High-resolution crystal structures of ribonuclease A complexed with adenylic and uridylic nucleotide inhibitors. Implications for structure-based design of ribonucleolytic inhibitors

Abstract

The crystal structures of bovine pancreatic ribonuclease A (RNase A) in complex with 3′,5′-ADP, 2′,5′-ADP, 5′-ADP, U-2′-p and U-3′-p have been determined at high resolution. The structures reveal that each inhibitor binds differently in the RNase A active site by anchoring a phosphate group in subsite P₁. The most potent inhibitor of all five, 5′-ADP (K_i = 1.2 μM), adopts a syn conformation (in contrast to 3′,5′-ADP and 2′,5′-ADP, which adopt an anti), and it is the β- rather than the α-phosphate group that binds to P₁. 3′,5′-ADP binds with the 5′-phosphate group in P₁ and the adenosine in the B₂ pocket. Two different binding modes are observed in the two RNase A molecules of the asymmetric unit for 2′,5′-ADP. This inhibitor binds with either the 3′ or the 5′ phosphate groups in subsite P₁, and in each case, the adenosine binds in two different positions within the B₂ subsite. The two uridilyl inhibitors bind similarly with the uridine moiety in the B₁ subsite but the placement of a different phosphate group in P₁ (2′ versus 3′) has significant implications on their potency against RNase A. Comparative structural analysis of the RNase A, eosinophil-derived neurotoxin (EDN), eosinophil cationic protein (ECP), and human angiogenin (Ang) complexes with these and other phosphonucleotide inhibitors provides a wealth of information for structure-based design of inhibitors specific for each RNase. These inhibitors could be developed to therapeutic agents that could control the biological activities of EDN, ECP, and ANG, which play key roles in human pathologies.

Bovine pancreatic Ribonuclease A (RNase A; EC 3.1.27.5) catalyzes the cleavage of P-O5′ bonds in RNA on the 3′ side of pyrimidines to form cyclic 2′,5′-phosphates (D’Alessio and Riordan 1997; Raines 1998). In recent years, RNase A has been the molecular target for the development of small molecule inhibitors. These inhibitors are being developed with the goal to restrain the biological activities of different RNase A homologs in a variety of pathological conditions. These homologs include human Angiogenin (Ang), a potent inducer of neovascularization involved in cancer and in vascular and rheumatoid diseases (Fett et al. 1985), eosinophil-derived neurotoxin (EDN), and eosinophil cationic protein (ECP), two neurotoxins implicated in hypereosinophilic syndromes and allergy (Gleich and Adolphson 1986). EDN has been also implicated in the soluble HIV-1 inhibitory activity by lymphocytes (Rugeles et al. 2003). The ribonucleolytic activity of these homologs is closely associated with their biological function because it is impaired when assayed in the presence of ribonucleolytic inhibitors or when residues critical for their enzymic activity are chemically modified or mutated (Shapiro and Vallee 1989; Shapiro et al. 1989; Sorrentino et al. 1992; Newton et al. 1994; Domachowske et al. 1998a,b). Development of small molecule inhibitors using RNase A as a model are quite amenable because the central region of the catalytic active site of RNase A encompassing subsites P₁, where the phosphate phosphodiester bond cleaveage occurs, and subsites B₁ and B₂, where the nucleotide bases on the 3′ and 5′ sides of the scissile bond bind, respectively, is highly conserved amongst RNase A homologs.

In the last years, a potent class of inhibitors, the adenosine 5′-pyrophosphate derivatives, has been discovered (Leonidas et al. 1997, 1999; Russo et al. 1997; Russo and Shapiro 1999; Jardine et al. 2001). Crystallographic studies on the binding of such derivatives like ppA-3′-p (K_i = 0.24 mM), ppA-2′-p (K_i = 0.52 mM; Leonidas et al. 1997^; Russo et al. 1997), dUppA (K_i = 11.3 μM; Jardine et al. 2001), and pdUppA-3′-p (K_i= 0.027 μM; Leonidas et al. 1999^; Russo and Shapiro 1999), with RNase A have shown that the phosphate groups constitute the anchor points for the binding of these inhibitors to RNase A. The 5′-pyrophosphate group binds with the β-phosphate at subsite P₁, while the 2′ or 3′ adenylic phosphates bind at the peripheral phosphate-binding subsites P_o and P₂, respectively. The adenosine adopts the unusual syn conformation and binds at B₂, while uridine is found at B₁. These inhibitors are also potent against EDN, RNase 4, and Ang (Shapiro 1998; Russo and Shapiro 1999; Russo et al. 2001). However, recent structural studies of EDN complexes with three adenylic phosponucleotide inhibitors (Leonidas et al. 2001a), ECP in complex with 2′,5′-ADP (Mohan et al. 2002) and Ang in complex with phosphate and pyrophosphate ions (Leonidas et al. 2001b) have shown that the binding mode of phosphonucleotides to these enzymes might differ considerably from the one observed in RNase A. Thus, although the interactions of a phosphate group at P₁ were still the driving force for the binding of phosphoadenosine inhibitors to EDN and ECP, the position of the nucleotide base differs considerably depending on the nucleotide (Leonidas et al. 2001a; Mohan et al. 2002). Furthermore, in Ang, the pyrophosphate ion binds with one of the phosphate groups in P₁ and the other directed towards the pyrimidine binding site B₁ in contrast to the binding mode of pyrophosphate derivatives to RNase A (Leonidas et al. 2001b).

In the present study we are investigating the mode of binding of three adenylic (3′,5′-ADP, 2′,5′-ADP, and 5′-ADP) and two uridylyl (U-2′-p, U-3′-p) inhibitors to RNase A at high resolution. These structures give a detailed picture of the specificity of the ribonucleolytic active site towards phosphates and nucleotide bases and show the degree of flexibility of subsites B₁ and B₂. The high-resolution structures also provide the basis for a comparative structural analysis of EDN, ECP, and RNase A phosphonucleotide complexes and assist the process of the development of ribonucleolytic inhibitors specific for each RNase. Furthermore, the crystallographic data of the RNase A complexed with 2′,5′-ADP and 5′-ADP at near atomic resolution (1.2 Å) revealed the positions of additional water molecules bound in the RNase A active site and indicated a complex water-mediated network of interactions between the inhibitors and the protein. This network may have important implications for the further development of new and more potent ribonucleolytic inhibitors.

Results and Discussion

Overall structures

The complexed structures of RNase A described here are similar to the free RNase A structure from monoclinic crystals reported previously (Leonidas et al. 1997), indicating that the binding of the inhibitors did not affect the overall structure of the protein. The RMSD between the structures of the free RNase A (PDB entry 1AFU) and RNase A in complex with either 3′,5′-ADP, 2′,5′-ADP, 5′-ADP, U-2′-p or U-3′-p is 0.79, 0.67, 0.76, 0.82, and 0.62 Å, respectively, for 248 equivalent C^α atoms. In all free RNase A structures reported so far the side chain of the catalytic residue His119 adopts two conformations denoted as A (χ₁ = ∼160°) and B (χ₁ = ∼−80°), which are related by a 100° rotation about the Cα–Cβ bond and a 180° rotation about the Cβ–Cγ bond (Borkakoti et al. 1982; Howlin et al. 1989; deMel et al. 1994). These conformations are dependent on the pH (Berisio et al. 1999) and the ionic strength of the crystallization solution (Fedorov et al. 1996). Conformation A is considered as the active conformation, which promotes catalysis, whereas conformation B is considered as the inactive conformation (Raines 1998). The side chain of His119 adopts conformation A in the 3′,5′-ADP, 5′-ADP, U-2′-p, and U-3′-p complexes while in the 2′,5′-ADP complex it is found in conformation B.

