Development of encoded Broccoli RNA aptamers for live cell imaging of alphavirus genomic and subgenomic RNAs

Sindbis virus (SINV) can infect neurons and cause encephalomyelitis in mice. Nonstructural proteins are translated from genomic RNA and structural proteins from subgenomic RNA. While visualization of viral proteins in living cells is well developed, imaging of viral RNAs has been challenging. RNA aptamers that bind and activate conditional fluorophores provide a tool for RNA visualization. We incorporated cassettes of two F30-scaffolded dimers of the Broccoli aptamer into a SINV cDNA clone using sites in nsP3 (genomic RNA), the 3′UTR (genomic and subgenomic RNAs) and after a second subgenomic promoter resulting in 4–28 Broccoli copies. After addition of the cell-permeable 3,5-difluoro-4-hydroxybenzylidene imidazolinone (DFHBI-1T) conditional fluorophore and laser excitation, infected cells emitted green fluorescence that correlated with Broccoli copy numbers. All recombinant viruses replicated well in BHK and undifferentiated neural cells but viruses with 14 or more Broccoli copies were attenuated in differentiated neurons and mice. The signal survived fixation and allowed visualization of viral RNAs in differentiated neurons and mouse brain, as well as BHK cells. Subgenomic RNA was diffusely distributed in the cytoplasm with genomic RNA also in perinuclear vesicle-like structures near envelope glycoproteins or mitochondria. Broccoli aptamer-tagging provides a valuable tool for live cell imaging of viral RNA.

Sindbis virus (SINV), an alphavirus in the Togaviridae family, is an enveloped plus-strand RNA virus that can cause seasonal outbreaks of rash and arthritis in humans and encephalomyelitis in experimentally infected mice [1][2][3] . On entry, nonstructural replicase proteins (nsPs) are translated from the capped and polyadenylated 11.7 kb genomic RNA as two polyproteins (P123 and P1234) to initiate infection. P123 and the nsP4 RNA-dependent RNA polymerase induce spherules at the plasma membrane and replicate the full-length minus-strand RNA in early replication complexes as a template for synthesis of plus-strand RNAs 4 . P123 is subsequently processed by the nsP2 protease to nsP1 and P23 and then to individual nsPs leading to cessation of minus-strand synthesis. With the switch to stable plus-strand replication complexes genomic and subgenomic (sg) RNAs are synthesized 5,6 . sgRNA is translated to produce the structural proteins E1, E2 and capsid required for assembly of virions containing genomic RNA 5,7-12 . To aid in studies of the regulation of viral RNA synthesis, we previously applied RNA aptamer technology to develop tagged fluorescent SINV RNA. RNA aptamers can bind and activate a conditional fluorophore that emits fluorescence in response to excitation with light at specific wavelengths [13][14][15][16] . The RNA aptamer Spinach2 within a tRNA scaffold was incorporated into a cDNA clone of the TE strain of SINV to produce recombinant viruses that expressed viral RNA tagged with the aptamer 17 . Spinach2 tagging allowed visualization of SINV RNA in live cells as green fluorescence after addition of the non-toxic cell-permeable conditional fluorophore DFHBI-1T ((Z)-4-(3,5-difluoro-4-hydroxybenzylidene)-2-methyl-1-(2,2,2-trifluoroethyl)-1H-imidazol-5(4H)-one) and excitation at 488 nm 17 . However, this system had several limitations. First, stop codons are present in all frames of tRNA-Spinach2 precluding insertion into viral protein coding regions and limiting the RNA tagging sites.
Therefore, for SINV the Spinach2 aptamers were only inserted into the 3′UTR and/or after an engineered second subgenomic promoter and did not allow differentiation between viral genomic and subgenomic RNAs. Second, the signal was too dim for visualization in a conventional epifluorescence microscope and even with laser-scanning confocal microscopy, the signal remained insufficiently bright for visualization in cells with restricted replication and background autofluorescence, such as differentiated neurons. , made from self-ligation between the complementary 5′ (BamHI -G^GATC_C) and 3′ end (BglII -A^GATC_T). The ligation forms a scar (GGATCT), which can no longer be cut by either BamHI or BglII. The tandem repeats can be inserted into 3 restriction sites of TEds plasmid in the specific order -nsP3, 3′UTR, and downstream of 2 nd subgenomic promoter.
Recently, Broccoli, a smaller and brighter aptamer in a different scaffold than Spinach2, has been developed 18,19 . Unlike the Spinach2 tRNA scaffold, the new more stable F30 Broccoli scaffold is not subjected to intracellular cleavage and accommodates two sets of dimeric Broccoli (dBroccoli) 20,21 . The F30-2xdBroccoli (4 copies of Broccoli in total) is compact (234 bases) and contains no in-frame stop codons. After binding to the Broccoli RNA aptamer and laser excitation, the fluorophore DFHBI-1T emits green fluorescence similar to green fluorescent protein (excitation/emission = 472/507) 15,18,21 (Fig. 1). To determine the applicability of this system to analysis of viral RNA dynamics, we constructed and characterized SINVs with different F30-2xdBroccoli numbers and insertion sites. These recombinant viruses were then used for analysis of RNA synthesis and distribution after infection of neural and non-neural cells.

