Host Retromer Protein Sorting Nexin 2 Interacts with Human Respiratory Syncytial Virus Structural Proteins and is Required for Efficient Viral Production

The present study contributes new knowledge to understand HRSV assembly by providing evidence that nonglycosylated structural proteins M and N interact with elements of the secretory pathway, shedding light on their intracellular traffic. To the best of our knowledge, the present contribution is important given the scarcity of studies about the traffic of HRSV nonglycosylated proteins, especially by pointing to the involvement of SNX2, a retromer component, in the HRSV assembly process.

for giantin was raised in rabbit, to overcome this limitation we used TGN46 antibody made in sheep. As negative control of PLA, cells were transfected with Vps4-GFP, and at 4 hours posttransfection, the cells were infected with HRSV. At 24 hpi the cells were fixed, and the PLA protocol for HRSV N and GFP was performed ( Fig. 2E to G). Since it has been known that the HRSV assembly process is independent of Vps4 (11), Vps4 becomes a good PLA negative control. As an additional PLA negative control, HRSV- This set of immunofluorescence images is a single plane, representative of at least three independent experiments. The Mander's coefficient was calculated from Z-stack images of at least five cells from three independent experiments. The P value was determined using analysis of variance (ANOVA) one-way Tukey's multiple-comparison test. *, P Ͻ 0.05; **, P Ͻ 0.01; ***, P Ͻ 0.001; ns, nonsignificant. Images were taken at 24 hpi infection with a Zeiss 780 confocal microscope. Magnification, ϫ63. All scale bars ϭ 10 m.
infected cells were probed for PLA omitting one of the two primary antibodies (Fig. 2H to J). Taken together, these results suggest that HRSV F and N proteins may interact with each other in the Golgi before reaching the plasma membrane.
HRSV N and M proteins partially colocalize with the trans-Golgi network marker TGN46. Next, we investigated whether HRSV N and M proteins reach the endosomal system and the plasma membrane via the trans-Golgi network (TGN). To this end, we labeled for TGN46, a transmembrane protein that cycles between early endosomes and the TGN, where it is mostly localized at steady state (21). As expected, TGN46 is concentrated in the juxtanuclear region in noninfected cells (Fig. 3A to C). In infected cells, TGN46 partially colocalized with HRSV N and M proteins that were apart from typical inclusion bodies (Fig. 3D to I and P). In addition, superresolution imaging demonstrated that it is possible to see HRSV N, which is not associated with inclusion bodies, appearing tightly associated with the TGN46 (Video S2). However, the colocalization of either HRSV M or N with TGN46 was significantly lower than that of HRSV F and TGN46 (Fig. 3J, L, P to R) at 24 hpi. This is in agreement with a previously published study (14) showing that HRSV N protein is localized in the TGN. However, our findings suggest that the colocalization of TGN46 with the HRSV M and N proteins is moderate. We conclude that the HRSV M and N proteins partially colocalized with TGN46. Interestingly, TGN46 also showed a distribution pattern at the cell periphery and surrounding structures that resemble HRSV inclusion bodies in ring-shaped structures that were apparent at 24 hpi ( Fig. 3D to I, pointed by arrowheads in Fig. 3E and H). To rule out immunofluorescence bleeding from the channels coming from the HRSV inclusion bodies, cells were infected with HRSV, and at 48 hpi they were stained for TGN46. To demonstrate that these cells are actually infected, we stained them for HRSV F protein, instead of staining them for HRSV proteins that are components of inclusion bodies ( Fig. 3M to O). It is possible to see that ring-shaped structures became more evident at 48 hpi ( Fig. 3M to O) and that they are not a result of the fluorescence signal bleeding at the moment of the imaging acquisition. Because the TGN localization of TGN46 is maintained by efficient recycling from endosomes (22), these results may suggest that this retrieval pathway could be compromised in late infection.
