Measles Virus Ribonucleoprotein Complexes Rapidly Spread across Well-Differentiated Primary Human Airway Epithelial Cells along F-Actin Rings.

The ability of viral particles to directly spread cell to cell within the airways without particle release is considered to be highly advantageous to many respiratory viruses. Our previous studies in well-differentiated, primary human airway epithelial cells suggest that measles virus (MeV) spreads cell to cell by eliciting the formation of intercellular membrane pores. Based on a newly generated ribonucleoprotein complex (RNP) “tracker” virus, we document by live-cell microscopy that MeV RNPs move along F-actin rings before entering a new cell. Thus, rather than diffusing through the cytoplasm of a newly infected columnar cell, RNPs take advantage of the cytoskeletal infrastructure to rapidly spread laterally across the human airway epithelium. This results in rapid horizontal spread through the epithelium that does not require particle release.

ized cells (17,35,36). Here, we used a recombinant MeV expressing GFP from an additional transcription unit ( Fig. 2A) to characterize the progression of MeV infection in HAE. Three days after inoculation, the F-actin along the lateral margins and the circumapical network disassembled in the center of infectious centers, but remained readily detectable in cells at the periphery (Fig. 2B, en face views, and Video S1 in the supplemental material). Because infectious centers form by radiating out from an initial infected cell (23,24,37), we postulate that peripheral cells are more recently infected than central cells. Unlike the F-actin along the lateral margins, the apical and basolateral F-actin networks remained readily detectable by phalloidin staining in all cells of infectious centers, including centrally located cells (Fig. 2B, Video S1). This observation was consistent for every infectious center from every donor tested.
To assess whether and how actin dynamics affects MeV spread, we treated HAE with three inhibitors of actin polymerization. Cytochalasin D (CytoD) binds the F-actin polymer and prevents further polymerization of monomers. CK-666 binds to the Arp2/3 protein complex that nucleates branched actin filaments (38). The small molecule inhibitor of formin homology 2 (SMIFH2) disrupts formin-dependent, but not Arp2/3 complex-dependent, actin cytoskeletal structures (39). The impact of the inhibitors on MeV spread in HAE was quantified by measuring the area of the infectious centers.
As shown in Fig. 2C, at the time of inhibitor application (36 h postinfection [hpi]), the sizes of the infectious centers were equivalent. Over the next 24 h (60 hpi), infectious center growth was significantly reduced in the CK-666 and SMIFH2 drug treatments compared to the dimethyl sulfoxide (DMSO) control. Of note, the SMIFH2 drug treatment had the greatest impact on infectious center growth, although the infectious center sizes after a 24-h treatment of CK-666 and SMIFH2 were not significantly different from each other. These findings potentially suggest a role for dynamic actin polymerization in the spread of MeV in HAE cells and may point to a specific role for formins. N, P, and M proteins localize near the apical and circumapical F-actin networks. Previous studies of MeV-infected immortalized cell lines suggest that F-actin is involved in particle assembly (40,41). To characterize this process in HAE cells, we used the recombinant virus MeV-nCFP (Fig. 3A). This virus expresses a variant of the cyan fluorescent protein that includes a nuclear localization signal (nCFP), allowing not only immediate identification of infected cells but also interference-free study of cytoplasmic assembly processes (42). Cells were fixed and permeabilized 3-days postinfection before labeling with antibodies against N ( For both N and P proteins, we monitored two distinct sites of accumulation in HAE cells. N and P were readily detectable in large perinuclear bodies ( Fig. 3B and D; see also Fig. S1A and B). These bodies may be the sites of viral transcription and replication (43,44). In addition, proximal to the apical plasma membrane, localization was observed ( Fig. 3B and D; see also Fig. S1A and B). These sites colocalize with the apical F-actin network. M protein was also occasionally found in perinuclear bodies ( Fig. 3F; see also Fig. S1C) but accumulated mainly at the apical membrane.
