Cap-snatching as a method of priming transcription is uniquely restricted to segmented negative strand viruses, notably orthomyxoviruses (influenza), bunyaviruses and arenaviruses. The recent structural characterisation of two functional domains relevant for cap-snatching by influenza polymerase, the cap-binding domain and the endonuclease, in respectively the PB2 and PA polymerase subunits, raise the question as to whether similar domains exist in the L-protein (polymerase) of bunya- and arenaviruses. The work presented here shows unequivocally that the extreme N-terminal 200 residues of LACV has a cap-snatching endonuclease activity with very close structural and biochemical features to that of the N-terminal domain of the influenza virus polymerase PA subunit. We do not yet know the context of the bunyavirus N-terminal endonuclease within the 3-dimensional structure of the complete polymerase. However it is likely that there is a cap-binding domain and probably other RNA binding domains within the polymerase (this is certainly true for influenza polymerase) that enhance affinity and provide specificity for capped cellular mRNAs. Also it is possible that, as with influenza virus, there are allosteric effects that activate or make accessible the endonuclease active site only upon cap-binding.
We next examined whether the endonuclease signature could be identified in the L-protein of other segmented RNA viruses. Sequence analysis gives strong evidence that a homologous endonuclease domain exists at the N-terminus of the L-protein of four Bunyaviridae
genera, orthobunya-, tospo, phlebo and hantaviruses, as well as tenuiviruses (which have four to six genome segments, 
. Tenuivirus) and orthomyxovirus (). In each case, the key metal binding and catalytic lysine residues can be identified. The sequence analysis shows that there are two sub-groups of these enzymes, with slightly different endonuclease signatures. Orthobunya- and Tospoviruses have the motif H....D...PD....DxK.....T, whereas Phlebo- and Hantaviruses have the motif H....E...PD....ExT.....K (although in Phleboviruses the first E is replaced by a D). The Hantavirus motif is identical to that found in orthomyxoviruses (). The first version has a preference for aspartates and the catalytic lysine emerges from beta-strand βb, whereas the second version has a preference for glutamates and the catalytic lysine emerges from alpha helix αd (see ). Interestingly, the catalytic lysine interchanges with an absolutely conserved threonine at the two alternative positions (). Nairoviruses are not included in this alignment as the location of the endonuclease is less certain. This genus of Bunyaviridae
, which includes Crimean-Congo hemorrhagic fever virus, has an unusually long L-protein (about 4000 residues, compared to 2100–2900 for most other bunyaviruses). The N-terminal half of nairovirus L-proteins (i.e. prior to the polymerase motifs which start around residue 2050) contains a putative ovarian tumour (OTU)-like cysteine protease at the beginning 
as well as other predicted motifs and domains 
. A putative endonuclease motif of the Phlebo/Hanta/Orthomyxo type exists in the residue range 630–710 (H(632)...PD(672)....E(686)F....K(699), numbering for Crimean-Congo virus) 
, but this needs to be confirmed by structural and functional data. It is interesting to note that the rice stripe tenuivirus also contains a predicted N-terminal OTU-like protease before the endonuclease motif 
. It has been suggested that the protease might release the viral polymerase and one or more additional proteins by autoproteolytic cleavage and/or have de-ubiquitination activity 
. Indeed de-ubiquitination activity of Crimean-Congo virus OTU domain has been shown to inhibit Ub- and ISG15-dependent antiviral pathways 
. Arenavirus L-proteins have a highly conserved N-terminal region of about 200 residues that contains the absolutely conserved sequence of residues PD(89)...E(102)xF....K(122)L (alignment not shown, numbering for Lassa virus). This closely resembles the Phlebo/Hanta/Orthomyxo endonuclease motif, although the histidine is clearly lacking. Very recently, systematic alanine mutation of conserved charged residues in Lassa virus L-protein outside the polymerase motifs have been performed and the effect on transcription and replication have been tested in a RNP reconstitution system 
. Seven charged residues in the N-terminal region, including Asp89, Glu102 and Lys122 and Asp129, were selectively important for mRNA synthesis but did not affect genome replication. The authors concluded from these results, combined with sequence similarities to type II endonucleases and influenza virus endonuclease, that this region of the L-protein was likely to be the cap-snatching endonuclease of arenaviruses, in full agreement with our analysis. Finally, the endonuclease signature is also clearly present in the L-proteins of two related but unclassified bunyaviruses (proposed to be called emaraviruses) which have four rather than the usual three genome segments, European mountain ash ringspot disease (Acc. No. YP003104764, 
) and fig mosaic virus (Acc. No. CAQ03479, 
). Both have the motif RH(105)D...PD(144)...E(158)xK(160) (numbering for mountain ash ringspot disease virus) and are thus most closely related to the Orthobunya and Tospoviruses, All these observations are summarised in which shows a schematic diagram of the architecture of polymerases from negative strand segmented RNA viruses.
