Human Antiviral Protein MxA Forms Novel Metastable Membraneless Cytoplasmic Condensates Exhibiting Rapid Reversible Tonicity-Driven Phase Transitions

Abstract

Phase-separated biomolecular condensates of proteins and nucleic acids form functional membrane-less organelles (e.g., stress granules and P-bodies) in the mammalian cell cytoplasm and nucleus. In contrast to the long-standing belief that interferon (IFN)-inducible human myxovirus resistance protein A (MxA) associated with the endoplasmic reticulum (ER) and Golgi apparatus, we report that MxA formed membraneless metastable (shape-changing) condensates in the cytoplasm. In our studies, we used the same cell lines and methods as those used by previous investigators but concluded that wild-type MxA formed variably sized spherical or irregular bodies, filaments, and even a reticulum distinct from that of ER/Golgi membranes. Moreover, in Huh7 cells, MxA structures associated with a novel cytoplasmic reticular meshwork of intermediate filaments. In live-cell assays, 1,6-hexanediol treatment led to rapid disassembly of green fluorescent protein (GFP)-MxA structures; FRAP revealed a relative stiffness with a mobile fraction of 0.24 ± 0.02 within condensates, consistent with a higher-order MxA network structure. Remarkably, in intact cells, GFP-MxA condensates reversibly disassembled/reassembled within minutes of sequential decrease/increase, respectively, in tonicity of extracellular medium, even in low-salt buffers adjusted only with sucrose. Condensates formed from IFN-α-induced endogenous MxA also displayed tonicity-driven disassembly/reassembly. In vesicular stomatitis virus (VSV)-infected Huh7 cells, the nucleocapsid (N) protein, which participates in forming phase-separated viral structures, associated with spherical GFP-MxA condensates in cells showing an antiviral effect. These observations prompt comparisons with the extensive literature on interactions between viruses and stress granules/P-bodies. Overall, the new data correct a long-standing misinterpretation in the MxA literature and provide evidence for membraneless MxA biomolecular condensates in the uninfected cell cytoplasm.IMPORTANCE There is a long-standing belief that interferon (IFN)-inducible human myxovirus resistance protein A (MxA), which displays antiviral activity against several RNA and DNA viruses, associates with the endoplasmic reticulum (ER) and Golgi apparatus. We provide data to correct this misinterpretation and further report that MxA forms membraneless metastable (shape-changing) condensates in the cytoplasm consisting of variably sized spherical or irregular bodies, filaments, and even a reticulum. Remarkably, MxA condensates showed the unique property of rapid (within 1 to 3 min) reversible disassembly and reassembly in intact cells exposed sequentially to hypotonic and isotonic conditions. Moreover, GFP-MxA condensates included the VSV nucleocapsid (N) protein, a protein previously shown to form liquid-like condensates. Since intracellular edema and ionic changes are hallmarks of cytopathic effects of a viral infection, the tonicity-driven regulation of MxA condensates may reflect a mechanism for modulation of MxA function during viral infection.

Keywords: antiviral granules; biomolecular condensates; hypotonic disassembly; interferons; isotonic reassembly; membraneless cytoplasmic organelles; myxovirus resistance protein A (MxA); regulation by tonicity; vesicular stomatitis virus.

FIG 1

Variation in MxA-positive cytoplasmic structures in different cells. Cultures of the indicated cells grown in 35-mm plates or 6-well plates were transiently transfected with pcDNA or the pHA-MxA expression vector (the same as that used by Stertz et al. [37]), fixed 1 day later, and permeabilized using a digitonin-containing buffer, and the distribution of MxA was evaluated using an anti-HA MAb or an anti-MxA rabbit PAb and immunofluorescence methods. (A) MxA structures in HEK 293T cells imaged using a 40× water immersion objective (scale bar, 20 μm). (B) MxA structures in HEK293T cells imaged using a 100× oil immersion objective (scale bar, 10 μm). Arrows point to large compact tubuloreticular MxA-positive structures, as observed in 25 to 35% of transfected cells. (C) MxA structures in Huh7 cells imaged using a 40× water immersion objective (scale bar, 10 μm) showing variably sized bodies (left image) and larger reticular formations in the cytoplasm, as observed in 25 to 35% of transfected cells in the cytoplasm (right image, arrow) (scale bar, 10 μm). (D) Huh7 cells were exposed to IFN-α2a (3,000 IU/ml) for 2 days or left untreated and then fixed and immunostained for endogenous MxA (in red) and endogenous DRP1 (in green) (scale bar, 5 μm).