We then used stimulated emission depletion (STED) microscopy to examine the extent of colocalization of the viral proteins N (Fig. 3C), P (Fig. 3E), and M ( Fig. 3G) with the apical F-actin network at higher resolution. Again, we used MeV-nCFP, whose fluorescent reporter protein is segregated to the nucleus. We quantified the colocalization of MeV proteins to apical F-actin by Mander's colocalization coefficient (MCC) analysis (Fig. 3H). This analysis of N, P, and M proteins (green) and F-actin (red) signals in dually labeled images resulted in MCC values of 0.773 Ϯ 0.07, 0.755 Ϯ 0.03, and The eGFP coding region was inserted between the leader sequence and the N gene as a separate transcription unit. (B) HAE cells were infected with MeV-GFP (MOI ϭ 1) and imaged using confocal microscopy 72 h later. Fixed and permeabilized HAE cells were stained for F-actin with rhodamine-conjugated phalloidin (red), and the nuclei were visualized with DAPI (blue). Both en face (upper panels) and vertical (lower panels) sections are shown. Arrows indicate the central regions of the infectious center where F-actin has disassembled. Scale bar, 20 m. Images are representative from n ϭ 6 samples (two technical replicates from three human donors [biological replicates]). (C) HAE cells were infected with MeV-GFP at an MOI of 1 and, 36 h later, the cultures were treated with the F-actin-disrupting agents cytochalasin D (CytoD), CK-666, or SMIFH2 or a control (DMSO) for 24 h. The areas of the infectious centers were quantified using ImageJ software at the time of drug delivery (36 h) and 24 h after drug delivery (60 h). Each dot represents the log-transformed value of the area of individual infectious centers. The data for each condition are pooled from six human donors. AU, arbitrary unit. Adjusted P value using one-way ANOVA corrected for multiple comparisons were determined (***, P Ͻ 0.001; ****, P Ͻ 0.0001). 0.793 Ϯ 0.06, respectively. A value of 1 is perfectly colocalized, and a value of Ϫ1 is perfectly mislocalized. These data confirm the close proximity of all three viral proteins to apical F-actin.
Generation and validation of recombinant MeV-RNP tracker . We previously monitored the intercellular transport of cytoplasmic GFP in MeV-infected HAE cells (24). To follow the intercellular spread of MeV RNP in real time, we generated here a recombinant MeV that incorporates GFP in its RNP, termed MeV-RNP tracker . In addition to the standard P protein, this virus expresses a GFP/P fusion protein from an additional transcription unit inserted between the H and L genes (Fig. 4A). The P and GFP/P proteins are expressed at similar levels ( Fig. 4B and C). As a control, we also generated MeV(GFP)H, which expresses soluble GFP from the same transcriptional unit between H and L; this virus differs from MeV-GFP ( Fig. 2A) only by the location of the GFP transcriptional unit within the MeV genome. Both recombinant MeV(GFP)H and RNP tracker are viable, and their replication kinetics are similar in Vero-hSLAM cells (Fig. S2A), but the replication kinetics of RNP tracker were slightly delayed in an epithelial cell line H358 (Fig. S2B). We further compared their replication in HAE cells. Infection of HAE cells with RNP tracker results in similar numbers and sizes of infectious centers compared to MeV(GFP)H ( Fig. S2C and D).
We then monitored the spread of these two viruses in HAE cells. The infectious centers of MeV(GFP)H (Fig. 4D) were characterized by diffuse cytoplasmic fluorescence, indistinguishable from what was observed with MeV-GFP (Fig. 2B). In contrast, in the MeV-RNP tracker infectious centers, GFP signals were punctate, concentrated along the intracellular surface of the apical membrane, and in perinuclear regions (Fig. 4E). These observations, which recapitulate the immunohistochemistry results using anti-P and anti-N antibodies in MeV-GFP-infected HAE cells (Fig. 3), are consistent with incorporation of the GFP/P protein in RNPs.
To confirm that GFP-tagged P protein produced from RNP tracker is incorporated into the RNP complex, we immunostained for P and N in RNP tracker -infected HAE cells ( Fig. 5A and Fig. S3A). Indeed, the GFP-tagged P protein colocalized with the N and P MeV RNPs spread to adjoining cells along the circumapical actin network. We previously suggested that MeV RNPs take advantage of intercellular pores and of the circumapical actin rings to rapidly spread across the human airway epithelium, a hypothesis that we can now test by MeV-RNP tracker -based microscopy. After infecting HAE cells with this virus, we focused on individual GFP-expressing cells at an early time point. These cells begin to appear ca. 30 to 36 h after inoculation. Remarkably, the MeV RNPs (green) encircle newly infected cells (Fig. 6, white arrow) and colocalize with F-actin (red) in the circumapical network. This colocalization of MeV RNPs and F-actin at apical and circumapical networks was also observed at later infection time points (72 hpi) at the periphery of infectious centers (Fig. S4, white arrows indicate MeV RNPs along the circumapical F-actin network in newly infected cells).
