in a righthanded sense, de®ning a convex face against which the a-helix is packed, and a concave surface which holds the active-site residues His50, Glu96, Arg121 and His137 (Figure 3(a)) with their sidechains projecting outwards from the cleft. In addition, residues 1 to 26 form a long b-hairpin that can be considered as two consecutive minor bhairpins connected by a hinge region. The ®rst one, closer to the open end of the hairpin, is composed by strands b1a (Thr2-Cys6) and b2a (Leu23-Asn26). This hairpin is stabilized by tertiary contacts; i.e. residue Ser149 makes b-sheet-like hydrogen bonds with Thr5 on b1a. The second sub b-hairpin is formed by two short strands b1b (Asp9-Asn12) and b2b (Lys17-Thr20) connected by a type I b-turn Pro13-Asn16. This last part of the N-terminal hairpin juts out as a solvent-exposed protuberance. The backbone hydrogen bonding pattern of all these regular secondary structure is essentially identical to that de®ned (Campos-Olivas et al., 1996a,b) on the basis of a qualitative evaluation of the NOE data. The remaining residues of the a-sarcin sequence form large loops connecting the secondary structure elements. These loops are classi®ed as follows: loop 1 (Pro38-Pro49) which connects the a-helix with the ®rst strand of the central b-sheet (b3), loop 2 (Thr53-Tyr93) joining the strands b3 and b4, loop 3 (Pro98-Pro119) which links the strands b4 and b5, loop 4 (Thr125-Cys132) connecting strand b5 and b6; and the short loop 5 (Lys139-Gly143) that joins strands b6 and b7. Despite the exposed character of these loops and their lack of repetitive secondary structure, their conformations are well de®ned. A structural analysis of these regions reveals that they are maintained by networks of intra- and interloop interactions: hydrogen bonds, hydrophobic interactions and salt bridges. Some structural features within the loops are notable. First, loop 2 is rich in glycine residues (55, 58, 60, 65, 72, 86 and 88) and is largely solvent exposed. Second, the hydrogen bond formed by the side-chain of Asn54 and the backbone oxygen atom of Ile69 could stabilize the conformation of the ®rst part of this

1065

Solution Structure of -Sarcin

loop. The Asn54 amide group side-chain protons are observable in the NMR spectra in 2H2O solution (weeks at 30  C; pH* 6.0), strongly suggesting that these hydrogen bonds seldom break. This interaction is conserved in microbial ribonucleases (Pfeiffer et al., 1997; SevcõÂk et al., 1991), and contributes signi®cantly to the overall stability (Hebert et al., 1998). In these proteins, the main-chain amide group of this Asn residue binds to the guanyl ring in the substrate and confers speci®city (Arni et al., 1988). The second longest loop, loop 3, is stabilized by a long-range hydrogen bond between the phenolic proton of Tyr106 and the backbone oxygen atom of Glu115. This interaction probably slows the Tyr's ring ¯ipping; its ring proton resonances are broad and could only be observed in spectra recorded at low temperature (5  C) (data not shown). Third, loop 5 forms many hydrophobic interactions, hydrogen bonds and salt bridges with other parts of the protein which help maintain the unique conformation adopted by this loop; Glu140 has unusual backbone torsional angles (Figure 2). In homologous ribotoxins from Aspergillus, restrictocin (LoÂpez-OtõÂn et al., 1984), mitogillin (FernaÂndez-Luna et al., 1985), and Aspf1 (Arruda et al., 1992), this position is occupied by a Gly residue, which can readily adopt positive f angles. Glu lacks the special ¯exibility of Gly, and our ®nding that Glu140 is forced to adopt the same local conformation as Gly shows that having a positive f angle at this position is a speci®c structural requirement of the ribotoxin fold. This suggests that mutating Glu to Gly would stabilize a-sarcin's loop 5. Loop interactions

Figure 3. Features of the a-sarcin solution structure. (a) Superposition of the active-site residues in 20 structures. Catalytic side-chains are labeled and depicted in red, side-chains for which an additional role has been suggested during the catalytic reaction are in green, the rest of side-chains in light gray, and backbone atoms are in dark gray. (b) Interface between loop 1 (backbone in yellow) and loop 3 (backbone in cyan) showing local interloop interactions that de®ne the relative orientation of both parts of the molecule (see the text). Broken lines represent hydrogen bonds. (c) Contacts of the a-helix (backbone in red) with other parts of the structure: loop 3 (backbone in cyan), with residue Val2 (backbone in violet) and residue Phe131 of the central b-sheet (back-

The long loops of a-sarcin show a variety of interactions with other parts of the protein that are distant in the sequence. For example, the relative orientation of the loop 1 with respect to loop 4 is largely determined by the hydrophobic interaction between just two residues, Leu39 and Pro127. On the other hand, there is a large number of hydrogen bonds and packing interactions between loop 1 and loop 2. These include the network of H bonds formed by the Asp41 side-chain with He2 of His82 and with the backbone amide group of Trp51. Loops 1 and 3 are linked by the hydrophobic packing of residues Pro38, Ser46, Tyr48, Pro49 and Phe108, and a hydrogen bond between Ser46 (NH) and Phe108 (CO) (Figure 3(b)).

bone in blue). Broken lines represent hydrogen bonds. (d) Interface between the hairpin N-terminal (backbone in blue) and loop 5 (backbone in orange) where some hydrogen bonds (broken lines) and the p/cation interaction between Tyr18/Lys139 is shown. All four parts were produced by the program MOLMOL.