HRSV glycosylated and nonglycosylated proteins depend on the Golgi integrity to reach the cell surface. To further explore the relationship between HRSV structural proteins and the Golgi complex, we treated infected cells with brefeldin A (BFA), a compound that binds Arf1-GDP protein and prevents attachment of coatomer protein complex I (COPI) to the Golgi. BFA treatment is known to reversibly induce Golgi stack disassembly and membrane redistribution to the ER (23,24). The expected effect of BFA was observed in HEp-2 cells (Fig. 4A to H). Golgi disassembly induced by BFA in HRSV-infected cells led to reduced staining of both HRSV F and N proteins at the cell periphery ( Fig. 4I to P). Also, the inclusion bodies appeared smaller in the presence of BFA (Fig. 4K and O and Fig. S2A to F), and the numbers of aggregates stained by HRSV N were significantly diminished in the presence of this drug ( Fig. 4K and O and Fig. S2G to I and S3A to U). To reinforce that the phenotype observed was due to the disassembly of the Golgi, BFA was washed out after 5 hours of incubation, and the cells were incubated with fresh medium until 24 hpi. Under these conditions, the normal Golgi apparatus display was recovered, the inclusion bodies became larger, and  labeling of the cell periphery filaments of the HRSV F and N proteins was restored ( Fig. 4Q to T, arrows). Since our results showed that the BFA had impact in the inclusion body size, we wanted to know if this impact was modulated depending on the timing of BFA addition. To investigate this, HEp-2 cells were infected with HRSV, and at 4, 8, 12, and 20 hpi, the BFA was added and kept in the cell medium. All the cells were fixed in 4% paraformaldehyde (PFA) after 24 hpi and stained for HRSV N and giantin (Golgi), and the size of the inclusion bodies was measured ( Fig. S3A to U). Interestingly, the size of inclusion bodies was dependent on the time of BFA addition ( Fig. S3A to U), and they were significantly different from the control (Fig. S3U). The colocalization graph showed that the treatment with BFA caused a lower colocalization of HRSV N with giantin when BFA was kept for 20 hours (Fig. S3V). The colocalization of N with giantin increased when BFA was kept for 12 and 16 hours. This was likely due to the giantin dispersion and to the number of HRSV N aggregates. Of note, these experiments do not enable us to state that BFA acts directly on the coalescence of inclusion bodies or impairing the (Q to T) The cell was treated with BFA for 5 hours, and then fresh medium was replaced without BFA; the sizes of the inclusion bodies (S) went back to being similar to the control (K), and it is possible to see filament formation budding from the plasma membrane, indicated by the arrows in panels Q, S, and T. The images represent a single focal plane in two independent experiments. Images were taken with Zeiss 780 confocal microscope. Magnification, ϫ63. All the scale bars ϭ 10 m. movement of the N protein that is associated with the Golgi. Taken together, these results suggest that normal Golgi integrity is required to deliver both glycosylated F and nonglycosylated N proteins to the cell periphery/surface and that the inclusion body size is affected by BFA treatment in unknown ways.
HRSV M and N proteins colocalize and interact with the early endosome protein sorting nexin 2. To assess whether HRSV N and M proteins are targeted to early endosomes, immunofluorescence (IF) microscopy for sorting nexin 2 (SNX2) was performed in HEp-2 cells at different times pi. SNX2 plays an important role in the formation of the retromer coat by shaping tubular structures out of the membranes of early endosomes (25), and therefore, it is an appropriate early endosome marker. Different from the normal distribution of SNX2 in noninfected cells ( Fig. 5A to C), there was evident colocalization of SNX2 with HRSV N (Fig. 5D to F and J) and M (Fig. 5G to I and K) at 24 hpi. Notably, SNX2 appeared accumulated in the inclusion bodies as confirmed by the plot profiles ( Fig. 5J and K). Furthermore, the Mander's coefficient of colocalization of HRSV N protein with SNX2 was determined at different times pi (Fig. 5L), and the peak of colocalization happened at 8 hpi. To investigate if the nonglycosylated proteins reach endosomes at early stages of infection when inclusion bodies are not apparent, IF was performed for HRSV F, M, and SNX2 at 4 hpi. The experiment revealed that while HRSV F and SNX2 have a