The intricate three-dimensional structure of the SARS s2m RNA, along with its rigorous sequence conservation, is compelling prima facie evidence for its biological importance in coronaviruses and astroviruses. The structure by itself, however, does not indicate what the function of this motif must be. Hence, comparison of this unique fold with those of known RNA structures is of particular value for formulating testable hypotheses regarding potential biological functions of the s2m RNA. In addition, identification of novel and rigorously conserved tertiary structures that are unique to the viral RNA is of critical importance for future rational design of anti-viral therapeutic agents that specifically target SARS and other coronaviruses and astroviruses.
Biological Relevance of the s2m Sequence and Crystal Structure
The s2m RNA sequence we crystallized was originally identified from the genomic sense strand within a rigorously conserved region of the 3′ UTR of the RNA. However, because RNA replication and transcription take place via a full-length negative-strand RNA intermediate, it is formally possible that the conserved sequence instead corresponds to a conserved structure at the 5′ end of the anti-sense RNA. We believe this to be improbable because of the energetically unfavorable tertiary structures that would be required to form from the sequence complement. For example, the variant of the energetically stable and rather common GNRA loop structure (GAGUA) would have to be replaced with an energetically unstable and rare CUCAU loop. Similar arguments apply to the other non-Watson–Crick regions of the structure.
Crystal packing interactions may potentially distort RNA structures. This effect is sometimes observed for small stem-loop sequences, which often crystallize as duplex dimers rather than as monomeric hairpins. The s2m RNA structure is sufficiently large, and apparently contains enough stabilizing secondary and tertiary interactions, to offset any energetic advantage that might come from crystallizing as a duplex. In addition, the 73% solvent content of the s2m RNA crystals ensures that most of the crystallized RNA is solvent-exposed, rather than involved in extensive packing interactions. At least three inter-molecular contacts are required to form a crystal. The most extensive contact is the base of residue G(11); it stacks upon that of its 2-fold symmetry mate (C). It is likely that these nucleotide bases become oriented in such a way as to optimize this stacking interaction. The nonessential nucleotide G(1) forms a weak (3.4-Å) hydrogen-bonding interaction with A(29) of an adjacent molecule, but most of this packing interaction appears to be due to shape complementarity and is thus expected to have little distorting effect. The remaining interaction is a nonspecific, presumably cation-mediated backbone parallel helical interaction, again unlikely to result in significant distortions.
Crystallographic temperature factors provide direct physical evidence for the relative flexibility or mobility of various regions of a macromolecule. D shows relative temperature factors color-coded on all non-hydrogen atoms (left) and on the RNA phosphate backbone atoms (right). Blue atoms have the lowest relative temperature factors and red atoms have the highest. Consistent with the observed electron density map, by far the most flexible region of the RNA is U(25). U(30) and the 5′-terminal triphosphate are also moderately disordered. Much of the rest of the structure appears to be rather rigid and well defined, including the three-purine asymmetric bulge and the seven-nucleotide asymmetric bubble, along with the hydrated magnesium complex ions that bind to the non-bridging phosphate oxygens of A(12). The phosphate backbone atoms of these non-Watson–Crick regions are among the most ordered in the structure.
Therefore, based on our chemical probing data, analysis of crystal packing interactions, and consideration of the crystallographic temperature factors, along with the ability to rationalize the sequence conservation pattern and intolerance for nucleotide insertions or deletions based on the structure, we conclude that the crystal structure of s2m is likely to be a close representation of the structure that forms in solution and in the context of the SARS virus RNA genome.