1066 Residues in the ®rst part of loop 2, together with residues from the central b-sheet form the major hydrophobic core of a-sarcin. This core is composed by residues Tyr56, Leu62, Pro63, Pro68, Ile69 and Phe71 from the loop 2 and Phe52, Leu94, Ile123, Ile135, Leu145 and Leu147 from the central b-sheet. The more external residues of this cluster (Leu145 and Leu147) have accessibilities of about 15 %, whereas the rest of the residues have accessibilities lower than 10 % or even close to 0 % (Ile135). This hydrophobic core, which also includes the disulphide bridge between Cys76 and Cys132, buries the convex face of the central b-sheet. The ®rst part of loop 3 (Pro98-Ser110) interacts with the a-helix, and a salt bridge is formed between side-chains of Asp105 and His36 (Figure 3(c)). Loop 4 also contacts the a-helix (Figure 3(c)). Residues from loop 3 and b6 make hydrophobic interactions with the open end of the N-terminal b-hairpin. Pro101 (loop 3) contacts Leu23 and Tyr25 (b2a), and Ile134 packs against Trp4 (b1a), which ¯anks the left and the right side of the a-helix, hindering the entry of solvent molecules between the helix and the b-sheet's back side. Several interactions de®ne the contact between the loop 5 and the N-terminal b-hairpin (Figure 3(d)). The side-chains of Lys139 and Glu140 in loop 5 form salt bridges with Asp9 and Lys11, respectively, in the N-terminal b-hairpin. The Glu140 to Lys11 salt bridge may be of special importance to the stability of the loop 5-b-hairpin interaction. As noted, Glu140 adopts an unfavorable backbone conformation and this site is occupied by glycine in other ribotoxins. This suggests that in a-sarcin, the stability of loop 5 has been sacri®ced to fortify the stability of the loop 5-b-hairpin contact. Additionally, Tyr18's ring and the side-chain of Lys139 form a p-cation interaction; this class of interactions has been described in detail in model systems and in protein structures (Scrutton et al., 1996; Gavillan & Dougherty, 1999). In addition to these charge-charge interactions, two hydrogen bonds (Asn8 Hd21-His137 O and Glu140 HN-Asp9 O) help maintain the relative orientation of loop 5 and the closed up of the N-terminal b-hairpin (Figure 3(d)). The hydrogen bond between the carbonyl group oxygen atom of Gly143 and the Hd1 of His137 (Figure 3(d)) could have important consequences for the activity, as discussed below. Surface properties of the a -sarcin structure a-Sarcin is shaped like a partially closed left hand with a thin wrist corresponding to the N-terminal hairpin (Figure 1(a), gray), a thumb corresponding to loop 3 (Figure 1(a), cyan), the palm being the concave side of the central b-sheet, and the exposed loop 2 (Figure 1(a), green) representing the ®nger tips. The exposed surface area is Ê 3. Ê 2 and its volume is about 20,000 A about 7000 A

Solution Structure of -Sarcin

While this volume corresponds to a sphere with a Ê diameter, the width of the protein is very 34 A Ê to 53 A Ê . This agrees variable, ranging from 38 A with the visual impression that this protein adopts a less globular shape than related proteins like RNase T1 (Pfeiffer et al. 1997). a-Sarcin's surface is highly charged: 39 % of the surface composed of charged side-chains and 26 % of polar side-chains. The protein structure is not absolutely compact Ê 3 cavity has been detected. The and a small 16 A volume of this cavity is large enough to accommodate a water molecule, although no such water molecule has been detected. If present, this water molecule could be stabilized by hydrogen bonds to the His92 side-chain and with peptide groups from residues 75 to 80 and 92 to 94. More evidence for this hypothetical water molecule could be provided in the future by NMR experiments speci®cally designed to detect bound water. The charge properties of the a-sarcin surface are likely to play a role in the protein's function. As can be seen in Figure 4(a) (left), the concave surface that contains the active site has a high positive charge, which probably contributes to the protein/ substrate interaction, since the target rRNA is negatively charged. The distribution of the isopotential ®eld lines in the plane of the active side (Figure 4b) illustrates how the electric ®eld created by the charges could serve to draw a-sarcin towards the rRNA and correctly orient the protein over the target site.

Figure 4. Surface electrostatic properties of the solution structure of a-sarcin. (a) Electrostatic potential in the protein surface: positive areas are represented in blue, negative in red and neutral in white. The concave face is shown on the left (same orientation as Figure 1) and the convex face on the right. (b) Distribution of the isopotential ®eld lines in the plane of the active center. In red, zones with negative potential and in blue zones with positive potential. Isopotential ®eld lines are drawn in blue (positive) and red (negative). Figure was generated with GRASP.

Solution Structure of -Sarcin

Discussion Overall structure quality The three-dimensional solution structures of the cytotoxic ribonuclease a-sarcin determined here are well de®ned and show satisfactory values for their internal energy and for the energetic terms corresponding to NOE restraints. The number and magnitude of the distance-constraint violations are extremely low. More than 60 % of these violations Ê and nearly the 90 % are below are below 0.1 A Ê . Moreover, the average sum of the distance 0.2 A violations and the maximum distance violation are very small; this is especially signi®cant if we consider the protein's size and the total number of restraints used in the calculations. These distance violations are randomly distributed in the sequence, and only ®ve residues (Thr3, Thr5, Tyr24, Leu62 and His150) have an average sum Ê . The average pairwise backbone greater than 0.6 A Ê , dropping to just 0.50 A Ê when less RMSD is 0.86 A de®ned regions are excluded from the comparison. When side-chains are taken into account, the Ê for the whole protein RMSD increases to 1.47 A Ê with excluded variable regions. and to 1.00 A These relatively low values for a protein of this size indicate that a great number of side-chains have a well-de®ned conformation. We note that buried, hydrophobic side-chains are particularly well de®ned. The local structural precision, measured by the local RMSD values and angular order parameters, is also excellent. What is more important is that the structure is very well de®ned in spite of having a very low percentage of residues in regular secondary structure elements. It might have been supposed in principle that the long connecting loops would be very ¯exible. However, it turns out that these loops are on the whole as well de®ned as the secondary structural elements, perhaps with the exception of three segments (a total of 25 amino acid residues). The backbone of the loops consists chie¯y of three or four residues consecutive b-turns, and their side-chains stabilized by different kinds of long-range interactions. Structural basis for the electrostatic interactions The characterization of the spectroscopic properties and the determination of pKa values for a-sarcin's titratable groups have been reported (MartõÂnez del Pozo et al., 1988, PeÂrez-CanÄadillas et al., 1998). In these studies, some aspartate residues with low pKa values (<3) were suggested to contribute to the denaturating transition below pH 2.5. Now, a more detailed analysis of the particular interactions observed between side-chain groups (such as ionic pairs) can be made in the light of the re®ned structure. We shall focus on aspartate residues 41, 77, 91, 102 and 105. The side-chains of 77, 102 and 105 have intermediate