better colocalization, HRSV M protein partially colocalized in some SNX2-containing vesicles ( Fig. 5M and N), suggesting that part of the HRSV nonglycosylated M proteins are localized in early endosomes at that time of the replication cycle. Seeking a better understanding of the relationship between SNX2 and HRSV inclusion bodies, superresolution imaging was performed for HRSV N and SNX2 ( Fig. 6A to Q). It was apparent that SNX2 was mostly located at the edges, but to a lesser extent also within HRSV inclusion bodies (Fig. 6D, H, and Q). In agreement with the data in Fig. 5, once again, HRSV F and N were seen in the same SNX2-containing vesicles ( Fig. 6E to H, arrowheads). It is noteworthy that intracellular filaments appeared to be emerging from the same small protein aggregates that contain HRSV F and N proteins and SNX2 ( Fig. 6E to H, arrows). As superresolution imaging demonstrated tight proximity between HRSV N and SNX2, PLA was performed to check for a possible interaction between SNX2 and HRSV M protein (Fig. 7). Because HRSV M and N proteins are known to interact with one another (26), they were used as positive controls for the PLA, revealing points of fluorescence distributed throughout the cytoplasm (Fig. 7E). Interestingly, PLA for HRSV M and SNX2 showed abundant intracellular puncta dispersed throughout the cytoplasm (Fig. 7F). To confirm that HRSV proteins interact with SNX2 by a different approach, a coimmunoprecipitation assay was performed using an anti-SNX2 antibody. A band corresponding to HRSV N protein was coimmunoprecipitated with SNX2 ( Fig. 7D and E, arrowhead). Taken together, the results suggest that HRSV structural proteins interact with SNX2 and may recruit this host protein to virus-induced compartments during assembly.
HRSV selectively recruits endosomal proteins to viral inclusion bodies. The association of SNX2 with SNX1, Vps26, Vps29, and Vps35 forms the retromer complex, a vesicle coat involved in retrograde transport of proteins from endosomes to the TGN (27). In order to verify whether another retromer component, besides SNX2, is present in structures containing HRSV proteins, IF for Vps26 was performed in HRSV-infected cells at 24 hpi (Fig. 8). Although less intense than for SNX2, and not localized in larger inclusion bodies ( Fig. 5 and 6), colocalization of N with Vps26 was clearly visible (Fig. 8A to D). Interestingly, Vps26 colocalization was restricted to tubular structures containing HRSV F and N proteins. These structures could be either tubular endosomal structures ( Fig. 8A to D, arrows) or filaments protruding from the plasma membrane ( Fig. 8E to H, arrowheads). Importantly, the results indicate that more than one component of the retromer is enriched in structures containing HRSV proteins. Colocalization between HRSV and retromer proteins may be explained by either the recruitment of these host proteins from the cytosol to virus-induced compartments or by the targeting of HRSV proteins to early endosomes. Therefore, to verify whether early endosome elements other than Vps26 were recruited to inclusion bodies, HRSV-infected cells were stained for the early endosome antigen-1 (EEA1). Different from what was observed for SNX2, EEA1 did not colocalize with HRSV inclusion bodies in infected cells at 24 hpi (Fig. S4A to I). This indicates that there is specific relocalization of SNX2 to HRSV inclusion bodies, independent from other early endosome elements.

Fractionation of HRSV-infected cells reveals repositioning of SNX2 in HRSVinfected cells.
To verify the partitioning of SNX2 in different compartments of HRSVinfected cells, cellular fractionation and Western blotting were performed. In HRSVinfected cells, SNX2 accumulates in fractions 1 to 7, which correspond to those that contain the nonglycosylated HRSV proteins that are known to compose the inclusion bodies (Fig. 9). This was in contrast to noninfected cells, in which SNX2 was detected throughout all the fractions. Interestingly, in HRSV-infected cells, TGN46 was detected in fractions 7 to 10, revealing a shift from the normal distribution seen in noninfected cells, in which TGN46 appeared in fractions 4 to 8, suggesting recruitment of TGN46 toward more dense fractions containing plasma membranes (Fig. 9). Importantly, the distribution of lysosomal-associated membrane protein-1 (Lamp-1), the early endosomal marker EEA1, and the plasma membrane protein EGFR appeared to be little affected by HRSV infection.
