Crystal structure at 1.5-A resolution of Pyrus pyrifolia pistil ribonuclease responsible for gametophytic self-incompatibility.

The crystal structure of the Pyrus pyrifolia pistil ribonuclease (S(3)-RNase) responsible for gametophytic self-incompatibility was determined at 1.5-A resolution. It consists of eight helices and seven beta-strands, and its folding topology is typical of RNase T(2) family enzymes. Based on a structural comparison of S(3)-RNase with RNase Rh, a fungal RNase T(2) family enzyme, the active site residues of S(3)-RNase assigned were His(33) and His(88) as catalysts and Glu(84) and Lys(87) as stabilizers of an intermediate in the transition state. Moreover, amino acid residues that constitute substrate binding sites of the two RNases could be superimposed geometrically. A hypervariable (HV) region that has an S-allele-specific sequence comprises a long loop and short alpha-helix. This region is far from the active site cleft, exposed on the molecule's surface, and positively charged. Four positively selected (PS) regions, in which the number of nonsynonymous substitutions exceeds that of synonymous ones, are located on either side of the active site cleft, and accessible to solvent. These structural features suggest that the HV or PS regions may interact with a pollen S-gene product(s) to recognize self and non-self pollen.

Many flowering plants have a self-incompatibility system that recognizes the self or non-self between the pistil and pollen (tube) after pollination and suppresses growth of the self-pollen tube to prevent self-fertilization (1,2). Gametophytic self-incompatibility (GSI) 1 is controlled genetically by a single locus (S-locus) with multiple alleles (1,2). When a pollen grain lands on a stigma of the pistil, a process that discriminates as to whether an S-allele of the pollen matches one of the two S-alleles of the pistil takes place. The pollen grain germinates on the stigma and grows into the style toward the embryo. If its S-allele matches one of the two S-alleles of the pistil, pollen tube growth is arrested in the style, and no fertilization takes place. In solanaceous, scrophulariaceous, and rosaceous plants that have GSI, the pistil glycoproteins that cosegregate with the S-alleles have been identified as ribonucleases of the RNase T 2 family (S-RNase) (3). McClure et al. (4) reported that pollen rRNA is degraded after self-pollination but not after crosspollination and suggested that GSI expression is mediated by degradation of the pollen rRNA of self-pollen tubes by S-RNase, leading to depletion of protein biosynthesis and the eventual arrest of tube growth. S-RNase has been confirmed necessary for GSI from results of gain-of-function and loss-of-function experiments on transgenic plants of solanaceous species (5,6). Transgenic experiments also have shown that the RNase activity of S-RNase is necessary for GSI (7), which in petunia the carbohydrate moiety is not responsible for GSI (8), and that mutant S-RNase, which has lost RNase activity, acts as a dominant negative for GSI (9).
Based on these findings, two models have been proposed to explain S-allele-specific inhibition of pollen tube growth; the receptor and inhibitor models (10). In the receptor model, the pollen S-gene product(s) is(are) the receptor that incorporates the S-RNase that matches the pollen S-allele. In the inhibitor model, S-RNase enters the pollen tube nonspecifically and is inhibited, except for the enzyme that has the same pollen S-allele. Recently, immunocytochemical research has shown that S-RNase enters and is accumulated in the cytoplasm of all pollen tube haplotypes without S-allele specificity, thus experimentally supporting the inhibitor model (11). The function of the GSI mechanism, however, is still not clear, because no pollen S-gene products that interact with S-RNase and cause self and non-self recognition between the pollen and pistil have yet been found.
