The N protein phase separates with RNA

We first asked whether the N protein forms bimolecular condensates in the presence or absence of viral RNA in vitro. To address this, we expressed wild-type (WT) N protein in HEK293 cells and purified it in a high salt buffer (1 M NaCl) to eliminate the retention of RNA from human cells [38] (S1A and S1B Fig). Consistent with recent studies, purified N protein eluted from gel filtration as an oligomer in 300 mM NaCl [26] and had a high affinity for binding to viral and nonviral RNA substrates [39,40] (S1C–S1F Fig). The protein was labeled with a fluorescent dye (LD655) at the carboxyl-terminal ybbR tag and introduced to a flow chamber in the presence or absence of RNA. Phase separation was monitored by the settling of N or N-RNA condensates onto the coverslip within 25 minutes under highly inclined and laminated optical sheet (HiLO) excitation (see Methods for details). In the absence of RNA, N protein did not form condensates in physiological salt (150 mM NaCl) (Fig 1C). Similarly, 2-kb long polyC RNA homopolymer did not form any condensates in the absence of N protein (S2A and S2B Fig). Mixing of 50 ng/μL polyC RNA and the N protein resulted in the formation of condensates in 150 mM NaCl (Fig 1C). LD655-N and Cy3-polyC colocalized in the droplets, suggesting that phase separation of N protein is mediated by RNA (S2C Fig). The analysis of the condensates settled on the coverslip revealed a saturation concentration (c_sat) of 3.3 ± 1.3 μM for N protein in the presence of 50 ng/μL polyC RNA (± SE, Fig 1C), lower than the concentration of N protein in infected cells [41]. The partition coefficient of N protein into condensates was 13 ± 2 (± SD).

Although N protein is unable to phase separate in the absence of RNA in 150 mM salt, 24 μM N protein efficiently formed condensates without RNA at lower salt with a half-maximal inhibition constant (IC₅₀) of 90 ± 1 μM NaCl (± SE) (Fig 1D). The addition of 50 ng/μL polyC RNA increased IC₅₀ to 152 ± 6 mM NaCl (Fig 1D). The ability of N protein to phase separate without RNA and sensitivity of these condensates to salt indicates that these condensates may be driven, in part, by electrostatic interactions among N proteins, as observed for other phase separating systems [42–45].

Condensate formation of N protein and RNA was dependent on protein–RNA stoichiometry. At 18 μM N protein, condensate formation was not observed in the presence of 0 to 5 ng/μL polyC RNA. Increasing the RNA concentration promoted phase separation with an optimal RNA concentration of 100 to 500 ng/μL, at which charge neutralization occurs with the positively charged N protein [46]. A further increase in RNA concentration dissolved these condensates (Fig 1E) [47,48]. This reentrant behavior is characteristic of heterotypic RNA and protein interactions in phase separating systems [47].

We also showed that N-polyC condensates exhibit liquid-like, rather than solid-like, material properties. First, these condensates were nearly spherical with an aspect ratio of 1.3 ± 0.6 (mean ± SD). Second, we observed the fusion of condensates with a mean fusion time of 90 ± 30 seconds (mean ± SD) after they come into contact (Fig 1F). Finally, fluorescence recovery after photobleaching (FRAP) experiments showed that 90 ± 10% of N protein in condensates can slowly exchange with the solvent with the half recovery time of 80 ± 30 seconds (mean ± SD). In comparison, polyC RNA exhibited slower recovery and a lower mobile fraction (Fig 1G and 1H, S3 Fig), indicating that the RNA may be stabilized by a network of interactions with multiple N proteins in the condensates.

Similar to polyC RNA, polyA and polyU RNA substrates produced spherical condensates at micromolar concentrations of the N protein (Fig 2A) [49]. However, combining N protein with polyG that forms Hoogsteen base pairing [50] or polyAU that forms Watson–Crick base pair interactions [51] led to the formation of nonspherical condensates (Fig 2A). These results indicate that the N protein forms liquid condensates with some of the homopolymeric RNA substrates, whereas other homopolymeric RNA substrates that form base pairing interactions result in asymmetric condensates [19,52–54].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. The N protein forms asymmetric condensates with viral RNA.

(A) The N protein forms spherical condensates with polyA and polyU RNA but forms asymmetric condensates with substrates that form Hoogsteen (polyG) and Watson–Crick (polyAU) base pairing interactions. The N protein concentration was set to 18.5 μM. (B) (Top) SARS-CoV-2 genomic RNA was divided into six 5-kb sections. (Bottom) The formation of asymmetric N condensates in the presence of 18 nM RNA. The N protein concentration was set to 18.5 μM. (C) The inverse cumulative distribution (1-CDF) of the aspect ratio of individual N condensates formed with different RNA substrates. The concentrations of N protein, RNA homopolymers, and 0–5 kb viral RNA were set to 18.5 μM, 50 ng/μL, and 18 nM, respectively. Solid curves represent a fit to exponential decay. (Insert) Decay constants of the exponential fits (± SE). (D) (Left) Representative FRAP imaging of an N and 0–5 kb viral RNA condensate. The image of a condensate before the time of photobleaching (0 second) shows colocalization of Cy3-labeled 0–5 kb viral RNA and LD655-N in the condensate. Rectangles show the photobleached area. (Right) N and 0–5 kb viral RNA do not exhibit fluorescence recovery in the bleached region (n = 16). Data underlying this figure can be found in S1 Data. CDF, cumulative distribution function; N, nucleocapsid; SARS-CoV-2, Severe Acute Respiratory Syndrome Coronavirus 2.