TGN46 and SNX2 colocalize with HRSV M and N proteins in virus filaments on the cell surface. In keeping with previous observations that HRSV budding forms typical filamentous shapes on the plasma membrane (28,29) that contain HRSV M and N proteins (Fig. 10A to F, arrowheads), TGN46 colocalized extensively with those viral proteins in the context of such filaments (Fig. 10G to S). This was reinforced by superresolution microscopy of filament-shaped structures on HRSV-infected cells ( Fig. 10T), showing an abundant localization of the TGN46 in the filamentous structures containing the HRSV proteins. Curiously, some of those filaments appeared to be detaching from the cells (Fig. 10T, arrowheads). SNX2 also colocalized with HRSV structural proteins F and M in viral filaments on the cell surface (Fig. 11A to I). Immunogold electron microscopy confirmed the presence of SNX2 in filament-shaped structures at the plasma membrane, as well as in structures apparently budded from HRSV-infected cells (Fig. 11J to M). Together, these results suggest that the HRSV particles may contain SNX2 and TGN46 to some extent.
Silencing of sorting nexins 1 and 2 partially impairs HRSV production and syncytium formation. Once it was demonstrated that SNX2 colocalizes and interacts with HRSV proteins, a knockdown (KD) experiment was done using small interfering RNA (siRNA) for sorting nexins 1 and 2 in HEp-2 cells (Fig. 12). It is well known that SNX1 and 2 form heterodimers; therefore, siRNA was performed for either SNX1 or SNX2 alone, and for both at the same time, to check for the impact of their silencing in the HRSV replicative cycle. The IF for HRSV N protein at 24 hpi demonstrated that the quantity of the inclusion bodies larger than 5 m 2 were slightly lower in cells KD for SNX2, than in those treated with the scrambled siRNA control (Fig. 12A to H). We chose FIG 9 Cell fractionation. HEp-2 cells uninfected or infected (HRSV) were harvested at 24 hpi, lysed, and loaded on the top of a discontinuous 12 to 30% glycerol gradient. After ultracentrifugation, fractions were collected and analyzed by SDS-PAGE and Western blotting. In HRSV-infected cells, SNX2 accumulated in the fractions of HRSV nonglycosylated proteins (fractions 1 to 7), whereas in noninfected cells, SNX2 was detected in all fractions. In HRSV-infected cells, the TGN46 protein is more abundant in the last fraction corresponding to the plasma membrane. Lamp-1 was not detected in the same fractions that accumulated HRSV N, P, M, and M2 proteins. Detection of EEA1 was similar between HRSV-infected and noninfected cells. Actin and EGFR are controls of cytosol and plasma membrane, respectively. The figure is a representation of two experiments. The membranes were visualized using a Bio-Rad Chemidoc. 5 m 2 as the threshold because Rincheval et al. (18) recently demonstrated that the HRSV mature inclusion bodies were larger than 5 m 2 . In addition, cells KD for SNX2 displayed a notable reduction of staining for HRSV N protein in the cell periphery ( Fig. 12D and G, arrowheads). Moreover, although syncytium formation in cells silenced for SNX1, SNX2, or SNX1 and 2 was not completely abrogated, there were significantly fewer syncytia in cells silenced for each or both SNX proteins (Fig. 12J to M), than in cells treated with scrambled siRNA (Fig. 12I and M). In agreement with the reduction in syncytium formation, Western blot analysis of HEp-2 cells at 24 hpi showed that the intracellular levels of HRSV G, F, and M proteins were reduced in cells KD for SNX1, SNX2, or SNX1 and 2 in comparison with the control (Fig. 12N). In addition, the immunoblotting of siRNA-treated cells demonstrated a reduction in HRSV protein levels in the supernatants (Fig. 12O). Consistently, there was a significant reduction in HRSV progeny production in cells silenced for SNX1 and 2 at 24 hpi (an approximate 70% reduction in the double KD) (Fig. 12P). Nevertheless, the viral genomic RNA production Superresolution image of an HRSV-infected cell, with arrowheads pointing to filaments budding from the cell, containing HRSV N and TGN46. All the images were taken at 24 hpi. Panels A to S are representative of a single plane from Z-stack imaging or a single focal plane of at least three independent experiments taken with a Leica SP5 confocal microscope. Magnification, ϫ63. Panel T was taken with a Nikon N-SIM microscope (superresolution imaging) and represents a single focal plane from Z-stack imaging. All the scale bars ϭ 10 m. quantified at 24 hpi was not significantly altered in cells and supernatants upon KD of SNX1, SNX2, or SNX1 and 2 ( Fig. 12Q and R), indicating that viral RNA replication was not significantly affected by the reduction in SNX1 and 2 levels in infected cells. Taken together, these data led us to conclude that the KD for the SNXs in HRSV-infected cells had an impact on the quantity of mature inclusion bodies, syncytium formation, HRSV protein amounts, and progeny production.