1067 solvent surface accessibilities (27 to 50 %) and form surface salt bridges with Lys70, His104 and His36, respectively. These three salt bridges could contribute to the protein's global stability and are surely responsible for the described perturbations from the nominal Asp pKa values. One salt bridge, 70 to 77, is located at the entrance of the small cavity detected within loop 2. In the crystal structure of the homolog restrictocin (Yang & Moffat, 1996), a chain of ordered water molecules is found in the interior of the corresponding loop, and interactions between the water molecules and side-chains might help stabilize this region. In a-sarcin, the Lys70-Asp77 bridge could act as a door to the cavity, which would open as a function of pH to allow communication between the bulk solvent and the hypothetical water molecule within the cavity. Experiments to detect this hypothetical structured water molecule in the cavity are in progress. In contrast with this surface interactions, Asp41 and Asp91 are buried (less than 7 % surface accessibility) within the protein. The carboxylate form of Asp41 (pKa < 3) is stabilized by the formation of an internal ion pair with His82's sidechain, which is also buried (solvent surface accessibility below 12 %) and also shows a perturbed pKa value (7.3). On the other hand, Asp91's side-chain (pKa <3) is distant from any charged group in the protein and the possibility that its low pKa is due to an ionic pair can be discarded. However, there is a hydrogen bond involving this carboxylate group and the amide group of Ser83 in half of the structures. This interaction likely affects the ionization constant of Asp91 and could also contribute to the rigidity of the last segment of loop 2. It is also known that the catalytic His137 has a depressed pKa value (5.8) (PeÂrez-CanÄadillas et al., 1998). The hydrogen bond found between the Hd1 of this imidazole ring and the carbonyl group oxygen atom of Gly143 (Figure 3(d)) could be one of the structural motifs that contributes to the low pKa. This hydrogen bond is also present in the restrictocin structure but absent in other microbial ribonucleases of known 3D structure (Pfeiffer et al., 1997; Noguchi et al., 1995). This hydrogen bond helps maintain the orientation of loop 5 which shields His137's side-chain from the solvent and de®nes a more hydrophobic environment for this residue compared to its functional equivalent in RNase T1. Comparison of a -sarcin with related structures The structure adopted by a-sarcin in solution is similar to the crystal structure of the homologous restrictocin (Yang and Moffat, 1996). In Figure 5, the lowest energy structure of a-sarcin is superimposed onto the restrictocin crystal structure. The restrictocin structure lacks electron density for residues in the central part of the N-terminal hairpin (11 to 16, Figure 5(b)), thus the structural comparison is limited to the rest of the structure. The global shape and the overall main-chain fold match clo-

1068

Solution Structure of -Sarcin

Figure 5. Structural comparison of a-sarcin and restrictocin. (a) Superposition of the lowest energy solution structure of a-sarcin (black, residues 1 to 150) and the crystal structure of restrictocin (RCSB PDB entry 1AQZ, chain A) (gray, residues 1 to 10, 17 to 149). (b) Detail of the superposition of the N-terminal hairpin; the discontinuity in the crystal structure of restrictocin is due to the lack of electron density for residues 11 to 16. (c) Superposition of loop 2 where the largest differences are found.

sely. The average RMSD between backbone atoms in the crystal structure compared to the 47 ®nal aÊ . When the more variable sarcin structures is 1.46 A regions (83 to 91 and 137 to 141) are excluded from Ê , and comparison, the RMSD goes down to 1.37 A when considering only regions of secondary strucÊ . It is clear that both proture, it descends to 0.93 A teins are very similar and that the largest structural differences are found mainly in loop regions, particularly in loop 2 (region 83 to 91) (Figure 5(c)) and loop 5. Both these structural regions correspond to stretches of high sequential variability in the Aspergillus ribotoxins. Intermolecular contacts in the restrictocin crystal could induce some rearrangements in loop 2 which shows a certain degree of ¯exibility in the a-sarcin structure. Even more substantial differences are observed in the orientation of loop 5. In the a-sarcin structure, this loop interacts with the N terminus by the salt bridge Lys11-Glu140. Glu140 is interesting from a protein stability view point, since it appears that this residue stabilizes the tertiary fold (loop 5-b-hairpin interaction) at the expense of the local stability of loop 5. The effect of Glu140 on the local and global stability could be studied by site-directed mutagenesis (Glu140 to Ala, Gly) in combination with hydrogen exchange measurements. In restrictocin, Lys11 and Glu140 changed to Leu10 and Gly139, respectively. The absence of the salt bridge probably affects the relative orientation of both loops in restrictocin and probably increases the mobility around the N-hairpin hinge region. This high level of mobility may explain the lack of electron density for this region in the crystal structure. RNase T1 (104 residues) (Arni et al., 1992; Pfeiffer et al., 1997) and RNase U2 (113 residues) (Noguchi et al., 1995) have the same structural topology as a-sarcin but have much shorter loops and

