Host protein kinases required for SARS-CoV-2 nucleocapsid phosphorylation and viral replication
Kinase substrate specificity profiling assigns the phosphorylation cluster in the SARS-CoV-2 nucleocapsid SR-rich domain to a cascade initiated by SRPK1 and SRPK2 and extended by GSK-3 and casein kinase 1, whose inhibition suppresses coronavirus replication.
- Senior authors
- Benjamin R. tenOever; Lewis C. Cantley; John Blenis; Nicholas S. Heaton
- Correspondence
- Nicholas S. Heaton; John Blenis; Lewis C. Cantley; Benjamin R. tenOever
Research areas & themes
Citation
Yaron TM, Heaton BE, Levy TM, Johnson JL, Jordan TX, Cohen BM, Kerelsky A, Lin TY, Liberatore KM, Bulaon DK, Van Nest SJ, Koundouros N, Kastenhuber ER, Mercadante MN, Shobana-Ganesh K, He L, Schwartz RE, Chen S, Weinstein H, Elemento O, Piskounova E, Nilsson-Payant BE, Lee G, Trimarco JD, Burke KN, Hamele CE, Chaparian RR, Harding AT, Tata A, Zhu X, Tata PR, Smith CM, Possemato AP, Tkachev SL, Hornbeck PV, Beausoleil SA, Anand SK, Aguet F, Getz G, Davidson AD, Heesom K, Kavanagh-Williamson M, Matthews DA, tenOever BR, Cantley LC, Blenis J, Heaton NS. Host protein kinases required for SARS-CoV-2 nucleocapsid phosphorylation and viral replication. Science Signaling. 2022. 15(757), eabm0808.
DOI 10.1126/scisignal.abm0808. PMID 36282911. PMCID PMC9830954.
Tomer M. Yaron, Brook E. Heaton, Tyler M. Levy, Jared L. Johnson and Tristan X. Jordan are marked as having contributed equally. Correspondence is addressed to Nicholas S. Heaton, John Blenis, Lewis C. Cantley and Benjamin R. tenOever.
One-sentence contribution
Kinase substrate specificity profiling assigns the phosphorylation cluster in the SARS-CoV-2 nucleocapsid SR-rich domain to a cascade initiated by SRPK1 and SRPK2 and extended by GSK-3 and casein kinase 1, whose inhibition suppresses coronavirus replication.
Executive summary
Host-directed antivirals are attractive because they can act across related viruses and are harder for a virus to escape by mutation, but they require knowing which host enzymes the virus actually depends on. This work targets the SARS-CoV-2 nucleocapsid protein, whose serine and arginine rich domain is heavily phosphorylated and unusually conserved across coronaviruses. Phosphoproteomics of infected ACE2-expressing A549 and Vero E6 cells found the nucleocapsid to be the most phosphorylated viral protein, with most sites in that domain and most shared with five previously published datasets. Conservation analysis across 82 coronaviruses showed the domain, its phosphoacceptor residues, and the motif residues around the proposed priming sites all constrained. Rather than screening kinases, the authors used measured substrate specificity motifs for the SRPK, GSK-3 and CK1 families to compute which kinase most plausibly acts at each site, and that analysis produced an ordered model in which SRPKs phosphorylate serine 206 and serine 188 first, GSK-3 walks toward the amino terminus in four-residue steps, and CK1 handles threonine 205 once serine 202 is phosphorylated. In vitro reactions with recombinant protein supported the order, since combining all three kinases gave more phosphate incorporation than the sum of individual reactions and mutating both priming serines abolished the effect. Knockdown of SRPK1 and treatment with SPHINX31, SRPIN340 or alectinib reduced viral RNA and infectious titer across several cell systems, and alectinib selectively reduced SR-rich domain phosphorylation.
