Ribosome-inactivating proteins (RIPs) are a group of ribotoxins, which inhibit mammalian protein biosynthesis by irreversible inactivation of their 60S ribosomal subunit.1 The physiological role of type II RIP in plant cells is still not completely understood. However, due to their cytotoxicity, it is generally believed that they comprise part of the plant's defense system towards animals. Of particular interest is their ability to selectively kill tumor cells in preference to normal cells.2 Thus, an interesting new approach using RIPs as immunotoxins has been proposed for cancer therapy. Based on their primary amino acid sequences, RIPs were divided into three categories. Type I RIPs consist of a single peptide chain with a molecular mass that varies from 11 to 30 kDa, whereas type II RIPs are composed of two polypeptide chains (A- and B-chain) linked by a disulfide bond. Both the type I RIPs and the A-chain of type II RIPs contain a single RNA N-glycosidase domain. The activity of the latter can remove an adenine base from the highly conserved sarcin/ricin domain in the largest RNA of the mammalian ribosome. This abolishes the ribosome's ability to bind elongation factors, thereby inhibiting protein biosynthesis.3, 4 The B-chain of type II RIP is a galactose-specific lectin that is responsible for the recognition of D-galactose of galactose-terminated glycoproteins and/or glycolipids found on the surface of eukaryotic cells, and facilitates the internalization of the toxic A-chain. Additionally, type III RIPs are composed of a type I RIP-like N-terminal domain covalently linked to a C-terminal domain of unknown function.5
Cinnamomin is a type II RIP isolated from Cinnamomum camphora seeds. Its physico–chemical, biochemical, and enzymatic properties have been documented.6–9 There are three functional genes encoding highly homologous cinnamomin isoforms (I,II,III) with a sequence identity of >98.0% between them.10 Cinnamomin shows inhibitory effects on cultured carcinoma cells2 and on the larva of the bollworm and mosquito.11 The functional protein is composed of two A and B chains with molecular mass of 30 kDa and 33 kDa, respectively. It has a total of 535 residues, 271 of which are in the A-chain and 264 of which are in the B-chain. Both chains are glycosylated with 0.3% sugar content in the A-chain and 4.3% in the B-chain.9 The N-linked oligosaccharide chains in the B-chain have been determined by NMR analysis.12 DTNB (5,5′-Dithio-bis 2-Nitrobenzoic Acid) titration experiments demonstrated that in cinnamomin, one cysteine is found in the A-chain and nine other in the B-chain.13 The B-chain is responsible for the binding to galactose-terminated receptors on cell membranes, whereas the A-chain displays specific activity as an RNA N-glycosidase that blocks protein synthesis.10 It has been shown that cinnamomin was two orders of magnitude less toxic to BA/F3B cells than ricin. The purified A-chains from the two RIPs had an almost equivalent IC50 for the inhibition of protein synthesis in rabbit reticulocytes.14 This suggests that the B-chains of the RIPs are implicated in cytotoxicity differences.
To understand the structure-function relationship of RIPs and differences between them, we report the first crystal structure of native cinnamomin. The structural information will be compared with demonstrated enzyme activity and specificity, facilitating the understanding of the mechanism of the protein with potential therapeutic application.
MATERIALS AND METHODS
Extraction and purification of cinnamomin from seeds
Cinnamomin was purified from Cinnamomum camphora seeds following a fast purification protocol as described earlier15.
Crystallization and data processing have been previously described.15 Cinnamomin was concentrated to 50 mg/mL in 50 mM Tris buffer pH 8.0. Protein crystals were grown using hanging drops in the presence of 0.15M CaCl2, 0.1M sodium cacodylate pH 6.5, and 18% (w/v) PEG 8000 at 298 K.15 The crystals were protected in 20% glycerol added to the reservoir solution for 1 min and frozen at −180°C during data collection.
Structure determination and refinement
The sequence of cinnamomin isoform III precursor (accession file AAK82460.1) from Cinnamomum camphora was used. The crystal structure was successfully determined by molecular replacement with the program MolRep16, using the structure of ricin (pdb code: 2AAI17) as a search model. Two molecules were located in the asymmetric unit; the phases were improved with noncrystallographic symmetry (NCS) averaging with the program DM18. The atomic model was built against a 2Fo-Fc electron density map with the program Coot.19 The structure refinement was carried out with the program Refmac20. Because of its low resolution, NCS restrained were introduced in the REFMAC protocol, medium for the main chain and side chains (REFMAC defined options). A few final cycles were performed without any NCS restraints. Small electron density patches corresponding to residues 255–271 at the C-terminus of A-chain were left unmodeled because any extension led to strong over fitting demonstrated by an increase in Rfree and a decrease in Rfact. These residues are therefore not included in the model. Water molecules and glycosylation sugars were added at the last stage of the refinement in Fo-Fc electron density map. The quality of the model was analyzed by the programs Procheck and Molprobity.21, 22 Schematic diagrams were generated with LIGPLOT.23 The Protein Data Bank accession number for cinnamomin is 2VLC.