Amino acid sequence analyses of solanaceous S-RNases have identified two hypervariable regions, HVa and HVb, with extremely high levels of divergence between allelic sequences; candidates for sequences that would be recognized as S-allelespecific by pollen (12). In fact, the S-allele specificity of the Solanum chacoense S 11 -RNase was changed to the S 13 -allele by swapping the HVa and HVb domains (13), and mutation in the HVa and HVb regions of S 11 -RNase produced a new S-RNase with dual specificity, S 11 -and S 13 -alleles (14). In contrast, only one hypervariable region (HV), corresponding to the HVa of the solanaceous S-RNase, has been detected in sequences of rosaceous S-RNases (15,16), the HVb region apparently being deleted (16,17). This structural feature suggests that the recognition mechanism of S-allele specificity may differ for solanaceous and rosaceous S-RNases.
Knowledge of the three-dimensional structure is required for further investigation of the recognition mechanism of rosaceous S-RNases at the molecular level. We therefore made a three-dimensional structural analysis of the S-RNases from Pyrus pyrifolia (Japanese pear), a member of the Rosaceae family. Seven S-RNases (S 1 -through S 7 -RNase) have been identified in and purified from the styles of P. pyrifolia (15,18). From these we chose S 3 -RNase for this x-ray crystallographic analysis, because it has only two N-glycosylation sites (19), and its amino acid sequence is highly homologous to that of P. pyrifolia S 5 -RNase (95.5% identity) (15). Identification of the three-dimensional structures of the S 3 -and S 5 -RNases will pave the way for a structural comparison that provides high resolution information on molecular recognition between these S-RNases and the pollen S-gene product(s). We report the crystal structure of P. pyrifolia S 3 -RNase at 1.5-Å resolution and discuss the recognition site(s) present on S 3 -RNase and the structural basis of its enzymatic activity. This is the first report on the three-dimensional structure of a rosaceous S-RNase.

EXPERIMENTAL PROCEDURES
Crystallization and Data Collection-P. pyrifolia S 3 -RNase was purified and crystallized as described previously (20). The crystals belong to the P2 1 space group with unit cell dimensions of a ϭ 45.65, b ϭ 52.59, c ϭ 47.57 Å, and ␤ ϭ 106.45° (20). Diffraction data were collected with an image plate detector Raxis IV (Rigaku) at the beam line of BL40B2 at SPring-8 to 1.5-Å resolution. Image data were processed by the programs DENZO and SCALEPACK (21) (Table I). Multiple isomorphous replacement was used to determine the crystal structure of S 3 -RNase. Two ethylmercurithiosalicylic acid derivatives (Hg1, Hg1Ј), a mercury(II) ammonium thiocyanate derivative (Hg2), and a lead(II) acetate derivative (Pb) were prepared by soaking the crystals in heavy atom reagents (Table I). All diffraction data on the derivative crystals were collected at the Photon Factory with a charge-coupled device detector Quantum 4R (ADSC) at the beam line of BL6A and processed by the MOSFLM (22) and SCALA (23) programs.
Crystal Structure Determination-Patterson maps were automatically interpreted by the program SOLVE (24) to locate the heavy atom coordinates of each derivative. By use of the solution coordinates, MIR phase calculation was done at 2.5-Å resolution by the maximum-likelihood refinement program SHARP (25) ( Table I). The experimental MIR phases were improved by the DM (23) and wARP (26) programs. First, the phases at 2.5-Å resolution gradually were extended to 1.7-Å resolution by DM with solvent flattening and histogram matching. The phases obtained were further refined and extended to 1.5-Å resolution by weighted averaging of multiple-refined dummy atomic models implemented in wARP. This technique for averaging the structure factors was very effective for improving the phases and provided substantial improvement in the resulting electron density map. The subsequent main-chain trace was done by wARP. Based on the automatically traced main chain coordinates, 200 amino acid residues were successfully built by program O (27). X-ray crystallographic refinement was done by the CNS program (28). One cycle of simulated annealing refinement, followed by positional and temperature factor refinements, was repeated several times. Throughout the refinements bulk solvent correction was applied, and the resolution range was gradually extended from 2.3 to 1.5 Å. At each stage of crystallographic convergence, the model was carefully checked and rebuilt into the simulated annealed omit map calculated by CNS. Solvent water molecules were picked up from a difference Fourier map by use of the automated scripts implemented in CNS. The lower cut-off level for picking up the solvent water molecules in the residual electron densities was set at 3. High resolution model coordinates of sugar chains were obtained from the Uppsala web server, HIC-Up (xray.bmc.uu.se/hicup/) (29). The crystal structure was refined to the respective R-and free R-factors of 17.2 and 20.2% in the resolution range of 500 -1.5 Å. The final model has 1643 protein atoms, 59 sugar atoms, and 266 water atoms. A Ramachandran plot from the program PROCHECK (30) shows that 91.5% of the residues are in the most favorable region, 8.5% in the additional allowed region, and none in the generally allowed and disallowed regions. The coordinates have been deposited in the Protein Data Bank with the accession code 1IQQ.

