Ethics

The patient study included in this work complies with the Declaration of Helsinki and national ethics guidelines and was approved by the Danish National Committee on Health Research Ethics (number 1-10-72-275-15), the Data Protection Agency and the institutional review board. Patients provided oral and written informed consent for participation in the study.

Cell line and reagents

Cell lines

SK-N-BE2 cells were kindly provided by R. Klein (MPI of Neurobiology). MeWo cells were kindly provided by A.-V. Borbolla (MHH). HEK293T (CRL-11268) cells were purchased from ATCC. HeLa Kyoto-expressing GFP-tagged GTF2B from the bacterial artificial chromosome (BAC) transgene was from I. Poser⁹⁴. HFF-1 (SCRC-1041) cells were a kind gift from M. Brinkmann (HZI). All cell lines were tested to be mycoplasma free. All cells were maintained in culture medium: DMEM medium (high glucose, pyruvate; Gibco Fisher Scientific), supplemented with 10% fetal calf serum (FCS), 100 µg ml⁻¹ streptomycin and 100 IU ml⁻¹ penicillin. Low passage cell lines were kept as frozen stock in liquid nitrogen after resuspension in culture medium supplemented with 45% FCS and 10% DMSO. Blasticidin (15205; Sigma-Aldrich), puromycin (P8833; Sigma-Aldrich) and zeocin (Invitrogen) were used for transduced cell selection in this study. For proteasome inhibition, we used MG-132 (474787; Sigma-Aldrich).

Plasmids

pDONR221 was purchased from Invitrogen. ICP0-pDONR201 was kindly provided by G. Superti-Furga (CEMM). pSicoR-SpCas9-ZeoR⁹⁵, pLenti6.3-GFP-blastR (number 40125; Addgene) and pWPI-tagBFP-blastR (generated in this study as described later) were used to generate lentiviruses, allowing the expression of Cas9 or reporter fluorescence proteins in SK-N-BE2 cells. The tagBFP expression cassette was amplified from pSCRPSY-PAC2A-tagBFP (kindly provided by C. Rice; Rockefeller University) and inserted into pWPI-blastR. The library of VZV ORFs were ligated into pCR8/TOPO entry vector (Invitrogen) and shuttled into pLenti6.3-TO-V5-DEST-BlastR (Invitrogen) via Gateway recombination, allowing expression of C-terminal V5-tagged proteins and expression of blasticidin resistance. The library of host gene gRNAs were ligated into pSicoR-U6-gRNA-EFS-PuroR⁹⁵, allowing expression of the individual gRNA and puromycin resistance. pENTR encoding for a codon-optimized NPHP4 gene was purchased from Twist Bioscience. pWPI-nHA-PuroR-GW, allowing the expression of amino-terminal (N-terminal) HA-tagged proteins and puromycin resistance was kindly provided by A. Plaszczyca (University of Heidelberg). The pWPI-nHA-PuroR encoding wild type NPHP4 and NPHP4(S862N), VZV ORF61, HSV-1 ICP0 and GFP expression were generated in this study as described later. Packaging vectors pMD2-VSV-G and pCMV-Gag-Pol or pCMVR8.91 (Didier Trono’s laboratory) were used to produce lentiviruses. pLVX, allowing lentiviral transduction of shRNA and expression of puromycin resistance, was kindly gifted by M. Friedrich (Technical University of Munich).

