The BUNV NSs protein has been widely studied in mammalian cells where it is has been shown to be a major virulence determinant. NSs counteracts the host innate immune response mainly by globally inhibiting RNA polymerase II-mediated transcription 
. On the other hand, BUNV NSs does not affect cellular transcription in infected mosquito cells 
, and for the related La Crosse orthobunyavirus (LACV), no specific function for its NSs protein was found in mosquito cell lines 
. However, our results presented above show that the BUNV NSs protein could be a crucial factor for efficient infection in certain cultured mosquito cells and in live mosquitoes.
Comparison of wtBUNV and rBUNdelNSs2 showed that NSs is nonessential for replication and establishment of persistent infection in Ae. albopictus
C6/36 and C7-10 cells. By contrast, rBUNdelNSs2 seemed unable to replicate productively in neither the Ae. albopictus
U4.4 cell line nor the Ae. aegypti
Ae cell line, indicating a requirement for the NSs protein. Unfortunately, attempts to express NSs exogenously in U4.4 cells, and thus enable rBUNdelNSs2 replication, have so far been unsuccessful. Comparison of the levels of released wtBUNV and the recombinant NSs deletion mutant suggests that NSs protein enables high-level virus replication in all cells except Ae. albopictus
C7-10, where expression or not of NSs had little effect. It is becoming clearer that, much like with mammalian cell lines, different mosquito lines differ in their response to viral infection and their ability to reproduce accurately events in whole mosquitoes 
. Our data suggest that of the Ae. albopictus
lines, the U4.4 cell line could similarly be a good tissue culture model to study BUNV replication and the role of NSs in infection of mosquito cells.
The three phases of infection, early, acute and late, described by Lopez-Montero and Risco 
in BUNV-infected C6/36 cells were also identified in the other two Ae. albopictus
cell lines. These phases were characterized by changes in the location of viral proteins and changes of the cell morphology throughout each stage. Microscopic observations of the wild type and NSs-deleted viruses showed that lack of NSs reduced the extent to which mosquito cells underwent morphological changes during the acute stage of infection. Possibly these changes are driven by a defense mechanism that allows the cells to cope with severe viral infection, and Lopez-Montero and Risco 
suggest that the filopodia-like projections could be involved in spreading “protective signals” among the cells. Branch-like projections have also been observed in cells infected with a rodent-transmitted bunyavirus, Sin Nombre hantavirus, and suggested to be sites where progeny particles were released 
. A release method that does not involve rupturing the cell membrane could explain why virus replication does not kill mosquito cells and persistence is maintained.
To date the only mosquito-borne bunyavirus that was shown to induce RNAi response in mosquito cells is La Crosse orthobunyavirus 
. We have also detected, by conventional Northern blotting analysis, virus-specific small RNAs (<30 nucleotides) in all Ae. albopictus
cell lines infected with wtBUNV (data not shown). Soldan et al. (2005) showed that La Crosse virus replication could be inhibited in C6/36 cells pre-treated with virus-specific small interfering RNAs 
. Here we showed that an RNAi response could be efficient in inhibiting BUNV infection too. When we transfected cells with long virus-specific dsRNA, BUNV replication was only reduced in U4.4 cells, which have previously been shown to have fully functional Dicer 2 
, suggesting that the dsRNA was processed efficiently to generate small inhibitory RNAs. Bunyaviruses efficiently avoid dsRNA-based RNAi responses by coating their RNA segments with the nucleoprotein, thereby avoiding the formation of dsRNA species 
. Our transfection experiments showed that if specific dsRNA species were produced in abundance, mosquito cells could overcome wtBUNV infection. Interestingly, the BUNV NSs deletion mutant was capable of efficient replication only in Dicer 2 incompetent cell lines. There are no studies showing involvement of any bunyavirus NSs protein in overcoming RNAi response in mosquito cells, but the NSs protein of La Crosse virus has been shown to inhibit RNAi antiviral activity in mammalian cells 
. Further work is required to investigate whether BUNV NSs has an effect on mosquito Dicer 2 activity, or if exogenous expression of NSs in U4.4 cells would render them permissive for rBUNdelNSs2 replication.
The lack of genomic sequence data for the Ae. albopictus
mosquito makes it a less attractive model in which to study host-virus interactions. Therefore, we investigated whether cells derived from Ae. aegypti
, whose sequence has been determined 
, were permissive for BUNV replication. Our results showed that BUNV growth in Ae cells resembled that in U4.4 cells, and that the NSs protein also proved to be necessary for efficient replication. Thus Ae. aegypti
cells could be a useful tool in studying BUNV infection and identification of cellular components that are important for viral replication.
There is only one study of BUNV replication in mosquitoes; Peers 
reported that BUNV multiplied in the gut, disseminated to salivary glands and was transmitted to suckling mice in Ae. aegypti
mosquitoes more efficiently than in Ae. vexans
and Ae. canadensis
. Our results for wtBUNV showed similar kinetics of viral replication and dissemination in Ae. aegypti
to those obtained by Peers. Fewer mosquitoes were infected with the NSs-deletion mutant. In addition, we showed that the NSs protein contributes to high-level virus replication in that the mutant virus lacking NSs grew to lower titres. Similarly, the wild-type virus disseminated to salivary glands more efficiently than rBUNdelNSs2, and the lower levels of rBUNdelNSs2 in salivary glands could affect the transmission potential of the virus. This requires further investigation.
Our experiments showed that BUNV replication in Ae. aegypti
mosquitoes resembled replication in Ae. aegypti
Ae cells, as well as Ae. albopictus
U4.4 cells. These data corroborate previous conclusions that these cell lines are the most appropriate mosquito cell culture models to study arbovirus infection. The NSs protein was required for efficient replication in both mosquito cells with a competent Dicer 2-RNAi system and in adult mosquitoes. In a proportion of mosquitoes, however, the mutant virus could eventually overcome host defences, though remained constrained as evidenced by lower virus titres. As BUNV is a relatively fast growing virus perhaps its reproduction rate is able to counteract the host's inhibitory responses. The NSs protein of Rift Valley fever phlebovirus (also in the family Bunyaviridae
), although being quite distinct in size, amino acid sequence and expression strategy from BUNV NSs, plays a similar role in mammalian cells in overcoming innate immune responses via global shut-down of cellular transcription 
. Two recent papers investigated the role of Rift Valley fever virus NSs in infection of mosquitoes, and neither observed any difference in infection or dissemination rates between wt and NSs-deleted viruses 
. Interestingly, deletion of another non-structural protein, NSm, from Rift Valley fever virus almost completely abolished its ability to replicate in mosquitoes 
. These results illustrate the diversity and complexity of virus-host interactions within the Bunyaviridae