RESULTS
Overall Structure-The crystal structure of P. pyrifolia S 3 -RNase was determined by the multiple isomorphous replacement (MIR) and refined to an R-factor of 17.2% at 1.5-Å resolution. Table I summarizes the data collected and gives the final refinement statistics. The geometry of the current model is such that the root mean square deviations from the ideal values are 0.022 Å for bond length and 2.02°for bond angles.
A stereo view of the three-dimensional structure of S 3 -RNase is shown in Fig. 1. The molecule's size is ϳ40 ϫ 50 ϫ 30 Å. This protein has an ␣ ϩ ␤ type structure consisting of eight helices (six ␣-helices (␣1, ␣2, ␣4, and ␣6-␣8), two 3 10 -helices (␣3 and ␣5)), and seven ␤-strands (␤1-␤7) (Fig. 1a). The folding topology of its main chain is very similar to the topologies of the RNase T 2 family enzymes: the RNase Rh from Rhizopus niveus (31, 32), RNase LE from cultured tomato cells (33), and RNase MC1 from bitter gourd seeds (34), all of whose tertiary structures have been determined (Fig. 1b). Two secondary structural element characteristics are present in S 3 -RNase; the length of the ␣2-helix is shorter than the lengths of the other RNases, and a very short 310-helix (␣3), Arg 74 -Lys 76 , not seen in the RNase T 2 family, is present (Fig. 2). The electron density map for S 3 -  (32), RNase LE in yellow (33), and RNase MC1 in green (34). c, stereo drawing of the space filling model of S 3 -RNase. Amino acid residues that constitute the P 1 , B 1 , and B 2 sites (46), respectively, are colored red, blue, and green. Orientations of the S 3 -RNase molecules in a, b, and c are identical. a and b were prepared by the programs MOLSCRIPT (54) and RASTER3D (55); c was prepared by the program GRASP (56).
Catalytic Site-The structure of S 3 -RNase was compared in detail with that of RNase Rh to search for catalytic site residues, because many RNase Rh studies on the structure-function relationship of this RNase T 2 family enzyme have been done by chemical modification (37)(38)(39), site-directed mutagenesis (40 -45), and x-ray crystallographic analysis (31,32). The main-chain frameworks, including the catalytic P 1 site residues of RNase Rh (␤2 and ␣4), were well superimposable on those of S 3 -RNase (Fig. 1b), indicating that their P 1 site topologies are similar. The P 1 site is here defined according to the nomenclature of the subsites of RNase A given by Richards and Wyckoff (46).