We then sought, using immunohistochemistry, to assess whether we can detect M protein in newly MeV-RNP tracker -infected cells. Figure S5 shows an infected cell next to a newly infected cell on its left. The vertical section of the left panel confirms that, like the P protein of MeV-nCFP (Fig. 3), the GFP-P protein of MeV-RNP tracker localizes to perinuclear bodies, as well as proximal to the apical plasma membrane. M protein is only detected at the plasma membrane. Within the apical region, M and P localization only partially overlaps, with the bulk of M-protein closer to the plasma membrane than the P protein. The horizontal sections on the right indicate that both the GFP-P and the M proteins of MeV-RNP tracker transfer to the newly infected cell.
To further investigate the cell-to-cell spread of MeV RNPs, we monitored MeV-RNP tracker spread in HAE cells by live-cell microscopy. For visualization of F-actin in live cells, HAE cells were transduced with an adenoviral vector expressing TagRFP-tagged LifeAct, a 17-amino-acid peptide that binds to F-actin in live cells without disruption of

DISCUSSION
We characterized MeV spread in well-differentiated, primary cultures of human airway epithelial cells. Based on a newly generated RNP "tracker" virus, we documented by video microscopy that RNPs move along the circumapical F-actin ring and enter a new cell. Thus, rather than diffusing through the cytoplasm of a newly infected columnar cell, RNPs take advantage of the cytoskeletal infrastructure to spread laterally across HAE. This results in rapid horizonal spread through the epithelium, as predicted by the intercellular-pore/fast-lane model (21,24). Viral spread through intercellular pores versus cell fusion. MeV infections have been studied mainly in monolayers of immortalized cells, where the virus spreads through cell fusion. Receptors on recipient cells trigger the viral membrane fusion apparatus on the surfaces of infected cells to form fusion pores. Rapid pore expansion results in the coalescence of plasma membranes and formation of multinucleated syncytia. However, in the airways of living hosts, syncytia are rarely detected (46)(47)(48).
Lack of syncytium formation in response to MeV infection is recapitulated in HAE cells, where large infectious centers are formed, while transepithelial resistance is maintained (23). Furthermore, plasma membranes of individual infected cells remain intact, suggesting that fusion pores do not expand (23,24). Even in the absence of cytopathology, MeV spread in the human airway epithelium is more rapid than that of other respiratory viruses (24). Thus, while extensive cell fusion occurs in infected monolayers of transformed cells, it may be counterproductive for infections of wildtype viruses in natural hosts.
F-actin and viral replication. We document how RNPs traveling on circumapical F-actin rings traverse individual columnar cells and cross over to the next one. During this process, some RNPs may be diverted from the apical rings and transported to a perinuclear location, where replication centers form. These RNPs could be inactive when they reach the adherens junction of the originally infected cell and become transcriptionally active in the new host cell. Alternatively, small, transcriptionally active replication centers could be transported intercellularly. In both cases, an almost synchronous wave of RNP infiltration would follow the opening of an intercellular fusion pore. This synchronous infection with multiple viral genomes may overwhelm the antiviral defenses of naive cells.
We consistently observe that the lateral regions of the F-actin network break down within the innermost cells of infectious centers but remains intact within peripheral cells. While the mechanistic cause of apical F-actin ring breakdown remains unclear, this process may favor directional movement of MeV RNPs from infected to uninfected cells, thereby preventing superinfection. In addition, disassembly of the circumapical network may redirect traversing RNPs to perinuclear locations most favorable for their replication.
