An Integrated Systems Biology Approach Identifies the Proteasome as A Critical Host Machinery for ZIKV and DENV Replication

Abstract

The Zika virus (ZIKV) and dengue virus (DENV) flaviviruses exhibit similar replicative processes but have distinct clinical outcomes. A systematic understanding of virus-host protein-protein interaction networks can reveal cellular pathways critical to viral replication and disease pathogenesis. Here we employed three independent systems biology approaches toward this goal. First, protein array analysis of direct interactions between individual ZIKV/DENV viral proteins and 20,240 human proteins revealed multiple conserved cellular pathways and protein complexes, including proteasome complexes. Second, an RNAi screen of 10,415 druggable genes identified the host proteins required for ZIKV infection and uncovered that proteasome proteins were crucial in this process. Third, high-throughput screening of 6016 bioactive compounds for ZIKV inhibition yielded 134 effective compounds, including six proteasome inhibitors that suppress both ZIKV and DENV replication. Integrative analyses of these orthogonal datasets pinpoint proteasomes as critical host machinery for ZIKV/DENV replication. Our study provides multi-omics datasets for further studies of flavivirus-host interactions, disease pathogenesis, and new drug targets.

Keywords: Chemical genetics screening; Multi-omics; Protein–protein interaction; RNAi screening.

Figure 1

Identification of ZIKV–and DENV–host PPIs A. Sample images of HuProt arrays showing human proteins bound by individual viral proteins. Each human protein was printed in duplicate. The orange, blue, and green boxes represent shared, ZIKV-specific, and DENV-specific interactions, respectively. B. Duplicate experiments performed for each viral protein probe showed high reproducibility. Pearson correlation coefficient analysis showed that signals of duplicate experiments based on DENV-NS5 have a high linear relationship (r = 0.961). C. Summary of numbers of unique and conserved virus–host interactions between each ZIKV and DENV homologous pair.

Figure 2

GO analyses of host proteins in the PPI networks A. Enriched GO terms in the categories of molecular function, biological process, and cellular component are found in both shared and virus-specific PPI networks. The folds of enrichment are color-coded by P value. As examples, interactions of six non-structural ZIKV proteins (NS2B, NS3, NS4A, NS4A + 2K, 2K + NS4B, and NS5) with proteasome complex (B) and interactions of eight non-structural ZIKV proteins (NS2A, NS2B, NS3, NS4A, NS4A + 2K, 2K + NS4B, SN4B, and NS5) with spliceosome complex (C) were shown respectively. Here, only the subunits capable of binding with ZIKV proteins were included. Circles with bright blue outlines indicate previously reported virus binding proteins. D. Co-IP of overexpressed FLAG-tagged ZIKV proteins (NS3 and NS5) and V5-tagged human proteasome subunits (PSMA1, PSMA3, and PSMB4) in HEK293FT cells. IP assays were performed with anti-FLAG mAb magnetic beads and eluted fractions were analyzed by Western blot using mouse anti-V5 antibodies. Mouse IgG magnetic beads were used as a negative control to evaluate any non-specific binding on the beads. Inputs correspond to 2% of total lysate incubating with anti-FLAG mAb magnetic beads.

Figure 3

Critical host proteins for ZIKV replication A. STRING analysis of genes that significantly affected ZIKV replication in RNAi screening. B. Percentage of genes with over 30% reduction of NS1 levels by siRNA-knockdown among all genes in a specific category. The “All” group indicates the collection of 10,415 siRNA-targeted druggable genes; the “ZIKV−host PPI” group indicates the 327 genes whose protein products were found to interact with ZIKV proteins in the PPI dataset. Note the high success rate (20 out of 47 members) in the category of “Proteasome complex.”

Figure 4

Small molecule inhibitors against ZIKV replication A. Flowchart of compound screening and confirmation with the ZIKV-NS1 TR-FRET assay. Precultured cells in 1536-well plates were treated with 6016 compounds for 1 h, and then infected with virus for 1 day, followed by the ZIKV-NS1 TR-FRET assay. Of the 6016 compounds, 256 were identified as preliminary hits and selected for secondary validation by the ZIKV-NS1 TR-FRET assay and cytotoxicity evaluation with the same cells. 217 of the preliminary hits were confirmed and 134 compounds exhibited greater than four-fold selectivity of ZIKV-NS1 inhibition over compound cytotoxicity. B. Summary of behaviors and IC₅₀ values of 12 groups of potent compounds categorized based upon their reported mechanisms of action. Values represent mean ± SD (n = 3 cultures). Curves represent best fits for calculating IC_50.

Figure 5

Integrative analysis of PPI and chemical genetics screen A. STRING analysis of 45 anti-ZIKV drug target human proteins that were found to interact with ZIKV proteins in our PPI analysis. B. Functional association networks among the proteins that interact with viral proteins and are targeted by effective compounds.

Figure 6

Experimental validation of the proteasome inhibitors A. PPI network analysis of virus proteins and human proteasome subunits reveals that most of the interacting proteasome subunits are part of the 20S core particle. B. Percentage of the ZIKV-binding subunits in 26S proteasome and its two sub-complexes, the 20S core particle and the 19S regulatory particle. C. Inhibition of ZIKV expression in human glioblastoma cell line SNB-19 by a panel of proteasome inhibitors. The SNB-19 cells were infected by ZIKV PRVABC59 (MOI = 1) in the presence of 1 µM of each inhibitor and then incubated for 48 h before the cultures were analyzed for ZIKV-E protein expression by immunostaining. Mock indicates cells without ZIKV infection. Scale bar: 100 µm. D. and E. Sample images (D) and quantification (E) of titer assay to assess the potency of the proteasome inhibitors against infectious ZIKV production in SNB-19 cells. All data were normalized to that of 0 µM for each compound. Dose-dependent antiviral activity is presented as fluorescent focus-forming units per ml (FFU/ml) and data are represented as mean ± SD (n = 6). Curves represent best fits for calculating IC₅₀ values (listed to the right). MOI, multiplicity of infection; ZIKV-E, ZIKV envelope.