Antibodies

TFIIB (western blot (WB) 1:1,000; 4169; Cell Signaling Technology), MPP8 (WB 1:1,500; 16796; Proteintech), ZNF280D (WB 1:1,000; PA5-56410; Invitrogen), NPHP4 (WB 1:3,000; A8934; Abclonal), UBXN7 (IF 1:500; HPA049442; Sigma-Aldrich), IFI16 (WB 1:1,000, IF 1:500; 14970; Cell Signaling Technology), V5-tag rabbit (WB 1:1,000, IF 1:1,000; 13202; Cell Signaling Technology), V5-tag mouse (WB 1:1,000, IF 1:400; R960-25; Invitrogen), HA-tag (WB 1:2,500, IF 1:100; 2367; Cell Signaling Technology), HA-tag-HRP (WB 1:1,000; H6533; Sigma-Aldrich), β-actin-HRP (WB 1:2,500; sc-47778; Santa Cruz), β-tubulin (WB 1:500; 2128; Cell Signaling Technology). For WB, secondary antibodies conjugated to HRP detecting rabbit IgG (1:2,500) and mouse IgG (1:5,000) were purchased from Dako and Sigma-Aldrich, respectively. For IF, DAPI (1:1,000) and secondary antibody detecting rabbit or mouse IgG conjugated to Alexa Fluor 488, Alexa Fluor 594 or Alexa Fluor 647 (1:200 to 1:500) were purchased from Invitrogen. GFP-DyLight-488 was purchased from Rockland (600-141-215; 1:1,000).

Virus strains and virus stocks preparation

VZV rOka was a gift from J. Cohen³⁰ (National Institutes of Health).

The recombinant VZV used in this study was generated based on pP-Oka, an infectious BAC clone of the pOka strain⁷⁴ using two-step red-mediated mutagenesis as previously described^96,97. The reporter RFP-tagged VZV variant was generated by insertion of the mRFP cassette into the pP-Oka BAC at the C terminus of both copies of the diploid VZV gene ORF63/70 encoding the immediate-early protein 63 (IE63). To detect the expression and localization of the VZV ORF61 protein, we fused an HA tag with a flexible glycine-serine linker to the C terminus of ORF61 to generate the VZV(pOka)-61–HA. Final clones were confirmed by restriction fragment length polymorphism analyses, PCR and DNA sequencing. Oligonucleotides used for the mutagenesis of ORF61 are given in Supplementary Table 8. The resulting BACs were transfected into MeWo cells, thereby generating the given recombinant VZVs.

VZVs were propagated in MeWo cells. Monolayers of infected cells were monitored microscopically for cytopathic effect and, when appropriate, mRFP expression. Cells that showed high levels of infection were detached and replated on seeded 70% confluent uninfected cells at appropriate ratios (1:2 to 1:4). Aliquots of a defined count of infected cells were cryopreserved in freezing medium (45% FCS, 10% DMSO), stored in liquid nitrogen and thawed for experiments.

VZV titre was determined by plaque assay: confluent monolayers of MeWo cells were infected for 3 h at 37 °C with serial 10-fold dilutions of 1 million freshly thawed VZV MeWo stock. At 3 h post-infection, the culture medium was changed and cells were kept in culture at 37 °C. For reporter fluorescence virus, RFP-fluorescent-infected foci were counted 2 days post-infection. For non-fluorescent virus, cells were fixed with 4% formaldehyde for 30 min at room temperature 4 days post-infection and stained with crystal violet (1% crystal violet, 10% ethanol) for 10 min at room temperature. Titres were defined by foci per plaque forming units per million cells.

DNA transfection

If not explicitly described, DNA transfection was performed as follows: plasmids were mixed with polyethylenimine (24765; Polysciences) at a DNA:polyethylenimine ratio of 1:3 in Opti-MEM (Gibco Fisher Scientific) for 20 min at room temperature. The DNA:polyethylenimine mix was added to cells and media was exchanged 6 h post-transfection.

Generation of lentivirus

The pLenti6.3 (expressing VZV ORFs or GFP), pWPI (expressing NPHP4 constructs or tagBFP) or pLVX (expressing shRNA) lentiviral expression plasmids, together with the packaging plasmids pCMV-Gag-Pol and pMD2-VSV-G, were transfected in HEK293T cells as described earlier. Viral supernatants were collected 48 h post-transfection, filtered on 0.45 µM PVDF membrane (Fisher Scientific) and stored at −80 °C. The pSicoR-based lentiviral expression plasmids (spCas9 or gRNAs) together with the packaging plasmids pCMVR8.91 and pMD2-VSV-G were mixed with TransIT-LT1 (Mirus Bio) in Opti-MEM (Gibco Fisher Scientific) for 20 min at room temperature. Supernatants were collected 72 h post-transfection and frozen at −80 °C. Lentiviruses were titred according to standard procedure.