Different observations suggest multiple roles for F-actin in MeV replication. F-actin was observed in close association with MeV RNPs in infected cells (40), and it was also reported that inhibitors of actin polymerization restrict MeV replication (49,50). Intriguingly, other studies documented large amounts of monomeric actin in MeV particles (36,51,52), far exceeding actin levels reported for other paramyxovirus particles (53). Although immune cells lack the same defined cytoskeletal organization of well-differentiated columnar epithelial cells, our observations may be consistent with how MeV spreads between monocytes and/or lymphatic cells. Indeed, MeV transmission from dendritic cells to lymphocytes relies on the formation of transient adhesive structures similar to virological synapses, where F-actin accumulates at the cellular junctions (54), again implicating F-actin in virus transmission. Future studies using RNP tracker and live-cell imaging of MeV-infected primary monocytes may further elucidate how it spreads systemically. Extensive colocalization of MeV RNPs to the circumapical F-actin network during infectious center formation in the human airway epithelium is consistent with previous observations and provides a mechanism for rapid cell-to-cell spread. Altogether, these observations are consistent with transport of MeV components on F-actin.
Our previous study highlighted the importance of the Nectin-4/Afadin complex during MeV spread in HAE (24), but thus far it remains unclear which F-actin binding and signaling proteins facilitate the transport of RNPs along the circumapical F-actin network. Additional cellular proteins are likely involved in RNP transport from infected to uninfected cells, as well as in switches between actin filaments (55,56).

Mechanisms of intra-and intercellular virus spread.
Microtubules are assumed to be the major cellular transporter highways (55). However, recent studies highlighted the importance of actin-dependent transport of Rab11-containing vesicles with several different cargos toward the plasma membrane of mammalian cells (57). Indeed, microtubule dependent Rab11A-positive endosomes mediate intracellular transport of MeV RNP complexes in Vero/hSLAM cells and polarized epithelial cells (58). A combination of microtubule-and actin-dependent movement may be responsible for intraand intercellular transport of MeV RNP complexes.
Using HAE cells, we previously characterized intercellular pores that allow the transfer of cytoplasmic contents from MeV-infected to naive neighboring columnar epithelial cells (24,37). Here, we address the critical question of what traffics through the pores. Opening membrane pores in HAE cells likely requires viral H and F proteins, implying that these proteins travel with the RNP. The M protein may be critical for bridging RNPs with the F and H proteins. Our data suggest that M protein also rapidly reaches naive cells but does not exclusively colocalize with RNPs. As such, partially assembled MeV particles composed of different combinations of viral proteins may spread cell to cell in HAE cells.
For viruses of several different families, a high concentration of components at sites of cell-cell contact may account for cell-to-cell transmission being more efficient than particle-dependent transmission (18)(19)(20). In polarized cells of the respiratory epithelium, MeV components are concentrated mainly to the narrow cytoskeletal ring connected to the adhesive and tight junctions and located near the apical surface. This concentration mechanism permits spreading at much reduced viral and cellular costs, allowing lower viral gene expression and extending cell viability. Other viruses use variations of this strategy. For example, in the case of HIV, particles on the cell surface are likely drawn to the virological synapse through cytoskeletal interactions (59,60). In the case of poxviruses, cytoskeletal structures forming below the cell surface push particles to "surf" toward noninfected cells (61). Herpes viruses spread in polarized neurons by a virus-directed mechanism that operates by coordinating opposing microtubule motors to favor sustained retrograde delivery of the virus to the peripheral ganglia (62). Thus, viruses can utilize existing sites of cell-cell contact, such as adhesive and tight junctions or neurological or immunological synapses, to most efficiently spread in the host.
In conclusion, our studies further develop a paradigm for the rapid intrahost spread of viruses that relies on cell-cell adhesion and on cytoskeletal driving forces and is particle independent. Such cell-associated mechanisms often result in the cotransmission of multiple viral genomes, further enhancing the efficacy of virus dissemination within hosts (63).

MATERIALS AND METHODS
Ethical statement. Primary cultures of human airway epithelial (HAE) cells were prepared from discarded tissue, autopsy, or surgical specimens. All specimens used in this study are obtained from the University of Iowa In Vitro Models and Cell Culture Core Repository. We were not provided with any information that could be used to identify a subject. All studies involving human subjects received University of Iowa Institutional Review Board approval.