As shown in Fig. 3a, the His 33 and His 88 of S 3 -RNase were superimposable geometrically on the general acid and base catalysts His (46) and His (109) of RNase Rh (amino acids numbered in parentheses are those of RNase Rh) (32,38,41), suggesting that His 33 and His 88 are the catalysts in S 3 -RNase. Glu 84 and Lys 87 are the respective counterparts of the Glu (105) and Lys (108) of RNase Rh, which are considered to stabilize a pentacovalent intermediate of substrate RNA in the transition state (32,33,(42)(43)(44). Glu 84 and Lys 87 therefore may have the same respective roles as Glu (105) and Lys (108) . Trp 36 is the counterpart of the Trp (49) of RNase Rh (Fig. 3a), which functions in the fixation of catalytically important Glu (105) and His (109) . Trp (49) N ⑀1 forms a hydrogen bond with the ␥-carboxyl group of Glu (105) , and the indole ring of Trp (49) has a partial stacking interaction with the imidazole ring of His (109) (32). In S 3 -RNase, a similar stacking interaction between Trp 36 and His 88 is possible, because the distance between the indole ring of Trp 36 and the imidazole ring of His 88 is about 3.50 Å, and both rings are nearly parallel (Fig. 3a). Trp 36 N ⑀1 cannot, however, interact directly with Glu 84 O ⑀1 by hydrogen bonding, because the distance between the two atoms is 4.59 Å. Possibly, they may interact via a water molecule. Indeed, a water molecule that can hydrogen bond to both Trp 36 N ⑀1 and Glu 84 O ⑀1 is present in the P 1 pocket (Fig. 3a).
Substrate Binding Sites-Two pockets, which correspond to the base binding B 1 and B 2 sites of RNase Rh (32,33), are present on either side of the P 1 site in S 3 -RNase (Fig. 1c). These B 1 and B 2 sites also have the nomenclature given by Richards and Wyckoff (46). The amino acid residues that constitute these sites could be superimposed geometrically on S 3 -RNase and RNase Rh (Fig. 3, b and c).
As for the B 1 site of S 3 -RNase, Trp 36 , Ser 38 , and Asp 44 are the respective counterparts of the Trp (49) , Asp (51) , and Tyr (57) of RNase Rh (Fig. 3b). The Trp (49) of RNase Rh has two important functions; to fix the active side residues (described above) and to recognize the base of the substrate RNA by an aromaticstacking interaction together with Tyr (57) (called double-sided stacking) (32,33). Trp 36 may bind to the base of the substrate in the same manner as Trp (49) , but the position corresponding to Tyr (57) is occupied by Asp 44 , consequently double-sided stacking recognition is impossible in S 3 -RNase (Fig. 3b). The base therefore probably is recognized at the B 1 site of S 3 -RNase by a single-sided stacking interaction with Trp 36 , unlike RNase Rh and RNase LE (32,33). X-ray crystallographic data on the RNase Rh/2Ј-AMP complex shows that Asp (51) contributes to the adenylic acid preference at the B 1 site of RNase Rh by hydrogen bonding to the adenine base of 2Ј-AMP (32,33). The position corresponding to Asp (51) , however, is occupied by Ser 38 in S 3 -RNase (Fig. 3b). Experiments on substrate specificity done with dinucleotide monophosphate showed that the base specificity at the B 1 site of S 3 -RNase is wider than that of RNase Rh (data not shown). This extended base specificity may be due to substitution of Asp (51) with Ser 38 , to single-sided stacking with Trp 36 , or both.
The B 2 site of S 3 -RNase seems to be comprised of Gln 9 , Pro 69 , Asn 70 , Val 71 , Phe 72 , and Phe 80 (Fig. 3c), which geometrically correspond to the Gln (32) , Pro (92) , Ser (93) , Asn (94) , Gln (95) , and Phe (101) of RNase Rh that have been identified as components of the B 2 site based on x-ray crystallographic data for the FIG. 2. Sequence alignment of RNase T 2 family enzymes and their secondary structural elements. Amino acid residue numbering is based on that for P. pyrifolia S 3 -RNase. Amino acid residues that form secondary structures are the ␣-helix (red), 3 10 -helix (red, slanted), and ␤-strand (blue). The secondary S 3 -RNase structures were defined by the program DSSP (57). complex with d(ApC) (33). In RNase Rh, the base is thought to be recognized at the B 2 site by an aromatic stacking interaction with Phe (101) and van der Waals contact with the side chains of Asn (94) and Gln (95) (33). Similar recognition should occur at the B 2 site of S 3 -RNase, because Val 71 , Phe 72 , and Phe 80 , respec-tively, can be superimposed geometrically on Asn (94) , Gln (95) , and Phe (101) (Fig. 3c).