https://doi.org/10.1371/journal.pbio.3001425.g002

Next, we sought to characterize how the N protein interacts with SARS-CoV-2 genomic RNA. The SARS-CoV-2 RNA genome was reverse transcribed and assembled into a DNA plasmid [55]. Using this plasmid, we generated six 5-kb fragments (Fig 2B) and two 1-kb fragments of the viral RNA genome via in vitro transcription [56]. In silico methods predict that these RNA fragments can form intra- and intermolecular base pairing interactions and contain extensive secondary structure elements (S4 Fig) [52,57,58]. The N protein formed nonspherical condensates with viral RNA fragments across a wide range of protein and viral RNA concentrations (Fig 2B and 2C, S4B–S4D Fig). These condensates were dissolved by increasing the salt concentration, but they neither changed shape over time, fused with each other, nor were strongly affected by raising the temperature from 20°C to 37°C (S5 Fig). We also did not detect fluorescence recovery of either the LD655-labeled N protein or Cy3-labeled viral RNA in FRAP assays (Fig 2D, S6 Fig). We concluded that N protein forms solid-like condensates with viral RNA in vitro and that base pairing interactions of the RNA may influence the material properties of the condensate [19,52,53].

CLMS identifies N protein interaction sites

To understand the mechanism of phase separation of the N protein, we performed CLMS to identify interactions between different domains of the full-length N protein in the absence of RNA [59]. CLMS detects protein–protein contacts by covalently capturing closely positioned lysine residues with bifunctional reagents. We first cross-linked the soluble (not phase separated) N protein in 300 mM KAc using a bifunctional cross-linker bis(sulfosuccinimidyl) suberate (BS3) (S7A Fig) [38]. We detected that the N-terminal half of the protein, including NTD, makes diverse contacts throughout the entire protein (S7A Fig). There was also an abundance of contacts between the regions immediately flanking the CTD on either side, referred to as R1 (amino acids 235 to 256) and R2 (amino acids 369 to 390).

Next, we performed quantitative CLMS measurements [60] comparing the soluble N protein in 300 mM KAc with phase-separated N protein in 100 mM KAc (Fig 3A). The soluble N protein was cross-linked with heavy (D12) BS3, whereas the phase-separated protein was cross-linked with light (H12) BS3. As a result, the cross-linked precursor ions from high and low salt conditions were spaced by 12 Da (Fig 3A). Interactions between specific regions that promote condensate formation are implicated by the ratio of the cross-linked precursor ion signal and its corresponding isotopic doublet (Fig 3A) [60]. This experiment was repeated by reversing the labels, such that (H12) BS3 was used to cross-link the soluble N protein and (D12) BS3 was used for the phase-separated N protein. Across 2 independent experiments, 29 unique cross-links were enriched, and 30 cross-links were depleted upon phase separation (Fig 3B–3C, S7B Fig). Although lysine residues are distributed throughout the N protein (S7A Fig) and depleted cross-links spanned the entire primary sequence, the analysis of the cross-link fold change (S1 Table) revealed that nearly all of the enriched interactions are concentrated in regions R1 and R2 (Fig 3C). These results suggest that the interactions involving the amino acids in regions R1 and R2 may be important in driving phase separation (Fig 3D, S7C Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. CLMS reveals interdomain interactions of the N protein.

(A) Schematic of the CLMS experiment. (Left) A high salt (300 mM KAc) buffer disrupts N condensates, whereas a low salt (100 mM KAc) buffer promotes condensate formation. (Right, top) Example of an individual cross-linked peptide in quantitative CLMS analysis. Precursor ions from the high salt (gray) and low salt (green) BS3 cross-linking conditions show the 12 Da shift between light (H12) and heavy (D12) cross-linkers. (Right, bottom) Ion chromatograms from the first 3 isotopes of each doublet were extracted and expressed as the ratio of peak areas. (B) The plot of cross-links depleted and enriched in the condensate condition. The width and transparency of the lines scale with the number of times the cross-link were detected across 3 independent experiments. (C) Fold changes of cross-link abundance upon condensate formation of N. As cross-links contain 2 positions, fold change information is plotted at both positions. Only cross-links with p-values less than 0.05 are included. Green and gray dots represent cross-links enriched and depleted in the condensate condition, respectively. The blue area represents a plot of median cross-link fold change. Data underlying this figure can be found in S1 Table. (D) Model for how multiple N dimers could phase separate via their disordered regions. (E) Phosphorylation sites detected by the CLMS experiment in 300 mM KAc. CLMS, cross-linking mass spectrometry; CTD, carboxyl-terminal domain; N, nucleocapsid; NTD, N-terminal domain.

https://doi.org/10.1371/journal.pbio.3001425.g003

Our mass spectrometry (MS) analysis also found phosphorylation sites on the N protein (Fig 3E). While some of these sites have been identified in previous studies [10,61,62], we also identified several novel sites (S2 Table). Although one of the phosphorylation sites (S176) is involved in a cross-link, the phosphorylated and unphosphorylated peptides were both strongly depleted in condensates, suggesting that S176 phosphorylation does not play a major role in phase separation (S1 and S2 Tables, S7B Fig). Additionally, MS identified native proteins that co-purified with the N protein in 1 M salt (S3 Table). Consistent with the recruitment of N to stress granules in cells [1,13,46, 63], 2 of the most frequently identified proteins were stress granule proteins G3BP1 and G3BP2, with R2 of the N protein interacting with G3BP1 (S7D Fig).