Structure based multiple alignment of the endonuclease of four genera of bunyavirus L-proteins together with the endonuclease of the PA subunit of selected orthomyxoviruses.
Schematic diagram of the polymerase architecture of negative strand segmented viruses.
It is well known that the 6 motifs characteristic of negative strand RNA-dependent RNA polymerases (pre-motif A and motifs A–E) are present in the central region of bunya and arenavirus L-proteins and in the PB1 subunits of orthomyxoviruses 
. The current work shows that the extreme N-terminal region of bunya-, tenui- and arenavirus L-proteins functionally corresponds to the N-terminal region of the PA subunit of orthomyxoviruses. Given that the three influenza A polymerase subunits total 2252 residues, very similar to the size of many bunyavirus complete L-proteins and all these viral enzymes have common mechanisms of transcription (cap-snatching) and replication, a natural hypothesis that follows is that the L-proteins might be architecturally, structurally and functionally equivalent to a concatemer of the three influenza polymerase subunits in the order PA-PB1-PB2 (). Some indirect support for the functional concatenation of the influenza polymerase subunits comes from the fact that the inter-subunit interactions are dominated by contacts between the C and N-terminal extremities of respectively PA and PB1 and PB1 and PB2 as visualised by recent crystal structures (reviewed in 
). The most significant implication of this hypothesis is that the C-terminal third of the L-protein might be structurally and functionally equivalent to PB2, which contains the cap-binding domain required for cap-snatching. Unfortunately, this region of the L-protein is the least well conserved and there are no obvious cross-genera conserved motifs that could point to a putative cap-binding site similar to that described for influenza A PB2 subunit 
. This is perhaps not surprising as the PB2-like subunits of, for instance, salmon anaemia and Quaranfil viruses, two non-influenza orthomyxoviruses, are highly diverged from influenza 
, even though both these viruses appear to possess an endonuclease at the N-terminus of the PA subunit (). Furthermore the fact that the distance of endonucleolytic cleavage from the 5′ cap is rather variable amongst cap-snatching viruses 
suggests that the location of the cap-binding domain might vary. In fact, there is no clear proof that any L-protein directly binds capped RNAs and even some evidence that in hantaviruses the viral N-protein may play this role 
. Clearly more experimental work is required to elucidate the complete mechanism of cap-snatching in bunya-, tenui- and arenaviruses and to validate or otherwise the hypothesis that L-proteins are architecturally equivalent to the concatenation of PA-PB1-PB2.
Finally it is important to note that for nearly two decades, influenza virus endonuclease has been targeted for anti-viral drug discovery and a number of specific endonuclease inhibitors have been described 
. Most of these compounds implicitly target the two metal binding site of the endonuclease, which is also the target for many HIV integrase inhibitors 
including the currently approved raltegravir 
. The recent structure determination of the endonuclease of influenza virus polymerase 
gives new impetus to structure-based optimisation of these inhibitors. The results described here show that bunyaviruses and arenaviruses, amongst which are several dangerous and emerging pathogens, contain a very similar endonuclease to influenza virus, which is also therefore a good target for anti-viral drug design. Indeed, the close similarities between influenza and bunyavirus endonucleases suggests that compounds targeting a broad spectrum of segmented negative strand RNA viruses could be envisaged. Our structure of DPBA bound to LACV endonuclease shows that this is indeed the case, although this compound is of low potency 
. In addition this structure provides the first concrete proof that these compounds do indeed chelate the two divalent cations in the endonuclease active site.