Signal intensity was brighter than Spinach2 and correlated with Broccoli copies in viral RNA.
To assess the effect of Broccoli copy number on signal intensity and compare the signal to that obtained with Spinach2, BHK cells infected with Spinach2-tagged and Broccoli-tagged SINVs were compared (Fig. 4). Viral www.nature.com/scientificreports www.nature.com/scientificreports/ RNA tagged with F30-2xdBroccoli is brighter than tRNA-Spinach2 although all viruses except TEds-2Sp replicate well in BHK cells 17 . For the Broccoli-tagged viruses, the signal intensity increased as numbers of Broccoli insertions increased from 4 to 14 copies. However, TEds28Br, which had tandem repeats of F30-2xdBroccoli, did not show a brighter signal than TEds14Br. Because the TEnsP3-4Br and TEnsP3-8Br had Broccoli only in genomic RNA, infected cells were not as bright as those infected with SINV that had Broccoli in the greatly amplified subgenomic RNA. These data indicated that the ability to insert more copies of the Broccoli aptamer compared to the Spinach aptamer into viral RNA improved visualization of RNA in infected cells.

Broccoli signal survived fixation and co-localized with SINV RNA fluorescence in situ hybridization (FISH).
To determine whether Broccoli-tagged RNAs were also suitable for imaging fixed cells, BHK cells infected with TE-ds6Br were fixed with 3.7% formaldehyde for 10 min (Fig. 5). Unlike Spinach2, fixation did not obliterate the Broccoli signal, but did decrease the signal intensity and increase autofluorescence. Thus, live cell imaging provided a better signal to noise ratio than fixed cell imaging, but fixed cell imaging was possible.
To compare imaging of SINV RNA by fluorescent in situ hybridization (FISH) with Broccoli-tagged SINV RNA, BHK cells were infected with TEUTR4Br and TEds10Br followed by FISH probes detecting sequences in E1 and E2 (Fig. 6). Post fixation imaging of infected BHK cells showed almost identical patterns between RNA FISH and Broccoli signals (R colocalization = 0.9049) with only rare discrepancies for TEds10Br (Fig. 7) perhaps due to Broccoli insertion behind a second subgenomic promoter. RNA FISH tended to be brighter than aptamer-tagged RNA.
Replication of SINV with a high Broccoli copy number was attenuated in differentiated neurons and mouse brain. To determine whether SINV replication was compromised by the Broccoli insertions, virus growth in BHK cells, undifferentiated cAP7 and differentiated dAP7 neural cells and in mouse brain was compared to TE (Fig. 8). All Broccoli-tagged SINVs replicated well in BHK cells (Fig. 8A). Viruses with Broccoli www.nature.com/scientificreports www.nature.com/scientificreports/  www.nature.com/scientificreports www.nature.com/scientificreports/ insertions in nsP3 replicated as well or better than untagged TE in cAP7 (Fig. 8B) and dAP7 (Fig. 8C) neuronal cells, but inconsistently in mouse brains with variable amounts of infectious virus detected 3 days after infection (Fig. 8D). TE3UTR4Br and TEds6Br replicated as well as untagged TE in neuronal cells (Fig. 8B,C) and in the brains of mice (Fig. 8D). On the other hand, TEds10Br, TEds28Br, and TEds14Br replicated less well in dAP7 cells (Fig. 8C) and mouse brains (Fig. 8D), but only marginally less well in cAP7 cells (Fig. 8B). Therefore, insertion site and number of insertions affected SINV replication in less permissive neural cells in vitro and in vivo more than in very permissive cells.
Broccoli-tagged SINV can be visualized in differentiated neurons. Spinach2-tagged SINV gave a very dim signal in differentiated neurons and spectral imaging (lambda mode) could differentiate the signal from autofluorescence in only a few of the brightest cells 17 . To assess the ability to visualize Broccoli-tagged SINV in differentiated neurons, we analyzed dAP7 cells infected with TEds10Br (Fig. 9). TEds10Br-infected dAP7 cells were bright enough for RNA to be visualized in most infected cells although lambda mode was still required. The Broccoli signal could also be observed in a brain slice obtained from a mouse 3 days after intracerebral infection with TEds10Br (Fig. 10A,B).

