Multi-proteomic profiling of the varicella-zoster virus-host interface reveals host susceptibilities to severe infection

Abstract

Varicella-zoster virus (VZV) infects most humans and causes chickenpox, shingles and central nervous system pathologies. The molecular basis for these phenotypes remains elusive. Here we conducted a multi-proteomic survey on 64 individual VZV proteins and infection-induced perturbations in a neuronal cell line, identifying 900 interactors and 3,618 regulated host proteins. Data integration suggested molecular functions of viral proteins, such as a mechanism for the ORF61-mediated IFI16 degradation via the recruitment of E3 ligase co-factors. Moreover, we identified proviral host factors (MPP8 and ZNF280D) as potential targets to limit infection. Integration of exome sequencing analysis from patients with VZV-associated central nervous system pathologies identified nephrocystin 4 as a viral restriction factor, and its S862N variant, which showed reduced activity and decreased binding to the regulatory proteins 14-3-3. Collectively, our study provides a comprehensive herpesvirus-host interface resource, which aids our understanding of disease-associated molecular perturbations and data-driven identification of antiviral treatment options.

Fig. 1. VZV modulates the proteome signature of infected neuronal cells.

a, Experimental design of the VZV–host proteomic survey. Neuroblastoma SK-N-BE2 cells were infected with VZV, and the effects of infection on their proteome were analysed by bottom-up MS to generate the infection dataset. SK-N-BE2 cells were transduced with individual V5-tagged VZV ORFs, and analysed by MS after AP of the tagged viral bait (interactome dataset) and on the proteome level (effectome dataset). b, Volcano plot of VZV-induced protein abundance changes in SK-N-BE2 cells infected with VZV for 48 h. Significant host protein changes (two-sided Student’s t-test, permutation-based FDR ≤ 5 × 10⁻², |median log₂-transformed fold change| ≥ 0.5, n = 4 independent experiments) are marked in dark grey or coloured according to their GO annotation as presented in panel c. Viral proteins are coloured in red. The plot displays one representative assay of two repeats, each including four independent experiments (Supplementary Table 1). The bigger circles highlight changes observed in the two repeats. Diamonds indicate truncated log₂-transformed fold change. c, GO terms enriched among the cellular proteins that are downregulated or upregulated in VZV-infected SK-N-BE2 cells as represented in panel b (one-sided Fisher’s exact test, Benjamini–Hochberg FDR ≤ 5 × 10⁻², enrichment factor ≥ 4.5). Regulated GO terms were grouped and coloured according to parental cellular functions, as defined in the legend in b.

Fig. 2. The VZV–host protein–protein interaction network in neuronal cells.

a, Assembled network of the individual V5-tagged VZV protein–host interactomes generated by AP–MS in neuroblastoma SK-N-BE2 cells. VZV baits and host preys are shown as squares and ellipses, respectively (n = 4 independent experiments) (Supplementary Table 3). Viral proteins are numbered according to their gene name. The prey border colour specifies the enrichment factor; for preys targeted by several baits, the strongest enrichment is displayed. Host proteins selected for CRISPR–Cas9 knockout screen are highlighted in pink. b, Bar plots of GO Cellular Components enriched among the host proteins interacting with individual VZV proteins (one-sided Fisher’s exact test, unadjusted P ≤ 10⁻⁴). Actual −log₁₀P values are indicated when truncated. Targeted GO Cellular Components were coloured according to parental cellular functions as defined in the legend.

Fig. 3. The effectome of viral proteins identifies individual functions.

a, The number of cellular proteins upregulated or downregulated in neuroblastoma SK-N-BE2 cells expressing individual VZV ORFs, as detected by full proteome MS analysis. ORFs are ranked according to their expression kinetics during viral replication, and viral proteins that are part of the virion are annotated with an asterisk (*) (n = 4 independent experiments) (Supplementary Table 4, ‘Effectome significant’ and ‘is effect’). b, Network of enriched pathways (GO Biological Processes), transcriptome factor target gene sets (OmniPath) and human protein–protein complexes (CORUM, IntAct) among the cellular proteins that are regulated in SK-N-BE2 cells expressing the indicated individual VZV proteins (one-sided Fisher’s exact test, unadjusted P ≤ 10⁻⁴). Viral proteins are numbered according to their gene name. Edge thickness indicates the P value. Regulated terms were coloured according to parental cellular functions as defined in the legend. dsDNA, double-stranded DNA.