Scientific context
Vaccines against SARS-CoV-2 had been deployed rapidly and effectively by the time of this work, but breakthrough infections and unvaccinated patients left a need for antivirals, and viral evolution places a limit on approaches directed at viral proteins. Targeting host factors offers breadth and a higher barrier to escape, and protein kinases are attractive because they are well characterized and widely druggable. On the viral side, the importance of nucleocapsid phosphorylation was already suggested by several lines of work. Studies of SARS-CoV had implicated GSK-3 and SRPK families in phosphorylating the SR-rich domain, SRPK1 had been shown to affect SARS-CoV nucleocapsid multimerization without a reported effect on viral growth, and GSK-3 inhibitors suppressed both SARS-CoV and mouse hepatitis virus. Other studies had linked general kinase families including CK2, CDKs and protein kinase C to SARS-CoV-2 replication without tying them to nucleocapsid phosphorylation. Separately, the SR-rich domain had been reported to drive gel to liquid phase transitions of nucleocapsid and RNA condensates. What was missing was an assignment of specific sites to specific kinases with an order of events, and a test of whether that cascade is required for replication.
Central question
Which host protein kinases phosphorylate the serine and arginine rich domain of the SARS-CoV-2 nucleocapsid protein, in what order do they act, and is that phosphorylation cascade required for coronavirus replication in human cells?
Experimental strategy
The strategy inverts the usual screening logic. Instead of testing candidate kinases against the substrate, the authors start from independently measured substrate specificity matrices for the relevant kinase families, obtained by combinatorial peptide library profiling, and use them to score every phosphoacceptor in the domain, which yields a prediction that is ordered in time because two of the three families require phospho-priming. GSK-3 prefers a phosphorylated residue four positions to the carboxyl side, CK1 prefers one three positions to the amino side, and the domain happens to contain three chains of serines and threonines spaced every fourth residue, so the geometry of the substrate and the geometry of the kinase preferences together dictate a direction of travel. Evolutionary conservation is used as independent evidence, comparing the phosphosites, the surrounding motif residues, and the remaining residues of each domain across 82 coronaviruses. The ordered model is then tested biochemically with purified components, using a Phos-tag gel for occupancy and radiolabeled ATP for total incorporation, with a double phospho-null mutant at the two proposed priming serines as the critical test. Requirement during infection is approached in three independent ways, knockdown, two research-grade SRPK inhibitors, and an approved drug with known off-target SRPK activity, each applied across an engineered line, a naturally infectable line, and primary human cells, with viability measured alongside so that antiviral effect is not confused with toxicity. Finally, phosphoproteomics after drug treatment closes the loop by asking whether the drug does to the nucleocapsid what the model says it should.
Key findings
- Phosphoproteomics of infected cells identified phosphosites across several viral proteins, with nucleocapsid by far the most phosphorylated, 14 sites in ACE2-A549 cells of which 11 were in the SR-rich domain and 26 sites in Vero E6 cells of which 15 were in that domain (Fig. 1B and 1C). Most SR-rich domain sites were also present in five previously published datasets.
- Across 82 coronaviruses the SR-rich domain was significantly more conserved than the linker domain and as conserved as the amino-terminal and carboxyl-terminal functional domains, with serine, threonine and arginine the most conserved residues (Fig. 1C and 1D). The SR-rich domain was the only domain in which detected phosphosites were significantly more conserved than the other residues of that domain (Fig. 1E).
- Measured substrate specificity showed SRPK1, SRPK2 and SRPK3 selecting arginine at positions minus three and plus three, serine at minus two and plus two, and proline at plus one, GSK-3 alpha and beta selecting a phosphorylated residue at plus four, and CK1 isoforms selecting a phosphorylated residue at minus three with partial preference for serine at minus four (Fig. 2A).
- Scoring each site produced an ordered model. SRPKs favor serine 206 and serine 188 as priming sites, GSK-3 then extends along two chains toward the amino terminus, and CK1 is predicted for threonine 205 once serine 202 is phosphorylated (Fig. 2B and 2C). This assignment is a computational prediction from specificity matrices and is stated as such.
- The motif residues predicted to direct SRPK to the two priming sites are as conserved across coronaviruses as the phosphoacceptors themselves (Fig. 2D). The authors read this as evidence that the cascade matters to the virus family, which is an inference from conservation rather than a functional test.
- In vitro, SRPK1 alone shifted recombinant nucleocapsid on a Phos-tag gel while GSK-3 or CK1 alone had only modest effects, and all three together increased phosphate incorporation more than the sum of the individual reactions (Fig. 2E), consistent with priming.