Distribution of SINV genomic and subgenomic viral RNAs in BHK cells. To localize genomic
and subgenomic viral RNAs, BHK cells infected with TEnsp3-4Br (genomic) and TEds6Br (genomic and subgenomic) were imaged 16-20 h after infection (Fig. 11). Broccoli-tagged subgenomic RNA remained diffusely distributed in the cytoplasm (Fig. 11A), while at later times after infection Broccoli-tagged genomic RNA accumulated near the nucleus and assumed a circular shape with an inside vacuole (Figs. 11B and 12). To determine the relationship of this RNA to viral structural proteins required for virion assembly, cells were co-infected with TEds14Br and SINV expressing mCherry-tagged E2 26 (Fig. 13A). Genomic RNA was present in close association with virion surface glycoprotein E2. To also determine the relationship of genomic RNA to mitochondria, cells infected with TEds14Br were stained with mitotracker dye to identify mitochondria (Fig. 13B). Co-staining showed genomic RNA associated with mitochondria. Time-lapse imaging of merged phase contrast and Broccoli showed many vesicles surrounded by SINV genomic RNA gathered near a perinuclear halo most likely to be the Golgi apparatus (Fig. 14).

Discussion
Live cell imaging of viral RNA provides many advantages over fixed cell imaging at single time points, but has been difficult to accomplish. Previous studies of SINV RNA labeled with the Spinach2 aptamer were limited by the availability of insertion sites and brightness of the signal 17 . The current study has shown that labeling SINV RNA with multiple copies of the F30-2xdBroccoli aptamer overcomes many of the obstacles previously encountered with the Spinach2 aptamer. F30-2xdBroccoli is smaller and much brighter than tRNA-Spinach2 and, unlike the tRNA scaffold, the F30 scaffold is not a target for intracellular cleavage and does not contain stop codons 21 .
Lack of stop codons in the cassette allowed insertion into the genomic RNA region that encodes the nonstructural polyprotein. Viruses with 4-8 Broccolis in the C-terminal coding region of nsP3 replicated well and allowed visualization of genomic RNA. Like Spinach2 17 , insertion of 4 Broccolis into the 3′UTR did not compromise virus replication. At this site, as well as after a second subgenomic promoter, the RNA that is visualized is primarily subgenomic RNA because of its greater abundance. However, also like Spinach2, insertion of too many Broccoli copies (14 or more) impaired virus replication in differentiated cells and in vivo.
The greater brightness of the Broccoli signal also improved the quality of the imaging. Spinach2 can only be used for live cell imaging of SINV sgRNA in highly permissive undifferentiated cells (~5 × 10 5 molecules/ undifferentiated cell 27 ). Multiple copies of Broccoli inserted into the nsP3 coding region, permitted visualization www.nature.com/scientificreports www.nature.com/scientificreports/ of genomic RNA (~1.6 × 10 5 molecules/undifferentiated cell 28 ) and viral RNA after infection of less permissive differentiated neural cells (Table 1).
In this study, we demonstrated at least two types of genomic RNA-containing vesicle-like arrangements in live cells. The one found earlier is most likely the endosome-derived cytopathic vacuole (CPV) type I, which is formed at the plasma membrane and contains the viral replication complex in spherules 29 . The genomic RNA near this vacuole has likely been exported from the spherule-associated replication complex that synthesizes a full-length minus-strand template for amplification of genomic viral RNA after initiation of infection 30,31 . Another vesicle  www.nature.com/scientificreports www.nature.com/scientificreports/ type, CPV type II, was found in the perinuclear area in association with the trans Golgi network and endoplasmic reticulum. This vesicle near sites of sgRNA translation contains envelope proteins, is surrounded by viral nucleocapsids and likely the preparatory site for virion assembly 32 . Genomic RNA was also located around mitochondria that might be another type of cytopathic vacuole 33 or reflect the interaction of viral RNA and protein with damaged mitochondria that accumulate in the perinuclear area prior to mitophagy 34 .
In summary, the production of virally encoded fluorescent RNA aptamers will facilitate new approaches for study of viral RNA dynamics in vitro and in vivo. The real-time measurement of the fluorescent signal intensity is faster, cheaper, easier and more informative than FISH or RT-qPCR. This technique is also suitable for high throughput anti-viral drug discovery that targets viral RNA replication.