FIG 2

Comparison of IFN-α-induced endogenous MxA expression with that from exogenous vectors and variation in GFP-MxA structures formed in Huh7 cells. (A) Western blot analyses showing comparable levels of MxA expression in Huh7 cells in 35-mm plates exposed to IFN-α (2,000 IU/ml for 2 days) or following transient transfection with 1 μg/plate of the GFP-MxA (in duplicate) and HA-MxA expression vectors. Protein-matched aliquots of whole-cell extracts were Western blotted and probed for MxA using a rabbit PAb or for GAPDH using a rabbit MAb. Fold increase in MxA is expressed in terms of the untreated control. (B to D) Huh7 cells in 35-mm plates were transfected with the pGFP-MxA expression vector and the cells imaged before (B and D) or after (C) fixation. (B) Live-cell wide-field 100× oil imaging of the periphery of one cell with GFP-MxA structures of variable size within the same cell 1 day after transfection; smaller-sized MxA bodies were at the cell periphery with larger structures closer to the cell center. (C) Portions of three adjacent cells 1 day after transfection showing GFP-MxA associated with structures in different configurations, including a reticular pattern of intersecting lines (arrows; left-most cell). Scale bars in panels B and C, 10 μm. (D) Open meshwork GFP-MxA reticulum in live Huh7 cells 2 days after transfection, imaged by placing a coverslip on the cells followed by time-lapse imaging using a 100× oil objective. Scale bar, 5 μm.

FIG 3

MxA structures were distinct from smooth endoplasmic reticulum tubules and the Golgi apparatus. (A and B) Huh7 cells were transfected with vectors for HA-tagged MxA or HA-tagged ATL3 as indicated, the cultures (in 35-mm plates) were fixed 1 day later, and the distribution of the HA tag and of RTN4 was sequentially evaluated using the mouse anti-HA MAb and then a goat anti-RTN4 PAb. Images were collected after placing a drop of PBS in each culture, overlaying it with a coverslip, and using a 100× oil immersion objective. In panel A, high-magnification imaging confirmed that HA-MxA did not colocalize with RTN4-positive tubules, while HA-ATL3 was colocalized with the RTN4-positive ER (both ATL3 and RTN4 are known ER-resident proteins, and peripheral ER tubules represent the smooth ER subcompartment [34–37]). Scale bar, 5 μm. Thin arrows point to three-way junctions that were variably positive for ATL3-HA and RTN4. White lines in the respective merged images in panel A show regions depicted in the line scans in panel B (subpanels a and b). (C) Huh7 cells in 35-mm plates were transiently cotransfected with the vectors for GFP-MxA and mCh-Sec61β (a subunit of the ER-resident translocon assembly [39–42]), fixed 2 days later, and evaluated for colocalization of the respective proteins using a high-resolution Zeiss confocal Airyscan microscopy system. Scale bar, 5 μm. (D) Huh7 cells in 35-mm plates were transiently transfected with the vector for HA-MxA, fixed 2 days later, and evaluated for colocalization of HA-MxA structures (probed using anti-HA mouse MAb) with the Golgi tether giantin (probed using anti-giantin rabbit PAb). Scale bar, 10 μm. R values in the respective merged images in panels A, C, and D correspond to Pearson’s correlation coefficient R with Costes’ automatic thresholding.

FIG 4

Thin-section EM of GFP-MxA structures in Huh7 cells carried out using a CLEM protocol and immuno-EM of MxA reticulum. (A to C) Huh7 cells plated sparsely in 35-mm MatTek gridded coverslip plates were transiently cotransfected with pGFP-MxA. Two days later the cultures were fixed with 4% paraformaldehyde for 1 h at 4°C. Confocal imaging was carried out using a tiling protocol to identify the location of specific cells with GFP-MxA structures on the marked grid. The cultures were then further fixed and embedded, and the previously identified grid locations were used for serial thin-section EM. The tiled light microscopy data were correlated with the tiled EM data to identify the ultrastructure of the GFP-fluorescent structures per the correlated light and electron microscopy (CLEM) procedure. (A) Thin-section EM image of GFP-MxA structures (arrows). (B) Thin-section EM image of a portion of a cell in the same section as that shown in panel A imaged at the same time, verifying the ability to detect ER membranes in this experiment. (C) Thin-section EM showing GFP-MxA structures (arrows) aligned along intermediate filaments (Int Filam; broken arrows). A higher magnification of inset a in panel C is shown in the lower left corner. Scale bars in panels A to C, 200 nm. (D and E) HA-MxA-expressing HEK293T cells (as shown in Fig. 1A) were prepared for immuno-EM analyses of MxA-positive structures using rabbit PAb to MxA and 15-nm protein A-gold particles as the secondary label. Panel D shows an example of MxA-reticulum (MxA-R), while panel E shows the boxed inset in panel D at a higher magnification. Scale bars, 500 nm (D) and 200 nm (E).