Recombinant MeV used and generation of MeV-RNP tracker . All four recombinant MeV used in this study are derivatives of wild-type strain IC-323 (64,65). The generation of MeV-GFP (65) and MeV-nCFP (42) were previously published. MeV-RNP tracker and its control virus MeV(GFP)H were generated for this study. MeV production was conducted as previously described (65,66). The design of MeV-RNP tracker is based on a previous study showing that a vaccine-lineage MeV expressing a GFP-P chimeric protein can replicate to high titers, provided that the original copy of the P gene is maintained (67). To generate a eGFP-P fusion construct, the P open reading frame was amplified from the p(ϩ)MeV323 plasmid (64) using the primers P 323 -fwd (GTA CCG GTG GAG CAG AAG AGC AGG CAC GCC ATG) and P 323 -rev (TTG CAT GCG CTA CTT CAT TAT TAT CTT CAT CAG CAT CTG GTG G) and cloned into pCG-eGFP-P NSe (67) using AgeI and SphI restriction sites, replacing the P protein of the vaccine-lineage strain (P NSe ). To express only GFP-P, without expressing GFP-V, the polypurine tract required for G-nucleotide insertion was mutagenized (68) using the primers P 323 -V KO  MeV production and titering. MeV production was conducted as previously described (65,66). Briefly, Vero cells stably expressing the MeV receptor SLAMF1 (Vero-hSLAM cells, kindly provided by Y. Yanagi) (69) were cultured in Dulbecco modified Eagle medium (DMEM; Thermo Fisher Scientific) containing 8% newborn calf serum (NCS; Thermo Fisher Scientific) and penicillin-streptomycin (100 mg/ ml; Thermo Fisher Scientific) were used to produce MeV. Recombinant MeV titers of approximately 10 7 50% tissue culture infective doses (TCID 50 )/ml were obtained.
Western blot analysis. For Western blot analysis, 10 6 Vero-hSLAM cells (8) were either infected with MeV-eGFP or MeV-RNP tracker at a multiplicity of infection (MOI) of 0.1 or left uninfected and then harvested at 24 hpi and lysed in 300 l of denaturing protein lysis buffer (62.5  . Samples were heated at 95°C for 10 min and 10 l/lane were loaded on 10% SDS-PAGE gels. Western blots were performed as previously described (44). The antibodies used for detection were polyclonal rabbit anti-N 505 (1:5,000 [70]), polyclonal rabbit anti-P 254 (1:5,000 [70]), monoclonal mouse anti-GFP (clone GF28R, 1:1,000, catalog no. 14-6674-82 Primary HAE cells. HAE cells were collected from tracheas and bronchi by enzymatic dispersion using established methods (22). Briefly, epithelial cells were dissociated and seeded onto collagencoated, semipermeable membranes (0.4-m pore size; surface area, 0.33 cm 2 ; Corning, Corning, NY). HAE cultures were maintained in Ultraser G (USG) media at 37°C and 5% CO 2 . Polyester transwell inserts were placed into 24-well plastic cell culture plates (Costar, Cambridge, MA). At 24 h after seeding, the mucosal medium was removed, and the cells were allowed to grow at the air-liquid interface as reported previously (22). Only well-differentiated cultures (Ͼ3 weeks old) were used in these studies. The presence of tight junctions was confirmed by measuring the transepithelial resistance using a volt-ohm meter (World Precision Instruments, Sarasota, FL; resistance, Ͼ500 ⍀·cm 2 ).
Infection of primary airway epithelial cells. Infection of HAE cells was performed as previously described (23,24). Briefly, to infect airway epithelia with MeV from the basolateral side, the cultures were inverted, and the virus was applied to the basolateral surface for 4 h in 80 l of serum-free medium. After the infection, the virus was removed, and the transwell insert was turned upright and incubated at 37°C under 5% CO 2 for the indicated times. For actin visualization in time-lapse microscopy, HAE cells were transduced with adenovirus-containing LifeAct-TagRFP (Ibidi, Munich, Germany) at a preoptimized MOI according to the manufacturer's protocol.