The carboxyl-terminal region and phosphorylation modulate phase separation

The quantitative CLMS experiments show that interactions between R1 and R2 are correlated with the formation of condensates. However, we could not exclude the possibility that some of the changes in pairwise interactions are due to differences in protein structure or electrostatic interactions induced by differences in salt concentrations used for the soluble and condensate phase. To directly test the predictions of the CLMS results, we determined how different domain deletions affected phase separation of the N protein with polyC RNA under the same salt concentration (Fig 4A, S8A Fig). The ΔSR, ΔPLD, and ΔR1 mutants formed spherical condensates with polyC RNA (Fig 4B, S9 Fig), suggesting that these regions are not essential for phase separation. In comparison, deletion of the R2 sequence fully abolished the formation of condensates with polyC RNA (Fig 4B, S9A–S9D Fig). Similarly, ΔR2 was unable to form condensates with viral RNA, whereas other deletion mutants phase separated with the same RNA fragments (S9E Fig). ΔR2 maintained a high affinity to bind polyC and viral RNA (S8C Fig), excluding the possibility that the disruption of phase separation is due to the lack of protein–RNA interactions. These results show that protein–protein interactions driven by the R2 motif are required for phase separation of N protein with RNA, consistent with the proposed role of this sequence in the oligomerization of N protein [26].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. The effect of domain deletions and phosphorylation on phase separation of the N protein.

(A) Deletion mutants of the N protein. (B) While ΔPLD, ΔSR, and ΔR1 phase separate, ΔR2 does not phase separate when mixed with polyC RNA. Protein concentration was set at 18 μM for all conditions. (C) Images of condensates formed by untreated, phosphorylated, and dephosphorylated N protein in 50 ng/μL polyC RNA. (D) The total volume of N-RNA condensates settled per micron squared area on the coverslip under different phosphorylation conditions (mean ± SD, n = 20 with 2 technical replicates). Linear fits (solid lines) reveal c_sat (± SE). (E) Images of condensates formed by untreated and dephosphorylated ΔR2 in 50 ng/μL polyC RNA. (F) The total volume of ΔR2-RNA condensates settled per micron squared area on the coverslip as a function of ΔR2 concentration (mean ± SD, n = 20 with 2 technical replicates). The linear fit (solid line) reveals c_sat (± SE). Data underlying this figure can be found in S1 Data. CTD, carboxyl-terminal domain; N, nucleocapsid; NTD, N-terminal domain; PLD, prion-like domain; SR, serine/arginine-rich; WT, wild-type.

https://doi.org/10.1371/journal.pbio.3001425.g004

Recent studies proposed that the N protein is phosphorylated early in infection, and this results in localization of N protein with the RTC, where it enhances transcription of subgenomic RNA [64]. However, nucleocapsid formation does not require phosphorylation and N protein in SARS-CoV viruses is hypophosphorylated [64,65]. The underlying molecular mechanism and kinases and phosphatases responsible for posttranslational modification of N protein remain poorly understood [8]. To understand how phosphorylation affects phase separation, we treated the full-length N protein with casein kinase 2 and λ protein phosphatase (see Methods). While kinase treatment did not alter the migration of N protein on a denaturing gel, phosphatase treatment resulted in a reduction in molecular weight (S8A Fig), suggesting that the N protein expressed in human cells is phosphorylated. Phosphatase treatment resulted in phase separation at slightly lower concentrations in comparison to kinase-treated N protein (Fig 4C and 4D). Similar to WT N, dephosphorylated ΔSR or ΔR1 had only a moderate increase in phase separation (S8 and S10 Figs). However, unlike untreated ΔR2, dephosphorylated ΔR2 exhibited robust phase separation with polyC RNA (Fig 4E and 4F). These results suggest that phosphorylated N is capable of phase separation with RNA, primarily due to interactions between R2 motifs. In the absence of the R2 motif, phosphorylation negatively regulates phase separation of the N protein.

Targeting phase separation of the N protein with small molecules

We next sought to identify small molecules that could interfere with the phase separation of the N protein. The condensates formed by the N protein in the absence or presence of polyC or viral RNA dissolve with the addition of 10% 1,6 hexanediol, indicating that phase separation is driven, at least partially, by hydrophobic interactions (Fig 5A) [45]. In comparison, lipoic acid that increases the liquidity of stress granules [66] did not affect phase separation (Fig 5A). Using a high-throughput microscopy platform, we also tested whether any of the 1,200 compounds in a Food and Drug Administration (FDA)-approved drug library alters phase separation of the N protein with polyC RNA in vitro. While none of the compounds fully dissolved the condensates, we identified several compounds that affected their number, size, and shape (S4 Table). Nelfinavir mesylate and LDK378 produced larger but fewer condensates, whereas crystal violet, tolcapone, and chlorhexidine enhanced phase separation by increasing the number and size of the condensates (Fig 5B, S11A and S11B Fig). Nilotinib resulted in a 50% increase in condensate volume and altered the shape of the condensates (Fig 5B), which may be due to changes in condensate fusion or maturation. While most drugs did not have a substantial effect on condensates formed by N and viral genomic RNA in vitro, nilotinib addition resulted in the formation of thread-like filaments (S11C and S11D Fig). Morphologically, these filaments were similar to those formed during the liquid-to-solid transition of FUS condensates [67], suggesting that nilotinib increases the viscosity of the condensates.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Identification of small molecules that alter phase separation of N in vitro and reduce viral titer in SARS-CoV-2 infected cells.