FIG 5

Association of HA-MxA structures with a cytoplasmic intermediate filament meshwork in Huh7 cells. Cultures of Huh7 cells were transiently transfected with the HA-MxA expression vector, and the localization of HA-MxA structures as well as the giantin-based intermediate filament meshwork unique to Huh7 cells (white arrows) was evaluated 2 days later by immunofluorescence assays. Panels A and B show two independent experiments using different microscopy methods (wide-field or confocal) in which the respective anti-HA MAb (to label HA-MxA) and anti-giantin PAb were displayed using different batches of respective Alexa Fluor-488 (green) or Alexa Fluor-594 (red) secondary antibodies. (C) Line scan in red and green of the white line in the triple merged image in panel B. Scale bars, 10 μm. R value in the merged image in panel B corresponds to Pearson’s correlation coefficient R with Costes’ automatic thresholding.

FIG 6

Homotypic association of GFP-MxA bodies to yield larger structures. (A) Huh7 cells in 35-mm plates were transfected with the pGFP-MxA expression vector. Two days later the live cells, under a coverslip with a drop of PBS, were imaged using a 100× oil immersion objective and time-lapse data collection (images were collected from the periphery of different cells at 5-s intervals for 1 to 3 min). Shown are images from one time-lapse sequence lasting approximately 3 min (Movie S1) showing four homotypic association events (arrows). The data also show persistence of the combined structure. Scale bar, 5 μm. (B) Thin-section EM images from the CLEM data, as shown in Fig. 4, which highlight the homotypic fusion event between small GFP-MxA bodies (arrows) with larger MxA structures, all of which lack an external limiting membrane. Scale bar, 200 nm.

FIG 7

Test of liquid-like properties of GFP-MxA condensates. (A) Hexanediol rapidly disassembled GFP-MxA condensates. Huh7 cells expressing GFP-MxA condensates were first imaged in PBS and then exposed to PBS containing 1,6-hexanediol (5%) and imaged 24 s and 108 s later. Scale bar, 10 μm. (B and C) FRAP analyses of replicate (n =8) internal regions of GFP-MxA condensates as summarized in Materials and Methods. (B) Location of two of the bleached spots evaluated. (C) Normalized bleaching and recovery plot for one of the spots in panel B. Overall (n = 8), the mobile fraction was 0.24 ± 0.02 and t_1/2 was 2.37 ± 0.35 s (means ± standard errors [SE]). Controls for normalization included unbleached spots in each image, as well as background recording away from the GFP-MxA condensate. As a positive control for mobility, Huh7 cells expressing GFP-STAT3 in the cytoplasm showed a mobile fraction of 0.70 (12, 70).

FIG 8

Metastability of GFP-MxA condensates: transition to filamentous reticulum. (A) Huh7 cells in 35-mm plates were transfected with the pGFP-MxA expression vector. Two days later the live cells were placed under a coverslip with a drop of PBS and imaged using a 100× oil immersion objective. By 30 to 40 min into the imaging session, GFP-MxA was observed to be in a fibrillar meshwork in most cells. (B) GFP-MxA-expressing cells were treated with the nitric oxide scavenger c-PTIO (300 μM for 7 h). Fibrillar conversion of GFP-MxA condensates was first observed using a 40× water immersion objective (not shown) and further confirmed by placing a coverslip and 100× oil imaging with time-lapse data collection. Image shown is one illustrative example of a 100× time-lapse experiment. A time-lapse movie of a different cell is shown in Movie S2. (C) GFP-MxA-expressing cells were treated with dynasore (20 nM) for 4 days. Fibrillar conversion of GFP-MxA condensates was first observed using a 40× water immersion objective (not shown) and further confirmed by adding a coverslip and using 100× oil imaging with time-lapse data collection. The image shown is one illustrative example of a 100× time-lapse experiment. A time-lapse movie of a portion of the same cell is shown in Movie S3. All scale bars, 5 μm.

FIG 9

Rapid and reversible tonicity-driven disassembly and reassembly of GFP-MxA condensates in the cytoplasm of Huh7 cells. (A) GFP-MxA-expressing cultures 2 days after transfection were sequentially imaged on a warm stage (37°C) using a 40× water immersion objective under isotonic conditions (RPMI 1640 medium), switched to hypotonic medium (erythrocyte lysis buffer [ELB] at approximately 40 mosM), imaged 3 min later and then switched to isotonic buffer (ELB supplemented with 0.3 M sucrose, net of approximately 325 to 350 mosM), and imaged 5 min later (cycle 1), followed by a second cycle of hypotonic and isotonic media as indicated. Scale bar, 10 μm. Numbers in parentheses are the separate objects counted in each using Otsu segmentation (Image J) and automatic investigator-independent thresholding. Scale bar, 10 μm. At the end of this experiment, the cultures were fixed and immunostained for the ER marker RTN4 to confirm that the reassembled GFP-MxA condensates were independent of the ER (data not shown). Movie S4 shows time-lapse images (5 s apart) of the first cycle of disassembly and then reassembly of GFP-MxA condensates in the same cell in a different experiment. Panels B and C show illustrative examples (labeled a, b, c, and d) of disassembly and reassembly of GFP-MxA condensates taken from time-lapse imaging experiments as generated by the medium/buffer changes indicated. Numbers in parentheses are the number of objects counted in each image using the Otsu segmentation protocol. Scale bar, 10 μm. (D) Quantitation of the speed of disassembly and reassembly using Otsu segmentation metrics in time-lapse movies corresponding to examples b, c, and d in panel B. Movie S5 shows the reassembly cell (c) in panels C and D.