Superposition of the structures of the RNase A–3′,5′-ADP complex (molecules I and II of the asymmetric unit) on the structure of free RNase A at 1.1 Å resolution (PDB code 1KF2; Berisio et al. 2002) reveals that 3′,5′-ADP displaces two (molecule I) and three (molecule II) water molecules from the active site. Similarly, two water molecules are replaced upon binding of 2′,5′-ADP (molec I and molec II) while three (molec I) and one (molec II) water molecules are replaced upon binding of 5′-ADP. Upon binding of U-2′-p, five (molec I) and four (molec II) water molecules are replaced while upon binding of U-3′-p, five water molecules are replaced in both molecules I and II of the asymmetric unit.

The RMSD between the positions of the C^α atoms of the 3′,5′-ADP complex and those of the 2′,5′-ADP, 5′-ADP, U-2′-p, and U-3′-p complexes are 0.46, 0.42, 0.46, and 0.54 Å respectively. The corresponding values within the complexes, 2′,5′-ADP against 5′-ADP, U-2′-p, and U-3′-p are 0.24, 0.30, and 0.22 Å, respectively, whereas the RMSD between the 5′-ADP and U-2′-p, and U-3′-p complexes are 0.26 and 0.31 Å, respectively. In addition, the RMSD between the positions of the C^α atoms of the U-2′-p and U-3′-p complexes is 0.35 Å. Thus, it appears that the differences between the five complex structures are very small, mainly concentrated on the loop regions and are not related to the binding of the different inhibitors. The structures of the 3′,5′-ADP, U-2′-p, and U-3′-p complexes are defined at 1.5 Å resolution and contain 389, 299, and 415 water molecules, respectively. The structures of the 2′,5′-ADP, and 5′-ADP complexes are determined at a higher resolution (1.2 Å), and contain 558 and 540 water molecules, respectively. Numbering scheme definitions for all five inhibitors are presented in Figure 1 ▶.

Figure 1.

(A–E) Diagrams of the sigma 2|F_o| − |F_c| electron density maps for 3′,5′-ADP, 2′,5′-ADP, 5′-ADP, U-2′-p, and U-3′-p, respectively. Electron density maps were calculated with CNS (Brünger et al. 1998) from the RNase A model before incorporating the coordinates of each inhibitor, are contoured at 1.0 σ level, and the refined structure of the inhibitor is shown. General numbering scheme for each inhibitor is also shown. The figure was prepared with BOBSCRIPT (Esnouf 1997).

The binding of 3′,5′-ADP to RNase A

The 3′,5′-ADP molecule is well defined in the electron density map in both molecules (I and II) of RNase A in the asymmetric unit (Fig. 1A ▶). The conformation of 3′,5′-ADP when bound to RNase A is similar to that observed previously for adenosine nucleotides bound at B₂ in the RNase A complexes with d(pA)₄ (McPherson et al. 1986), d(ApTpApApG) (Fontecilla-Camps et al. 1994), and d(CpA) (Zegers et al. 1994), as well as to those frequently observed in the unbound and protein bound adenosines (Moodie and Thornton 1993). The glycosyl torsion angle χ adopts the anti conformation, and the rest of the backbone and phosphate torsion angles are in the preferred range for protein bound adenosines (Moodie and Thornton 1993). The γ torsion angle of 3′,5′-ADP in RNase A molecule II is in the unusual sp range, but its value (18°) is close to the favorable +sc range (30°–90°; Table 1). The ribose is found in both molecules at the C2′-endo conformation.

Table 1.

Torsion angles for 3′,5′-ADP, 2′,5′-ADP, 5′-ADP, U-2′-p, and U-3′-p when bound to RNase A

	3′,5′-ADP		2′,5′-ADP		5′-ADP	U-2′-p	U-3′-p
RNase A molecule	I	II	I	II	I	I	I
Backbone torsion angles
O5′-C5′-C4′-C3′ (γ)	32 (+sc)	18 (sp)	57 (+sc)	46 (+ac)	−75 (−sc)	60 (+sc)	59 (+sc)
C5′-C4′-C3′-O3′ (δ)	138 (+ac)	138 (+ac)	149 (+ac)	134 (+ac)	142 (+ac)	83 (+ac)	141 (+ac)
C5′-C4′-C3′-C2′	−104	−104	−92	−109	−100	−157	−100
C4′-C3′-C2′-O2′	−152	−151	−158	−152	−157	−76	−151
Glycosyl torsion angle
O4′-C1′-N9′-C4 (χ)	49 (anti)	−62 (anti)	−83 (anti)	−66 (anti)	73 (syn)
O4′-C1′-N1-C2 (χ′)					−137 (anti)		−134 (anti)
Pseudorotation angles
C4′-O4′-C1′-C2′ (v0)	−29	−28	−15	−31	−32	6	−23
O4′-C1′-C2′-C3′ (v1)	39	38	34	38	42	−28	35
C1′-C2′-C3′-C4′ (v2)	−34	−34	−39	−31	−36	38	−33
C2′-C3′-C4′-O4′ (v3)	18	19	31	14	18	−36	21
C3′-C4′-O4′-C1′ (v4)	6	5	−10	11	8	19	1
Phase	152	154	176	145	150	10	159
	C2′-endo	C2′-endo	C2′-endo	C2′-endo	C2′-endo	C3′-endo	C2′-endo
Phosphate torsion angles
O3A-PA-O5′-C5′ (α)					59 (+sc)
P2-O5′-C5′-C4′ (β)	−158 (−ap)	−167 (−ap)	178 (+ap)	166 (+ap)	164 (+ap)
PB-O3A-PA-O5′A (ζpp)					−94 (+ap)
C4′-C3′-O3′-P1 (ɛ)	−111	−95					120
P1-O3′-C3′-C2′	135	152					−126
P1-O2′-C2′-C3′ (ɛ′)			−117 (ac)	−97 (ac)		−92 (−ac)
P1-O2′-C2′-C1′			130	150		157

Definitions of the torsion angles are according to the current IUPAC-IUB nomenclature (IUPAC-IUB and [JCBN] 1983), and the phase angle of the ribose ring is calculated as described previously (Altona and Sundaralingam 1972). The α torsion angle for 3′,5′-ADP and 2′,5′-aDP cannot be defined because the position of atom O3A is ambiguous (the phosphate group is free). For atom definitions see Figure 1 ▶.

The binding mode of 3′,5′-ADP to both molecules of RNase A in the asymmetric unit is almost identical. The inhibitor binds to the P₁-B₂-P₂ region of the catalytic site (Fig. 2A ▶) with the 5′-phosphate group in P₁ involved in hydrogen bond interactions with Gln11, His12, Lys41, and Phe120, while the 3′-phosphate group is bound at the P₂ subsite forming a hydrogen bond with Lys7 (Fig. 2A ▶). The adenine is bound at the B₂ subsite in a manner similar to that observed previously for other adenylic nucleotides bound to RNase A (McPherson et al. 1986; Fontecilla-Camps et al. 1994; Zegers et al. 1994). The adenine ring packs with its five-member ring against the imidazole ring of His119, which adopts conformation A (χ₁ = 158°). The N6 group of adenine is in hydrogen-bonding distance from the side chain atoms of residues Gln69 and Asn71 (Table 2, observed only in RNase A molecule I).

Figure 2.

(A, B, D, E, and F) Diagrams of the interactions between RNase A molecule I of the asymmetric unit and 3′,5′-ADP, 2′,5′-ADP, 5′-ADP, 2′ UMP, and 3′ UMP, respectively. (C) Diagram of the interactions between RNase A and 2′,5′-ADP in molecule II of the asymmetric unit. Molecules I and II refer to the two RNase A molecules in the asymmetric unit. Residues are drawn as ball-and-stick models and water molecules as white spheres. Hydrogen bonds are indicated as dashed lines.

Table 2.