(A) Condensates formed in the presence of 7.8 μM N and either 50 ng/μL polyC or 18 nM 0–5 kb viral RNA were not affected by 10 mM lipoic acid but dissolved in the presence of 10% 1,6 hexanediol. (B) (Left) Examples of drugs that affected N phase separation with polyC in vitro. The N protein and polyC RNA concentrations were set to 7.8 μM and 50 ng/μL, respectively. (Right) The percent change on the number (blue) and total volume (black) of N-polyC condensates settled per micron squared area on the coverslip under different drug concentrations (mean ± SD, n = 8 with 2 technical replicates). Solid curves represent a fit to a dose–response equation (see Methods). (C) (Left) Percent CPE inhibition (black, mean ± SD, 2 technical replicates) and cell viability (blue) of SARS-CoV-2–infected Vero-E6 cells treated with serial dilutions of drugs. Solid curves represent a fit to a dose–response equation (see Methods) to determine a half-maximal response constant EC₅₀ (S4 Table). (Right) Viral titer in SARS-CoV-2–infected Vero-E6 cells treated with serial dilutions of drugs as measured by a TCID₅₀ assay (mean ± SD, 3 technical replicates). Data underlying this figure can be found in S1 Data. CPE, cytopathic effect; N, nucleocapsid; SARS-CoV-2, Severe Acute Respiratory Syndrome Coronavirus 2; TCID, tissue culture infectious dose.

https://doi.org/10.1371/journal.pbio.3001425.g005

We then tested these drugs in human pulmonary epithelial (Calu-3) and African Green Monkey kidney (Vero-E6) cells infected with SARS-CoV-2 (see Methods), as these cell lines supported high levels of infection [68]. The cells were treated with different concentrations of drugs 1 hour before infection, and the inhibition of SARS-CoV-2–mediated cell death under drug treatment was quantified using the cytopathic effect (CPE) inhibition assay [68]. The toxicity of the drugs was quantified by measuring the viability of uninfected cells under drug treatment. Among the molecules we identified, nelfinavir mesylate resulted in the highest percent CPE inhibition in both cell lines without significantly affecting cell viability (Fig 5C, S12A Fig), with the exception that the highest dose (40 μM) was toxic to the cells.

We also tested the effectiveness of these drugs using a median tissue culture infectious dose (TCID₅₀) assay. The supernatant from Vero-E6 cells that had been cultured with virus and drugs was added to uninfected cells, and the viral titer was measured by observing CPE in these cells. In this assay, nelfinavir mesylate, nilotinib, and LDK378 each reduced SARS-CoV-2 titer viability, with nilotinib and LDK378 producing an inhibitory effect similar to remdesivir, which strongly inhibits proliferation of SARS-CoV-2 in infected cells (Fig 5C, S12B Fig) [69]. Collectively, our drug screen identified compounds that affect the condensate formation of N protein and RNA in vitro, inhibit virus-mediated cell death, and reduce viral titer.

The N protein phase separates in mammalian cells

To test whether N protein also phase separates in mammalian cells, we expressed N-GFP in HEK293T cells. We observed the formation of distinct puncta in the cytoplasm of N-GFP–expressing cells, which were not observed in control cells expressing GFP only (Fig 6A). FRAP imaging of these puncta revealed a recovery signal with 60% mobile fraction and 6.3 ± 0.1 seconds recovery lifetime, suggesting that N protein is capable of forming liquid condensates in cells (Fig 6B, S13A–S13D Fig). The recovery of N-GFP was substantially slower than GFP only (Fig 6C–6E), but an order of magnitude faster than that of N-polyC condensates in vitro (Fig 1H) [8,38], suggesting that these condensates are less viscous than N-polyC condensates formed in vitro. This difference is not due to the absence or low stoichiometry of RNA in N-GFP puncta in live cells because N condensates formed in the absence of RNA in vitro also exhibited an order of magnitude slower recovery than N-GFP puncta in live cells (S14 Fig). We next tested whether deletion of the R2 region disrupts phase separation of N protein in cells. Surprisingly, cells expressing the R2 deletion mutant (ΔR2-GFP) still formed puncta and did not display significantly different FRAP recovery times than N-GFP in live cells (Fig 6A–6E). This may be due to macromolecular crowding of the cellular environment because the addition of 10% polyethylene glycol (PEG) triggered phase separation of ΔR2 in vitro (S15 Fig).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Dynamics of N condensates in vivo.

(A) Example images of HEK293T cells expressing N-GFP, ΔR2-GFP, and GFP stained with DRAQ5. (B) Representative FRAP imaging of cells exhibiting N-GFP or ΔR2-GFP puncta or high GFP expression. Circles show the photobleached area. (C) Fluorescence recovery of the GFP signal in the bleached versus the control regions. The solid curve represents a single exponential fit to reveal the recovery lifetime (τ, ±95% confidence interval). (D) The distribution of fluorescence recovery lifetimes of cells expressing N-GFP (n = 28) or ΔR2-GFP (n = 15) exhibiting puncta and GFP only (n = 15). The center and edges of the box represent the median with the first and third quartiles. The p-values were calculated from a 2-tailed t test. (E) The maximum fractional recovery after photobleaching of cells expressing N-GFP (n = 28), ΔR2-GFP (n = 15), or GFP only (n = 15). The center and edges of the box represent the median with the first and third quartiles. The p-values were calculated from a 2-tailed t test. (F) U2OS cells stably expressing N-Clover form condensates in the cytoplasm. (Left) Cells co-transfected with an N-MS2 expression plasmid exhibit colocalization of N-Clover and Cy3-MS2 FISH signal in N-condensates (N = 10 out of 55 cells, 2 technical replicates). (Right) The Cy3-MS2 FISH probe does not partition into N condensates in untransfected cells (N = 55 cells, 2 technical replicates). Inset scale bar is 2 μm. (G) Model for remodeling of viral RNA genome by the SARS-CoV-2 N protein. The N protein packages viral genomic RNA through phase separation, which may facilitate efficient replication of the genomic RNA and the formation of the enveloped virus. Data underlying this figure can be found in S1 Data. FRAP, fluorescence recovery after photobleaching; N, nucleocapsid; SARS-CoV-2, Severe Acute Respiratory Syndrome Coronavirus 2.