Fig. 4. Multi-proteomic data integration.

a, The orthogonal analysis of infection and effectome datasets (1) or effectome and interactome datasets (2) provides hypotheses on the molecular mechanism involved in VZV functions. b, WB analysis of IFI16 in HEK293T cells after co-transfection with HA-GFP, HA-ORF61 or HA-ICP0. Representative of n = 3 independent experiments. Data are summarized in Extended Data Fig. 5c. c, IF analysis of HFF cells mock infected or infected with recombinant HA-ORF61 VZV for 8 h, treated or not with the proteasome inhibitor MG-132. Cells were stained for IFI16 and VZV (ORF61) and with DAPI. More than 120 nuclei per condition were analysed at ×10 magnification. Minimum and maximum, first and last quantiles, and the median log₁₀ intensity of IFI16 or VZV are indicated (two-sided Mann–Whitney test, unadjusted). Non-infected cells, defined by the maximal observed VZV signal in mock infected cells (grey dashed line), were excluded from the infected conditions in the IFI16 plot. Representative of n = 3 independent experiments. Images are presented in Extended Data Fig. 5d. d, IF analysis of the subcellular localization of UBXN7 and VZV ORF61 in HFF cells infected with recombinant HA-ORF61 VZV for 8 h. Cells were stained for UBXN7 and ORF61 and with DAPI, and analysed at ×63 magnification. Each channel and the merge of ORF61 and UBXN7 are displayed for one representative cell. Scale bar, 10 µm. The line profiles represent UBXN7 and ORF61 intensities extracted as indicated by the white arrow in the merge image. Representative of three independent experiments. e, IF analysis of HFF cells, either control or UBXN7-depleted by shRNA expression, and transduced with V5-ORF61 or V5-GFP. Cells were stained for IFI16 and V5 tag and with DAPI. More than 500 nuclei per condition were analysed at ×20 magnification. Minimum and maximum, first and last quantiles, and the median intensity of IFI16 are indicated (two-way ANOVA, adjusted with Tukey’s method). Representative of n = 2 independent experiments. f, Interactome and effectome data were mapped onto the cellular gene network ReactomeFI and submitted for network diffusion analysis (Methods and Extended Data Fig. 5f) to generate individual viral ORF-altered gene subnetworks. g,h, Featured subnetworks resulted from network diffusion predictions of VZV ORF12 (g) and ORF9 (h). Edges indicate ReactomeFI connections that passed the random walk transition probability threshold (0.05). Complete HotNet output subnetworks are shown in Extended Data Fig. 5g,h, and interactive versions are given in Supplementary Data 1. Source data

Fig. 5. Loss-of-function screen identifies VZV restriction and dependency factors.

a, Host gene knockout screen for VZV replication. Cas9-free GFP-expressing and Cas9 BFP-expressing SK-N-BE2 cells were co-cultured and co-transduced with sgRNA targeting host genes. Following selection, cells were infected by co-culture with mRFP-VZV(pOka)-infected MeWo cells and analysed by flow cytometry. b, Flow cytometry gating strategy of the knockout screen presented in a allows for the exclusion of the MeWo inoculum cells and individual gating of the target (BFP+) and control (GFP+) cell populations, as shown for representative NTC mock and infected wells. The number of gated cells is indicated. The histogram below shows the overlaid distribution of the RFP intensities within target and control cells, indicating their respective infection level. MRIs are indicated as values and dashed lines. c, Array knockout screen was performed on 116 host genes selected from the VZV proteomic survey. The heat map shows the z-scored target-to-control normalized MRI for each sgRNA (number 1 to 4) per gene, averaged across duplicates. Targeted host genes are sorted according to their most potent sgRNA. d, Validation by flow cytometry analysis of the function of MPP8 and ZNF280D. Knockout (KO) or NTC BFP-expressing SK-N-BE2 cells were infected via co-culture with mRFP-VZV(pOka)-infected MeWo cells. Gating the BFP+ SK-N-BE2 population allows for exclusion of the inoculum MeWo cells. The number of gated cells is indicated. The histograms below show the overlaid distribution of the RFP intensities within SK-N-BE2 cells, NTC (grey) or knockout for the indicated gene (green) (representative well). MRIs are indicated as values and dashed lines. e, Fold change of the MRI within SK-N-BE2 cells, NTC or knockout for the indicated gene, and infected with mRFP-VZV(pOka) as presented in d, compared with the estimate (Methods). Mean ± s.e.m. is indicated (n = 3 independent experiments) (one-sided Student’s t-test, unadjusted).