Cloning and cell line generation

Generation of SpCas9 and reporter SK-N-BE2 cell lines

SK-N-BE2 cells were transduced with the SpCas9-ZeoR cassette, allowing the expression of a human codon-optimized nuclear-localized Streptococcus pyogenes cas9 gene in the absence of a U6 promoter–sgRNA and the zeocin resistance gene (ZeoR) using a lentivirus generated with the pSicoR-SpCas9-ZeoR⁹⁵ construct and polybrene at 8 µg ml⁻¹. To generate the pWPI-tagBFP-blastR, the tagBFP cassette was amplified by PCR from pSCRPSY-PAC2A-tagBFP and transferred into the lentiviral expression plasmid pWPI-blastR, allowing expression of blasticidin resistance, via BamHI-HF and AscI restriction digest (New England Biolabs). Oligonucleotides used for PCR are given in Supplementary Table 8. SK-N-BE2 wild-type or SK-N-BE2-spCas9(ZeoR) cells were transduced with GFP-blastR or tagBFP-blastR cassettes, respectively, using polybrene at 8 µg ml⁻¹. Cells were selected with 10 µg ml⁻¹ blasticidin for 10 days and maintained with 5 µg ml⁻¹ blasticidin to generate the SK-N-BE2-GFP(blastR) and the SK-N-BE2-spCas9(ZeoR)-tagBFP(blastR) cell lines. SK-N-BE2-spCas9(ZeoR)-tagBFP(blastR) cells were sorted for high BFP expression by flow cytometry (FACSAria III; BD Bioscience).

Cloning the VZV ORF expression plasmid library and generation of VZV ORF-expressing SK-N-BE2 cell lines

The generation of the library of VZV ORF expression plasmids has been described previously¹⁴. Briefly, individual VZV ORFs were amplified by PCR on reversed-transcribed viral RNA extracted from VZV rOka-infected cells, allowing the insertion of an upstream Kozak sequence (GCCGCC) and the removal of the stop codon for subsequent fusion with a C-terminal tag. Fragments were ligated into the pCR8/TOPO entry vector and shuttled via Gateway recombination into the lentiviral expression vector pLenti6.3-TO-V5-DEST-blastR, allowing the fusion of a C-terminal V5 tag and the expression of blasticidin resistance. pLenti6.3-TO-V5-GFP-blastR was used as control. SK-N-BE2 cells (70% confluent) were transduced with individual lentiviruses using polybrene at 8 µg ml⁻¹. Cells were selected with 8 µg ml⁻¹ blasticidin and expanded for at least 10 days and collected for MS analysis or kept as frozen stock. VZV ORF2, ORF22, ORF27, ORF31, ORF42-45 and ORF62 are not included in this study. Oligonucleotides used for VZV ORF amplification from the viral genome are available in ref. ¹⁴.

Cloning of the host gene gRNA expression plasmid library

Two to four gRNA sequences per gene were designed using the CRISPOR online tool (http://crispor.tefor.net). Forward and reverse oligonucleotides, allowing the generation of BsmBI overhang, were purchased from IDT. Each complementary oligonucleotide (100 µM) was annealed in annealing buffer (1 mM EDTA, 50 mM NaCl in 10 mM Tris pH 8.0) by denaturation at 95 °C, followed by progressive cooling down to 5 °C. Annealed oligos were ligated into BsmBI.v2-digested (New England Biolabs) pSicoR-U6-gRNA-EFS-PuroR with T4 DNA ligase (Thermo Fisher Scientific). gRNA sequences are listed in Supplementary Table 8.