lack their larger cousin's special biological properties. Structurally, they may be considered as unelaborated versions of a-sarcin. The similarities in the sequence (ManchenÄo et al., 1995), basic fold and the ribonucleic activity (Weaver et al., 1985) of these proteins suggest that they evolved from a common ancestor. The superposition of secondary structural elements from a-sarcin with those of RNase U2 and T1 leads to backbone RMSD values Ê and 1.15 A Ê , respectively. The catalytic of 1.23 A residues of the three ribonucleases appear in equivalent structural positions. As for the differences between these structures, the N-terminal hairpin is very reduced in RNases T1 and U2, and in U2, the N terminus is extended by ®ve residues without a complementary increase in the length of the C terminus. Loop 2 appears very reduced in RNases T1 and U2, but some residues superimpose quite well with discontinuous segments of the a-sarcin loop. This could indicate that loop 2 of a-sarcin evolved by gene insertion and lengthening of a structure similar to that now possessed by the smaller microbial RNases. The other long loops 1 and 3 have similar lengths in the three proteins, but their residue composition and spatial orientation are different. Finally, the orientations of loop 5 in a-sarcin and the homologous loops in RNases T1 and U2 are different. The interaction between loop 5 and the N-terminal hairpin in a-sarcin is missing in RNases T1 and U2 and could explain this structural difference. Structural basis of biological activity General ribonucleolytic activity The high degree of structural homology shared by a-sarcin's and RNase T1's active sites is well

1069

Solution Structure of -Sarcin

documented (Campos-Olivas et al., 1996b), and it has been proposed that both proteins share a common ribonucleolytic mechanism (PeÂrez-CanÄadillas et al., 1998). The essential catalytic residues of asarcin are His50, Glu96 and His137, (Lacadena et al., 1999) and those of RNase T1 are His40, Glu58 and His92. Apart from these essential residues, three other side-chain groups (Tyr38, Arg77 and Phe100) in RNase T1 have been proposed to play a role in the stabilization of the different reaction intermediates (for a review, see Steyaert, 1997). Tyr38 and Arg77 in RNase T1 correspond to Tyr48 and Arg121 in a-sarcin; they occupy the same structural positions and likely perform an equivalent function. Phe100 has no equivalent in a-sarcin, and it had been suggested that this difference probably affects the electrostatic properties of the catalytic His137 (PeÂrez-CanÄadillas et al., 1998). Now, on the basis of the re®ned structure, some additional differences in the environment of the catalytic center, such as the orientation of loop 5, can be analyzed. In a-sarcin, loop 5 (Thr138Glu144) folds towards the active center, whereas in RNase T1 (Thr93- Asn99) it folds in the opposite direction. This different loop orientation has two important consequences for the catalytic histidine residues, His137 in a-sarcin and His92 in RNase T1. First, His137 is much more buried in a-sarcin than His92 in RNase T1 (%SA ˆ 7 % in a-sarcin and 22 % in RNase T1). Second, the hydrogen bond formed by the His137's side-chain with Gly143 in a-sarcin (Figure 3(d)) is not formed in RNase T1, so the hypothetical communication between the active site and the N-terminal hairpin via this hydrogen bond in a-sarcin would be impossible in RNase T1. The structural differences around the active centers also affect the electrostatic properties of these catalytic histidines. His137 in a-sarcin is more acidic (pKa ˆ 5.8) (PeÂrez-CanÄadillas et al., 1998) than His92 in RNase T1 (pKa ˆ 7.7) (Inagaki et al., 1981), which changes the activity pro®le and reduces the enzymatic ef®cacy. RNase T1 is a guanyl-speci®c ribonuclease. On the basis of the crystal structure, it has been proposed that the recognition of the purine base takes place by a complex network of hydrogen bonds involving residues Asn43, Asn44, Tyr45, Glu46 and Asn98 (Koepke et al., 1989). Of note is Glu46's side-chain, which has the appropriate orientation to interact with the guanine amino group by forming two hydrogen bonds. Sequence alignment suggests that a-sarcin and RNase T1 could recognize substrates in the same way, since they have similar residues in equivalent positions. The similarity observed at the sequence level, is, however missing in the structure. Asp57's side-chain (a-sarcin), which is suggested by the sequence alignment to act like Glu46 in RNase T1, points off in a different direction. This residue, therefore, can not recognize or discriminate the base. The structural difference arises from the different length of loop 2 in the two proteins. Loop 2 in RNase T1 is very

short, and just suf®cient to permit the backbone to link two consecutive b-strands, while, loop 2 in a-sarcin is very long. In joining the same structurally equivalent consecutive strands, loop 2 folds upon itself and the backbone repeatedly changes course. One of these turns projects the backbone and the side-chain of Asp57 outward, and away from the substrate. The lack of a residue in a-sarcin which pays the role of Glu46 in RNase T1 could explain in part the protein's different substrate speci®cities. Specific cytotoxic activity Perhaps the most interesting property of a-sarcin is its exquisite substrate speci®city. To identify the structural elements responsible for this speci®city, a model of the a-sarcin-rRNA complex was built on the basis of the a-sarcin structure and the crystal structure of a 20-mer RNA substrate analog (Correll et al, 1998) (Figure 6). To build the model, the scissile phosphodiester bond (between bases G4325 and A4324) was placed in the active site, close to the catalytic side-chains. Then, pivoting on this point, the RNA structure was rotated and translated to obtain the best ®t between both surfaces and to place the phosphate group in a position compatible with all NMR evidence found in the a-sarcin/20 GMP complex (data not shown).

Figure 6. Model of the interaction between the 28S rat SRD (Sarcin Ricin Domain) and a-sarcin. The protein is represented by the electrostatic surface and the RNA fragment with a stick model. The phosphodiester chain is colored in red, the recognition guanine G4310 in green and the bases adjacent to the scissile bond, A4324 and G4325, in yellow. Recognition sites and the N-terminal hairpin in the protein structure are labeled. Figure generated with GRASP.

