Targeting the coronavirus nucleocapsid protein through GSK-3 inhibition

Abstract

The coronaviruses responsible for severe acute respiratory syndrome (SARS-CoV), COVID-19 (SARS-CoV-2), Middle East respiratory syndrome-CoV, and other coronavirus infections express a nucleocapsid protein (N) that is essential for viral replication, transcription, and virion assembly. Phosphorylation of N from SARS-CoV by glycogen synthase kinase 3 (GSK-3) is required for its function and inhibition of GSK-3 with lithium impairs N phosphorylation, viral transcription, and replication. Here we report that the SARS-CoV-2 N protein contains GSK-3 consensus sequences and that this motif is conserved in diverse coronaviruses, raising the possibility that SARS-CoV-2 may be sensitive to GSK-3 inhibitors, including lithium. We conducted a retrospective analysis of lithium use in patients from three major health systems who were PCR-tested for SARS-CoV-2. We found that patients taking lithium have a significantly reduced risk of COVID-19 (odds ratio = 0.51 [0.35-0.74], P = 0.005). We also show that the SARS-CoV-2 N protein is phosphorylated by GSK-3. Knockout of GSK3A and GSK3B demonstrates that GSK-3 is essential for N phosphorylation. Alternative GSK-3 inhibitors block N phosphorylation and impair replication in SARS-CoV-2 infected lung epithelial cells in a cell-type-dependent manner. Targeting GSK-3 may therefore provide an approach to treat COVID-19 and future coronavirus outbreaks.

Keywords: COVID19; GSK-3; coronavirus; lithium; nucleocapsid.

Fig. 1.

The SARS-CoV-2 N protein is phosphorylated by GSK-3 in two conserved consensus sites. (A) The RS domains of SARS-CoV-2 (amino acids 176 to 206) and SARS-CoV N proteins are 90% identical and contain tandem sets of SXXXS motifs, labeled “a” and “b.” Consensus site serines and threonines are in bold; red indicates sites shown previously by mass spectroscopy to be phosphorylated (1, 8). (B) Alignment of RS domains in N proteins from pathogenic CoVs showing conservation of repeated SXXXS motifs (“S” in motif represents serine or threonine) and a highly conserved arginine 3 residues before the putative priming sites (SRXXS). Blue indicates β-CoV; orange: α-CoV; green: γ-CoV. (C) SARS-CoV-2 N was expressed in 293T cells. Cells were treated 24 h after transfection with LiCl for an additional 18 h and then harvested and subjected to SDS/PAGE and immunoblotting for N protein (CoV2-N), phosphorylated glycogen synthase (pGS), or β-actin as a loading control. Alkaline phosphatase (AP) treatment of cell lysates increases electrophoretic mobility. LiCl inhibits N phosphorylation with IC₅₀ ∼10 mM. Ctl, nontreated control; NT, nontransfected; “phos” indicates phosphorylated N protein; “unphos” indicates dephosphorylated N. (D) SARS-CoV-2 N expressing 293T cells were treated at 24 h posttransfection with BIM-I or CHIR99021 (CHIR) at the indicated concentrations. Cell lysates were harvested after 18 h and immunoblotted as in C.

Fig. 2.