Cloning of UBXN7 shRNA and generation of HFF knockdown cells

Scramble (control) and UBXN7-targeting shRNA sequences were identified in the MISSION TRC library (Sigma-Aldrich). Forward and reverse oligonucleotides, allowing shRNA expression with overhangs for cloning into BamHI and EcoRI sites, were purchased from Eurofins. Each complementary oligonucleotide (100 µM) was annealed in annealing buffer (1 mM EDTA, 50 mM NaCl in 10 mM Tris pH 8.0) by denaturation at 95 °C, followed by progressive cooling down to 5 °C. Annealed oligonucleotides were ligated into BamHI-EcoRI-digested (New England Biolabs) pLVX with T4 DNA ligase (Thermo Fisher Scientific). Final constructs were confirmed by DNA sequencing. Oligonucleotides are available in Supplementary Table 8. HFF cells were transduced with the respective shRNA and selected with 1.5 µg ml⁻¹ puromycin after 2 days. Cells were used for experiments after another 3 days in culture.

Generation of stable knockout SK-N-BE2 cell lines

Stable knockout SK-N-BE2 cell lines for selected host genes were generated to replicate the effect on VZV spread observed in the knockout screen and to perform functional assays. SK-N-BE2-spCas9(ZeoR)-BFP(blastR) cells were transduced with the following host gene sgRNA, selected from the same library designed for the knockout screen, using polybrene at 8 µg ml⁻¹: pooled sgRNA targeting MPP8 or ZNF280D, an sgRNA targeting NPHP4 (AAGGCTGGCGCGCTCTCTGT) or an empty vector as NTC. After 2 days, cells were selected with 2 µg ml⁻¹ puromycin and kept in culture for another 12 days to allow efficient knockout and to increase the chances of clearing the remaining expressed protein. Efficient knockout was validated relative to the NTC by genotyping (the targeted regions were amplified by PCR, sequenced and analysed by Synthego ICE analysis⁹⁸, WB and MS analysis).

Cloning of HA-tagged expression cassettes

ORF61 and GFP cassettes were inserted into pDONR221 by PCR from pLenti6.3 expression plasmids generated as described earlier. The NPHP4, ORF61, ICP0 and GFP cassettes were shuttled from pENTR or pDONR into the lentiviral expression plasmid pWPI-nHA-PuroR-GW by Gateway recombination. pWPI-nHA-PuroR-NPHP4(S862N) encoding for the expression of the NPHP4 variant was generated by two-step PCR mutagenesis from the wild-type construct. In brief, the 5′ and 3′ fragments of NPHP4 were amplified by PCR using the Phusion Hot Start II DNA Polymerase (Thermo Fisher Scientific) and the primers A and B for 5′ fragments and C and D for 3′ fragments. Primers were designed to allow for the incorporation of the c.2585G>A mutation, complementarity within their respective 3′ and 5′ ends, and the insertion of the NdeI restriction site before the start codon and the SpeI site after the stop codon. The complete gene cassette was then amplified by a second PCR using the two fragments as templates and the primers A and D. The resulting product was then shuttled into pWPI-nHA-PuroR using NdeI and SpeI-HF restriction digests (New England Biolabs), followed by T4 DNA ligation (Thermo Fisher Scientific).

Final constructs were confirmed by DNA sequencing.