1070 The resulting model is similar to that described for restrictocin (Yang & Moffat, 1996), and both the shape of the two molecular surfaces and their charges are highly compatible. The interactions that likely account for the substrate recognition include the lysine-rich region of a-sarcin's loop 3 (Ser110 to Lys114). This region would be able to interact with the negatively charged phosphodiester chain around the bulged guanine G4319 which is a key nucleotide for a-sarcin activity (Endo et al., 1983). The non-speci®c ribotoxic activity shown by the Lys111Gln mitogillin mutant (Kao & Davies, 1999) is consistent with this residue being involved in substrate recognition. It has been proposed recently that a-sarcin also recognizes the GAGA loop (Takeda et al. 1997). The model is also compatible with this proposal, since the GAGA loop lies near the a-sarcin's loop 5 and the base-recognition site, which have structural elements necessary for the speci®c recognition of adenine or guanine, or both. In spite of the limitations of our model of the complex (for example it is assumed that the molecules do not change conformation on binding), the above inferences seem reasonable considering all the data presently available. Used with caution, this model can aid the rational design of new variants of a-sarcin with enhanced or different biological functions.

Conclusions This work focused on understanding the structural requirements for the general ribonucleolytic and cytotoxic activities of the protein a-sarcin. With this aim, we have determined the solution structure of this enzyme to a very high resolution using NMR methods. The structure is well de®ned, not just in the secondary structural elements, but also in the long loops characteristic of cytotoxic ribotoxin structures. The generally high precision of the backbone is also observed for many sidechains. The data allow a detailed discussion at a residue level that is absolutely necessary to draw conclusions about how a-sarcin recognizes and cuts the rRNA. A docking model has been constructed with the protein and a substrate mimic. The molecular interface is highly complementary both in shape and charge, and the RNA scissile bond is within reach of the catalytic side-chains. Two different regions in the protein have been found to possess the correct geometry and electrostatic properties to interact with the target nucleic acid. Region 110 to 114 could interact electrostatically with the phosphate chain at the G4319. Distant from this region, two other segments formed by residues, 51 to 55 and loop 5 are able to interact speci®cally with rRNA structure near the scissile phosphodiester bond (A4324-G4325). These two Ê recognition regions, separated by more than 11 A in the protein structure, appear to account for a-sarcin's very high level of substrate speci®city.

Solution Structure of -Sarcin

Materials and Methods NMR samples Unlabeled recombinant a-sarcin was obtained and puri®ed as described (Lacadena et al., 1994; CamposOlivas et al. 1996a,b). [U-15N] and [U-20 % 13C] samples were obtained by growing E. coli RB791 cells in M9 minimal media, using either 15NH4Cl or 13C-glucose (20 %) as the primary source of 15N or 13C, respectively. Samples for NMR experiments were 1.0 ÿ1.5 mM. NMR spectra for natural isotopic abundance samples were acquired in either 2H2O (pH* 4.0) or 90 % H2O/10 % 2H2O (pH 6.0) at 40  C (2H2O sample) or 35  C (H2O sample). To measure 3JHN,aH coupling constants, NMR experiments were performed with a [U-15N] sample in 90 % H2O/10 % 2H2O at pH to 6.0 and 35  C, whereas the [U-20 % 13C] sample in 2H2O at pH* 3.2 and 35  C was used to obtain stereospeci®c assignments of Val, Leu and Ile methyl groups. Sodium trimethylsilyl (2,2,3,3,-2H4)propionate was used as the internal chemical shift reference. NMR spectroscopy All spectra were acquired in the phase-sensitive mode with time proportional phase incrementation mode (Marion & WuÈthrich, 1983) on a Bruker AMX-600 spectrometer equipped with a triple resonance probe and pulsed ®eld gradients. Pulse gradients were utilized for suppression of the water signal and undesired coherence pathways. The homonuclear and heteronuclear NMR experiments used for the structure calculation were performed as described (Campos-Olivas et al., 1996a,b; PeÂrez-CanÄadillas et al, 1998). Homonuclear 3JHN,aH coupling constants were measured with a modi®ed 15N-HSQC experiment based on the method described (Neri et al., 1990). All spectra were processed and analyzed using XWINNMR (Bruker, Rheinstetten), ANSIG (Kraulis, 1989) and XEASY (Bartels et al., 1995) software on an IRIS Indigo workstation. Structural restraints Two sources of experimental restraints consisting of upper limit distances and dihedral angle constraints were used as input for the structure calculations. Two 2D NOESY spectra with mixing times of 50 ms, recorded in different solvents (H2O, pH 6.0 at 35  C; and 2H2O, pH* 4.0 at 40  C) provided the NOE-derived upper limit restraints. Both spectra were integrated and the peak volumes and chemical shift lists were outputted to a DYANA-compatible format (GuÈntert et al., 1997). For each NOESY spectrum, the automatic calibration method implemented in DYANA was run to transform the peak volumes into distance constraints. The upper limits so obtained were used without modi®cation for the initial structure calculations. Cross-peak assignment in the NOESY spectra was carried out in several steps. First, a preliminary NOE data set was used to calculate initial structures with DYANA. The resulting conformers with lowest energies were then used to assign additional peaks in the NOESY spectra, and to obtain additional stereospeci®c assignments using the program GLOMSA (GuÈntert et al., 1991). This cycle of calculations and assignments was repeated until no further assignments were possible. The quality of the restrictions was checked by analyzing the restraint violations of the cal-