Hypervariable Region-One hypervariable (HV) region with an allele-specific sequence is present in the rosaceous S-RNases (Fig. 4a) and is a candidate for the recognition site of pollen S-gene product(s) (15). The HV region of S 3 -RNase (Pro 49 -Gln 63 ) is made up of a loop (Pro 49 -Glu 57 ) and half of an ␣2-helix (Lys 58 -Gln 63 ) and is exposed on the molecule's surface (Figs. 1a and 4b). Although the exposed loop appears to be somewhat flexible, three of its nine amino acid residues (Asn 52 , Arg 54 , and Arg 56 ) interact with the amino acid residues inside the molecule by hydrogen bonding (Fig. 5) (Fig. 5).
In contrast, the half of the ␣2-helix (Lys 58 -Gln 63 ) located within the HV region has six amino acid residues. Of these, the side chains of Leu 60 and Glu 61 interact with other amino acids (Fig. 5). The side chain of Leu 60 is packed into a hydrophobic space composed of the side chains of Leu 64 , Asn 100 , Phe 103 , and Ile 107 and the main chain of Arg 56 -Glu 57 . The O ⑀1 and O ⑀2 of Glu 61 , respectively, are hydrogen-bonded to Arg 56 N ⑀ and Asn 78 N ␦2 (Fig. 5). These described interactions may fix and stabilize the conformations of the loop and helix; i.e. the HV region.
This HV region is comprised of 15 amino acid residues, ten of which, Pro 49 , Ile 50 , Lys 51 , Ile 53 , Lys 55 , Glu 57 , Lys 58 , Leu 59 , His 62 , and Gln 63 , are widely exposed to solvent. The side chains of Ile 53 , Lys 55 , Glu 57 , Lys 58 , Leu 59 , and His 62 , in particular, prominently extend into the solvent (Fig. 5). Given that the HV region is in contact with the pollen factor(s), these residues may have importance in pollen factor binding.
Positively Selected Regions-Recognition sites in some proteins (e.g. the major histocompatibility complex (47)(48)(49), antigenic surface proteins of parasites and viruses (49), and acrosomal proteins of the abalone (50,51)) are reported to be regions in which the number of nonsynonymous nucleotide substitutions (dN) exceeds that of synonymous substitutions (dS), and positive selection probably takes place in these regions. Window analysis of the dS and dN in rosaceous S-RNases detected four regions with an excess of dN over dS, in which positive selection may operate (PS1-PS4) (Fig. 4) (52). PS1 nearly duplicates the HV region. It is interesting that the four positively selected (PS) regions were detected by window analysis of the rosaceous S-RNase genes but that there is only a single region with an allele-specific amino acid sequence. PS1 (HV) and PS2, as well as PS3 and PS4, respectively, are close in the three-dimensional structure of S 3 -RNase, on either side of the active site cleft (including the P 1 , B 1 , and B 2 sites) and accessible to solvent (Figs. 1c and 4b).