Sample preparation and analysis of MS-based proteomic and transcriptomic experiments

Host proteome changes induced by VZV infection of SK-N-BE2 cells

As VZV is a highly cell-associated pathogen that releases few infectious particles into the extracellular media^99,100, SK-N-BE2 cells were infected by co-culture with VZV rOka-infected MeWo inoculum cells using a transwell system to prevent contamination with inoculum cells³¹. A ratio of 2:1 or 5:1 of uninfected:infected MeWo cells, depending on the virus lot, was seeded on 1 side of porous transwells (insert 6-well, PET, 1 µm pores; Sarstedt) in culture medium. After 24 h, the transwell was turned upside down into a six-well plate filled with culture medium, and SK-N-BE2 cells were seeded on the second side of the transwell in culture medium. At 48 h post-infection, SK-N-BE2 cells were scraped from the transwell membrane, washed in ice-cold PBS, lysed in SDS lysis buffer (50 mM TRIS-HCl pH 7.6, 10 mM DTT, 4% SDS), boiled for 5 min at 95 °C, flash frozen and kept at −80 °C. Frozen pellets were thawed, sonicated (5 min, 4 °C, 30 s on and off, high frequency; Bioruptor; Diagenode), boiled for 5 min at 95 °C and alkylated with 55 mM iodoacetamide for 20 min at room temperature. Proteins were precipitated and cleaned from SDS by adding 4 volumes of acetone and incubating for 2 h at −20 °C. The pellet was resolubilized in denaturation buffer (6 M urea, 2 M thiourea, 10 mM HEPES pH 8.0) and frozen at −20 °C. Total proteins (50 µg) were thawed and predigested with 1 µg LysC (Wako Chemicals) at room temperature for 4 h, followed by a 1:5 dilution in ABC buffer (50 mM NH₄HCO₃, 100 mM Tris-HCl pH 8) and digested for 15 h at 30 °C with 1 µg trypsin (Sigma-Aldrich). The digest was stopped and the peptides were solubilized by the addition of 0.6% trifluoroacetic acid (TFA) and 2% acetonitrile (ACN). Samples were spin-centrifuged, and the cleared peptide supernatant was transferred into new tubes before peptide purification. Peptides were purified on StageTips with 3 layers of C18 Empore filter discs (3M). MS analysis was performed on an EASY-nLC 1200 system (Thermo Fisher Scientific), coupled with the mass spectrometer (Q Exactive HF-X; Thermo Fisher Scientific) via a nano-electrospray source as previously described²⁵. Briefly, peptides were eluted on a 50 cm reverse-phase analytical column (75 µm diameter; ReproSil-Pur C18-AQ 1.9 µm resin; Dr. Maisch) using a gradient of ACN in 0.1% formic acid at a flow rate of 300 nl min⁻¹ (sequential linear gradients of 80% ACN: 5–30% for 95 min, 30–60% for 5 min, 60–95% for 5 min, followed by a stationary step at 95% for 5 min to elute the most hydrophobic peptides and re-equilibration of the column at 5%). To avoid carryover of remaining peptides across conditions, the column was washed with 95% of 80% ACN for 15 min between quadruplicates. The mass spectrometer was operated and MS spectra acquired using the XCalibur software (Thermo Fisher Scientific) with data-dependent acquisition (DDA) mode. Full MS scans (300–1,650 m/z, resolution (R) = 60,000) were acquired at an ion target of 3 × 10⁶. The top 15 most abundant precursor peptides were fragmented by higher-energy collisional dissociation (HCD) with a normalized collision energy of 27% and MS/MS scan (R = 15,000) acquired at an ion target of 1 × 10⁵ and a maximum injection time of 25 ms. Isolation and fragmentation of the same peptide precursor were eliminated by dynamic exclusion for 20 s. The whole process (infection, sample preparation and MS measurement) was repeated twice, each with four independent experiments.

Host proteome changes and AP of SK-N-BE2 cells expressing V5-tagged VZV proteins

SK-N-BE2 cells transduced with individual expression cassettes for VZV ORF or GFP fused to the C-terminal V5 tag were expanded in quadruplicates to reach 2 confluent 15 cm dishes per replicate. Control cell lines (GFP, ORF60 and ORF66) were expanded to reach 16 replicates. Cells were gently washed in ice-cold PBS, scraped, pooled per replicate and washed twice in ice-cold PBS by centrifugation at 600g at 10 °C for 10 min. Before the last wash, an aliquot of 1 × 10⁶ cells from each replicate was kept for full proteome analysis. All samples were flashed frozen in liquid nitrogen before storage at −80 °C.