DISCUSSION
It is well known that the HRSV envelope glycoproteins G, F, and SH traffic through the Golgi apparatus, where the F protein undergoes maturation cleavages (2). However, little is known about the traffic of the HRSV nonglycosylated proteins to virus assembly sites. The present study has shown that the nonglycosylated HRSV structural proteins M and N are detected in association with the secretory pathway and retromer elements, and these are novel findings that contribute to understanding the HRSV assembly process.  In the present study, HRSV M and N proteins partially colocalized with giantin, a specific marker for cis and medial Golgi, mainly at the edges of this organelle. This finding that the traffic of the M protein is associated with membranes is in agreement with results by Henderson et al. (30), who reported an association of the M protein with membrane-enriched fractions of HRSV-infected cells. We have also shown that the HRSV F protein, in the Golgi, partially colocalized with inclusion bodies located at the vicinity of cis and medial Golgi cisternae. Since the N terminus of HRSV F glycoprotein is in the lumen of the Golgi (31), the cytosolic C-terminal portion of F could interact with the nonglycosylated HRSV M and N proteins within inclusion bodies. This was reinforced by results of superresolution microscopy and PLA (Fig. 2) and agrees with what was previously proposed by Ghildyal et al. (32), who suggested that the HRSV G and M proteins interact with one another in the Golgi. Céspedes et al. (14) also addressed the issue of HRSV N protein colocalization with the cis and medial Golgi using another Golgi marker, GALNT2. In the present study, we have expanded that by showing that in addition to the cis and medial Golgi, HRSV M and N proteins only partially colocalized with the TGN46, and this colocalization was significantly less than that of HRSV F with TGN46.
The importance of the Golgi apparatus in the HRSV protein traffic was also supported by experiments in which cell treatment with BFA resulted in smaller HRSV inclusion bodies and reduced N and F proteins in the cell surface, an effect that was reversible by BFA wash out. These observations are also in keeping with data by Céspedes et al. (14), who showed a BFA-induced reduction of N protein in the plasma membrane by flow cytometry. The present study added to that by showing that both HRSV glycosylated F and nonglycosylated N proteins depend on the integrity of the Golgi to traffic to the plasma membrane. Even though we are prone to believed that the lower quantity and size of inclusion bodies in BFA-treated cells is due to the dependency of the Golgi for its traffic, it is not possible to conclude that the inclusion body size was impacted only because of the Golgi disassembly. These phenomena could be due to the effect of BFA directly on the inclusion bodies, and an in-depth study should be done to fully understand it. Another explanation for the partial colocalization of HRSV inclusion bodies with giantin is that, in the HRSV entry process, vesicles containing the virus reach intracellular places near the Golgi region, and perhaps this is the first place where the inclusion bodies will appear, and therefore, as they get bigger they happen to colocalize with the Golgi. However, once again, this does not explain why the inclusion body size and delivery to the plasma membrane are altered by BFA treatment.