- The double phospho-null mutant carrying serine 188 to alanine and serine 206 to alanine abolished the SRPK-driven Phos-tag shift and reduced radioactive incorporation by all three kinases (Fig. 2F). This is the strongest direct support for the priming step of the model.
- Small interfering RNA knockdown of SRPK1 reduced SRPK1 messenger RNA and protein, reduced viral RNA, and shifted the ratio of phosphorylated to unphosphorylated nucleocapsid toward the unphosphorylated form (Fig. 3A to 3D).
- The SRPK1 and SRPK2 inhibitors SPHINX31 and SRPIN340 reduced viral RNA and infectious titer in a dose-dependent manner at concentrations tolerated by the cells, in ACE2-A549 cells, in naturally infectable Calu-3 cells, and in primary human type II pneumocytes from four donors (Fig. 3E to 3N).
- Alectinib, an approved anaplastic lymphoma kinase inhibitor with known SRPK1 and SRPK2 activity, reduced viral RNA and infectious titer dose-dependently in ACE2-A549 cells, Calu-3 cells and primary pneumocytes (Fig. 4A to 4F), and reduced replication of the distantly related alphacoronavirus HCoV-229E in Huh7 cells by more than a thousandfold (Fig. 4G).
- Phosphoproteomics after alectinib treatment showed reduced phosphorylation at most SR-rich domain sites while sites outside that domain did not decrease (Fig. 4I), and phosphosites scoring above the ninetieth percentile for SRPK specificity were significantly overrepresented among sites reduced by the drug (Fig. 4J).
- Proteomic analysis also showed that some interferon-stimulated gene products increased in abundance after alectinib treatment. The authors state that their experiments do not discriminate between SRPKs being needed for normal antiviral gene expression and reduced nucleocapsid phosphorylation compromising viral immune suppression.
Mechanistic model
The model is a directional, priming-dependent cascade. SRPK1 and SRPK2 phosphorylate serine 206 and serine 188, each in a locally favorable arginine and serine context. Those marks license GSK-3, which requires a phosphorylated residue four positions to the carboxyl side and therefore propagates along the four-residue chains toward the amino terminus, covering serine 202 through serine 186 and serine 184 through serine 176. Casein kinase 1, which requires a phosphorylated residue three positions to the amino side, accounts for threonine 205 once serine 202 has been modified. Several parts of this are demonstrated and several are not. The site assignments are predictions from measured specificity matrices rather than direct mapping of each site to each kinase in cells. The requirement for the two priming events is shown biochemically by mutation, and the cooperativity among the three kinases is shown by comparing combined against individual reactions on purified protein. The requirement of SRPK activity for replication is shown genetically and pharmacologically, and the drug is shown to reduce phosphorylation specifically in the targeted domain. What is not established is why the phosphorylation is needed. The authors say directly that future work is required to establish the functional role of nucleocapsid phosphorylation in the replication cycle, and they note that the inhibitors may act on aspects of the life cycle unrelated to nucleocapsid, that other kinases may phosphorylate other domains and other viral proteins, and that effects on host protein phosphorylation cannot be ruled out as contributors to the antiviral phenotype.
Conceptual or technical advance
The work shows that a dense, ambiguous cluster of phosphorylation sites can be resolved into an ordered sequence using measured kinase substrate preferences rather than candidate testing, and that the ordering falls out of the combination of substrate geometry with the priming requirements of two kinase families. That is a general method, applicable wherever phosphosite spacing is regular and specificity matrices exist. Practically, it nominates SRPK1 and SRPK2 as host targets whose inhibition suppresses two coronaviruses from different genera, and it identifies an already approved drug that reaches them, which shortens the distance between a mechanistic finding and something testable in people. The conservation analysis extends the relevance beyond currently circulating viruses, since the same sites and surrounding motifs are conserved in bat coronaviruses.