Functional Implications of the s2m Three-Dimensional Structure
The several unique features and unanticipated tertiary contacts we identified in the SARS s2m RNA crystal structure allowed us to reexamine genomic sequences and previously determined RNA tertiary structures for similar motifs with additional constraints imposed by knowledge of the tertiary structure. Our analysis of the human genome, other animal and viral genomes, and the currently available database of RNA three-dimensional structures revealed that the s2m element is found only in astroviruses and coronaviruses; no cellular homologs are immediately apparent. The G(11) to A(33) tertiary contact in the s2m RNA is homologous to the G(1,452) to A(1,486) contact in Domain III of the 23S ribosomal RNA, but the context of the interaction in the ribosome is completely different, and the sequence is not conserved between Escherichia coli
and Thermus thermophilus
. However, if we relax the sequence constraints and focus attention upon the conformation of the RNA backbone, we find that the phosphodiester backbone fold accompanying the 90° kink in s2m RNA mimics that found in the 530 stem-loop of 16S ribosomal RNA [19
] (A). The latter binds to the S12 protein found at the interface between the small and large ribosomal subunits. The 530 stem-loop, and the S12 protein that binds to it, have been implicated in EF–G-independent ribosomal translocation [20
]. Remarkably, superposition of the s2m RNA upon the 530 stem-loop within the 30S ribosome in which prokaryotic initiation factor 1 (IF-1) has been added [21
] reveals plausible modes of s2m RNA binding to both the S12 protein and to IF-1 (B). Both S12 and IF-1 have eukaryotic homologs; the structure of IF-1 and its eukaryotic analog, eIF-1A, possess almost identical RNA oligomer binding (OB) folds [22
]. Based upon these structural homology arguments, we propose that the SARS s2m RNA is a functional macromolecular mimic of the 530 loop of the small subunit ribosomal RNA (which is conserved in eukaryotes). Mechanisms of translation and protein synthesis regulation via macromolecular mimicry are in fact well established [24
]. We propose, on the basis of the similarity between the 530-loop fold and the s2m fold, that the s2m RNA of SARS may be capable of binding one or more eukaryotic proteins whose structures resemble S12 or the OB folds typical of these ribosomal proteins, and that each would do so in a manner similar to that shown in B. This proposal leads us to formulate two separate, testable hypotheses regarding the function of the s2m RNA in SARS.
SARS Virus RNA Macromolecular Mimicry
Does s2m Macromolecular Mimicry Facilitate Viral Hijacking of Protein Synthesis?
eIF-1A, like IF-1, possesses an OB fold. Our first hypothesis is that eIF-1A may bind to the 90° bend of the SARS s2m RNA. In addition, we suggest that the function of the s2m RNA of SARS and related viruses might involve viral hijacking [26
] of the cell's protein synthesis machinery, either facilitating mRNA circularization and ribosome re-initiation, in gross analogy to viral internal ribosomal entry site–mediated mechanisms [27
], or perhaps even more simply by titrating eIF-1A away from the host initiation complexes and thus inhibiting host cell protein synthesis in favor of viral protein synthesis by sequestering a factor required by the host.
Does s2m Bind to the nsp9 SARS Protein to Facilitate Virus Transcription?
Recently, two protein structural genomics investigations of SARS revealed the structure of a so-called nonstructural protein, nsp9, that is believed to be involved in viral RNA synthesis and to interact with the viral polymerase in an unspecified manner [29
]. The crystal structure of nsp9 reveals it to be a variant of the OB fold, a protein structural motif not previously recognized to be involved in viral replication. The authors demonstrate nonspecific single-strand RNA binding affinity for nsp9. We propose that nsp9, by virtue of its OB fold, may bind specifically to s2m in a manner similar to that illustrated in B, and may thus facilitate viral polymerase RNA transcription, translation, or replication.
From Structure to Functional Predictions
Our structural genomics analysis of the SARS RNA has thus enabled us to formulate specific, experimentally testable hypotheses regarding the function of a highly conserved RNA motif whose importance has been evident [2
] but whose biological activity hitherto was completely unknown. The possibility that the 90° bend of the s2m RNA binds to an OB-like protein permits us to propose two potential mechanisms of interaction relevant to the two main functions of the SARS virus (protein synthesis and viral replication). The possibility of additional interactions with proteins at the S12-like site and in the highly structured and rigorously conserved tunnel region formed by the three-purine bulge and the seven-nucleotide bubble should also not be overlooked, as these both are likely sites for RNA–protein or RNA–RNA interactions that are crucial to the function of the SARS virus, and therefore also merit further attention.
The s2m RNA Tunnel Is an Attractive Target for the Design of Anti-SARS Drugs
C and D dramatically illustrates the most striking and unique structural feature within the SARS s2m RNA. A tunnel is created by the tertiary contacts between A(33) of the purine asymmetric bulge (red), G(11) and A(12) of the seven-nucleotide bubble (green), and the helical region between them (purple). The non-bridging phosphate oxygens of G(11) and A(12) line the surface of the cavity, creating a negatively charged region into which Mg2+ ions are observed to bind. It is likely that in the context of the virus, this invariant feature of the s2m structure is involved in binding interactions with highly conserved proteins or other components of the host cell that interact specifically with the negatively charged cavity. Because this tunnel structure is unique to coronaviruses and astroviruses and because the sequence comprising this structure is invariant, it is reasonable to propose that by designing a drug that specifically targets this structural feature and binds tightly to it, an anti-SARS therapeutic might be obtained that avoids the pitfall of being toxic to uninfected host cells while escaping the usual problem of drug resistance that develops in rapidly mutating RNA viruses.