It has been previously shown that the HRSV N protein reaches late endosomes, as indicated by colocalization of N with LAMP1, a late endosome marker (14). The present study showed a considerable colocalization of HRSV M and N proteins with SNX2, which is a protein commonly found in association with early endosomes (Fig. 5). This suggests that HRSV inclusion bodies that concentrate nonglycosylated N and M proteins are sites of preferential accumulation of SNX2. Also, the accumulation of SNX2 in HRSV inclusion bodies is not uniform, as seen in Fig. 5E and H, and Fig. S4. We do not know yet what that means; however, Rincheval et al. (18) recently demonstrated that inclusion bodies are dynamic and heterogeneous, and this different pattern of SNX2 accumulation could be due to the inclusion bodies' heterogeneity and maturation during the course of the HRSV infection. The HRSV N and M proteins are known to interact with one another (2,26), which was confirmed by the present PLA results. Also, the present study has shown that the HRSV M and N proteins interact with SNX2, as evidenced by PLA for the M protein, and immunoprecipitation for the N protein. The findings of the present study indicate that there is an association of HRSV nonglycosylated proteins with elements of the secretory pathway. This observation is in agreement with previous demonstration of the recruitment of the adaptor protein complex-3 (AP-3), known to be involved in protein traffic between endosomes and the Golgi (33)(34)(35), to HRSV inclusion bodies and its interaction with the HRSV M protein (36). Furthermore, SNX2 was shown to colocalize with the AP-3 complex, which in turn, colocalizes with Vps26 (34,37). Moreover, a recent study has pointed out the colocalization of HRSV glycosylated G and nonglycosylated proteins in the same vesicles as the glycoprotein G recycles back from the plasma membrane (38).
It was noteworthy that the PLA of the HRSV M protein with SNX2 was seen as numerous puncta throughout the cytoplasm of infected cells. While our results do not confirm that all the PLA dots of M and SNX2 colocalized with HRSV inclusion bodies, considering that they were much more abundant than the viral inclusion bodies at the same time postinfection, it is tempting to speculate that the HRSV M protein traffics through early endosomes before reaching the larger more mature inclusion bodies or the plasma membrane. Since the HRSV M protein is a link between glycosylated and nonglycosylated proteins to form viral particles, its trafficking in association with endosomes could be a help to bridge the glycosylated proteins with RNP-complexed N protein in an endosome, which would facilitate the viral traffic and assembly. However, more studies should be performed to test this hypothesis, such as, for example, expression of the F, M, and N proteins in different combinations to check their capacity to associate with membranes in vitro. In addition, this movement of M protein is reminiscent of what was proposed by Cifuentes-Muñoz et al. (39), who showed that the formation of inclusion bodies of human metapneumovirus, another virus in the family Pneumoviridae, is dynamic, resulting from the coalescence of smaller elements.
It has been previously shown that in HRSV-infected cells, long filamentous structures containing HRSV proteins protrude from intracellular vesicles toward the cytosol (38). Our results showed that such intracellular filaments contain SNX2, suggesting that SNX2 may be involved in their formation. SNX2 and Vps26 are known components of the retromer complex that shapes filaments out of membranes (37), and we have detected Vps26 and SNX2 in vesicles containing HRSV N and F proteins as well as in filamentous structures pinched out from their surfaces, which reinforces that retromer components can be present in vesicles and filaments containing HRSV proteins.
It is well known that HRSV traffic is endosome dependent (2), but exactly how the HRSV nonglycosylated proteins interact with endosomes is not understood. Considering that SNX2 and Vps26 were detected in vesicles containing both glycosylated and nonglycosylated HRSV proteins, we asked if these vesicles could be early endosomes. We did not find detectable recruitment of EEA1, a marker of early endosomes, to HRSV inclusion bodies (Fig. S3), suggesting that SNX2 recruitment to inclusion bodies is selective and that inclusion bodies are not canonical early endosomes. Corroborating these data, cell fractionation showed recruitment of SNX2 to fractions that correspond to inclusion bodies, while EEA1 distribution in cell fractions was not different between HRSV infected and noninfected cells. Moreover, Lamp1, a marker of late endosomes and lysosomes, was very little detected in fractions corresponding to HRSV inclusion bodies, suggesting that these organelles are not components of HRSV inclusion bodies. It is noteworthy that actin was concentrated in cell fractions corresponding to HRSV inclusion bodies, in agreement with previous findings by Shahriari et al. (40).