1071

Solution Structure of -Sarcin culated conformers. The NOE cross-peaks corresponding to restrictions that were consistently violated in a signi®cant number of structures were checked for possible overlap, and the corresponding restraints were consequently modi®ed. This cycle was repeated until no conÊ was detected. Only 7 % of sistent violation over 0.05 A the initial restraints were modi®ed. Obtaining f angle constraints from 3JHN,aH values is hindered by the fact that the Karplus equation (Karplus, 1963), which relates these two variables, is degenerate. Thus, multiple values of f are consistent with a single value of 3JHN,aH. Different strategies have been used to overcome this problem (Pfeiffer et al., 1997). Here, the very large number of total structural restraints permitted the following approach to discriminate between multiple solutions. First, preliminary structures calculated using only NOE-derived distance restraint were used to restrict the range of the f angle. Then, this restricted f angle range together with the sequential NOE intensities (HN(i)-Ha(i ÿ 1)), were used to discriminate between the multiple solutions of the Karplus equation. The uncertainty in the measurement of 3JHN,aH values is about 0.5 Hz. Structure calculations and analysis The 3D structure of a-sarcin was determined using the program DYANA (GuÈntert et al., 1997). According to the standard procedure, default weighting factors were attributed to the different restraints categories for the different types of protons. Pseudoatom corrections were added to those upper limit distances involving stereospeci®c atoms which were not stereospeci®cally assigned. The disulphide bridges linking Cys6 to Cys148 and Cys76 to Cys132 were included in the molecular topology. All peptide bonds were ®xed as trans, except for those preceding Pro49, Pro113 and Pro127, for which cis peptide bonds had been identi®ed from the inspection of their sequential NOE cross-peak intensities. All histidine, arginine and lysine residues were regarded as positively charged, while the glutamate and aspartate side-chains were treated as negatively charged. We have used a simulated annealing protocol in which 60 randomly generated structures are cooled down during 5000 torsion angle dynamics (TAD) cycles. Then, a second simulated annealing step of 800 TAD cycles was performed at a low temperature with the minimized structures. Following the torsion angle dynamics calculation, the 47 conformers with the lowest target-function values were subjected to restrained energy minimization using the GROMOS force ®eld (van Gunsteren & Berendsen, 1987). These energy-minimized structures are very similar to the input structures produced by DYANA, which indicates that the large number of experimental restraints allows only small structure rearrangements during minimization. The resulting 47 energy minimized conformers represent the solution structure of the cytotoxic ribonuclease a-sarcin. The program MOLMOL (Koradi et al., 1996) was used to visualize and globally characterize the structures. The electrostatic potential surface and other surface characteristics of a-sarcin were calculated using the program GRASP (Nicholls et al., 1991). The quality of the structures was evaluated with the program PROCHECKNMR (Laskowski et al., 1996) and the secondary structural elements were analyzed using the program PROMOTIF (Hutchinson & Thornton, 1996). All RMSD values reported here are mean pairwise RMSD. Global RMSD

values are obtained from the best superposition of all residues, whereas local RMSD values result from superposing tripeptide segments.

Acknowledgments We thank Dr D. V. Laurents and Dr J. L. Neira for helpful discussions and suggestions. This work was supported by the DireccioÂn General de InvestigacioÂn Cientõ®ca y TeÂcnica (Spain) (PB93-0189) and by the DireccioÂn General de EnsenÄanza Superior (Spain) (PB96-0601).

References Arni, R., Heinemann, U., Tokuoka, R. & Saenger, W. (1988). Three-dimensional structure of the riboÊ resolution. nuclease T1 20 -GMP complex at 1.9-A J. Biol. Chem. 263, 15358-15368. Arni, R. K., Pal, G. P., Ravichandran, K. G., Tulinsky, A., Walz, F. G., Jr. & Metcalf, P. (1992). Threedimensional structure of Gln25-Ribonuclease T1 Ê resolution: structural variations at the base 1.84 A recognition and catalytic sites. Biochemistry, 31, 3126-3135. Arruda, L. K., Mann, B. J. & Chapman, M. D. (1992). Selective expression of a major allergen and cytotoxin, Asp fI in Aspergillus fumigatus. J. Immunol. 149, 3354-3359. Bartels, C., Xia, T., Billeter, M., GuÈntert, P. & WuÈthrich, K. (1995). The program XEASY for computersupported NMR spectral analysis of biological macromolecules. J. Biomol. NMR, 6, 1-10. Campos-Olivas, R., Bruix, M., Santoro, J., MartõÂnez del Pozo, A., Lacadena, J., Gavilanes, J. G. & Rico, M. (1996a). 1H and 15N nuclear magnetic resonance assignment and secondary structure of the cytotoxic ribonuclease a-sarcin. Protein Sci. 5, 969-972. Campos-Olivas, R., Bruix, M., Santoro, J., MartõÂnez del Pozo, A., Lacadena, J., Gavilanes, J. G. & Rico, M. (1996b). Structural basis for the catalytic mechanism and substrate speci®city of the ribonuclease a-sarcin. FEBS Letters, 399, 163-165. Correll, C. C., Munishkin, A., Chan, Y. L., Ren, Z., Wool, I. G. & Steitz, T. A. (1998). Crystal structure of the ribosomal RNA domain essential for binding elongation factors. Proc. Natl Acad. Sci. USA, 95, 13436-13441. Endo, Y. & Wool, I. G. (1982). The site of action of a-sarcin on eukaryotic ribosomes. The sequence at the asarcin cleavage site in 28 S ribosomal ribonucleic acid. J. Biol. Chem. 257, 9054-9060. Endo, Y., Hubber, P. W. & Wool, I. G. (1983). The ribonuclease activity of the cytotoxin a-sarcin. The characteristics of the enzymatic activity of a-sarcin with ribosomes and ribonucleic acids as substrates. J. Biol. Chem. 258, 2662-2667. FernaÂndez-Luna, J. L., LoÂpez-OtõÂn, C., Soriano, F. & MeÂndez, E. (1985). Complete amino acid sequence of the Aspergillus cytotoxin mitogillin. Biochemistry, 24, 861-867. FernaÂndez-Puentes, C. & Carrasco, L. (1980). Viral infection permeabilizes mammalian cells to protein toxins. Cell, 20, 769-775. Gasset, M., OnÄaderra, M., Thomas, P. G. & Gavilanes, J. G. (1990). Fusion of phospholipid vesicles pro-