Fig. 6. Rare variant analysis in patients with VZV CNS infection identifies a mutation in the restriction factor NPHP4.

a, Prioritization of genes identified by WES variant analysis from 13 patients affected by VZV CNS infection who were otherwise immunocompetent. WES identified rare and predicted deleterious variants in patients, to which two biological filters were applied: the proteomic profiling of VZV–host interactions and the results of the VZV replication functional assay. b, Validation by flow cytometry analysis of the function of NPHP4 and characterization of the NPHP4(S862N) variant. KO or NTC BFP-expressing SK-N-BE2 cells, transduced with an empty vector (EV), wild type (WT) NPHP4 or NPHP4(S862N) as indicated, were infected via co-culture with mRFP–VZV(pOka)-infected MeWo cells. Gating the BFP+ SK-N-BE2 population allows exclusion of the inoculum MeWo cells. The number of gated cells is indicated. The histograms below show the overlaid distribution of the RFP intensity within SK-N-BE2 cells across rescue conditions compared with the knockout non-rescue cells (KO or EV) (representative well). MRIs are indicated as values and dashed lines. c, Fold change of the MRI within SK-N-BE2 cells that were NTC or knocked out for NPHP4 and infected with mRFP–VZV(pOka), as presented in b, following the transduction of an EV, NPHP4-WT or NPHP4(S862N), compared with the NTC or EV control. Mean ± s.e.m. is indicated (n = 3 independent experiments) (one-way ANOVA test). d, Interactomes of HA–NPHP4 wild-type and S862N generated by AP–MS in neuroblastoma SK-N-BE2 cells. log₂ enrichment factors are shown compared with the HA–GFP control. Differential binding partners that were identified from statistical analysis of the comparison between the WT and the variant constructs are indicated: increased enrichment by the WT is shown in light blue circles; increased enrichment by the variant is shown in pink circles; grey circles indicate shared enrichment by both WT and the variant; binders that pass the statistical threshold for the comparison of WT versus mutant are indicated by squares; and 14-3-3 proteins are indicated in dark blue (n = 4 independent experiments) (Supplementary Table 5). e, Review of the virus–host interactions between VZV and NPHP4 as identified in our multi-proteomic and functional analysis, and the functional characterization of the S862N variant. NS, not significant.

Extended Data Fig. 1. VZV infection of SK-N-BE2 cells.

(a) Schematic representation of the infection by transwell co-culture with VZV-infected MeWo cells inoculum. (b) Histogram of fluorescence intensity of SK-N-BE2 cells mock- (grey) or VZV-infected (red) with rOka VZV for 48 h as described in (a), and stained for FITC-conjugated antibodies against VZV immediate-early ORF62 protein and late glycoprotein E. Percentage of FITC-positive cells are indicated. (n = 3 independent experiments).

Extended Data Fig. 2. Evaluation of the V5-tagged VZV proteins’ expression and host interactome.

(a) Expression of V5-tagged VZV proteins, in stably transduced SK-N-BE2 cells used in interactome and effectome analysis. Blue arrows indicate the bands used for quantification. n = 4 independent experiments. (b) Subcellular localization analysis by immunofluorescence of the indicated transduced V5-tagged VZV ORFs in SK-N-BE2 cells. Cells were fixed and stained with V5+rbAlexa488 (green) and DAPI (cyan), and subjected to confocal microscopy at 63x magnification. Images are representative of n = 3 independent experiments. Scale bar = 10μm. (c) Number of host interactor (target) targeted by one or several VZV ORFs, as represented in Fig. 2a. The percentages of total interactors are indicated on top of the bar. (d) Correlation analysis between the number of targets (Fig. 2a) and the expression level of individual VZV ORF as quantified from Extended Data Fig. 2a. (e) Intersection of the measured VZV-host protein-protein interactions with known VZV and HSV-1 interactions from public databases (Biogrid, Intact, VirHostNet 2.0). (Supplementary Table 2). VZV proteins are numbered according to their gene name. The gene name of the corresponding HSV-1 homologous protein and the common short protein name (bracket) are indicated. (f) Western-blot analysis of the V5 immunoprecipitation of SK-N-BE2 cells expressing VZV ORF32-V5, VZV ORF23-V5 or non-transduced. Total lysates and immunoprecipitation samples (IP) are shown (n = 1 independent experiment). (g) Western blot analysis of GFP immunoprecipitation of Hela Kyoto cells expressing GFP-GTF2B, 24 h after transfection of VZV ORF32-V5, VZV ORF23-V5 or non-transfected. Total lysates and immunoprecipitated (IP) samples are shown (n = 1 independent experiment). (h) Immunofluorescence analysis of HeLa Kyoto cells stably expressing GFP-GTF2B and transiently transfected with VZV ORF32-V5. Cells were stained with V5+rbAlexa594 (magenta), GFP-DyLight-488 (green) and DAPI (cyan), and subjected to confocal microscopy at 63x magnification. Images are representative of n = 2 independent experiments. Scale bar = 10μm. Source data