RESULTS AND DISCUSSION
Overall cinnamomin structure
In this study, we have determined the crystal structure of this type II RIP that has been purified from the Cinnamomum camphora seeds. This natural source contains three isoforms of cinnamomin that have similar molecular mass as they appear as a single band in gel electrophoresis and therefore could not be separated during the purification procedure reported earlier.15 However, after the crystallization and subsequent structure determination, we found that the cinnamomin sequence in the crystal structure matches that of isoform III. This suggests that the crystallization conditions15 have favored the selective crystal growth of only this isoform. Such crystallogenesis may be related to crystal packing that is worth for further study. The structure of cinnamomin III has been determined at 2.95 Å resolution. There are two copies of cinnamomin in the crystallographic asymmetric unit. The final model contains 518 residues for each molecule in the asymmetric unit, totalling 1036 residues. Each molecule has an A-chain with 254 out of 271 residues and a B-chain with 264 residues (see Fig. 1). The latter contains four intrachain disulfide bonds between Cys305-Cys324, Cys348-Cys365, Cys436-Cys449, and Cys475-Cys493. The statistics for the data collection and structure refinement are summarized in Table I. The overall structure of cinnamomin is very similar to that of abrin (PDB code 1ABR) and ricin (PDB code 2AAI). The root mean square deviations (r.m.s) calculated for the structurally related α-carbon atoms of cinnamomin and ricin is 1.08 Å and 1.17 Å between cinnamomin and abrin. These values are larger than those calculated separately for the A-chains namely 0.91 Å for ricin and 1.00 Å for abrin and for the B-chains: 0.88 for ricin and 0.89 Å for abrin. These differences indicate that different interactions occur between their A and B-chains, although the overall folding is very similar between these RIPs.
Table I. Crystal and Refinement Statistics of Data Collected from Cinnamomin Crystal
In cinnamomin, the A and B-chains are connected by a single disulfide bond, between A-chain Cys250 and B-chain Cys289, that can be reduced without loss of toxicity.14 The area of contact between cinnamomin's A and B-chains is very similar to that of ricin (1575 Å2) and is smaller than that observed in abrin (1770 Å2). The two chains are held together by several polar and nonpolar interactions (see Fig. 2) with a concentration of the hydrophobic interactions at the center of the contact area. The interactions between the two chains seem to be important in maintaining chain association at very low toxin concentrations.
A-chain structure-function analysis
The cinnamomin A-chain can be divided into three domains: an N-terminal domain, a middle domain, and a C-terminal domain. The N-terminal domain is composed of residues 1–114 consisting of a six-stranded β-sheet (βa through βf) and two helices αA and αB. The helices and strands alternate in the order aAbcdeBf. The first strand (βa) and the last (βf) lie parallel to the neighboring strands, βb and βe, respectively. The four central strands of β-sheet, b to e, are antiparallel. Among the structural differences between cinnamomin, ricin and abrin is the distinct folding of loops bc (residue 57–63) and eB (residue 90–93).
The middle domain consists of five α-helices, αC to αG, between residues 115–191. Overall folding is similar to abrin and ricin except the loop de (residues146–151) in which the three structures present distinct folding.
The C-terminal domain is composed of residues 192–254 and consists of two α-helices and two β-strands: αE (residues 192–211) is kinked at the active site residue Trp201 in the three structures, αF: 239-243, βg (residues 221−224) and βh (residues 230–233). The two antiparallel strands βg and βh are arranged as EghF.
The cinnamomin A-chain active site residues have roughly the same position and side chain orientation as those of ricin and abrin. One exception is that the side-chain of Tyr75 in cinnamomin is rotated in a position more similar to that found in abrin. In ricin, it is necessary to rotate the tyrosine side chains to accommodate the adenine between them.24
The location of the active site region of the cinnamomin A-chain is illustrated in Fig. 3. A larger area of density, even with the highest contour level (up to 5.8 σ in the (2Fo–Fc) map) is located between tyrosines 75 and 115 (see Fig. 3), where an adenine molecule is expected to be in the molecule in complex as demonstrated in ricin structure.24 Water molecules wat118 and wat119 occupy this position in their respective asymmetric unit in the present structure.