https://doi.org/10.1371/journal.pbio.3001425.g006

Additionally, we sought to determine whether the N RNA transcript was recruited to the N condensate. We transfected U2OS cells stably expressing N-Clover [38] with a construct producing N-MS2 RNA (see Methods) and performed fluorescence in situ hybridization (FISH) assays against the MS2 RNA sequence. In the absence of N-MS2 transfection, the MS2 probe did not localize to the N-Clover puncta in any of the cells (N = 55 cells, 2 independent experiments) (Fig 6F). In comparison, the MS2 probe was strongly localized to all of the N-Clover condensates in approximately 18% of the cells transfected with N-MS2 (N = 10 in 55 cells, 2 independent experiments, Fig 6F). In other cells, we observed that the MS2 probe was uniformly distributed inside and outside of the N puncta, suggesting that these cells were not expressing N-MS2 at high enough levels to recruit the probe. We also performed RNA FISH using a d(T)20 probe that binds the polyA tails of RNA transcripts. Consistent with a previous report [38], the d(T)20 probe was not recruited to N-Clover condensates (S13E Fig). These results suggest that N-Clover condensates selectively recruit viral RNA. Differences between in vitro and in vivo phase separation properties of N protein might be attributed to macromolecular crowding of the cytosol or the interaction of the N protein with other cellular proteins, such as G3BP1 (S7D Fig) [1,46,63,70].

Protein purification

A construct for expressing N protein with a carboxyl-terminal Strep-tag was obtained from the Krogan lab (UCSF). A carboxyl-terminal yBBr labeling site was added for labeling with a fluorescent dye. Deletion mutants, ΔPLD (amino acids 2 to 30: SDNGPQNQRNAPRITFGGPSDSTGSNQNG), ΔSR (amino acids 183 to 195, SSRSSSRSRNSSR), ΔR1 (amino acids 235 to 256, SGKGQQQQGQTVTKKSAAEASK), and ΔR2 (amino acids 369 to 390, KKDKKKKADETQALPQRQKKQQ) were generated using Gibson cloning (New England Biosciences, Massachusetts, USA). Sufficient amounts of DNA were obtained by growing 1 L of transfected XL1 Blue Escherichia coli cells overnight and performing a gigaprep (Zymo, California, USA). For protein expression, HEK293S GNTI- cells (RRID: CVCL_A785) were grown in suspension in Freestyle media (Thermo Fisher Scientific, Massachusetts, USA) supplemented with 2% fetal bovine serum (FBS, VWR, Pennsylvania, USA) and 1% penicillin-streptomycin (PS, Gemini Bio Products, California, USA) to 2 million cells/mL. Cells were spun down for 10 minutes at 1,200 g and resuspended in fresh, antibiotic-free media. For 250-mL cells, the transfection solution was created by mixing 1.8-mL polyethylenimine (1 mg/mL, pH 7.0 in PBS) dissolved in 20-mL Freestyle media and 0.66-mg DNA dissolved in 20-mL Freestyle media. The mixture was incubated at room temperature for 15 minutes before being adding it to the cell culture. Transfected cell culture was grown for 72 hours at 125 rpm at 37°C with 5% CO₂ and 5% humidity.

Cells were then harvested at 4,000 g for 10 minutes and resuspended in 50 mL lysis buffer (50 mM HEPES pH 7.4, 1 M NaCl, 1 mM PMSF, 1 mM DTT, and 1 tablet of protease inhibitor (Sigma-Aldrich, Missouri, USA)). Lysis was performed using 15 loose and 15 tight plunges of a Wheaton glass dounce. The lysate was clarified using a 45 minutes, 360,000 g spin in a Ti70 rotor. The supernatant was incubated with 1-mL Streptactin sepharose beads (IBA Life Sciences, Goettingen, Germany) for 1 hour. Beads were washed with 40 mL of lysis buffer followed by 30 mL labeling buffer (50 mM HEPES pH 7.4, 300 mM NaCl, 10 mM MgCl₂, 1 mM EGTA, 10% glycerol, 1 mM DTT). Beads were then collected and incubated with purified Sfp phosphopantetheinyl transferase and an LD655 dye functionalized with CoA (Lumidyne, New York, USA) at room temperature for 30 minutes. Beads were washed with 30-mL labeling buffer. If proteins were to be kinase or phosphatase treated, they were additionally washed with 30 mL kinase (20 mM HEPES pH 7.5, 300mM NaCl, 10 mM MgCl2, 200 μM ATP, 10% glycerol, 1 mM DTT) or phosphatase (20 mM HEPES pH 7.5, 300mM NaCl, 1mM MnSO4, 10% glycerol, 1 mM DTT) buffer. Protein was eluted in 1 mL fractions in its final buffer supplemented with 10 mM desthiobiotin and concentrated using Amicon Ultra 30K concentrators. For kinase and phosphatase treatment, 5 μL of caseine kinase 2 (New England Biolabs, Massachusetts, USA) or 2.5 μL λ phosphatase (New England Biolabs, Massachusetts, USA) was added per 50 μL concentrated protein, respectively, and the samples were incubated at 30°C for 1 hour. Final protein concentration was measured using Bradford reagent, and aliquots were snap frozen in liquid nitrogen.

In vitro transcription and RNA labeling

For in vitro transcription of long viral RNA, the region of interest was first PCR amplified from a plasmid (N plasmid was a generous gift from the Krogan lab [1], 5′ UTR plasmid was a generous gift from the Gladfelter lab [56], and SARS-CoV-2 cDNA plasmid was a generous gift of the Thiel lab [55]) using a forward primer with a T7 polymerase binding site (TAATACGACTCACTATAGGG). The amplified DNA was tested for purity on a 0.8% agarose gel and cleaned up using GlycoBlue and ethanol precipitation. RNA was generated using the HiScribe T7 Quick kit (New England Biolabs, Massachusetts, USA) and extracted using trizol and isopropanol precipitation. RNA was Cy3 labeled using a Label IT kit (Mirus Bio, Wisconsin, USA), and RNA purity was verified using a 0.8% agarose gel. All RNA structure predictions were done on the RNAfold server [57,58].