Potential hydrogen bonds of 3′,5′-ADP with RNase A (this study) and eDN (Leonidas et al. 2001a)

			Distance (Å)
3′,5′-ADP	RNase A	EDN	RNase A mol. I	RNase A mol. II	EDN
O1P	Lys7 Nζ		—	2.8	—
O3P		Met0 N	—	—	2.7
O4P	His12 Nɛ2		2.7	2.8	2.9
O4P	Phe120 N	Leu130 N	3.1	3.0	3.0
O5P	Gln11 Nɛ2		2.9	3.2	3.0
O5P	Lys41 Nζ		3.1	—	—
O6P		His129 Nδ1	—	—	2.3
N1		Asn70 Nδ2	—	—	2.6
N6	Gln69 Oɛ1		3.3	(3.4)	—
N6	Asn71 Oδ1	Asn70 Oδ1	3.0	(3.4)	2.7
O5′	His119 Nδ1		2.7	—	—
N1	Water		3.0	—	—
N7	Water		3.0	2.9	—
N3	Water		2.9	—	—
O2′	Water		3.3	—	—
O2′	Water		3.1	—	—
O2′	Water		3.0	—	—
O2′	Water		2.7	—	—
O2′	Water		2.9	—	—
O3′	Water		3.1	3.0	—
O4′	Water		2.7	—	—
O1P	Water		2.7	—	2.7
O1P	Water		2.8	—	—
O2P	Water		2.7	2.8	—
O2P	Water		2.8	—	—
O3P	Water		—	2.6	—
O4P	Water		2.9	3.0	2.9
O5P	Water		2.7	2.6	2.8
O6P	Water		2.4	3.2	—
O6P	Water		2.8	—	—

Hydrogen bond interactions were calculated with the program HBPLUS (McDonald and Thornton 1994). Atom names are shown in Figure 1 ▶.

There are 10 and 11 water molecules forming hydrogen bonds with 3′,5′-ADP in RNase molecules I and II, respectively. Furthermore, six and four of them mediate polar interactions between the inhibitor and RNase A molecules I and II, of the asymmetric unit, respectively (Fig. 2A ▶). RNase A residues and 3′,5′-ADP atoms are involved in 35 and 48 van der Waals contacts in molecules I and II, respectively.

The binding of 2′,5′-ADP to RNase A

The 2′,5′-ADP molecule is well defined within the electron density map in both RNase A molecules of the crystallographic asymmetric unit (Fig. 1B ▶). The conformation of 2′,5′-ADP is similar in the two RNase A molecules of the asymmetric unit and to that of 3′,5′-ADP with the glycosyl torsion angle χ in the most commonly found anti range and the ribose at the C2′-endo conformation (Table 1).

The binding mode of 2′,5′-ADP in RNase A molecules I and II differs considerably, as different phosphate groups of the inhibitor are bound in subsite P₁. In molecule I, it is the 2′-phosphate that binds at P₁ (Fig. 2B ▶), while in RNase A molecule II the 5′-phosphate occupies the P₁ subsite (Fig. 2C ▶). In RNase A molecule I the side chain of His119 is found in conformation B (χ₁ = −55°), while in RNase A molecule II adopts conformation A (χ₁ = 157°). The adenine packs with its five-member ring against the imidazole ring of His119 and it is involved in π–π stacking interactions only in molecule II (Fig. 2C ▶).

In molecule I, the adenine ring is in a different orientation but still parallel to the imidazole ring of His119 (Fig. 2B ▶). The 5′ and the 2′-phosphate groups of 2′,5′-ADP in molecules I and II, respectively, point away from the protein, occupying two distinct positions 5 Å apart. These groups are exposed to the solvent and are not involved in any hydrogen bond interactions with RNase A. However, in molecule I, the 5′ phosphate is oriented towards Lys7, the sole component of subsite P₂. Although the closest distance of the phosphate oxygens to atom Nζ of Lys7 is ∼7 Å, it is sufficient for long-range coulombic interactions, which have been shown to play an important role in catalysis by RNase A (Fisher et al. 1998a,b).

The binding mode of 2′,5′-ADP in molecule I resembles that of 3′,5′-ADP, and it is similar to that observed previously for adenosine nucleotides bound at B₂ in the RNase A complexes with d(pA)₄ (McPherson et al. 1986), d(ApTpApApG) (Fontecilla-Camps et al. 1994), and d(CpA) (Zegers et al. 1994). The binding mode of 2′,5′-ADP in molecule II is a novel mode observed for the first time. Upon superposition of the two RNase A molecules, the 5′-phosphate group of 2′,5′-ADP in molecule I passes through the adenine ring of 2′,5′-ADP in molecule II (Fig. 3A ▶). However, apart from the side chain of His119, the rest of the active site residues are structurally conserved between molecules I and II (Fig. 3 ▶). The positions occupied by the 2′ and the 5′-phosphate groups at P₁ in the two binding modes are only 0.5 Å apart. In both binding modes the inhibitor is involved in hydrogen bond interactions with His12, His119, and Phe120, although through different atoms (Table 3). In addition, in molecule II, 2′,5′-ADP is involved in hydrogen bonds with Gln11 and Asn71, whereas in molecule I it forms an additional hydrogen bond with Lys41 (Table 3).

Figure 3.

Superimposed structures of (A) the 2′,5′-ADP complex in the two RNase A molecules of the asymmetric unit (B) the RNase A–U-2′-p (gray) and the RNase A–U-3′-p complex structures (white), (C) the RNase A–U-3′-p (gray) and the K7A/R10A/K66A RNase A–U-3′-p (PDB entry 4RSK; white) complexes, (D) the RNase U–2′-p (gray) and the RNase A–2′CMP (PDB entry 1ROB; white), and (E) the RNase A–U-3′-p (gray) and the RNase A–3′CMP (PDB entry 1RPF; white) complexes.

Table 3.

Potential hydrogen bonds of 2′,5′-ADP with RNase A (present study), EDN (Leonidas et al. 2001a) and ECP (Mohan et al. 2002)

				Distance (Å)
2′,5′-ADP	RNase A	EDN	ECP	RNase A mol. I	RNase A mol. II	EDN	ECP
O1P	His119 Nδ1	His129 Nδ1	Lys38 Nζ	2.7	—	2.7	2.8
O2P	Lys41 Nå	Gln14 Nɛ2	Gln14 Nɛ2	2.9	—	3.3	3.0
O3P	His12 Nɛ2	His15 Nɛ2	His15 Nɛ2	2.6	—	2.6	3.0
O3P	Phe120 N	Leu130 N		2.9	—	2.8	—
O5P	His12 Nɛ2			—	2.9	—	—
O5P	Phe120 N			—	2.8	—	—
O6P	Gln11 Nɛ2		Arg34 Nη2	—	2.9	—	2.7
O3′		Leu130 O		—	—	2.6	—
O5′	His119 Nδ1			—	2.7	—	—
N3			His128 Nɛ2	—	—	—	3.3
N6	Asn71 Oδ1			—	3.1	—	—
N3	Water			2.9	3.2	2.9	—
N3	Water			—	2.8	—	—
N6	Water			3.2	—	—	—
N7	Water			—	3.1	—	—
O3′	Water			2.7	—	—	—
O3′	Water			3.2	—	—	—
O4′	Water			—	2.7	—	—
O5′	Water			—	2.2	—	—
O1P	Water			—	2.6	—	—
O2P	Water			2.8	3.0	—	3.1
O3P	Water			2.9	2.0	2.8	2.4
O4P	Water			2.5	2.5	2.9	2.7
O4P	Water			—	2.6	—	—
O5P	Water			2.6	2.8	—	—
O6P	Water			2.2	2.5	—	2.7
O6P	Water			—	3.0	—	—

Hydrogen bond interactions were calculated with the program HBPLUS (McDonald and Thornton 1994). Atom names are shown in Figure 1 ▶.