AP of V5-tagged VZV ORF

Samples were processed in 3 immunoprecipitation (IP) batches of 19 VZV ORFs. To account for batch effect, the three controls (GFP, ORF60 and ORF66) were included within each batch. Frozen cell pellets were thawed and lysed on ice for 30 min in lysis buffer (0.2% NP-40, 100 mM NaCl, 5% glycine, 1.5 mM MgCl₂, 50 mM Tris-HCl pH 7.5) supplemented with 1% in-house benzonase and EDTA-free Complete Protease Inhibitor (Roche). Samples were sonicated (5 min, 4 °C, 30 s on and off, high frequency; Bioruptor; Diagenode) and centrifuged at 15,000g at 4 °C for 30 min. Supernatants were collected in 96-well deep-well plates with randomized positions to minimize plate position effects. Total protein concentrations were measured by Pierce 660 nm Protein Assay (Thermo Fisher Scientific) and normalized to 6 mg in 750 µl lysis buffer supplemented with EDTA-free Complete Protease Inhibitor (Roche) during transfer into 4 24-well deep-well plates. Cleared lysates were mixed with 30 µl of anti-V5 magnetic bead slurry (MBL M215-11), previously equilibrated in lysis buffer, and agitated for 2 h at 4 °C. After incubation, the samples were transferred into a 96-well deep-well plate. To favour intra- and inter-batch reproducibility, IPs were automatized on a Freedom EVO 200 robotic platform (Tecan) equipped with an 8-needle liquid handling station, a plate magnet position and a plate shaker. Immune complexes attached to the magnetic beads were allowed to collect on the magnet for 5 min. The flow-throughs were aspirated, the plate was moved to the shaker position, 480 µl of lysis buffer was added for washing and the plate was agitated at 1,200 rpm for 2 min. The wash was repeated six times in lysis buffer to reduce unspecific binding, followed by eight additional washes in wash buffer (lysis buffer without NP-40) to eliminate remaining detergents. Excess buffer was removed while the plate was positioned on the magnet. Beads were resuspended in 20 µl of 1:10 diluted guanidinium chloride buffer (0.6 M GdmCl, 1 mM tris (2-carboxyethyl)phosphine (TCEP), 4 mM chloroacetamide (CAA) in 0.1 M Tris-HCl pH 8.0) to allow for denaturation, reduction, and alkylation of the enriched proteins, and then transferred into a 96-well microplate and frozen at −20 °C. To prevent further batch effects, protein digest and peptide purification of the three IP batches were processed simultaneously. Sample plates were thawed at room temperature and predigested with 1 µg LysC (Wako Chemicals) for 4 h at 37 °C, followed by a 1:5 dilution in 0.1 M Tris-HCl pH 8.0 and digestion for an additional 15 h at 30 °C with 1 µg trypsin (Sigma-Aldrich). The digestion was stopped and peptides solubilized in 0.6% TFA and 2% ACN. Beads were sedimented by using the plate magnet, peptides were transferred onto a new 96-well microplate and frozen at −20 °C. Peptides were purified on StageTips with 3 layers of C18 Empore filter discs (3M). MS analysis was performed on an EASY-nLC 1200 system (Thermo Fisher Scientific), coupled online with the mass spectrometer (Q Exactive HF-X; Thermo Fisher Scientific) via a nano-electrospray source as previously described²⁵. Briefly, peptides were eluted on a 20 cm reverse-phase analytical column (75 µm diameter; ReproSil-Pur C18-AQ 1.9 µm resin; Dr. Maisch) using a gradient of ACN in 0.1% formic acid at a flow rate of 300 nl min⁻¹ (sequential linear gradients of 80% ACN: 5–30% for 85 min, 30–60% for 12 min, 60–80% for 3 min, and 80–95% for 1 min, followed by a stationary step at 95% for 5 min to elute the most hydrophobic peptides and re-equilibration of the column at 5%). As samples were measured in a randomized order, the column was washed with 95% of 80% ACN for 15 min after each run to avoid carryover of peptides between samples. The mass spectrometer was operated and MS spectra acquired using the XCalibur software (Thermo Fisher Scientific) with DDA mode. Full MS scans (300–1,650 m/z, R = 60,000) were acquired at an ion target of 3 × 10⁶. The top 15 most abundant precursor peptides were fragmented by HCD with a normalized collision energy of 27% and MS/MS scan (R = 15,000) acquired at an ion target of 1 × 10⁵ and a maximum injection time of 25 ms. Isolation and fragmentation of the same peptide precursor were eliminated by dynamic exclusion for 20 s.