The results of immunofluorescence and cell fractionation experiments indicate that TGN46 accumulated in the last fraction, which is enriched for plasma membrane ( Fig. 9 and 10) in HRSV-infected cells compared to noninfected cells. In addition, TGN46 was shown to be present in HRSV at the plasma membrane by superresolution microscopy, including seemingly detached HRSV buds. Taken together, these results suggest that TGN46 is carried in budded virus particles. Fluorescence findings similar to those obtained for TGN46 were observed for SNX2, which was detected in filamentous structures budding from the plasma membrane. It is interesting also that not all filaments containing HRSV proteins contained TNG46 and SNX2, suggesting that the particles that bud from the plasma membrane are not homogeneous.
One should keep in mind that TGN46 recycling from the plasma membrane back to the trans-Golgi could be compromised in HRSV-infected cells, since SNX2, which is part of the retromer complex, is recruited by inclusion bodies.
Importantly, the knockdown of SNX1 and 2 had a negative impact on several HRSV processes. First, upon cell silencing for these two proteins, alone or in combination, the amounts of intracellular HRSV F, G, and M proteins were reduced, and the HRSV inclusion bodies were smaller than in control cells at 24 hpi. Moreover, the knockdown for SNX1, 2, or both resulted in significantly smaller syncytia (Fig. 12). While the knockdown for sorting nexins did not affect HRSV replication as indicated by genome copy quantities at 24 hpi, the titers of infectious virus in the supernatant were significantly reduced in cells simultaneously silenced for SNX1 and 2. It is possible that the silencing of SNX1 and 2 may have altered the final content of structural proteins in progeny virions compared to controls, which might have affected the quantification of progeny production by plaque assay. These findings should be interpreted with caution since previous studies have shown that even HRSV particles lacking envelope G and SH glycoproteins retain infectivity (12,13), which could help to explain the moderate reduction seen in virus titers produced in cells silenced for SNX1, SNX2, and SNX1 and 2. SNX1 and 2 are part of the retromer complex, which is involved in the transport of cargo from endosomes to the trans-Golgi network (41). It has also been shown that in the absence of SNX1 and 2, some proteins that depend on the retromer complex for their traffic are addressed to lysosomal compartments for degradation (25). This could explain why in cells silenced for SNX1 and SNX2, there was a reduction in amounts of F protein and smaller syncytia at 24 hpi.
In summary, the present study contributed findings that help to establish the importance of the secretory pathway in the traffic of HRSV nonglycosylated structural proteins and showed for the first time the involvement of retromer proteins in the biogenesis of HRSV particles.

Cells and virus.
HEp-2 cells were grown in minimal essential medium (MEM) with 10% fetal bovine serum (FBS) and maintained in MEM with 2% FBS in 5% CO 2 at 37°C. HRSV-A of the strain Long (ATCC VR-26) was propagated in HEp-2 cells, and the stock was titrated by plaque assay using routine methods (42).
HRSV in vitro infection. Infection assays were done on glass coverslips inside 24-well plates, using HRSV stock diluted in phosphate-buffered saline (PBS) to reach a multiplicity of infection (MOI) of 1, with 1 hour of incubation at 4°C to synchronize virus entrance. Then, coverslips were washed three times in cold PBS, replenished with MEM with 2% FBS at 37°C, and then incubated in 5% CO 2 at 37°C. At 0, 4, 8, 12, 24, and 48 hpi the coverslips were fixed. For the immunoprecipitation and cell fractionation assays, HEp-2 cell monolayers were prepared in 75-cm 2 flasks, infected at MOIs of 0.1 and 1, and harvested at appropriate times.