Several water molecules mediate hydrogen bond interactions between 2′,5′-ADP atoms and residues Ala4, Gln11, Lys41, Pro42, Val43, Lys66, Asn67, Val118, Ph120, and Asp121 in molecule I (Fig. 2B ▶). In molecule II, Ala4, Phe8, Gln11, His12, Lys41, Lys66, Asn67, Glu111, Gly112, Asn113, Val118, and Asp121 are involved in hydrogen bonds with 2′,5′-ADP atoms, mediated by seven water molecules (Fig. 2C ▶). The inhibitor and the protein participate in 22 van der Waals contacts, in molecule I, through residues His12, Lys41, Asn67, and His119. In molecule II, there are 36 van der Waals contacts between 10 protein residues and 14 inhibitor atoms.

The binding of 5′-ADP to RNase A

The 5′-ADP molecule is bound at the P₁-B₂ region of the RNase A active site. The electron density is of high quality, and all atoms of 5′-ADP are well defined (full occupancy) in molecule I of the crystallographic asymmetric unit (Fig. 1C ▶). In molecule II, despite the 1.2 Å resolution data, the electron density for the inhibitor is relatively poor (occupancy 0.6). This partial binding, which has also been observed for U-2′-p and U-3′-p, as well as in previous binding studies with RNase A (Leonidas et al. 1997), might be attributed to the lattice contacts which limit access to the active site of molecule II of the asymmetric unit. Therefore, only molecule I will be considered for this complex throughout our discussion.

The conformation of 5′-ADP is similar to that of the adenosine 5′-pyrophosphate derivatives bound to RNase A (Leonidas et al. 1997, 1999; Jardine et al. 2001) and very different from all the other adenosine phosphonucleotides complexed with RNase A. The glycosyl torsion angle χ is at the unusual syn conformation, in contrast to 3′,5′-ADP and 2′,5′-ADP (Table 1; Fig. 1C ▶). The ribose is at the C2′-endo conformation and the rest of the backbone and phosphate torsion angles have values frequently observed in protein bound nucleotides (Table 1). The syn conformation is rare in protein-bound adenosine nucleotides (Moodie and Thornton 1993), and has been previously observed in adenosine 5′-pyrophosphate derivatives bound to RNase A (Leonidas et al. 1997, 1999; Jardine et al. 2001).

The adenine packs with the six-member ring against the imidazole of His119 (conformation A, χ₁ = 158°), in contrast to 3′,5′-ADP and 2′,5′-ADP, and the N6 group is involved in hydrogen bond interactions with Asn67 and Asn71 (Table 4; Fig. 2D ▶). 5′-ADP binds at the RNase A active site with the β- rather than the α-phosphate group at P₁, and forms hydrogen bonds with Gln11, His12, Lys41, His119, and Phe120 (Fig. 2D ▶). This mode of binding is reminiscent of the binding of the adenosine 5′-pyrophosphate derivatives to RNase A (Russo et al. 2001).

Table 4.

Potential hydrogen bonds of 5′-ADP with RNase A (present study) and EDN (Leonidas et al. 2001a)

			Distance (Å)
5′-ADP	RNase A	EDN	RNase A	EDN
O1A	Lys7 Nζ	Lys38 Nζ	2.9	2.9
O1A	Gln11 Nɛ2		3.1	—
O3A	Gln11 Nɛ2		3.0	—
O1B	His12 Nɛ2	His129 Nδ1	2.7	2.5
O1B	Phe120 N		3.0	—
O2B	Lys41 Nζ	Gln14 Nɛ2	2.7	2.8
O3B	His119 Nδ1	His15 Nɛ2	2.6	2.8
O3B		Leu130 N	—	2.8
O2′		Gln40 O	—	3.1
O2′		Gln40 N	—	3.1
O5′	Lys7 Nζ		2.9	—
N6	Asn67 Oδ1		3.3	—
N6	Gln69 Oɛ1		3.3	—
N6	Asn71 Oδ1		3.0	—
N1	Water		2.8	—
N3	Water		2.8	—
N7	Water		3.0	—
O2′	Water		2.8	—
O3′	Water		2.8	—
O3′	Water		2.8	—
O1A	Water		2.7	—
O2A	Water		2.5	—
O2A	Water		2.9	—
O2A	Water		2.9	—
O3A	Water		3.2	—
O1B	Water		2.9	—
O2B	Water		2.9	—
O2B	Water		2.8	—
O3B	Water		2.4	3.1
O3B	Water		3.2	—

Hydrogen bond interactions were calculated with the program HBPLUS (McDonald and Thornton 1994). Atom names are shown in Figure 1 ▶.

There are 12 water molecules hydrogen bonded to 5′-ADP out of which six mediate hydrogen bond interactions to residues Glu2, Ala4, Lys7, Gln11, Arg39, Pro42, Val43, Asn67, Gln69, Glu111, Val118, Phe120, and Asp121. Ten RNase A residues and 17 atoms of 5′-ADP are involved in 38 van der Waals contacts.

The binding of U-2′-p and U-3′-p to RNase A

Both U-2′-p and U-3′-p bind to the P₁-B₁ region of the RNase A active site adopting similar conformations and in a manner similar to that of uridine vanadate (UVan, a transition state analog) and other uridylyl nucleotides to RNase A (Gilliland 1997; Ladner et al. 1997). The two inhibitors, are well defined within the electron density map only in RNase A molecule I of the asymmetric unit (Fig. 1D,E). The electron density for both inhibitors is poor for molecule II, like in the RNase A–5′-ADP complex; hence, our study has been focused only in the complex of the inhibitors with molecule I.

The glycosyl torsion angle of uridine, in both inhibitors, is in the anti range, as found in most free and protein-bound pyrimidine nucleotides (Moodie and Thornton 1993). Torsion angles γ and δ are in the +sc and +ac range, respectively (Table 1). The ribose rings of U-2′-p and U-3′-p adopt the C3′-endo and the C2′-endo puckering, respectively, the two furanose puckerings most frequently observed in free and bound nucleotides (Moodie and Thornton 1993). The C2′-endo puckering has also been observed in 2′-deoxyuridine in the RNase A–pdUppA-3′-p and dUppA–3′-p complexes (Leonidas et al. 1999; Jardine et al. 2001), in cytidine in the RNase A–d(CpA) complex (Zegers et al. 1994), and in thymidine in the RNase A–d(ApTpApApG) complex (Fontecilla-Camps et al. 1994), while the C3′-endo has also been found in uridine in the RNase A–UVan complex (Wladkowski et al. 1998). This is different from the ribose puckerings of uridine in the d(UpcA) (C3′-exo) complex (Gilliland 1997) and in cytidine in the 2′,5′-CpA (C4′-exo) and 3′,5′-d(CpA) (C1′-exo) complexes (Toiron et al. 1996).

In both U-2′-p and U-3′-p complexes the uridine ring is packed against the phenyl ring of Phe120 and participates in hydrogen bond interactions with Thr45, the primary functional component of this subsite (Fig. 2E,F ▶). The phosphate groups are in the P₁ subsite occupying two distinct positions 2.0 Å apart (Fig. 3B ▶). However, in both complexes, the phosphate group forms hydrogen bonds with residues Gln11, Lys41, and His119 (Table 5; Fig. 2E,F ▶). In addition, the phosphate of U-2′-p forms two hydrogen bonds with residues His12 and Phe120 that might explain differences in the potency of the two uridylyl molecules (U-2′-p and U-3′-p are potent inhibitors of RNase A with K_i values of 7.1 and 82 μM, respectively; Anderson et al. 1968). There are 6 and 10 water molecules hydrogen-bonded to U-2′-p and U-3′-p, respectively (Table 5; Fig. 2E,F ▶). Among them there is a water molecule that mediates interactions between atom O4 of uracil of U-2′-p and residues Asp83 and Ser 123. The same interactions are mediated through two water molecules in the U-3′-p complex. These water molecules might correspond to those proposed previously that adjust their donor/acceptor hydrogen bonding roles depending on whether uracil or cytosine is bound at the B₁ subsite (Gilliland et al. 1994).

Table 5.

