**4. Discussion**

Many hantaviruses establish persistent infections within their rodent hosts by manipulating the host cellular environment to allow the abundant synthesis of viral components for an extended time period, while also evading or minimizing the cellular innate immune response [6,7]. To better understand the intracellular events during long-term hantavirus infections, we used immunofluorescent microscopy to investigate the intracellular distribution of TULV NP and RNA, alongside multiple host cell components during an extended time course of up to 30 days, during which time the persistent infection by TULV has been previously demonstrated [23]. This was combined with EM imaging to understand the morphological changes that TULV induces in infected cells. Our findings revealed that as the infection progressed, the NP formed increasingly large ultrastructural assemblies, mostly with an overall filamentous and tubular appearance, but also as discrete puncta (Figure 1). At all the time points examined, these NP assemblies closely co-localized with TULV RNA, regardless of its negative or positive sense polarity (Figure 6). This close association between all hantavirus RNA and NP is consistent with the known structure of the hantavirus RNP segments, in which hantaviral vRNAs and cRNAs are always found in close association with NP, forming extended helical assemblies [31]. Interestingly, our observations also sugges<sup>t</sup> that TULV mRNAs, which will likely comprise the majority of positive sense TULV-specific RNA in the infected cell [30], also remain in close proximity with NP. This may reflect the organisation of the replication factories, with newly transcribed S, M and L segmen<sup>t</sup> mRNAs remaining in the vicinity of their respective NP enwrapped vRNA templates. Alternatively, it may reflect the previous findings that the hantavirus NP binds the 5 mRNA cap at a separate RNA binding site [32], and this interaction is proposed to enable mRNA circularization, so as to promote eIF4F independent translation [33]. The lack of distinction between the sites of positive sense and negative sense RNA staining at the resolution used here also suggests that TULV RNA replication and mRNA transcription processes are not physically separated into distinct compartments.

Co-staining experiments using markers for intracellular compartments revealed that NP and TULV RNA co-localized within structures that were closely associated with the Golgi (Figure 5). At the early stages of infection (36 hpi), NP, RNA and Golgi components localized within both punctate and irregular filamentous structures within and surrounding compartments that exhibited a canonical Golgi morphology. At the later time points of 7 dpi and 30 dpi, NP and RNA were detected with increased abundance within a single filamentous/tubular structure of large dimensions, which also closely co-localized with the Golgi (Figure 5B–C). The morphology of this filamentous/tubular NP/RNA/Golgi stained structure was atypical of the Golgi compartment, with a clear loss of the typical fused compact ribbon structure associated with Golgi cisternae, suggesting that the Golgi was re-modelled, most likely by TULV-related activities. As both TULV NP and RNA components localised to these compartments (Figure 7), it is di fficult to escape the conclusion that this remodelled Golgi is the site of TULV RNA synthesis and accumulation, and most likely represents TULV replication factories.

The visualization of long filamentous structures at later 7 dpi and 30 dpi time points in TULV-infected cells only, by TEM, was in agreemen<sup>t</sup> with our findings using light microscopy (Figure 2). These TULV-infected cell sections revealed the increased detail of the filamentous/tubular NP/RNA/Golgi stained structures, and confirmed their atypical morphology. Their apparent absence in sections of 36 hpi infected cells likely reflects their low abundance and the correspondingly high likelihood that they would not appear in the selected sections. Few studies have employed EM to study hantavirus-infected cells, with most analyses focusing on the morphology of cell-associated hantavirus particles [34–36]. However, the filamentous structures we observed here are reminiscent of filamentous and granular structures induced by the Sin Nombre virus [21], which were also in close proximity to the Golgi. However, in this study we showed that these filamentous structures also harbour abundant viral RNA.

If the filamentous/tubular NP-stained structures are the sites of RNA synthesis and accumulation, then the function of the punctate NP structures that are seen throughout the infection time course remains unclear. As they mostly stain weakly for RNA, we propose that they are unlikely to be sites of on-going abundant RNA synthesis. One possibility is that they may represent Golgi-derived particles, released by Golgi fragmentation, although an alternative possibility that we cannot rule out, is that the smallest of these structures may represent exocytic vesicles containing assembled viruses in transit towards the plasma membrane for release.

In contrast to the SG marker TIA-1, the levels of co-localization between the NP and the PB marker DCP1a were considerably lower (Figure 4), and this was unexpected given the previous reports of a close association between DCP1a and the NP of the NW SNV, where NP is proposed to bind and protect capped RNAs for later delivery to the transcribing RdRp [9]. This may represent fundamental differences between the host–pathogen interactions of NW and OW hantavirus species, and may sugges<sup>t</sup> that OW hantaviruses such as TULV do not acquire or process host cell cap moieties in the same way. Given the high degree of association we report between TULV NP and TIA-1, an alternative cap resource may be SGs.