Immunofluorescence. At appropriate times, coverslips were fixed with 4% paraformaldehyde (PFA) in PBS for 10 minutes at room temperature and then washed once with PBS, permeabilized with 0.01% Triton in PBS for 20 minutes, and then blocked with PBS containing 3% bovine serum albumin (BSA) for 30 minutes. After 10 washes with PBS, coverslips were incubated with the primary antibody for one hour at 37°C and then washed 10 times with PBS and incubated with the secondary antibody for 1 hour at 37°C. Coverslips were washed 10 times in PBS and once in distilled water and then mounted on glass slides using Flouromount. Nuclei were stained with DAPI, and images were acquired on Leica SP5 or Zeiss 780 confocal microscopes or a Nikon N-SIM microscope (super resolution imaging) and were analyzed using ImageJ software Fiji.  Small interfering RNA assays for SNX1 and SNX2. HEp-2 cells were seeded in 12-well plates (containing coverslips) in 40% of confluence; the second siRNA shot (40 nM each siRNA per well) was performed and 48 hours after the first shot, and then the cells were infected with HRSV (MOI ϭ 1). Then, 24 and 48 hours postinfection, the supernatant and cells were harvested for plaque assay, IF, real-time PCR, and Western blot analyses as described above. The sequences of each siRNA were siSNX1 (AUG AAG AAC AAG ACC AAG AGC CCA C) (IDT, Inc., Brazil) and siSNX2 (AAG UCC AUC AUC UCC ACC AAG AGC CAC). The scramble sequence was acquired from Sigma-Aldrich, Brazil.
Drug treatment. HEp-2 cells were seeded on coverslips and infected with HRSV (MOI ϭ 1), and 4 hours postinfection, fresh medium was replenished (2% FSB in MEM) containing 10 M brefeldin A (BFA) in DMSO or DMSO alone. The media containing BFA or DMSO vehicle were maintained for 5 hours, and then BFA or vehicle was washed out, while other wells were maintained to complete 19 hours of drug treatment and 24 hours of infection. Then, after 24 hours of infection, all the cells were fixed with 4% PFA in PBS. After fixation, the cells were subjected to IF as described above. For the experiment reported in Fig. S3, the HEp-2 cells were infected with HRSV at an MOI of 1. Next, 4, 8, 12, and 20 hours postinfection, the cells were incubated with medium (2% FSB in MEM) containing 10 M BFA. Then, 24 hpi, the cells were fixed with 4% PFA, and stained for HRSV N, giantin, and DAPI.
Cell fractionation. This protocol was based on a previous paper by Perez-Victoria (22,46). HEp-2 cells were infected with HRSV (MOI ϭ 1). After 24 hours, the supernatant from infected and mock-infected cells was discarded, and the monolayers were washed three times with PBS, followed by detaching with PBS with 0.5M EDTA. The cells were subjected to centrifugation at 200 ϫ g for 5 minutes; the pellets were washed with PBS and with STE buffer (250 mM sucrose, 20 mM Tris-HCl pH 7.4, 1 mM EDTA, with protease inhibitors) to produce an osmotic shock. Cells were passed 20 times through a 25G needle for cell lysis in 0.5 ml of STE buffer without sucrose. The cell homogenates were centrifuged at 1,000 ϫ g for 2 minutes to remove nuclear contents, and the post-nuclear supernatants were loaded on top of a discontinuous 10 to 30% (wt/vol) glycerol gradient in STE buffer, laid on 0.5 ml of 80% sucrose cushion in ultracentrifuge tubes. After that, tubes were centrifuged at 29,000 rpm for 2 hours in a Thermo TH629 rotor. After centrifugations, fractions (1.250 ml) were carefully collected from top to bottom of the ultracentrifuge tubes. Then, trichloroacetic acid (TCA) was added to each collected fraction for protein concentration, and the pellet was suspended in sample buffer with 4% beta-mercaptoethanol. The samples were heated at 95°C, loaded in 10% SDS-PAGE gel, transferred to nitrocellulose membrane, and analyzed by Western blotting.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. VIDEO S1, AVI file, 0.4 MB. VIDEO S2, AVI file, 3.9 MB.