Potential hydrogen bonds of U-3′-p and U-2′-p with the wild type RNase A (present study), and K7A/R10A/K66A RNase A variant (Fisher et al. 1998b) with U-3′-p in the crystal

		Distance (Å)
Inhibitor atom	RNase A	U-3′-p	U-3′-p-variant	U-2′-p
N3	Thr45 Oγ1	2.8	2.8	2.9
O2	Thr45 N	2.9	2.8	3.0
O2′	Lys41 Nζ	—	2.7	—
O3′	Lys41 Nζ	3.2	—	—
O1P	His119 Nδ1	2.8	2.7	2.7
O2P	Lys41 Nζ	—	—	3.1
O2P	Gln11 Nɛ2	3.2	—	2.9
O3p	His12 Nɛ2	—	3.0	2.6
O3P	Phe120 N	—	2.8	2.8
O4	Water	3.1	2.9	2.8
O4	Water	2.8	—	—
O2′	Water	3.0	—	—
O4′	Water	3.1	—	3.0
O5′	Water	3.1	2.7	2.8
O5′	Water	3.1	—	—
O1P	Water	2.6	—	2.7
O2P	Water	2.6	2.9	2.7
O3P	Water	2.6	—	3.3
O3P	Water	2.5	—	—

Hydrogen bond interactions were calculated with the program HBPLUS (McDonald and Thornton 1994). Atom names are shown in Figure 1D and E ▶.

Comparison of the two uridylyl complexes with the RNase A–UVan complex (Ladner et al. 1997) reveals that the uridine moieties bind almost identically in all three complexes. Furthermore, the position of the phosphate group of U-2′-p is similar to that of the vanadate ion, while in the case of U-3′-p it is in close distance (2.0 Å).

The binding of U-3′-p in the crystal of an RNase A variant, where Lys7, Arg10, and Lys66 were replaced by alanines, was studied previously (Fisher et al. 1998b). Superposition of the wild-type RNase A–U-3′-p complex structure of this study onto the K7A/R10A/K66A RNase A variant–U-3′-p complex structure reveals that U-3′-p binds in a similar mode to both enzymes (Fig. 4B ▶). However, the location of U-3′-p in the active site, in the two complexes, differs by approximately 1.0 Å (Fig. 3C ▶). In the variant complex structure His119 is in the B conformation (χ₁ = −64°), in contrast to the native complex structure where it is found in conformation A. The rest of the active site residues in the two complexes superimpose well. The two uracil rings bind similarly in the B₁ subsite, they are involved in stacking interactions with the phenyl ring of Phe120, and they are hydrogen-bonded to Thr45. In contrast, the position of the 3′-phosphate group is different in the two complexes (Fig. 3C ▶). The ribose of U-3′-p in the variant complex structure adopts the C3′-endo puckering (C2′-endo in the native complex structure) orienting the 2′-hydroxyl and the 3′-phosphate groups in positions 2.5 Å and 2.2 Å away from their respective positions in the wild type complex structure (Fig. 3C ▶). This difference in the ribose puckering could be attributed to steric hindrances imposed by the alternative conformations of His119. Thus, the 3′-phosphate group in the variant complex structure is placed between the side chains of His12 and His119, forming hydrogen bonds with their imidazoles and with the main chain nitrogen of Phe120. In the wild-type complex, the 3′ phosphate forms a hydrogen bond with the side chain of His119 but no hydrogen bond interactions exist with His12 and Phe120 (Table 5). Instead, there is only a water-mediated interaction of the 3′-phosphate group to the side chain of His12. In addition, the 2′-hydroxyl group in the variant complex structure is hydrogen bonded to His12 Nɛ2 and Lys41 Nζ. In the wild-type complex structure, Lys41 Nζ forms a hydrogen bond with O3′ while O2′ does not participate in any hydrogen bond interactions with the protein (Table 5). Overall, upon binding, U-3′-p forms five and six direct hydrogen bonds with the wild and variant RNase A, respectively (Table 5).

Figure 4.

Superimposed structures of (A) the RNase A–3′,5′-ADP and EDN–3′,5′-ADP (PDB entry 1HI4) complexes; (B) the RNase A–2′,5′-ADP (molecule I) and EDN–2′,5′-ADP (PDB entry 1HI3) complexes; (C) the RNase A–5′-ADP and EDN–5′-ADP (PDB entry 1HI5) complexes, (D) the RNase A–2′,5′-ADP (molecule I) and ECP–2′,5′-ADP (PDB entry 1H1H) complexes. Residues and inhibitor molecules from the RNase A complexes are shown in gray and labeled using the one-letter code, while residues from EDN or ECP are shown in white and are labeled using the three-letter code.

Structural comparison of the RNase A K7A/R10A/K66A variant–U-3′-p complex with the wild-type RNase A–3′-CMP complex (Zegers et al. 1994) showed that the two structures are very similar, and that replacing Lys7, Arg10, and Lys66 by alanines did not affect the binding of U-3′-p (Fisher et al. 1998b). However, U-3′-p binds to the variant protein with three- to fivefold weaker affinity than the one exhibited on binding to wild-type RNase A, and this has been attributed to the long-range coulombic interactions of the mutated residues with the ligand (Fisher et al. 1998b). Our results from the wild-type complex structure also support this hypothesis, and emphasize even further the importance of the long-range coulombic interactions of residues Lys7, Arg10, and Lys66 with the ligand, because although U-3′-p binds with a weaker affinity to the variant protein, there are less hydrogen bond interactions in the wild-type complex structure than in the variant complex structure.

The different conformations adopted by the side chain of His119 in the two complex structures might also contribute to the difference in the affinity. Conformation A, found in the wild-type complex structure and also in the RNase A–UVan complex structure (Ladner et al. 1997), triggers binding of the phosphate group 2.2 Å away from the position it occupies in the variant–U-3′-p complex structure, towards the B₁ subsite. It also promotes a C2′-endo puckering of the ribose, in contrast to a C3′-endo, in the variant. These conformations drive the uridine deeper, by approximately 1.0 Å, in the B₁ pocket, in the native structure than in the variant structure, and allow it to form closer van der Waals contacts with the side chain of Phe120.

Superposition of the structures of U-2′-p and 2′-CMP RNase A (Lisgarten et al. 1993) complexes reveals that the nucleotide and phosphate moieties bind in similar locations to RNase A (Fig. 3D ▶). Their riboses share the same C3′-endo conformation, and both inhibitors participate in a similar hydrogen-bonding network with RNase A. Structural comparison of the structures of U-3′-p and 3′-CMP (Zegers et al. 1994) complexes shows that their binding mode to RNase A is different (Fig. 3E ▶). Their differences are similar to those observed when comparing the structures of the present RNase A–U-3′-p complex to the K7A/R10A/K66A RNase A–U-3′-p complex (Fig. 3C ▶; Fisher et al. 1998b). Thus, in contrast to the U-3′-p complex where His119 is found in conformation A, in the 3′-CMP complex His119 adopts conformation B (Fig. 3E ▶). The puckering of the riboses is also different: C2′-exo in 3′-CMP and C2′-endo in U-3′-p. The pyrimidine moieties of U-3′-p and 3′-CMP (both in anti conformation) bind in close positions (N4–O4 distance 1.0 Å) but their phosphate groups in the P₁ subsite are quite apart (phosphorous-phosphorous distance: 2.4 Å). Thus, major differences are observed again in the interactions of the phosphate group in the active site, and as in the 3′-CMP complex structure, there are two additional hydrogen bonds to His12, and Phe120 and no hydrogen bonds with Gln11 when compared to the U-3′-p complex. The conformation of the side chain of Asp83 in the uridine complexes is different from its conformation in cytidine complexes (Lisgarten et al. 1993; Zegers et al. 1994). In the present uridine complexes, atom Oγ of Thr45 is hydrogen bonded to atom Oδ1 of Aps83, in agreement with previous observations with uridine vanadate (Ladner et al. 1997) and mutation results (delCardayre and Raines 1994, 1995), which indicated that this hydrogen bond strengthens preferentially the binding of uracil.