FIG 10

Dissecting the mechanism of disassembly and reassembly of GFP-MxA condensates. (A) A culture of Huh7 cells in a 35-mm plate expressing GFP-MxA kept continuously at 37°C was sequentially imaged in 5-min steps in the indicated media (warm PBS or warm ELB ± 0.3 M sucrose). Images 1 to 8 illustrate representative examples of cells at each stage (out of 10 to 20 images per step). (B) Huh7 cells expressing GFP-MxA condensates kept in PBS were imaged just before and just after exposure to saponin (0.03%)-containing PBS using time-lapse imaging as indicated. (C) Quantitation of the average fluorescence per cell in GFP-MxA-expressing cells without or with treatment of cultures with saponin (0.03%) in PBS for 2 to 5 min, expressed as means ± SE. n, number of cells evaluated; *, P < 0.001. (D) Huh7 cells expressing GFP-MxA condensates first were imaged in PBS (left) and then switched to saponin (0.03%)-supplemented PBS for 2 min with imaging at the same exposure setting (middle). At the conclusion of the experiment (5 min), the same cells were imaged using a 3× exposure setting. Solid white arrow indicates a cell in which saponin solubilized all of the GFP-MxA, while the dotted white arrow indicates a cell with residual saponin-resistant GFP-MxA structures. All scale bars, 10 μm.

FIG 11

Tonicity-driven phase transitions of condensates of IFN-α-induced endogenous MxA in Huh7 cells. (A and B) Huh7 cultures transfected with GFP-MxA were used to test the ability of a 10-min fixation with methanol-acetone (1:1, vol/vol) (Meth-Ac) at −20°C to preserve the disassembled state after exposure of cells to hypotonic buffer (ELB) and the reassembled state after shift to isotonic buffer (PBS). Live and fixed cells were imaged in cultures as indicated. (C) Replicate Huh7 cultures were treated with IFN-α (2,000 IU/ml) for 2 days and then fixed with methanol-acetone for 10 min directly from full medium (DMEM), after 5 min in hypotonic medium (DMEM>ELB), or after reversal back to isotonic medium (DMEM>ELB>DMEM). The fixed cultures then were immunostained for endogenous MxA using an anti-MxA PAb and the fluorescence imaged using a 100× oil objective in z-stack mode. Respective slices were deconvolved and merged for illustration. All scale bars, 10 μm.

FIG 12

Antiviral effect of MxA. VSV nucleocapsid (N) protein associated with GFP-MxA condensates. Huh7 cells (approximately 2 × 10⁵ per 35-mm plate), transfected with the pGFP-MxA expression vector 1 day earlier, were replenished with 0.25 ml serum-free Eagle’s medium, and then 10 μl of a concentrated VSV stock of the wild-type Orsay strain was added (corresponding to an MOI of >10 PFU/cell). The plates were rocked every 15 min for 1 h, followed by addition of 1 ml of full culture medium. The cultures were fixed at 4 h after the start of the VSV infection, and the extent and localization of N protein expression in individual cells was evaluated using immunofluorescence methods (using the mouse anti-N MAb) and Image J for quantitation. (A) N protein expression in GFP-MxA-negative (Neg) and -positive (Pos) cells in the same culture in arbitrary units (AU) per cell; values are means ± SE; *, P < 0.001. (B) Field of GFP-MxA-negative cells expressing high levels of N protein surrounding a GFP-MxA-positive cell with limited N protein expression. (C) Higher-magnification image of a different GFP-MxA-positive cell in the same culture as that shown in panel B expressing some N protein and the colocalization of N with GFP-MxA in spherical condensates (white arrows) as well as lack of colocalization in irregularly shaped GFP-MxA condensates (white double arrows). A total of 19 cells with overlap between GFP-MxA and VSV-N condensates, similar to the cell shown in panel C, were observed in this experiment. As a negative control, uninfected GFP-MxA-expressing Huh7 cells did not immunostain with anti-VSV-N MAb (data not shown). All scale bars, 20 μm.