Extended Data Fig. 3. Prediction of the subcellular localisation of the VZV proteins.

(a) For each VZV ORF interactome (Fig. 2a), the host preys were categorised according to their subcellular localisation annotations (Gene Ontology Cellular Component, Uniprot, Protein Atlas), as indicated (Supplementary Table 3). (b) Systematic analysis of the individual VZV ORF as described in (a). n indicates the total number of preys for a given ORF. The numbers in grey indicate the number of preys assigned to the given subcellular category for each VZV ORF.

Extended Data Fig. 4. Host protein abundance changes identified in the effectome of VZV proteins.

(a) Number of host protein regulated and number of total up- and down-regulation (effect) observed in the effectome of one or several VZV ORFs, as represented in Fig. 3a. The percentages of total effects are indicated. (b) Correlation analysis between the number of effects (Fig. 3a) and the expression level of individual VZV ORF as quantified from Extended Data Fig. 2a. (c, d) log₂ LFQ intensity of the indicated host protein measured in SK-N-BE2 cells expressing the given VZV ORF. The background is defined by VZV ORF expressing cells from the same measurement batch (See Methods). Median and 95% confidence intervals are indicated (n = 4 independent experiments per VZV ORF). p-value of the effect as compared to the background is indicated (two-sided Wilcoxon rank-sum test; Benjamini-Hochberg adjusted). (e-h) Change of protein abundance of the components of (e) the peptidyl-diphthamide metabolic pathway following ORF61 expression, (f) the tight junction complex following ORF9A expression, (g) the adherens junction complex following ORF10 or ORF38 expressions and (h) the cell cycle cyclin-kinase complexes following ORF8 or ORF12 expression. log₂ LFQ intensity of the indicated host protein were measured in SK-N-BE2 cells expressing the given VZV ORF by LC-MS/MS. The background is defined by VZV ORF expressing cells from the same measurement batch (See Methods). Bars are coloured according to the regulated cellular function the given protein belongs to, as represented in Fig. 3b. Median and 95% confidence intervals are indicated (n = 4 independent experiments per VZV ORF). p-value of the effect as compared to the background is indicated (two-sided Wilcoxon rank-sum test; Benjamini-Hochberg adjusted).

Extended Data Fig. 5. Hypothesis-driven and systematic data integration.

(a, b) Intensity of the indicated host protein measured by mass spectrometry in SK-N-BE2 cells, expressing the given VZV ORF or from the background(Methods). Median and 95% confidence intervals are indicated (n = 4 independent experiments per ORF; two-sided Wilcoxon rank-sum test; Benjamini-Hochberg adjusted). (c) Western blot analysis of IFI16 in HEK293T cells after co-transfection with HA-GFP, -ORF61 or -ICP0. Mean +/− SEM are indicated (n = 3 independent experiments; paired one-way ANOVA test, Bonferroni adjusted). A representative blot is presented in Fig. 4b. (d) Immunofluorescence analysis of HFF cells mock- or recombinant HA-ORF61 VZV-infected for 8 h, treated or not with the proteasome inhibitor MG132. HFF cells were stained for IFI16, VZV and with DAPI, and analysed at 20x magnification. Representative of n = 3 independent infection experiments. Analysis of nuclear intensities is summarized in Fig. 4c. Scale bar = 50 µm (e) Validation of the knockdown of UBXN7 in HFF cells by immunofluorescence analysis. Scale bar = 100 µm (f) Network diffusion analysis of VZV ORFs. Interactome and effectome data were mapped onto the cellular gene network ReactomeFI. Effects were used to weight the random diffusion, resulting in several subnetworks of given minimal edge weight and average effect-target path length. Comparison between the ‘real data’ and randomly permuted networks of given minimal edge weight defines the optimal subnetwork with the best proximity between targets and effects. The displayed subnetwork connects targeted, affected and intermediate proteins and provides p-value which evaluate the significance of the connections as compared to the random network. (g, h) Altered subnetworks of ORF12 (g) and ORF9 (h) resulting from the network diffusion analysis. Connections featured in Fig. 4g and h are indicated in blue. Gene names and nature of the connections (for example protein-protein complex; activation; expression control) can be browsed on interactive versions in Supplementary Data.