GSK3 is required for N phosphorylation. (A) Control 293T cells (WT), 293T cells with CRISPR/Cas9 KO of GSK3A (GSK3A^−/−), and GSK3A^−/− cells with siRNA knockdown of GSK3B (GSK3A^−/−-;siGSK3B) were treated with LiCl at indicated concentrations and lysates were immunoblotted for N protein, phospho-β-catenin, GSK-3α/β, or β-actin. Combined loss of GSKA and GSK3B impairs phosphorylation of N and β-catenin and enhances sensitivity to LiCl. (B) GSK3B was deleted in GSK3A^−/− cells using CRISPR (GSK3 DKO). N protein was expressed in both wild-type and DKO cells in the presence of increasing concentrations LiC for 18 h as above and immunoblotted for N protein, phospho-GS (pGS), total β-catenin, GSK-3α/β, and β-actin. N is not phosphorylated in DKO cells and mobility is not affected by LiCl treatment. Total β-catenin protein accumulates in absence of GSK-3 (DKO) (34) or upon inhibition with LiCl (35). (C) Serine-188 and serine-206 were mutated to alanine by site directed mutagenesis and single- and double-mutant N proteins were expressed in 293T cells in the presence of vehicle or 10 mM LiCl and immunoblotted for N protein, pGS, or β-actin. The double-mutant N^S188A;S206A migrates similar to dephosphorylated wild-type N. Single mutants are more sensitive to LiCl. (D and E) N protein was immunoprecipitated from wild-type HEK293T cells treated with or without 10 mM LiCl for 18 h (indicated by “Ctl” or “LiCl” below each lane in D) or from GSK3 DKO cells (E). Immunoprecipitated N protein was added to an in vitro kinase reaction with recombinant GSK-3β. GSK-3β phosphorylates N from LiCl treated wild-type and DKO cells as indicated by slower electrophoretic mobility (“phos” in E). (F) N protein immunoprecipitated from DKO cells was added to an in vitro kinase reaction with recombinant GSK-3β as in E, except that γ-[³²P]ATP was included and gels were fixed, dried, and exposed to X-ray film. WB, Western blot.

Fig. 3.

Enzastaurin inhibits N phosphorylation: (A) N expressing 293T cells were treated with DMSO or increasing doses of Enzastaurin. Enzastaurin inhibited phosphorylation of N and GS in a dose-dependent manner. Inhibition of PKC in these samples is described in SI Appendix, Fig. S3. (B) In vitro GSK-3 kinase assay using tau protein as substrate. Unphosphorylated Tau migrates more rapidly (“unphos”) than Tau phosphorylated by GSK-3 (“phos”). Enzastaurin inhibits GSK-3 activity directly at 0.5 µM. (C) N protein was immunoprecipitated from HEK293T cells treated with 20 mM LiCl as in Fig. 2D and added to an in vitro kinase reaction with recombinant GSK-3β. Phosphorylation of N protein was inhibited in the presence of Enzastaurin (10 µM) and CHIR99021 (2 µM).

Fig. 4.

GSK-3 inhibitor blocks replication in SARS-CoV-2 infected cells: (A) Dose–response analysis of Calu-3 cells treated with GSK-3 inhibitors CHIR99021 or Enzastaurin (University of Pennsylvania). Cells were treated with drug at the indicated concentrations and then inoculated with SARS-CoV-2. Cells were fixed at 48 hpi and total cell count (green) and percent viral infection (blue) detected by immunofluorescence for dsRNA were assessed. (B) Calu-3 cells were treated with vehicle or the indicated concentrations of CHIR99021, inoculated with SARS-CoV-2, fixed at 48 hpi, and Spike protein was detected by immunofluorescence (University of California, Los Angeles). Enzastaurin had no effect on viral infection in Calu-3 cells. Magnification, 100×. (C) Calu-3 cells were treated with vehicle or CHIR99021 (10 µM), inoculated with SARS-CoV-2 at t = 0, and supernatants were sampled at 48 hpi for TCID50 quantification. Median titers are indicated within the boxes (note log₁₀ scale). P = 0.05 (one tailed Mann–Whitney U test). (D) Calu-3 cells were treated with vehicle or CHIR99021, inoculated with SARS-CoV-2 at t = 0, and cell lysates were harvested for immunoblotting for N protein (Upper) or tubulin (Lower) at the indicated times after infection. Abundance of phosphorylated N (slower mobility form) relative to total N protein was assessed densitometrically. Values normalized to 2-h control sample are shown below each lane. A longer exposure of this image is provided in SI Appendix, Fig. S4.

Fig. 5.

Overview of EHR analysis workflow. The flow diagram depicts the steps for sample selection, quality control, and statistical approach to study association between lithium and COVID-19 susceptibility using EHR.