Relationship to the broader research program
The tenOever contribution here is one component of a multi-laboratory study led from Duke and Weill Cornell, and it lies in the SARS-CoV-2 infection and phosphoproteomic work carried out at the New York sites, with several co-authors from that group. The broader corpus interest it connects to is the host determinants of coronavirus replication and the use of host-directed rather than virus-directed intervention, a theme that recurs in the laboratory's COVID-19 work. The observation that alectinib treatment coincided with increased abundance of interferon-stimulated gene products touches the laboratory's long-standing interest in how SARS-CoV-2 suppresses the interferon response, though the paper does not resolve the direction of that relationship. Any statement linking this study to the laboratory's other coronavirus work would be category 3 synthesis and should be assembled centrally.
Related publications
- Blanco-Melo et al. 2020 and associated tenOever laboratory work establishing the ACE2-expressing A549 infection system. Methodological foundation, the cell system used for the New York arm of this study is cited as previously described.
- Prior reports from other groups on GSK-3 and SRPK phosphorylation of the SARS-CoV nucleocapsid, and on suppression of SARS-CoV and mouse hepatitis virus replication by GSK-3 inhibitors. Predecessors, the direct antecedents of the kinase families examined here.
- Five previously published SARS-CoV-2 phosphoproteomic datasets from other groups. Companion, used here for cross-validation of detected sites.
- Published work on gel to liquid phase transition of nucleocapsid and RNA condensates. Conceptual extension from other groups, cited as a candidate function for the phosphorylation described here but not tested in this study.
Limitations and boundaries
The kinase assignments to individual sites are computational predictions from substrate specificity matrices, validated in aggregate by in vitro reactions and by drug-induced changes in phosphorylation, not by direct site-by-site demonstration in infected cells, and the third priming site threonine 205 is assigned to CK1 largely by elimination. The biochemical reconstitution uses purified recombinant protein and single representative isoforms, SRPK1, GSK-3 alpha and CK1 epsilon, so isoform redundancy in cells is untested. The authors state that they cannot rule out effects of the inhibitors on aspects of the viral life cycle unrelated to nucleocapsid, that other kinases may contribute to phosphorylation of other domains and other viral proteins, and that altered phosphorylation of host proteins may contribute to the antiviral effect. The functional consequence of nucleocapsid phosphorylation for replication is not established here. All infection work is in cell culture, immortalized lines and primary pneumocytes, and the authors say explicitly that whether targeting SRPK1 and SRPK2 in vivo produces a similar magnitude of effect requires future study. The clinical observation offered in support, favorable COVID-19 outcomes in two lung cancer patients receiving alectinib, is described by the authors as not definitive and amounts to two case reports. Conservation across coronaviruses is sequence evidence for importance and not a functional demonstration in any virus other than the two tested.
Audience summaries
25 words
Three host kinase families phosphorylate the SARS-CoV-2 nucleocapsid in a set order, and blocking the first of them, including with the approved drug alectinib, suppresses coronavirus replication.
75 words
The SARS-CoV-2 nucleocapsid protein carries a cluster of phosphorylation sites in a region conserved across coronaviruses. By matching each site to the kinases whose sequence preferences fit it, this work reconstructed an ordered cascade in which SRPK enzymes act first and GSK-3 and casein kinase 1 extend the modification. Removing or inhibiting the initiating kinases reduced viral replication in several human cell systems, and an already approved cancer drug that hits them worked too.
150 words
Host-directed antivirals need well-defined host dependencies. This study focuses on the serine and arginine rich domain of the SARS-CoV-2 nucleocapsid protein, which phosphoproteomics identified as the most heavily modified region of any viral protein and which is unusually conserved across 82 coronaviruses. Using measured substrate specificity matrices rather than candidate screening, the authors predicted that SRPK enzymes phosphorylate serine 206 and serine 188 first, that GSK-3 then propagates along chains of residues spaced every four positions toward the amino terminus, and that casein kinase 1 covers threonine 205 after serine 202 is modified. In vitro reactions with recombinant protein supported the order, and mutating both priming serines abolished the cascade. Knockdown of SRPK1 and treatment with SPHINX31, SRPIN340 or the approved inhibitor alectinib reduced SARS-CoV-2 replication in engineered A549 cells, Calu-3 cells and primary human pneumocytes, and alectinib also suppressed HCoV-229E. The functional role of the phosphorylation itself remains undefined.
Documented publication relationships
- Imbalanced Host Response to SARS-CoV-2 Drives Development of COVID-19 — methodological foundation.