1072 duced by the anti-tumour protein a-sarcin. Biochem. J. 265, 815-822. Gasset, M., OnÄaderra, M., Thomas, P. G. & Gavilanes, J. G. (1991). Acid phospholipid vesicles produce conformational changes on the antitumor protein a-sarcin. Biochem. Biophys. Acta, 1080, 51-58. Gasset, M., ManchenÄo, J. M., Lacadena, J., Turnay, J., Olmo, N., Lizarbe, M. A., MartõÂnez del Pozo, A., OnÄaderra, M. & Gavilanes, J. G. (1994). a-Sarcin, a ribosome-inactivating protein that translocates across the membrane of phospholipid vesicles. Curr. Top. Pept. Protein. Res. 1, 99-104. Gallivan, J. P. & Dougherty, D. A. (1999). Cation-pi interactions in structural biology. Proc. Natl Acad. Sci. USA, 96, 9459-9464. GuÈntert, P., Braun, W. & WuÈthrich, K. (1991). Ef®cient computation of three-dimensional protein structures in solution from nuclear magnetic resonance data using the program DIANA and the supporting programs CALIBA, HABAS and GLOMSA. J. Mol. Biol. 217, 517-530. GuÈntert, P., Mumenthaler, C. & WuÈthrich, K. (1997). Torsion angle dynamics for NMR structure calculation with the new program DYANA. J. Mol. Biol. 273, 283-298. Hebert, E. J., Giletto, A., Sevick, J., Urbanikova, L., Wilson, K. S., Dauter, Z. & Pace, C. N. (1998). Contribution of a conserved asparagine to the conformational stability of ribonucleases Sa, Ba and T1. Biochemistry, 37, 16192-16200. Hutchinson, E. G. & Thornton, J. M. (1996). PROMOTIF: a program to identify and analyze structural motifs in proteins. Protein Sci. 5, 212-220. Inagaki, F., Kawano, Y., Shimada, I., Takahashi, K. & Miyazawa, T. (1981). Nuclear magnetic resonance study on the microenvironments of histidine residues of ribonuclease T1 and carboxymethylated ribonuclease T1. J. Biochem. 89, 1185-1195. Karplus, M. (1963). Vicinal proton coupling in nuclear magnetic resonance. J. Am. Chem. Soc. 85, 2870-2871. Kao, R. & Davies, J. (1999). Molecular dissection of mitogillin reveals that the fungal ribotoxins are a family of natural genetically engineered ribonucleases. J. Biol. Chem. 274, 12576-12582. Koepke, J., Maslowska, M., Heinemann, U. & Saenger, W. (1989). Three-dimensional structure of ribonuclease T1 complexed with guanylyl-20 ,50 -guanosine Ê resolution. J. Mol. Biol. 206, 475-88. at 1.8 A Koradi, R., Billeter, M. & WuÈthrich, K. (1996). MOLMOL: a program for display and analysis of macromolecular structures. J. Mol. Graph. 14, 51-55. Kraulis, P. J. (1989). ANSIG: a program for the assignment of protein 1H 2D NMR spectra by interactive computer graphics. J. Magn. Reson. 24, 627-633. Lacadena, J., MartõÂnez del Pozo, A., Barbero, J. L., ManchenÄo, J. M., Gasset, M., OnÄaderra, M., LoÂpezOtõÂn, C., Ortega, S., GarcõÂa, J. & Gavilanes, J. G. (1994). Overproduction and puri®cation of biologically active native fungal a-sarcin. Gene, 142, 147151. Lacadena, J., MartõÂnez del Pozo, A., MartõÂnez-Ruiz, A., PeÂrez-CanÄadillas, J. M., Bruix, M., ManchenÄo, J. M., OnÄaderra, M. & Gavilanes, J. G. (1999). Role of histidine-50, glutamic acid-96 and histidine-137 in the ribonucleolytic mechanism of the ribotoxin a-sarcin. Proteins: Struct. Funct. Genet. 37, 474-484. Laskowski, R. A., Rullmann, J. A. C., MacArthur, M. W., Kaptein, R. & Thornton, J. M. (1996). AQUA and PROCHECK-NMR: programs for checking the qual-