Sugar Chains-P. pyrifolia S 3 -RNase has two potential Nglycosylation sites, Asn 18 and Asn 116 (Fig. 1a). The Asn 18 site is specific to the S 3 -and S 5 -RNases, whereas the Asn 116 site is conserved among all the rosaceous S-RNases (Fig. 4a). A twodimensional sugar map and mass spectrometry of fragmented peptides showed that both sites have heterogeneous N-glycans, Asn 18 mainly glycosylated by a chitobiose (GlcNAc␤1-4GlcNAc) and Asn 116 primarily occupied by a xylomannose type sugar chain (19). The core structure, Man␤1-4GlcNAc␤1-4GlcNAc, of the xylomannose type sugar chain at Asn 116 , clearly visible in the difference Fourier map, was modeled to fit the densities (data not shown). No clear electron densities corresponding to the chitobiose at Asn 18 and the mannose and xylose moieties of the sugar chain at Asn 116 were observed, indicating that their conformations are highly disordered. Because Asn 18 and Asn 116 , respectively, are located at the end of and on the opposite side of the active site cleft (Fig. 1, a and c), the S 3 -RNase sugar chains probably are not involved in its enzymatic properties and the recognition of the self and non-self. DISCUSSION The structure of the active site (the catalytic P 1 site and substrate binding B 1 and B 2 sites) of S 3 -RNase is typical of the structures of the RNase T 2 family enzymes. Probably, His 33 and His 88 function as general acid and base catalysts, and Glu 84 and Lys 87 stabilize the pentacovalent intermediate in the transition state. One marked difference between the P 1 sites of S 3 -RNase and RNase Rh is that the imidazole ring of His 33 is rotated about 90°from that of His (46) (Fig. 3a), even though the distance between His 33 N ⑀2 and His 88 N ⑀2 (7.15 Å) is nearly equal to that between His (46) N ⑀2 and His (109) N ⑀2 (6.71 Å) (32). Because S 3 -RNase has ribonuclease activity (20), rotation of the ring of His 33 is not crucial for its catalytic activity. Another difference is that the His (104) of RNase Rh, which is expected to bind to the phosphate group of the substrate RNA (32,41), is replaced by Lys 83 in S 3 -RNase (Fig. 3a). Because the ⑀-amino group of Lys 83 extends outside the P 1 site and is about 8 Å from the imidazole ring of His (104) , it is unlikely that it interacts with the phosphate group of the substrate unless a large induced fit occurs on the side chain of Lys 83 during catalysis. A more detailed x-ray crystallographic study of the S 3 -RNase complex with nucleotides is required to clarify how the rotation of the imidazole ring of His 33 and substitution of His (104) for Lys 83 affect the ribonuclease activity of S 3 -RNase.
Although the in vivo substrate of S-RNase during the GSI reaction has not been identified experimentally, it must be a pollen (tube) rRNA according to the receptor and inhibitor models (10). Because the overall shape of the active site of S 3 -RNase is very similar to the active sites of the other RNase T 2 family enzymes (except for a few amino acid substitutions), S 3 -RNase is not likely to have a strict substrate specificity corresponding to the S-alleles. Actually, S 3 -RNase can hydrolyze dinucleotide monophosphate nonspecifically (data not shown); therefore, it is reasonable that S-allele specificity in the inhibition of pollen tube growth is not expressed due to the enzyme's restricted substrate specificity but to its interaction with one or more pollen S-gene products, consistent with either the receptor or inhibitor model.
Domain swapping experiments proved that, in the solanaceous S-RNases, two hypervariable regions, HVa and HVb, essentially are responsible for S-allele-specific pollen recognition (13,14). Only one hypervariable (HV) region appears to be present in the rosaceous S-RNases, and it most likely is a recognition site for the pollen factor(s) (Fig. 4a). As compared with the sequences of the solanaceous S-RNases (16,17), the HVa and HVb regions correspond, respectively, to the HV region and vicinity of the ␣3-helix (Arg 74 -Lys 76 ), tentatively called the HVbЈ, of S 3 -RNase. Because HVbЈ is much shorter than HVb, it has not been identified as an allele-specific sequence (Fig. 4a). The three-dimensional structure around the ␣3-helix, therefore, is also assumed to differ in the solanaceous and rosaceous S-RNases.