Comparison to eosinophil ribonucleases

The three central subsites, P₁, B₁, and B₂ of the active site of RNase A, are conserved in EDN (Leonidas et al. 2001a) and ECP (Boix et al. 1999). Structural comparisons have shown that RNase A residues Gln11, His12, Lys41, and His119 of P₁ are conserved as Gln14, His15, Lys38, and His129 in EDN and as Gln14, His15, Lys38, and His128 in ECP. Residues of the pyrimidine specific subsite B₁ Thr45 and Phe120 in RNase A are conserved as Thr42 and Leu130 in EDN and as Thr42 and Leu129 in ECP. Furthermore, the purine binding subsite B₂, constituted by residues Asn71, Glu111, and His119 in RNase A, is formed by the homologous residues Asn70, Asp112, and His128 (His129) in ECP and EDN, respectively (Boix et al. 1999; Leonidas et al. 2001a). The conformation of all the active site residues is conserved between RNase A, EDN, and ECP.

3′,5′-ADP is a potent ribonucleolytic inhibitor with a K_i value of 5 μM for RNase A (Russo et al. 1997), 32 μM for EDN (Leonidas et al. 2001a), and 6–7 μM for ECP (Mohan et al. 2002). Superimposition of the crystal structure of the RNase–3′,5′-ADP complex onto the structure of the EDN–3′,5′-ADP complex (Leonidas et al. 2001a) reveals that the inhibitor binds similarly to the active site of these two RNases, although it occupies two distinct locations approximately 1.5 Å apart (Fig. 4A ▶). The inhibitor adopts similar conformations when bound to either RNase A or EDN, adopting the energetically favored C2′-endo anti conformation (Moodie and Thornton 1993). The 5′ phosphates bind at the P₁ subsite but they occupy different locations with the phosphorous atoms located 1.0 Å apart, upon superposition of the two complex structures (Fig. 4A ▶). Despite their different locations, the 5′ phosphate groups are involved in analogous interactions with residues of the P₁ subsite in the two enzymes. The only difference is the atom of the inhibitor that is hydrogen-bonded to His119 (EDN: His129). In the RNase A complex it is the 5′ oxygen, while in the EDN complex it is one of the phosphate oxygens (Table 2). The adenine rings pack against the side chains of His119 or His129 with both histidines adopting conformation A (Fig. 4A ▶). Differences in the hydrogen bond interactions of two adenines include a hydrogen bond of the N1 group with Asn70 in the EDN complex, not found in the RNase A complex, and a hydrogen bond of the N6 group with Gln69 of RNase A not observed in the EDN complex (Table 2). The 3′ phosphates are located towards the N-terminal ends of RNase A and EDN, respectively (Fig. 5 ▶). Overall, the inhibitor is involved in seven hydrogen bond interactions with RNase A and EDN residues in the corresponding complexes (Table 2).

Figure 5.

A schematic representation of RNase A in complex with 3′,5′-ADP (in green). The inhibitors 2′,5′-ADP (in gray and cyan from molecules I and II of the asymmetric unit, respectively), 5′-ADP (in red), U-3′-p (in magenta) and U-2′-p (in yellow) are also shown.

2′,5′-ADP inhibits the enzymic activities of RNase A, EDN, and ECP with K_i values of 8 μM, 64 μM (Leonidas et al. 2001a), and 6–7 μM (Mohan et al. 2002), respectively at pH 5.5–6.5. The conformation of the inhibitor when bound to RNase A, EDN, and ECP is similar (Fig. 4B,D). The backbone torsion angles γ and δ have similar values, the glycosyl torsion angles χ have values between −66° to −112°, and the nucleotide in all three complexes adopts the anti conformation. The pseudorotation angles of the ribose in 2′,5′-ADP also have similar values in all three complexes, and the ribose is found at the C2′-endo conformation.

The mode of binding of 2′,5′-ADP to RNase A is different in the two molecules of the crystallographic asymmetric unit, as mentioned above. 2′,5′-ADP binds to RNase A molecule I, as it binds to EDN and ECP (Fig. 4B,D). The 2′-phosphate occupies the P₁ subsite and forms similar hydrogen bonds with protein residues in all three complexes (RNase A molecule I, EDN, and ECP; Table 3). The side chains of His119 in RNase A molecules I and II, (ECP: His128, EDN: His129) adopt different conformations. In the RNase A molecule I and in EDN, His119 adopts conformation B (Fig. 4B,D). In these complexes the adenine moiety of 2′,5′-ADP forms stacking interactions with the imidazole ring of His119 (EDN: His129; Fig. 4B ▶) and places its 2′-phosphate group in P₁. In the ECP and RNase A molecule II 2′,5′-ADP complexes, His128 (RNase A: His119) is in conformation A. In the ECP–2′,5′-ADP complex the adenine moiety does not pack against the side chain of His128, but the 2′-phosphate still binds at P₁ (Fig. 4D ▶; Mohan et al. 2002). It is not clear why the adenine of 2′,5′-ADP does not pack against the imidazole ring of His128 in ECP. Interestingly, the adenine ring of 2′,5′-ADP packs against the imidazole ring of His119 in the RNase A molecule II–2′,5′-ADP complex, but then it is the 5′-phosphate and not the 2′-phosphate that binds to subsite P₁.

The 5′-phosphate groups of 2′,5′-ADP in the RNase A molecule I and in the EDN complexes are oriented similarly pointing to the region of subsite P₂ (Lys7 in RNase A and Trp10 in EDN). However, they are not involved in any hydrogen bonding interactions with the two proteins. In the ECP–2′,5′-ADP complex (Mohan et al. 2002) the 5′-phosphate group makes a hydrogen bond interaction with Arg34, which is part of a novel P₋₁ subsite identified recently in EDN(Leonidas et al. 2001a) and ECP (Boix et al. 1999), and is located upstream of the cleavable phosphodiester bond (Table 3). There are also some other differences in the binding of 2′,5′-ADP to RNase A, EDN, and ECP. In the complex with RNase A molecule II, atom N6 of the adenine forms a hydrogen bond with the side chain of Asn71 and in the EDN - 2′,5′-ADP complex atom O3′ of the ribose, forms a hydrogen bond with the main chain oxygen of Leu130 (Table 3).

The RNase A inhibitor, 5′-ADP, is more potent than 3′,5′-ADP and 2′,5′-ADP, with a K_i value of 1.2 μM at pH 5.9 (Russo et al. 1997). However, in EDN, it is less potent than 3′,5′-ADP and 2′,5′-ADP (K_i = 92 μM at pH 6.5; Leonidas et al. 2001a), while in ECP, it is almost as potent as 3′,5′-ADP and 2′,5′-ADP (K_i ∼ 2–3 μM at pH 6.0; Mohan et al. 2002). The inhibitor adopts two different conformations when bound to RNase A and EDN. In RNase A the adenosine is at the syn conformation (χ = 73°), whereas in the EDN it is found at the anti conformation (χ = −100°; Leonidas et al. 2001a). In both complexes the ribose adopts the C2′-endo conformation and the pseudorotation angles v₀–v₅ are broadly conserved. The backbone torsion angles α, β, and δ have also similar values, whereas the γ and ζ_pp differ, with those in the RNase A complex in the −sc and +ap range and those in the EDN complex in the +sc and +ap range, respectively. The mode of binding of 5′-ADP to RNase A and EDN is also different (Fig. 4C ▶). In EDN, 5′-ADP is bound at the active site with the β-phosphate at subsite P₁, like in RNase A, but the adenosine is located in a region away from the B₂ subsite close to, but not at B₁ (in RNase A the adenosine binds at B₂). Thus, the inhibitor is found in two opposite directions within the active site of the two enzymes.