The observation that the SG marker TIA-1 showed a punctate distribution at the early stages of virus infection, but was filamentous/tubular distribution at later times, suggests that it was re-distributed into the Golgi-derived filamentous/tubular compartments (Figure 4A–D). To the best of our knowledge, SGs and the Golgi have not been previously reported to extensively co-localize for a native function, and thus we propose that this redistribution is in response to TULV infection. Whether this represents a pro-viral or anti-viral response is currently unknown. Considerable evidence suggests that SGs represent innate immune signalling platforms [37] and one possibility is that their sequestration into TULV replication factories may prevent or delay a host cell anti-viral response to infection. Such a mechanism of immune evasion may contribute to the establishment of the characteristic state of hantavirus persistence.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2073-4409/9/7/1569/s1, Table S1. DNA oligonucleotide probes designed to hybridize with TULV S segmen<sup>t</sup> RNAs in FISH experiments. Sequences shown are written 5 to 3 and represent the actual probe sequences, Movie S1. TULV-infected Vero E6 cells at 30 dpi were stained using TULV NP antisera and DAPI, and successive Z plane images taken, which were compiled into a movie, with rotation about a single axis, Figure S1. The effect of cytoskeletal disassembly and reorganisation on the localisation of TULV NP in Vero E6 cells persistently infected with TULV. Vero E6 cells were infected with TULV at an MOI of 0.5, and at 30 dpi cytoskeleton filaments were disassembled by treatment with A) 400 nM okadaic acid (OA) for periods of 60 and 90 min. Distribution of vimentin and TULV NP were examined using LSCM at 40 X magnification, stained with specific antisera NP antisera (green) and vimentin (magenta) with nuclei stained with DAPI (blue). Vimentin filaments were allowed to reassemble by the removal of OA, and the distribution of NP and vimentin examined as above. The scale bar represents 30 μm. B) The effect of microtubule depolymerisation on TULV NP localisation was also examined in Vero E6 cells at 30 dpi. Microtubules were depolymerised by treating Vero E6 cells with 17 μM nocodazole (NOC) for 60 min, with the distribution of tubulin and TULV NP examined using LSCM at 40 X magnification, stained with specific antisera NP antisera (green) and tubulin (magenta) with nuclei stained with DAPI (blue). The scale bar represents 30 μm, Figure S2. Okadaic acid treatment of TULV-infected Vero E6 cells decreases viral genome copies. TULV-infected Vero E6 cells at 30 dpi were treated with 400 nM okadaic acid (OA) for 30, 60 or 90 min or 17 μM nocodazole (NOC) for 60 min. After harvesting, the TULV genome copies in cell lysates were quantified by qRT-PCR., Figure S3. Co-localisation of cellular trafficking proteins with TULV NP in TULV-infected Vero E6 cells during early and persistent time points. The spatial distribution of Rab5, Rab7, Rab11, LAMP1 and clathrin (magenta) was observed alongside TULV NP (green) in TULV-infected Vero E6 cells at A) 36 hpi and B) 30 dpi by LSCM at 40 X magnification using specific antisera against TULV NP, Rab5, Rab7, Rab11, LAMP1 and clathrin. Nuclei are stained with DAPI (blue). The scale bar represents 30 μm. Fluorescent line scans taken using Fiji software.

**Author Contributions:** Conceptualization: K.A.D., B.C., J.N.B., J.M., J.F. and R.H.; methodology: K.A.D. and B.C.; validation: K.A.D. and B.C.; formal analysis: K.A.D. and BC.; investigation: K.A.D. and B.C.; resources: J.N.B., J.M., J.F. and R.H.; data curation: K.A.D. and B.C.; writing—original draft preparation: J.N.B.; writing—review and editing; J.N.B., K.A.D., B.C., J.M., J.F. and R.H.; supervision: J.N.B., J.F.; project administration: J.N.B.; funding acquisition, J.N.B., R.H. and J.F. All authors have read and agreed to the published version of the manuscript.

**Funding:** KAD was funded by a faculty PhD studentship from The University of Leeds. BC is supported by a studentship from the MRC Discovery Medicine North (DiMeN) Doctoral Training Partnership (MR/N013840/1). This work was supported by the Academic Fellow scheme at the University of Leeds (to J. Fontana and J.M.). Electron Microscope and ultra-microtome were funded by Wellcome Trust (090932/Z/09/Z and 208395/Z/17/Z).

**Acknowledgments:** We would like to thank Martin Fuller for support with resin embedding and ultra-thin sectioning.

**Conflicts of Interest:** The authors declare no conflict of interest.
