The structural and functional results presented here show that the LCMV NL1 domain is an RNA endonuclease. The uncoupling of RNA replication from transcription and selective disappearance of mRNA when NL1 active site residues are mutated strongly suggests that this activity is involved in cap-snatching.
The identification of the arenavirus endonuclease is in line with the recent discovery of the PAN
endonuclease domain of influenza virus. Whereas the active site of influenza virus features a cluster of three acidic residues, the active site of arenavirus contains four acidic residues (E51, D89, E102 and D119), as well as two important lysine residues K115 and K122 neighboring D119 (). The NL1 active site resembles but is clearly distinct from that of influenza PAN
. Indeed, there is no histidine in the catalytic center, and the arenavirus NL1 nuclease has some specific features both upstream and downstream of the PD signature sequence. We define the arenavirus endonuclease motif as E-X38
-K. The most obvious difference with the only known related RNA endonuclease, that of influenza virus PAN
, is a divergence upstream the PD motif in structural elements carrying the E51 residue (), and the presence of a triad K…D…K at the distal side of the latter signature sequence (). Contrary to PAN
which shares a conserved and essential histidine involved in the binding of both the metal ion and a nucleotide onto helix α3 
, NL1 does not possess this conserved histidine residue. Instead, NL1 has a glutamic acid residue E51, which might reflect a different nucleobase specificity as detected in our nuclease assays (). Likewise, residues downstream the PD motifs are distinct from the consensus sequence, and differently organized into a triad including two lysines. The presence of water molecules and previous structural models for influenza PAN
allows to propose putative positions of metal ions, coordinated by D89 and E102.
The first step in the general mechanism for phosphodiester hydrolysis is the preparation of the attacking nucleophile by deprotonation, usually involving a general base deprotonating a water molecule. Lysine is often considered as this general base candidate in endonucleases but is not strictly conserved 
. Here, there are no indications against D119 being this general base. Alternately, it could well be either lysine K115 or K122. Both are oriented towards the active site, and they could well have their pKa lowered by D119 in order to initiate the reaction. Reverse genetic studies provide evidence for K122, not K115. Indeed, mRNA production is selectively abolished and clearly uncoupled from RNA synthesis in the case of K122A mutant, while the K115A mutant was completely defective preventing interpretation of its role in the endonuclease catalytic site. Although it is not known if uncapped mRNAs are synthesized and degraded for the transcription-null mutants, the most plausible scenario is that primer shortage prevents significant capped mRNA synthesis. Overall, the replicon data presented here closely match those obtained on the closely related Lassa arenavirus using a similar replicon system 
. Arenaviruses may thus use two clearly independent and distinct RNA synthesis priming mechanisms: one is dependent on an active endonuclease carried by the N-terminus of the L protein, and the other might be linked to the observation that an extra G residue is found at the 5′-end of arenavirus genomes and antigenomes. The latter G bases would thus reflect a yet-uncharacterized priming mechanism unrelated to the U/A cleavage sequence preference of NL1.
NL1 also binds nucleotides, but the NTP binding site should differ from that of PAN. Indeed, the influenza PAN histidine 41 is involved in binding the nucleobase of the presumed incoming RNA substrate. The NL1 endonuclease does not share the same sequence specificity, and E51 is positioned at a spatially equivalent position.
The cap structure does not seem to be a direct RNA binding determinant (), as endonucleolytic cleavage is not directed to cleavage sites preferentially in the vicinity of the cap. We thus infer that an independent cap-binding site way exist elsewhere in viral proteins to bind and select cellular mRNAs, a possibility reminiscent of influenza for which PA carries the endonuclease activity and PB2 the cap binding site 
Structure and sequence alignment studies show that the N-terminal endonuclease domain of the L protein is also conserved in the Bunyaviridae family, although the Nairovirus endonuclease domain is not located into the N-terminal end of the protein. These findings were confirmed by the endonuclease activity of the N-terminal end of the L protein of TOSV (). Thus, we provide evidence that all three segmented negative single-strand RNA virus species share an endonuclease domain probably involved in the cap-snatching process during the viral life cycle. These data raise the question of a possible common ancestor for these viruses. Indeed, these three virus families use a cap-snatching mechanism involving binding and cleavage of cellular mRNA caps subsequently used by a large primer-dependent RNA-dependent RNA polymerase. It seems more plausible that the L gene has evolved by divergence over time, rather than by multiple acquisitions of several activities converging into a common structure, at least in the case of the endonuclease. Furthermore, our study raises the interesting possibility that other activities involved in RNA replication/transcrition might be discovered by comparative analysis of Orthomyxoviridae PB1, PB2, PA and Arenaviridae/Bunyaviridae L proteins.
To our knowledge, a single crystal structure of a functional arenavirus protein is currently available, that of the Machupo virus glycoprotein GP1 in complex with its human receptor, TfR1 
. Our results provide an arenavirus L domain structure, with a role consistent with the hypothesis of a cap-snatching mechanism suggested for arenaviruses 
. The strategy used here to produce individually active domains might be useful to further characterize the Arenaviridae/Bunyaviridae
large L protein which had so far resisted all biochemical characterization attempts.
The influenza, Arenaviridae
endonucleases are so far the only three examples of RNA endonucleases similar to type II DNA restriction endonucleases. The presence of such an endonuclease suggests that it could serve as a fruitful target for antiviral strategies against these two families, since such kind of inhibitors have been reported in the case of the influenza virus