Inhibitor treatment. Inhibitor stocks were prepared in DMSO at concentrations of 4 mM for CytoD (Sigma-Aldrich), 10 mM for CK-666 (Sigma-Aldrich), and 3 mM for SMIFH2 (Sigma-Aldrich). Different dilutions of the inhibitors were assessed (1 to 4 M for CytoD, 100 to 500 M for CK-666, and 100 to 500 M for SMIFH2) to determine the appropriate working concentrations that have maximal effects on the actin cytoskeleton (see Fig. S6 in the supplemental material) without causing a loss of transepithelial resistance below 500 ⍀·cm 2 . At 36 h after MeV-GFP infection, HAE cells were treated apically and basolaterally with final concentrations of 4, 500, and 100 M for CytoD, CK-666, and SMIFH2, respectively, for 24 h. After treatment, the HAE cells were imaged using an inverted UV fluorescence microscope. The data are presented as means Ϯ the standard deviations of the log-transformed value of individual data points. Statistical significance as adjusted P values between groups was determined by using one-way analysis of variance (ANOVA) with corrections for multiple comparisons.
Immunostaining and microscopy. At the indicated times postinfection, HAE cultures were processed for confocal microscopy. For processing, the cells were fixed in 2% paraformaldehyde, permeabilized in 0.2% Triton X-100, and blocked in 1% Superblock (Thermo Fisher Scientific) for 1 h. For immunostaining, HAE cells were incubated with the primary antibodies anti-N 505 (1:200 [24]), polyclonal rabbit anti-P 254 (1:200 [44]), and anti-M (1:200, catalog no. MAB8910 [Millipore-Sigma]) overnight at 4°C. Secondary antibodies were either Alexa 568-labeled goat anti-mouse and Alexa 568-labeled goat anti-rabbit secondary antibody (Life Technologies) for confocal imaging or Alexa 532-labeled goat anti-mouse and Alexa 532-labeled goat anti-rabbit secondary antibody (Life Technologies) for STED imaging. Actin was stained with either rhodamine-phalloidin (1:100, catalog no. R415; Thermo Fisher Scientific) or phalloidin-Alexa 647 (1:40, catalog no. A22287; Thermo Fisher Scientific) for 30 min. The transwell inserts containing the cells were separated from the plastic cylinder by cutting the edges with a razor blade. The cells were mounted on a slide with either Vectashield with DAPI (4=,6=-diamidino-2phenylindole; Vector Laboratories, Inc., Burlingame, CA) for confocal imaging or with ProLong Diamond with DAPI antifade reagent (catalog no. P36971; Thermo Fisher Scientific) for STED imaging. Unless otherwise noted, immunostaining was performed on at least two cultures (technical replicates) from three human donors (biological replicates) for all microscopy figures. Images were captured from each donor, and representative images are shown. We note that the sizes and numbers of infectious centers varied from donor to donor; however, the expression pattern of viral proteins and F-actin remained remarkably consistent.
Confocal images were acquired on a Leica TCS SP3 confocal microscope (Leica Microsystems, Inc., Buffalo Grove, IL) using a 40ϫ or 63ϫ oil immersion objective. STED images were acquired on a Leica SP8 STED 3ϫ (Leica Microsystems, Inc.) platform equipped with lasers for the depletion of fluorophores emitting in the orange (660 nm; Laser Quantum) and red/far red (775 nm; OneFive, GmbH) using a 100ϫ NA:1.4 oil immersion lens objective. Confocal and STED images were obtained at 1,024 ϫ 1,024 or at 2,048 ϫ 2,048 pixels, respectively. A pair of depletion wavelength lasers (nm ϭ 660/592) or a single wavelength laser (nm ϭ 660) was applied for multicolor STED imaging. To optimize resolution without bleaching, the STED lasers were applied at the lowest power that can provide sufficient improvement in resolution compared to confocal imaging. STED images were deconvolved with Huygens essential software (Scientific Volume Imaging) using the CMLE algorithm. We analyzed 50 to 60 optical z-stacks (step size, 1 m) using ImageJ 1.47v software.
To determine the colocalization of red and green signals, the images were analyzed as z-stacks using channel 1 for green and channel 2 for red. The region of interest was analyzed using the Coloc2 plugin with the Costes method for automatically estimating threshold values for identifying background levels (71,72).