The active-site residue His129 in EDN is found in the B conformation, while His119 in RNase A adopts conformation A (Fig. 4C ▶). The β-phosphate groups are engaged in similar hydrogen bonding networks with RNase A and EDN, involving residues His12, His119, and Phe120 (EDN: His15, His129, and Leu130; Table 4). There are also differences in the binding of the β-phosphate group. In the EDN complex the β-phosphate makes a hydrogen bond to Gln14 (RNase A: Gln11), whereas in the RNase A complex it is hydrogen-bonded to Lys41 (EDN: Lys38). The α-phosphate group makes a single hydrogen bond interaction with Lys38 (RNase A: Lys41) in the EDN complex, while in the RNase A complex forms four hydrogen bonds.

In the EDN complex the adenine does not form any direct hydrogen bonding interaction with the protein, in contrast to the RNase A complex (Table 4). The binding of the adenosine in EDN is stabilized through two hydrogen bonds of the ribose with Gln40 and through van der Waals interactions of the adenine ring with Lys38, Gln40, and His82 (Leonidas et al. 2001a). The difference in the potency of 5′-ADP against RNase A and EDN (5′-ADP exhibits a 75 times lower K_i in RNase A than in EDN) might be attributed to differences in the polar interactions made by 5′-ADP in the two complexes.

Conclusions

The binding of five different phosphonucleotide inhibitors to RNase A was studied. Each inhibitor binds at the catalytic site by anchoring a phosphate group at the P₁ subsite and adopts a completely different conformation (Fig. 5 ▶). The order of potency against RNase A for the five inhibitors studied is 5′-ADP > 3′,5′-ADP > U-2′-p ≥ 2′,5′-ADP > U-3′-p. The occluded surface area of RNase A, calculated with the program GRASP (Nicholls and Honig 1991), by 5′-ADP, 3′,5′-ADP, U-2′-p, 2′,5′-ADP, and U-3′-p, is 638 Ų, 576 Ų, 500 Ų, 455 Ų (molec I), 577 Ų (molec II), and 497 Ų, respectively. Its size follows the potency of the inhibitors against the activity of RNase A. The extensive number of interactions of 5′-ADP with RNase A can explain its higher potency. Although 3′,5′-ADP is a fragment of the physiological substrate of RNase A (RNA) and its binding mode should represent the binding of RNA in the active site, it is involved in fewer interactions with RNase A than 5′-ADP because only one of its phosphate groups is involved in hydrogen bond interactions with the enzyme. The diversity in the binding mode of 2′,5′-ADP in RNase A (it is found in two different binding modes) may explain its high affinity for RNase A despite the smaller number of direct hydrogen bonds with RNase A in comparison to those of 3′,5′-ADP. Significant differences on the phosphate binding of U-2′-p and U-3′-p in subsite P₁ (two distinct locations), may contribute to the difference in the potency between the two inhibitors (U-2′-p has a 10 fold lower K_i for RNase A than U-3′-p; Anderson et al. 1968). The position in the U-2′-P complex allows the phosphate group to form more interactions with P₁ residues than in the U-3′-p complex. Interestingly, similar difference in the locations of the phosphate exists also between the RNase A–2′CMP and the RNase A–3′CMP complexes (Howlin et al. 1987; Zegers et al. 1994) where 2′CMP has a 10-fold lower K_i for RNase than 3′CMP (Anderson et al. 1968).

The RNase A complex structures, described in this work, suggest how further development of tight binding inhibitors of RNases could be achieved through structure driven design. The high resolution of the crystal structures in the present study, allowed the identification of a large number of water molecules in the active site with bound inhibitors. Several of these water molecules mediate interactions between the inhibitors, and the protein and their positions could be the starting point for the design of more potent inhibitors. The adenine of 5′-ADP is involved in water-mediated interactions through two water molecules with the main chain of Asn67 and the side chain of Gln69. The 3′ hydroxyl group of 2′,5′-ADP in its complex with RNase A molecule II, is involved in water-mediated interactions with Glu111 and Val118. Two water molecules link the 5′-phosphate group in the same complex with the main-chain nitrogen of Ala4. The uracyl of U-2′-p is involved in water-mediated interactions with Ser123 and the side chain of Asp83. These water-mediated interactions are also conserved in the U-3′-p complex. All these water molecules indicate positions with high potential to form hydrogen bonds with the protein. Introduction of various substituents to the respective lead molecules could be used to decorate these ligands with groups that will allow them to reach those water positions and exploit their interactions with the protein, thus increasing their potency. The water network of the RNase A active site is partially conserved in the EDN active site as revealed by comparison of the atomic resolution structure of EDN (Swaminathan et al. 2002) with the structures reported here, further emphasizing its importance in the structure-based design of EDN inhibitors.

The structural comparison of the binding of analogous inhibitors in RNase A, EDN, and ECP has also implications for further design efforts of potent specific ribonucleolytic inhibitors of EDN and ECP. Interactions at P₁ are broadly conserved for the three adenylic and the two uridylyl complexes with RNase A, for the three adenylic with EDN (Leonidas et al. 2001a), and for 2′,5′-ADP with ECP (Mohan et al. 2002). Subsite P₁ shows a clear preference for 5′ phosphate groups over 3′ because the binding of 3′,5′-ADP in RNase A (both molecules) and EDN is very similar with the 5′-phosphate group always bound at P₁. This preference is also supported by the binding mode observed by different 3′,5′-ADP derivatives to RNase A, which is very similar to that for 3′,5′-ADP. However, it is not clear whether a 5′- over a 2′-phosphate group is preferable, as the enzyme binds both phosphate groups equally well in different complexes. For example, in the case of 2′,5′-ADP in RNase A molecule II, the 5′- is preferred while in RNase molecule I the 2′-phosphate group binds at P₁.

The adenine base occupies a different location in each adenylic complex with RNase A, EDN, and ECP. The position observed in the 3′,5′-ADP complex and in the 2′,5′-ADP complex (RNase A molecule II) seems to be a substrate analogous position because it has also been observed in previous RNase A and EDN nucleotide complexes (Leonidas et al. 2001a). The additional positions observed in the RNase A–5′-ADP, the RNase A molecule I–2′,5′-ADP, and the ECP–2′,5′-ADP complexes seem to represent equivalent locations within the B₂ subsite that can accommodate an adenine base, and in each case there are sufficient protein interactions to support the binding. Moreover, it seems that there is a correlation between the binding of a specific phosphate group (5′-or 2′-) and the position occupied by the adenine. Thus, when the 5′-phosphate group binds to P₁, the nucleotide adopts a conformation that allows packing of the five-member ring of adenine against the imidazole of His119 (3′,5′-ADP in RNase A, and in EDN; 2′,5′-ADP in RNase A molecule II). In contrast, when the 2′-phosphate is bound to subsite P₁ the nucleotide adopts various conformations placing the adenine in different locations in RNase A, EDN, and ECP.

The binding mode of 5′-ADP to RNase A confirms the preference of subsite P₁ for the β- rather than the α-phosphate group when a pyrophosphate group is present in the inhibitor. This preference has been previously shown in complexes of RNase A with ppA-2′-p, ppA-3′-p, dUppA, and pdUppA-3′-p (Leonidas et al. 1997, 1999; Jardine et al. 2001). The binding of the β-phosphate seems to force the nucleotide to adopt the unusual syn conformation and to place the six-member ring against the imidazole of His119. The only case where it adopts the most preferred conformation anti (Moodie and Thornton 1993) is in the EDN–5′-ADP complex (Leonidas et al. 2001a), and the adenosine is placed outside subsite B₂ to an opposite direction, towards subsite B₁. All these observations reveal a high flexibility of subsite B₂, which can accommodate an adenine base in a variety of locations.

In contrast, subsite B₁ seems to be more confined, and bound uridines are always placed in almost identical positions with the uridine ring being packed against the phenyl ring of Phe120 and the 2′-hydroxyl group in hydrogen bonding distance from the side chain atoms of Thr45. This has been observed in the current complexes of RNase A with U-2′-p and U-3′-p as well as in RNase A in complex with other uridylyl nucleotides (Leonidas et al. 1999; Jardine et al. 2001). However, the phosphate group binds to P₁ despite its position in the ligand (either 2′- or 3′-). The binding of either the 2′- or the 3′-phosphate group to P₁ does not seem to affect the binding mode of uridine to subsite B₁, as shown in the RNase A complexes with U-2′-p and U-3′-p in contrast to adenylic inhibitors.