Sample preparation and microscopy

Purified protein and RNA samples were diluted into the imaging buffer (50 mM HEPES pH 7.4, 150 mM NaCl, 5 mM MgCl₂, 1 mM EGTA, 1 mM DTT, 1% pluronic) to their final concentration and introduced into the flow chamber. Samples were settled onto the coverslip for 25 minutes before imaging. We confirmed that all of the condensates were settled to the surface within 25 minutes, as we did not observe an increase in the number of condensates per viewing area of the coverslip, and we could not detect freely diffusing condensates in the flow chamber after 25 minutes. Different buffers were used for phosphatase-treated (50 mM HEPES pH 7.4, 150 mM NaCl, 0.5 mM MnSO4, 1 mM DTT, 1% pluronic) and kinase-treated (20 mM HEPES pH 7.5, 150 mM NaCl, 5 mM MgCl₂, 100 μM ATP, 1 mM DTT, 1% pluronic) samples.

Imaging was performed using a custom-built fiber-coupled Nikon Ti-E Eclipse microscope equipped with an objective-type total internal reflection fluorescence (TIRF) illuminator and 100X 1.49 N.A. Plan Apo oil immersion objective (Nikon, New York, USA). The samples were excited in near-TIRF using 561 and 633 nm laser beams (Coherent, California, USA). The fluorescent signal was detected by Andor Ixon electron-multiplying CCD (EMCCD) camera (512 × 512 pixels). The effective pixel size was 160 nm after magnification. 10 single-frame images were acquired for each condition in each replicate.

Image analysis

To calculate the saturation concentration (c_sat), the area of each condensate was quantified with Fiji using the Phansalkar function with a 30-pixel radius and a minimum condensate size of 10 pixels. The volume of the condensates was estimated from 2D projections by taking the semi-principal axis in the z-plane as the geometric average of semi-principal axes in the xy plane. The total volume of the condensates settled per micron squared on the coverslip was quantified. Conditions that resulted in measurable condensate volumes were fit to linear regression in Origin. The x-intercept of the linear regression represents c_sat, the minimum protein concentration that results in condensate formation. The aspect ratio was calculated for individual condensates in Fiji using the Phansalkar function with a 30-pixel radius and a minimum condensate size of 10 pixels. To measure fusion times, condensates were visualized as they settled on the imaging surface from 5 to 25 minutes after mixing. Movies were recorded at 5 frames per second. Fusion times were calculated as the time between the last frame where 2 condensates appear separated (i.e., no overlap) and the first frame where the fused condensate appears nearly spherical (aspect ratio = 1.1).

In vitro FRAP

In vitro FRAP assays were performed with a Zeiss 880 Confocal Laser Scanning Microscope equipped with a 100×, 1.4 NA oil immersion objective. Samples were prepared as described above. All in vitro FRAP experiments used the 488 nm laser at 100% for bleaching conditions. For N-polyC at 150 mM NaCl, 5 bleaching iterations were used, and images were taken every 2 seconds for 232 seconds. For viral RNA at 150 mM NaCl, 100 bleaching iterations were used, and images were taken every 10 seconds for 8 minutes. For the experiments conducted at 50 mM NaCl, droplets containing polyC RNA or viral RNA were bleached with 20 iterations, and images were taken every 5 seconds for 5 minutes. For droplets containing N protein only, 20 bleaching iterations were used, and images were collected every 2 seconds for 6.5 minutes. All data were acquired using Zeiss Zen 2.3 SP1 FP3 (black) (64bit) Version 14.0.20.201. Data were background corrected, converted to normalized intensity, and fit to an exponential decay function in Origin.

MS

The cross-linking analysis was performed as described by McGilvray and colleagues [90] with deviations outlined here. 15 μL 90 μM N protein in the labeling buffer was diluted 1:3 v/v into either water or labeling buffer. The diluted protein formed condensates in 100 mM NaCl in the final buffer, but not in the labeling buffer that contains 300 mM NaCl. Isotopically coded light (H12) and heavy (D12) BS3 (bis(sulfosuccinimidyl)suberate) cross-linkers (Creative Molecules, California, USA) were immediately added to the diluted solution. Final concentration of BS3 was 0.8 mM for 100 mM NaCl dilution and 2.5 mM for 300 mM NaCl dilution. Cross-linking was performed for 30 minutes at room temperature and then quenched with 1 M Tris pH 8.0 buffer. This experiment was performed in duplicate, once with D12-BS3 cross-linking N protein in condensates and H12-BS3 cross-linking soluble N protein and once with the labels reversed. The cross-linked proteins from both channels were pooled before acetone precipitation. Protein was precipitated with acetone overnight at −20°C. Protein was pelleted, the supernatant was removed, and the samples were air-dried for 10 minutes. The pellets were brought up in 8 M urea, reduced with TCEP, alkylated with iodoacetamide, diluted 4-fold, and digested with 2 rounds of trypsin. Peptides were desalted on a C18 MacroTrap column (Michrom Bioresources, California, USA) and fractionated on a Superdex Peptide (GE Life Sciences, Illinois, USA) size exclusion column. Fractions enriched in cross-linked peptides were dried and resuspended in 0.1% formic acid for MS. Later eluting fractions were used for phosphorylation analysis. Each of the replicates was fractionated by size exclusion chromatography (SEC), and each fraction was injected twice, generating 16 MS files for analysis.