Cinnamomin A-chain active site is composed of the invariant residues Tyr75 in loop cd, Tyr115 on helix αC, Glu167 and Arg170 on αE, Trp201 on αH, and conserved residues Glu70, Asn73, Arg126, Gln163, Glu198, and Asn199. There are two hydrogen bonding interactions Glu198 to Gln163 and Asn199 to Arg126 that link helix αG to both helix αE and loop cd. These interactions appear to have a structural role in stabilizing helix αE and the position of catalytic residues Arg170 and Glu167. Other interactions that maintain active site structural stability are observed as Tyr115 makes a hydrogen bond to Glu167 which is also hydrogen-bonded to Arg170. The role of the invariant active site residues Glu167, Arg170 was confirmed by mutagenesis studies demonstrating that the cinnamomin Glu167Asp variation led to an almost 50-fold loss in enzyme activity, whereas Arg170Lys resulted in a 45-fold loss of activity.25 According to the existing concept of protein inhibition activity of RIPs, these residues are directly involved in the catalytic process of NC glycosidic bond hydrolysis. Modest reduction in glycosidase activity occurred for variants Tyr75Phe (20fold), Tyr115Phe, and Trp201Phe (8 fold each, see Ref.25) confirming their role in binding rather than in catalysis. In the prevailing reaction mechanism transition state established from ricin studies,24 the oxycarbonium ion on the ribose is stabilized by Glu167. Argl70 stabilizes anion development on the departing adenine by protonation at N-3 and may activate a water molecule that is the ultimate nucleophile in the depurination. The mechanism of action proposed for ricin is also applicable to cinnamomin due to the similarity of the active site structures, especially at positions of Glu167 and Arg170 in the cinnamomin, abrin and ricin A-chains.
Structural comparison with the abrin and ricin structures reveals two characteristic sites on the surface of the A-chain of cinnamomin. The first site is similar in both cinnamomin and ricin and is composed of three aligned phenylalanine, Phe86 on strand βe, Tyr109 on βd, and Tyr107 on βf, oriented in an energetically favorable position (see Fig. 4). In Abrin, the site is twice the length and include a stack of 6 aromatic and imidazole rings between histidines 105 and 136. The second site is perpendicular to the first and is similar in ricin and cinnamomin. It is formed by four stacked aromatic rings that are nearly aligned on an axis with a distance of 4.5–5.5 Å between the phenyl centroids (see Fig. 4). In cinnamomin, the aligned residues are Tyr79, Phe109, Phe111, and active site residue Tyr75 oriented differently. One aromatic ring missing in abrin's site and is replaced by Leu167. The positions of these aromatic residues suggest a role in the interaction with ribosomal RNA by stacking, to orient the ribosomal sarcin/ricin stem loop bearing the GAGA tetraloop to the active site. These differences between cinnamomin, abrin, and ricin A-chains indicate a varying platform for RNA interaction that may influence ribosome recognition and binding.
There is one glycosylation site on the A-chain sequence at Asn10-Ala-Thr. Close examination of the electron density map revealed an extended area of electron density where a di-N-acetyl-glucosamine (GlcNAc) can be fitted as follows: GlcNAcβ1,4GlcNAcβ1-Asn10-Ala-Thr.
This site corresponds to glycopeptide GP2 identified by NMR analysis.12 Interestingly, sequence alignment (data not shown) reveals that this site is absent in ricin and abrin but is present in saporin (Asp10-Pro-Thr), a type I RIP.26
B-chain structure-function analysis
Although the overall folding of the cinnamomin B-chain is similar to the abrin and ricin B-chain, the differences are located on the conformation of 7 loops whose locations are: L1: 326 to 331, L2: 386 to 392, L3: 396 to 401, L4: 425 to 429, L5: 451 to 455, L6: 467 to 471, and L7: 481 to 488. Sequence alignment of these loops reveals that differences in L1, L3, and L7 conformation are likely due to the sequence. On the other hand, the sequences of loops L2, L4, L5, and L6 are highly conserved but folded differently.
The cinnamomin B-chain is composed mainly of β-strands (see Fig. 1) and contains 4 intrachain disulfide bonds and two homologous domains. The main body of each domain formed by the α, β, and γ subdomains is situated around a pseudo-three fold axis. The amino acid sequence alignment for subdomains α, β, and γ of both domains based on structural similarity shows that all these α, β, and γ subdomains are homologous (Table II). Sequence comparison with other RIP sequences (data not shown) shows that subdomain 2β has the lowest sequence similarity with analog domains in ricin and abrin.
Table II. Structure Based Superimposition of the Homologous Subdomains of the Cinnamomin B-Chain with Conserved Residues Highlighted in Grey
A large area of electron density on 2Fc-Fo and Fo-Fc maps protrude from to asparagines 380 and 421 indicating their glycosylation. The two N-linked oligosaccharide sequences were partially located in the current electron density map.
GP1: Xylβ1-2 Manβ1-4 GlcNAcβ1-4GlcNAcβ1-Asn421
GP: Manβ1-4 GlcNAcβ1-4GlcNAcβ1-Asn380
These sites correspond to glycopeptide GP and GP1 previously identified by NMR analysis.12
Future studies may address the modulation of the toxicity of this class of proteins by varying the residues involved in activity and binding. The potential for designing the most effective combination of toxicity and binding ability could prove to be valuable.