The HV region of S 3 -RNase is located on the molecule's surface, easily accessible to external molecules such as one or more pollen factors. As shown in Fig. 1b, the main-chain frameworks of RNase T 2 family enzymes, including S 3 -RNase, superimpose well on one another; in particular their core structures composed of three ␣-helices (␣4, ␣6, and ␣7) and four ␤-strands (␤1, ␤2, ␤4, and ␤5). The backbone structures, however, are notably very variable in what corresponds to the HV region (Figs. 1b and 4b) even though it is composed of the same secondary elements, a loop and an ␣-helix, as in all the RNases (Fig. 2). This suggests that the HV region of S 3 -RNase has an allele-specific conformation.
Five basic amino acids, Lys 51 , Arg 54 , Lys 55 , Arg 56 , and Lys 58 and two acidic amino acids, Glu 57 and Glu 61 , are present in the HV region of S 3 -RNase (Fig. 4a). Although they seem to form a positively or negatively charged cluster on the molecule's surface, no such cluster was clearly detected, because the side chains of the basic or acidic amino acids are oriented in different directions (Figs. 5 and 6). Because the ␦-guanidino groups of Arg 54 and Arg 56 are neutralized by the formation of hydrogen bonds, the surface of HV carries a weak rather than strong positive charge (Figs. 5 and 6). If the other six P. pyrifolia S-RNases (S 1 , S 2 , and S 4 through S 7 ) have the same mainchain framework as S 3 -RNase, their HV regions also would similarly have a weak positive charge, which might be important for binding to one or more pollen factors (Figs. 4 and 5).
PS regions with an excess of dN over dS, like the HV region, which is a candidate for the binding site to one or more pollen factors, are located on either side of the active site cleft (Fig.  4b). Although there is no marked charge cluster in any of these regions, PS1 (HV), PS2, and PS3 are hydrophilic and weakly basic (Fig. 6). In contrast, PS4 is neutral and hydrophobic despite its exposure to solvent. Why the rosaceous S-RNase has been positively selected on such a wide area of its molecular surface is not clear, but location topology of the PS regions suggests that S 3 -RNase simultaneously interacts with multiple pollen factors and the substrate RNA.
The amino acid sequence identity between the P. pyrifolia S 3 -and S 5 -RNases is 95.5%, and there are only nine substitutions in their 200 amino acid residues, all concentrated on the N-terminal half of S 3 -RNase (Fig. 4a). Two of the nine substitutions, Lys(Arg) 51 and His(Pro) 62 , are the most likely to contribute to recognition between the two alleles, because they are located on the molecular surface of the HV region (amino acids in parentheses are those of S 5 -RNase) (Fig. 4b). The side chain of Lys(Arg) 51 is fully exposed to solvent and, therefore, has no interaction with the other residues (Fig. 5). In contrast, His-(Pro) 62 is close to Glu(Ala) 65 in the three-dimensional structure, although at separate locations in the primary structure. His 62 N ⑀2 forms a water-mediated hydrogen bond with Glu 65 O ⑀1 , and both are appreciably exposed to solvent (Fig. 5). Two possibilities are suggested as to how one or more pollen factors discriminate S 3 -RNase from S 5 -RNase. The pollen factor(s) must come in contact with Lys 51 , the site consisting of His 62 and Glu 65 , or both to recognize the difference between the FIG. 6. Electrostatic surface potential of the HV and PS regions of P. pyrifolia S 3 -RNase. a, surfaces of the HV (PS1) and PS2 regions; b, surfaces of the PS3 and PS4 regions. The potential displayed represents a range of Ϫ10 to ϩ10 k B T, red being negative and blue positive. This graphic was prepared by using the program GRASP (56). S 3 -and S 5 -RNases. Or, substitution of His 62 for Pro produces a large conformational change in the ␣2-helix and the pollen factor(s) recognize such a conformational change rather than the amino acid substitutions. In general, substitution for proline tends to bend the helix structure (53). An x-ray crystallographic study of P. pyrifolia S 5 -RNase is in progress in our laboratory. If the three-dimensional structure of S 5 -RNase can be determined, a detailed structural comparison can be made of the two S-RNases, which should shed light on the recognition mechanism that operates between S 3 -and S 5 -alleles.