Extended Data Fig. 6. CRISPR-Cas9-based knockout of identified host genes and their functional evaluation.

(a) Monitoring of cell density at infection for the host gene knockout screen of VZV replication. Per well correlation analysis between cell density at infection and the raw median RFP intensity of the whole SK-N-BE2 cell population (GFP+ and BFP+) (left), or the target-to-control normalized median RFP intensity (BFP + / GFP+) (right). (b) Resazurin viability assay of the NTC and indicated knockout (KO) cell lines. Mean and standard error to the mean are indicated (n = 3 independent experiments) (c,f) Percentage of nucleotide insertion/deletion (indel) within the sgRNA-targeted regions of the (c) MPP8 and (f) ZNF280D genes in the MPP8-KO cells as compared to NTC. (d-h) Abundance of (d,e) MPP8 and (g,h) ZND280D proteins assessed by western blot (d,g; exact band indicated by an arrow) and mass spectrometry (e,h; mean normalized intensity +/− SD) analysis in KO and NTC cells. (i) Viral growth kinetics within SK-N-BE2 cells NTC or knockout for the indicated gene, and infected with mRFP-VZV(pOka), monitored by live-imaging. Representative of n = 3 independent experiments. (j) Area under the curve (AUC) analysis of the VZV-mRFP growth kinetics in SK-N-BE2 cells NTC or knockout for the indicated gene, as presented in (f). Mean and standard error to the mean are indicated (n = 3 independent experiments; paired one-way ANOVA test, Bonferroni adjusted). (k) Volcano plot of protein abundance changes induced by MPP8 knockout in SK-N-BE2 cells as compared to NTC. Significant host protein changes (Bayesian linear model-based unadjusted two-sided p-value ≤ 5.10⁻², |median log2 fold change| ≥ 0.5, n = 4 independent experiments) are represented with black dots. Members of the HUSH complex are indicated in red. (Supplementary Table 5). h = hour. Source data

Extended Data Fig. 7. Functional characterisation of NPHP4 and its S862N variant during VZV infection.

(a) The percentage of nucleotide insertion/deletion (indel) within the sgRNA-targeted region of the NPHP4 gene in the NPHP4-KO cells as compared to NTC. (b) The abundance of NPHP4 protein was assessed by western blot analysis (exact band indicated by an arrow). (c) Resazurin viability assay of the NTC and NPHP4 knockout cell line. Mean and standard error to the mean are indicated (n = 3 independent experiments) (d) Western blot analysis of the expression level of the HA-NPHP4 and -NPHP4-S862N constructs in NTC and NPHP4-knockout (KO) cells. (e) Volcano plot of differentially regulated transcripts in SK-N-BE2 cells infected with VZV, as compared to mock cells. Significant host protein changes (two-sided adjusted p-value ≤ 5.10⁻², |median log₂ fold change| ≥ 1, n = 3 independent experiments) are represented with black dots. (Supplementary Table 5) (f) Comparison of differentially regulated transcripts (log₂ fold change) following VZV infection in NTC and NPHP4 knockout (KO) SK-N-BE2 cells. (n = 3 independent experiments) (Supplementary Table 5) (g, h) Normalized transcript counts of the WNT target genes FOSL1 and ZEB2 in SK-N-BE2 cells NTC or NPHP4-KO following VZV-infection as displayed in (f) (two-sided Wald test unadjusted p-value are indicated; n = 3 independent experiments) (i) Interactomes of HA-NPHP4 wild-type generated by affinity-purification coupled to mass-spectrometry in neuroblastoma SK-N-BE2 cells. Significant host protein association are marked in black (two-sided Welch t-test, permutation-based FDR ≤ 1.10⁻², S0 = 4, n = 4 independent experiments) (Supplementary Table 5). Associated cilia basal body proteins (Reactome 267965) are indicated in light blue, including the NPHP1-NPHP4-NPHP8(RPGRIP1L) complex which is labelled. Associated 14-3-3 proteins are indicated in dark blue. Source data