Discussion
We used a mass spectrometry-based approach to study perturbations in protein abundance and phosphorylation during SARS-CoV-2 infection. Viral proteins increased, starting 8 h after infection, indicative of viral replication, whereas only small changes in host protein abundance were observed within 24 h. In contrast, large changes were observed in protein phosphorylation, highlighting the degree by which the virus makes use of the host post-translational regulatory systems to promote rapid changes in cellular signaling.
Changes in phosphorylation reflect altered activities of kinases that are hijacked during the infection. Based on changes in phosphorylation of their annotated substrates, we estimated changes in activity of 97 of the 518 human kinases. The changes in kinase activity offer insights into the biology of viral infection. Kinases represent ideal drug targets; here, we identified kinases and pathways altered by SARS-CoV-2 infection that can be targeted by 87 FDA-approved drugs and compounds in clinical trials or in preclinical development (Table S8).
The most strongly regulated kinases fall into a set of signaling pathways that include p38/MAPK signaling, AKT and ERK signaling, Rho GTPase and CK2 cytoskeleton signaling, and cell cycle regulation. The downregulation of ROCK and PAK kinase activity and upregulation of CK2 cytoskeleton-related targets suggest virus-induced changes in cytoskeleton organization. Imaging of infected cells revealed formation of actin-rich filopodia containing viral proteins. Higher-resolution electron microscopy data confirm the presence of assembled viral particles in these structures. Many viruses, including vaccinia, Ebola, and Marburg, hijack the host cell cytoskeleton to promote egress and rapid cell-to-cell spread across epithelial monolayers. Vaccinia promotes Arp2/3-dependent actin assembly, producing a filopodial protrusion with a virus at the tip (Leite and Way 2015). In contrast, Marburg virus hijacks the unconventional motor protein Myosin X, which promotes filopodium formation and trafficks the virus along the filopodium shaft. The SARS-CoV-2 protein clusters peppered throughout the length of filopodial protrusions more closely resemble Marburg than vaccinia, but additional work is required to understand whether SARS-CoV-2 makes use of either Myosin X motor activity or actin filament assembly to move along filopodia. CK2 is known to phosphorylate myosin proteins at endocytic sites to drive actin polymerization (Fernández-Golbano et al., 2014). Furthermore, CK2 has been found to regulate actin tail formation during vaccinia virus infection, enabling efficient cell-to-cell spread of the virus (Alvarez and Agaisse 2012; Smith and Law 2004). Here the CK2 inhibitor silmitasertib displayed robust antiviral activity, suggesting a role of this kinase in regulating the SARS-CoV-2 life cycle.
In addition, kinase activity profiling analysis shows that CDK1/2 activities are significantly reduced by SARS-CoV-2 infection, leading to a S/G2 phase arrest that is similar to infectious bronchitis virus (IBV), a prototypical coronavirus (Dove et al., 2006; Li et al., 2007a), and other RNA viruses (Lilley et al., 2007; Ariumi et al., 2008). Arresting cells in S/G2 phase may provide benefits for viral replication and progeny production by ensuring an abundant supply of nucleotides and other essential host DNA repair/replication proteins (Chaurushiya and Weitzman 2009).
The predicted increase in p38/MAPK activity led us to investigate the effects of p38/MAPK inhibition on pro-inflammatory cytokine production and viral replication in SARS-CoV-2-infected cells. Recent immunological studies have indicated that increased IL-6, IL-10, and TNF-α and lymphopenia are associated with severe COVID-19 cases (Pedersen and Ho 2020). The p38/MAPK pathway responds to and controls production of potentially harmful pro-inflammatory cytokines. Several pathogenic viral infections induce a p38/MAPK signaling state that exhibits uncontrolled positive feedback regulation, leading to excessive inflammation associated with severe disease. Inhibition of p38/MAPK signaling suppressed the overproduction of inflammatory cytokines induced by several viral infections, including SARS-CoV, Dengue virus, and influenza A virus, improving survival in mice (Fu et al., 2014; Growcott et al., 2018; Jimenez-Guardeño et al., 2014). However, p38/MAPK inhibition did not directly impair the virus in these cases but, instead, the host’s immune response to the infection. In contrast, during SARS-CoV-2 infection, p38/MAPK inhibition suppressed cytokine production and impaired viral replication by a still unknown mechanism, suggesting that p38/MAPK inhibition may target multiple mechanisms related to COVID-19 pathogenesis.
We tested 68 drugs and compounds and found antiviral activity for several that are FDA approved, in clinical testing, or under preclinical development for various diseases, including silmitasertib (CK2, phase 2), gilteritinib (AXL, FDA approved), ARRY-797 (p38, phase 2/3), MAPK13-IN-1 (p38, preclinical), SB203580 (p38, preclinical), ralimetinib (p38, phase 2), apilimod (PIKFYVE, phase 1), and dinaciclib (CDK, phase 3), among others (Figure S5; Table S8). Silmitasertib, a small molecule undergoing clinical trials for various cancers, is now being considered for testing in humans to combat COVID-19. Although the effectiveness of CK2 inhibition may be attributed to its regulation of stress granules (Gordon et al., 2020), viral egress and dissemination could be facilitated by CK2-mediated remodeling of the extracellular matrix (Figure 5).
Ralimetinib is currently in phase 2 clinical trials for treatment of ovarian cancer (Patnaik et al., 2016), and ARRY-797 is in phase 3 clinical trials for treatment of cardiomyopathy. The antiviral activity observed for gilteritinib, an FDA-approved drug for treatment of acute myeloid leukemia, is supported by involvement of another AXL inhibitor, bencentinib, in the RECOVERY COVID-19 clinical trial in the United Kingdom. AXL is known to regulate various intracellular signaling pathways (Allen et al., 2002; Hafizi and Dahlbäck 2006), including Ras/ERK, PI3K, and p38 (Allen et al., 2002); AXL inhibition here may contribute to the downregulation of p38 signaling. Apilimod, a PIKFYVE inhibitor, has been described in a recent study to have antiviral capacity (Ou et al., 2020). Here we expand this into a mechanism of regulation by phosphorylation of PIKFYVE upon SARS-CoV-2 infection.
Similar to successful antiretroviral therapy for HIV, a combinatorial drug cocktail may be a viable treatment option for SARS-CoV-2 infection. Specifically, combining remdesivir with the kinase inhibitors identified in this study as well as with translation inhibitors and/or modulators of sigma-1 receptor (Gordon et al., 2020) warrants further testing. Furthermore, pairing genetic and pharmacological perturbations in a systematic fashion could identify new combination therapy approaches and illuminate disease mechanisms.
The unbiased, global phosphoproteomics approaches used here highlight cellular processes hijacked during SARS-CoV-2 infection. To address the need for improved therapeutic strategies to fight COVID-19, we employed a data-driven approach by mapping phosphorylation profiles of dysregulated signaling pathways to drugs and compounds targeting those signaling pathways. We hope this paradigm can be employed in the future to find additional therapies for COVID-19 and other infectious diseases.

Limitations of Study
A limitation of the current study is the use of a non-human cell line for proteomics analysis upon SARS-CoV-2 infection; here, an African green monkey cell line (Vero E6) was used because it has been shown previously to be highly permissible to SARS-CoV-2 infection (Harcourt et al., 2020). However, pharmacological inhibition of SARS-CoV-2 was assessed in human lung A549-ACE2 cells in addition to Vero E6 cells. The majority of drug effects were found to be replicated between cell lines.