MS was acquired on an Orbitrap Fusion Lumos coupled with an Easy-Spray nanoelectrospray ion source, a 15 cm × 75 μm PepMap C₁₈ column (Thermo Fisher Scientific, Massachusetts, USA) and an M-class NanoAcuity UPLC system (Waters, Massachusetts, USA). Liquid chromatography–mass spectrometry (LC–MS) runs were 90 minutes long. Precursor ions were measured in the Orbitrap at 120 k resolution. Selected precursor ions (triply charged and higher) were isolated, and the product ions were measured in the Orbitrap at 30 k resolution. Samples that were analyzed for phosphorylation were run similarly except only HCD product ions spectra were collected and doubly charge precursors were included.

Cross-linked spectra were identified with Protein Prospector 6.2.23 [91] using the combination of DSS/DSS:2H12 at uncleaved Lys residues and protein N-terminus as the cross-linking reagents. The corresponding light and heavy dead-end modifications, incorrect monoisotopic peak assignment (+1Da neutral loss), N-terminal pyroglutamate formation, methionine oxidation, and acetylation and loss of the protein N-terminal methionine were set as variable modifications with 3 variable modifications per peptide allowed. Trypsin with 2 missed cleavages was the digestion enzyme, and the mass tolerances were 20 and 30 ppm for precursor and product ions. The search database comprised the 14 most abundant proteins (sorted by spectral abundance factor, SAF) found in the linear peptide SEC fractions, alongside a decoy database that was 10 times longer. Cross-link spectral matches were classified at a 1% false discovery rate (FDR) threshold.

Protein Prospector was used to search for phosphopeptides of the linear peptide fractions using the following parameters: tryptic specificity with 2 missed cleavages, 7 and 15 ppm precursor and product ion tolerances, and carbamidomethyl (C) fixed modification. Variable modifications were as above except that Phospho (STY) was included, and only dead-end cross-link modifications were included. Peptides were reported with a maximum expectation value of 0.001, and a minimum prospector score of 15. A site localization in peptide (SLIP) score threshold of 6 was used to determine site localization. All phosphopeptide spectra reported were manually inspected for evidence of correct site assignment. Where S or T residues are adjacent, the location of the phosphosite is ambiguous (see S2 Table). All phosphorylation data can be accessed here (or on the MS-Viewer website with search key “jyovxenjny”).

A supplemental search of the cross-linked peak lists was made for cross-links that also contained a phosphorylation site. This search was identical to the previous cross-linking searches except that Phospho (ST) was included as a variable modification, and the search was limited to the N protein sequence. Search results containing a phosphorylation site were manually assessed. Only phosphorylation sites that had been discovered in the phospho search of the linear peptides were considered.

For quantitating the isotopically labeled cross-links, peak areas were measured from the extracted precursor ion chromatograms (XICs) using the small molecule interface of Skyline (v20.1.0.155). A Skyline transition list was generated containing the elemental composition of each distinct peptide pair with both light and heavy BS3 modification and in each charge state detected in the Prospector search. Retention times were present in the transition list to help with peak detection. Peptide level measurements were imported into R and summarized at the level of unique residue pairs (“cross-links”). For each precursor ion, the log₂ ratios of heavy to light peptides were calculated and mean normalized to 0 for each of the 2 biological replicates and then transformed into log₂ (water/salt) ratios. Weighted t tests were performed in R.

Drug screening and image processing

For the FDA-approved drug screen, 75 mL of N protein at 16 μM was purified from HEK293S GNTI- cells. polyC RNA was obtained from Sigma. The FDA-approved drug library (TargetMol) has 1,200 compounds of well-characterized biological activity. A total of 10 mM compounds were stored in 100% dimethyl sulfoxide (DMSO) in 384-well plates. For screening plates, 20 μL imaging buffer (50 mM HEPES pH 7.4, 150 mM NaCl, 5 mM MgCl₂, 1 mM EGTA, 1 mM DTT, 1% pluronic) was aliquoted into 384-well, glass-bottomed plates (Greiner Bio-One, Kremsmunster, Austria) and incubated for 5 minutes. Buffer was removed, and 0.5 μL of 2 mM compounds were stamped into 384-well plates with an Analytik-Jena Cybio Well Vario liquid handler. The final concentration of 7.8 μM N protein and 50 ng/μL polyC RNA were added to each plate. Each well contained a 25 μL mixture and 40 μM compound in the primary screen.

Wells were homogenized with a Bioshake 3000 ELM orbital shaker at 2,400 rpm for 45 seconds, and condensates were allowed to settle for 1 hour before imaging. Moreover, 4 images (each 224 × 167 μm) were taken at 40× with ImageXpress Micro High Content Imaging System (Molecular Devices, California, USA). Images were analyzed with Metamorph Imaging software calculating condensate count and area. The volume of the condensates was calculated as described above. Data were then uploaded to CCD Vault for normalization to DMSO vehicle control wells. Wells that exhibit 3 SDs from the untreated sample and as well as others that produced qualitative morphological changes in N condensates were selected for dose–response screening. The same N protein and RNA were tested against candidate drugs using a 10-point serial dilution starting at 40 μM using the same procedure. Dose–response curves were generated by fitting to the equation where C is the concentration of the drug, n is the Hill coefficient, EC₅₀ is the half-maximal response concentration, and y is either the volume of the N condensates or the number of N condensates settled onto per micron squared area on the coverslip.