Comparative structural analysis of different complexes of nucleotide inhibitors to RNase A, EDN and ECP shows that the interactions at the P₁ subsite are the main driving force for the binding of all the phosphonucleotides to RNases. Our results suggest that the specificity of P₁ surpasses the specificity of the other subsites, and therefore, subsite P₁ should be the target for the design of new RNase inhibitors. We also showed that each inhibitor does not necessarily bind to RNase A, EDN, and ECP in the same mode as in the case of 2′,5′-ADP and 5′-ADP, and that subtle differences in the enzymic active sites can lead to profound alterations in the binding modes of an inhibitor to these RNases. These findings emphasize further the importance of obtaining direct structural information on the binding of each new inhibitor with RNase A and each RNase homolog in the process of structure-based inhibitor design.

Materials and methods

Bovine pancreatic RNase A (type XII-A) and phosphonucleotides were obtained from Sigma. Crystals of RNase A were grown at 16°C using the hanging drop vapor diffusion technique as described previously (Leonidas et al. 1997) ; they belong to space group C2, with two molecules per asymmetric unit and cell dimensions a = 99.7 Å, b =32.5 Å, c = 72.2 Å, and β = 90.03°. Crystals of the inhibitor complexes were obtained by soaking the RNase A crystals in 20 mM sodium citrate, pH 5.5, 25% PEG 4000, and 50 mM, 2′,5′-ADP, 3′,5′-ADP, U-2′-p, U-3′-p, or 5 mM 5′-ADP for at least 4 h prior to data collection. Diffraction data for the 3′,5′-ADP, 2′ UMP, and 3′ UMP complexes, were collected to 1.5 Å resolution at 100 K using an R-AXIS IV image plate mounted on a RIGAKU RU-H3RHB rotating anode X-ray source with CuKα radiation. Data for the 5′-ADP and 2′,5′-ADP complexes were collected to 1.2 Å resolution at 100 K at The European Synchrotron Radiation Facility (ESRF) at Grenoble (France) on station ID14-2 (λ = 0.933 Å) using an ADSC Q4 CCD detector, and at the Protein Diffraction Beamline (λ = 1.0006 Å) at Elettra-Trieste (Italy) using a MAR345 image plate, respectively. Three different data sets (589 frames in total) were collected for the 2′,5′-ADP complex, to 1.2 Å, 1.48 Å, and to 2.3 Å resolution, to ensure completeness at the low resolution shell (92% for the 30.0–3.26 Å resolution shell). Similarly for the 5′-ADP complex, two different data sets (to 1.2 Å and 2.6 Å, 376 frames in total) were collected to ensure completeness at the low resolution shell (30.0–3.26 Å: 87.8%). Raw data were processed using the HKL program suite (Otwinowski and Minor 1997), and intensities were truncated to amplitudes by the TRUNCATE (French and Wilson 1978) program. Phases were obtained using the structure of free RNase A from monoclinic crystals (Leonidas et al. 1997) as a starting model. Alternate cycles of manual building with the program O (Jones et al. 1991), and refinement by torsion angle dynamics, conjugate gradient minimization, restrained individual B-factor refinement, and bulk solvent correction using the maximum likelihood target function as implemented in the program CNS v1.1 (Brünger et al. 1998) improved the model. Ligand molecules were included during the final stages of the refinement by use of their respective coordinates from the Protein Data Bank (ID codes: 2′,5′-ADP, 1HI3; 3′,5′-ADP, 1HI4; 5′-ADP, 1HI5 [Leonidas et al. 2001a]; U-3′-p, 4RSK [Fisher et al. 1998b]). A model of U-2′-p, was generated using the program SYBYL (Tripos Associates Inc.). Furthermore, the 1.2 Å resolution data from the 2′,5′-ADP and 5′-ADP complexes enabled the anisotropic refinement of temperature factors with the program REFMAC (Murshudov et al. 1997). Details of data processing and refinement statistics are provided in Table 6.

Table 6.

Crystallographic statistics

RNase A complex	2′,5′-ADP	3′,5′-ADP	5′-ADP	U-2′-p	U-3′-p
Resolution (Å)	30–1.2	30–1.5	30–1.2	30–1.5	30–1.5
(Outermost shell) (Å)	1.22–1.20	1.55–1.50	1.22–1.20	1.55–1.50	1.55–1.50
Reflections measured	949,552	423,949	623,242	278,544	239,794
Unique reflections	67,242	36,948	70,118	37,206	37,930
Rsymma	0.038	0.062	0.053	0.052	0.061
(Outermost shell)	0.226	0.369	0.130	0.390	0.294
Completeness (outermost shell) (%)	92.9	97.9	96.5	98.3	99.4
	71.2	95.5	96.9	95.6	99.6
<I/σI> (outermost shell)	30.9 (10.8)	15.1 (3.6)	20.9 (5.8)	14.1 (2.5)	15.4 (2.6)
Rcrystb	0.194	0.229	0.190	0.211	0.220
(Outermost shell)	0.261	0.446	0.195	0.442	0.412
Rfreec	0.228	0.271	0.219	0.24	0.240
(Outermost shell)	0.277	0.452	0.202	0.446	0.436
Number of solvent molecules	558	389	540	399	415
R.m.s. deviation from ideality
in bond lengths (Å)	0.011	0.004	0.011	0.004	0.004
in angles (°)	1.7	1.2	1.8	1.3	1.3
Average B factor
Protein atoms (Å2) (molecule I/molecule II)	18.6/16.7	18.8/23.6	18.2/17.7	19.0/18.7	19.0/20.2
Solvent molecules (Å2)	30.9	33.0	33.1	30.7	30.8
Ligand atoms (Å2) (molecule I/molecule II)	17.8/31.1	22.4/34.4	13.6/21.3	18.0/30.6	19.9/26.8

^aR_symm = ∑_h∑_i│I(h) − I_i(h)/∑_h∑_iI_i(h) where I_i(h) and I(h) are the ith and the mean measurements of the intensity of reflection h.

^bR_cryst = ∑_h│F_o − F_c│/∑_hF_o, where F_o and F_c are the observed and calculated structure factors amplitudes of reflection h, respectively.

^cR_free is equal to R_cryst for a randomly selected 5% subset of reflections not used in the refinement (Brünger 1992).

The atomic coordinates and diffraction data for the five RNase A complexes have been deposited in the Protein Data Bank, Research Collaboratory for Structural Bioinformatics, Rutgers University (http://www.rcsb.org; PDB ID Codes are 1O0F, 1O0O, 1O0H, 1O0M, and 1O0N for the 3′,5′-ADP, 2′,5′-ADP, 5′-ADP, U-2′-p, and U-3′-p RNase A complexes, respectively). All figures were prepared with the program Molscript (Kraulis 1991), and rendered with RASTER3D (Merritt and Bacon 1997).

Abbreviations

• RNase A, bovine pancreatic Ribonuclease A

• Ang, angiogenin

• EDN, eosinophil derived neurotoxin

• ECP, eosinophil cationic protein

• 3′,5′-ADP, adenosine-3′,5′-diphosphate, 2′,5′-ADP, adenosine-2′,5′-diphosphate, 5′-ADP, adenosine-5′-diphosphate, pdUppA-3′-p, 5′-phospho-2′-deoxyuridine 3-pyrophosphate (P′→5′) adenosine 3′-phosphate

• ppA-3′-p, 5′-diphosphoadenosine 3′-phosphate

• ppA-2′-p, 5′-diphosphoadenosine 2′-phosphate

• dUppA-3′-p

• 2′-deoxyuridine 3′-pyrophosphate (P′→5′) adenosine

• PEG, poly(ethylene glycol)