Solution Structure of -Sarcin ity of protein structures solved by NMR. J. Biomol. NMR, 8, 477-486. LoÂpez-OtõÂn, C., Barber, D., FernaÂndez-Luna, J. L., Soriano, F. & MeÂndez, E. (1984). The primary structure of the cytotoxin restrictocin. Eur. J. Biochem. 143, 621-634. ManchenÄo, J. M., Gasset, M., Lacadena, J., MartõÂnez del Pozo, A., OnÄaderra, M. & Gavilanes, J. G. (1995). Predictive study of the conformation of the cytotoxic protein a-sarcin: a structural model to explain a-sarcin-membrane interaction. J. Theor. Biol. 172, 259-267. Marion, D. & WuÈthrich, K. (1983). Application of phasesensitive two-dimensional correlated spectroscopy (COSY) for measurement of proton-proton spinspin coupling constants. Biochem. Biophys. Res. Commun. 113, 967-974. MartõÂnez del Pozo, A., Gasset, M., OnÄaderra, M. & Gavilanes, J. G. (1988). Conformational study of the antitumor protein a-sarcin. Biochim. Biophis. Acta, 953, 280-288. Neri, D., Szyperski, T., Otting, G., Senn, H. & WuÈthrich, K. (1989). Stereospeci®c nuclear magnetic resonance assignments of methyl groups of valine and leucine in the DNA-binding domain of the 434 repressor by biosynthetically directed fractional 13C labeling. Biochemistry, 28, 7510-7516. Neri, D., Otting, G. & WuÈthrich, K. (1990). New nuclear magnetic resonance experiment for measurements of the vicinal coupling constants 3JHNa in proteins. J. Am. Chem. Soc. 112, 3663-3665. Nicholls, A., Sharp, K. & Honig, B. (1991). Protein folding and association: insights from the interfacial and thermodynamic properties of hydrocarbons. Proteins: Struct. Funct. Genet. 11, 281-296. Noguchi, S., Satow, Y., Uchida, T., Sasaki, C. & Matsuzaki, T. (1995). Crystal structure of Ustilago Ê resolution. sphaerogena ribonuclease U2 at 1.8 A Biochemistry, 34, 15583-15591. Olson, B. H. & Goerner, G. L. (1965). Alpha sarcin, a new antitumor agent. I. Isolation, puri®cation, chemical composition, and the identity of a new amino acid. Appl. Microbiol. 13, 314-321. Olson, B. H., Jennings, J. C., Roga, V., Junek, A. J. & Schuurmans, D. M. (1965). Alpha sarcin, a new antitumor agent. II. Fermentation and antitumor spectrum. Appl. Microbiol. 13, 322-326. Otero, M. J. & Carrasco, L. (1988). Proteins are cointernalized with virion particles during early infection. Virology, 160, 75-80. Parente, D., Raucci, G., Celano, B., Pacilli, A., Zanoni, L., Canevari, S., Adobati, E., Colnaghi, M. I., Dosio, F., Arpicco, S. & De Santis, R. (1996). Clavin, a type-1 ribosome-inactivating protein from Aspergillus calvatus IFO 8605. cDNA isolation, heterologous expression, biochemical and biological characterization of the recombinant protein. Eur. J. Biochem. 239, 272-280. PeÂrez-CanÄadillas, J. M., Campos-Olivas, R., Lacadena, J., MartõÂnez del Pozo, A., Gavilanes, J. G., Santoro, J., Rico, M. & Bruix, M. (1998). Characterization of pKa values and titration shifts in the cytotoxic ribonuclease a-sarcin by NMR. Electrostatic interactions, structure and catalytic function. Biochemistry, 37, 15865-15876. Pfeiffer, S., Karimi-Nejad, Y. & RuÈterjans, H. (1997). Limits of NMR structure determination using variable target function calculations: ribonuclease T1, a case-study. J. Mol. Biol. 266, 400-423.

Solution Structure of -Sarcin Sevcik, J., Dodson, E. J. & Dodson, G. G. (1991). Determination and restrained least-squares re®nement of the structures of ribonuclease Sa and its complex Ê resolution. Acta Cryswith 30 -guanylic acid at 1.8 A tallog. sect. B, 47, 240-253. Scrutton, N. S. & Raine, A. R. (1996). Cation-p bonding and amino-aromatic interactions in the biomolecular recognition of substituted ammonium ligands. Biochem. J. 319, 1-8. Steyaert, J. (1997). A decade of protein engineering on ribonuclease T1. Atomic dissection of the enzymesubstrate interaction. Eur. J. Biochem. 247, 1-11. Takeda, E., Bi, X., Yoshinari, S. & Endo, Y. (1997). Mechanism of substrate recognition by the ribotoxin alpha-sarcin. Nucl. Acids Symp. Ser. 37, 131-132. Turnay, J., Olmo, N., JimeÂnez, A., Lizarbe, M. A. & Gavilanes, J. G. (1993). Kinetic study of the cytotoxic effect of a-sarcin, a ribosome-inactivating protein from Aspergillus giganteus, in tumor cell lines: protein biosynthesis inhibition and cell binding. Mol. Cell. Biochem. 122, 39-47. van Gunsteren, W. F. & Berendsen, H. J. C. (1987). Groningen Molecular Simulation (GROMOS) Library Manual, Biomos, Groningen. Wawrzynczak, E. J., Henry, R. V., Cumber, A. J., Parnell, G. D., Derbyshire, E. J. & Ulbrich, N.

1073 (1991). Biochemical, cytotoxic and pharmacokinetic properties of an immunotoxin composed of a mouse monoclonal antibody Fib75 and the ribosome-inactivating protein a-sarcin from Aspergillus giganteus. Eur. J. Biochem. 196, 203-209. Weaver, L. H., Grutter, M. G., Remington, S. J., Gray, T. M., Isaacs, N. W. & Matthews, B. W. (1985). Comparison of goose-type, chicken-type, and phage-type lysozymes illustrates the changes that occur in both amino acid sequence and threedimensional structure during evolution. J. Mol. Evolution, 21, 97-111. Wirth, J., MartõÂnez del Pozo, A., ManchenÄo, J. M., MartõÂnez-Ruiz, A., Lacadena, J., OnÄaderra, M. & Gavilanes, J. G. (1997). Sequence determination and molecular characterization of gigantin, a cytotoxic protein produced by the mould Aspergillus giganteus IFO 5818. Arch. Biochem. Biophys. 343, 188-193. Wool, I. G. (1984). The mechanism of action of the cytotoxic ribonuclease a-sarcin and its use to analyze ribosome structure. Trends Biochem. Sci. 9, 14-17. Yang, X. & Moffat, K. (1996). Insights into speci®city of cleavage and mechanism of cell entry from the crystal structure of the highly speci®c Aspergillus ribotoxin restrictocin. Structure, 4, 837-852.

Edited by M. F. Summers (Received 22 February 2000; received in revised form 20 April 2000; accepted 25 April 2000)