Dose–response with SARS-CoV-2–infected cells

Vero-E6 (ATCC, CRL-1586) and Calu-3 cells (ATCC HTB-55) were cultured in high glucose DMEM (Gibco) supplemented with 10% FBS (R&D Systems), 1X GlutaMAX (Gibco), and 1X PenStrep (Gibco) at 37°C and 5% CO₂. For screening selected drugs against infected cells, 2,500 Vero-E6 (12 μL/well) or 10,000 Calu-3 (12 μL/well) were seeded in 384-well white optical bottom tissue culture plates (Nunc) with the Multidrop Combi liquid handling instrument (Thermo Fisher Scientific). Cells were allowed to recover for 24 hours for Vero-E6 and 48 hours for Calu-3 at 37°C and 5% CO₂. Dose responses were generated by diluting the compounds using a Cybio Well Vario liquid handler (Analytik Jena, Jena, Germany), leading to a final concentration of DMSO at 0.4% in the assay plate (v/v). Cells were incubated at 37°C and 5% CO₂ for 1 hour before infection. The viral inoculum was prepared such that the final multiplicity of infection (MOI) was 0.05 upon addition of 6 μL/well viral inoculum. After complete CPE was observed in DMSO-treated, infected wells (72 hours postinfection (hpi) for Vero-E6 and 96 hpi for Calu-3), the plates were developed with the CellTiter-Glo 2.0 reagent (Promega, Wisconsin, USA) according to the manufacturer’s instructions. For Vero-E6, the reagent was diluted 1:1 (v/v) in PBS. Luminescence of developed plates was read on a Spectramax L (Molecular Devices). Each plate contained 24 wells uninfected/DMSO treated cells (100% CPE inhibition) and 24 wells infected/DMSO treated cells (0% CPE inhibition). Average values from those wells were used to normalize data and determine % CPE inhibition for each compound well. To determine the cytotoxicity of the compounds, the same protocol was used but with 6-μL growth media added instead of viral inoculum. The data were plotted and analyzed in Origin and fit to the dose–response function above.

Dose–response determination by TCID₅₀ assay

Vero-E6 cells were plated at 10,000 cells per well in Vero growth medium as above in 96-well plates and allowed to recover for 24 hours. Drug treatments were prepared in the same medium at 1.33× of final concentration and 0.53% v/v DMSO. Growth medium was removed from cells and replaced with 150-μL medium with drug treatment and incubated for 1 hour. Virus in 50-μL medium was added to each well to produce a final MOI of 0.05, final drug concentrations as indicated, and 0.4% v/v DMSO per well. Vehicle-only treatment was used as a positive control for productive viral infection and CPE, and remdesivir at 50 μM was used as a positive control for viral restriction by drug treatment. Each condition was plated in triplicate. Virus titer following drug treatment was determined by TCID₅₀ assay in Vero-E6 cells. At 24 hours post infection, 100 μL of supernatant from drug-treated infected cells was removed from each well and used to prepare a 10-fold dilution series in Vero growth medium. Moreover, 50 μL of the dilution series was added to Vero cells in 96-well plates, prepared as above, in 100 μL medium. Cells were observed for CPE for 3 days. TCID₅₀/mL results were calculated using the Spearman and Kärber method [92].

HEK293T cell culture and transfection

HEK293T (RRID: CVCL_0063) cells were cultured in phenol-negative DMEM media supplemented with 10% FBS and 1% PS at 37°C with 5% CO₂. Prior to imaging, cells were transferred to glass-bottomed plates (Nunc Lab-Tek, 0.4 mL working volume) at approximately 25% confluence and allowed to recover for 24 hours. Media was exchanged into phenol-negative DMEM media supplemented with 10% FBS, and cells were transfected with N-GFP, NΔR2-GFP, or GFP only constructs. For each well, 400 ng of DNA was added to 20 μL DMEM media. A total of 1.2 μL FuGENE HD transfection reagent (Promega) was added, and the mixture was incubated at room temperature for 15 minutes before being added to the cell culture. Cells were allowed to express the constructs for 4 days. One hour prior to imaging, DRAQ5 stain at 1:1,000 was added to the media.

In vivo imaging and FRAP assays

In vivo imaging and FRAP measurements were performed with a Zeiss 880 Confocal Laser Scanning Microscope equipped with a 63×, 1.4 NA oil immersion objective. Samples were prepared as described above. All in vivo FRAP experiments used the 488 nm laser at 100% for bleaching conditions. Cells expressing N-GFP, ΔR2-GFP, or GFP only were bleached with 30 iterations, and images were collected continuously at 6.7 Hz for 30 seconds. All data were acquired using Zeiss Zen 2.3 SP1 FP3. Data were converted to the normalized intensity and fit to an exponential decay function in Origin.

U2OS cell culture, transfection, and RNA FISH

U2OS cells stably expressing SARS-CoV-2 N protein with a carboxyl-terminal Clover tag were a gift from the Cleveland lab at the University of California San Diego (UCSD). Cells were cultured in phenol-negative DMEM media supplemented with 10% FBS and 1% PS at 37°C with 5% CO₂. Prior to imaging, cells were transferred to glass-bottomed plates (Nunc Lab-Tek, 0.4 mL working volume) at approximately 25% confluence. Cells were transfected with a construct expressing N-MS2 RNA (the N cDNA sequence with 6 upstream stop codons and 3 carboxyl-terminal MS2 repeats) as described above. Cells were induced with doxycycline at a final concentration of 1,000 ng/mL and allowed to express for 14 hours. FISH was performed following the Stellaris RNA FISH protocol for adherent cells. Cells were fixed with 3.7% formaldehyde, permeabilized with 70% ethanol, and incubated in the dark at 37°C overnight with hybridization buffer (Biosearch Technologies, Middlesex, UK) supplemented with 10% deionized formamide and either 125 nM Cy3-(d)T20 oligonucleotides (Gene Link, Florida, USA) or 1 μM Cy3-MS2 probe (IDT (Iowa, USA), sequence AGGCAATTAGGTACCTTAGG) [19]. Cells were washed with wash buffer A (Biosearch Technologies, Middlesex, UK) for 30 minutes, then incubated with DRAQ5 1:1,000 in wash buffer A for 30 minutes and exchanged into wash buffer B (Biosearch Technologies) before imaging.

Disclaimers

The content is solely the responsibility of the authors and does not represent the official views of funding institutions.