Materials and Methods
Viruses. SINV Toto E2-mCherry was a gift from Richard Kuhn (Purdue University, West Lafayette, IN) 26 .
SINV cDNA clones TE 22 and TEds 23 were the background for all aptamer-tagged recombinant viruses. After sequencing to confirm the correct insertions and orientations, each constructed plasmid was linearized with XhoI (New England Biolabs) and transcribed into mRNA by Sp6 mMessage mMachine (Ambion). Recombinant viruses were recovered by transfection of mRNA into BHK cells and collection of supernatant fluids after 24 h as previously described 17 . Titers of virus stocks were determined by plaque assay in BHK cells.  www.nature.com/scientificreports www.nature.com/scientificreports/ Cell culture, differentiation and infection. BHK-21 cells were cultured at 37 °C 5% CO 2 in DMEM-10% FBS. AP7 odora cells (gift from Dale Hunter, Tufts University, Boston), derived from rat olfactory neurons and immortalized with a temperature-sensitive SV40 T antigen 35 , were either passaged as cycling cells (cAP7) in DMEM-10% FBS at 33 °C 7% CO 2 , or were differentiated for 7 days into neurons at 39 °C 5% CO 2 (dAP7) by adding 1 μg/ml insulin (Sigma), 100 μM ascorbic acid (Sigma), and 20 μM dopamine (Sigma) into the medium as previously described 36 . Cells were infected with the recombinant viruses in DMEM-1% FBS at multiplicities of infection (MOI) of 5 to 100. Replication was assessed in BHK, cAP7, and dAP7 cells after infection at an MOI of 5 by assaying supernatant fluids collected in triplicate by plaque assay in BHK cells.
Live cell imaging. Cells were grown in 35 mm dishes, glass bottom 24-well plates (MatTek), or 8-well chamber cover glasses (Lab-Tek). Infected cells were imaged 30 min after replacing the culture media with phenol-red-free imaging medium (Fluorobrite, Gibco) containing 20 μM DFHBI-1T (Lucerna), 25 mM HEPES (Gibco) at 37 °C, 1% FBS, 1% PS and 2 mM glutamine. Autofluorescence was assessed relative to SINV TE (no aptamer)-infected cells and was minimized by adjusting the laser power and gain in each experiment. The laser scanning confocal microscope (Zeiss LSM780FCS, Axio Observer Z1, Zen Software) with A 488 nm laser excitation was used to activate the Broccoli-DFHBI-1T complex. In dAP7 cells, the true signal was differentiated from autofluorescence with the lambda scan that captured an emission spectrum in each pixel. The Broccoli-DFHBI-1T signal had an emission peak at 507 nm and appeared green in lambda mode, while the autofluorescence signal had no peak, extended from green to far red in its spectrum, and appeared yellow. In time-lapse experiments, temperature and CO 2 were kept at the same level as the culture condition for each cell type with a heated chamber at 24% humidity. ImageJ Software was used for image analysis.