Full proteome

Frozen pellets of 1 × 10⁶ cells were thawed on ice and lysed in guanidinium chloride lysis buffer (6 M GdmCl, 10 mM TCEP, 40 mM CAA in 0.1 M Tris-HCl pH 8.0) for 30 min. Samples were boiled at 99 °C with shaking at 500 rpm for 15 min and then sonicated (5 min, 4 °C, 30 s on and off, high frequency; Bioruptor; Diagenode). Supernatants were collected after centrifugation at 15,000g at 4 °C for 30 min. Total proteins (50 µg) were predigested with 1 µg LysC (Wako Chemicals) for 3 h at 37 °C, followed by a 1:5 dilution in 0.1 M Tris-HCl pH 8.0 and digestion for an additional 15 h at 30 °C with 1 µg trypsin (Sigma-Aldrich). The digest was stopped and the peptides solubilized by the addition of 0.6% TFA and 2% ACN. Samples were spin-centrifuged, and the cleared peptide supernatant was transferred into new tubes before peptide purification. Peptides were purified on StageTips with three layers of C18 Empore filter discs (3M). MS analysis was performed on an EASY-nLC 1200 system (Thermo Fisher Scientific), coupled online with the mass spectrometer (Q Exactive HF-X; Thermo Fisher Scientific) via a nano-electrospray source as previously described²⁵. Briefly, peptides were eluted on a 50 cm reverse-phase analytical column (75 µm diameter; ReproSil-Pur C18-AQ 1.9 µm resin; Dr. Maisch) using a gradient of ACN in 0.1% formic acid at a flow rate of 300 nl min⁻¹ (sequential linear gradients of 80% ACN: 5–30% for 95 min, 30–60% for 5 min, 60–95% for 5 min, followed by a stationary step at 95% for 5 min to elute the most hydrophobic peptides and re-equilibration of the column at 5%). To avoid carryover of peptides across conditions, the column was washed with 95% of 80% ACN for 15 min between quadruplicates. The mass spectrometer was operated and MS spectra acquired using the XCalibur software (Thermo Fisher Scientific) with DDA mode. Full MS scans (300–1,650 m/z, R = 60,000) were acquired at an ion target of 3 × 10⁶. The top 15 most abundant precursor peptides were fragmented by HCD with a normalized collision energy of 27% and MS/MS scan (R = 15,000) acquired at an ion target of 1 × 10⁵ and a maximum injection time of 25 ms. Isolation and fragmentation of the same peptide precursor were eliminated by dynamic exclusion for 20 s.

Proteome changes induced by MPP8 gene depletion in SK-N-BE2 cells

One million MPP8-knockout or NTC SK-N-BE2-spCas9(ZeoR)-BFP(blastR) cells were collected in triplicate, and the pellets were flash frozen in liquid nitrogen for subsequent MS analysis of the full proteome. Frozen cell pellets were thawed on ice and lysed in guanidinium chloride lysis buffer (6 M GdmCl, 10 mM TCEP, 40 mM CAA in 0.1 M Tris-HCl pH 8.0) for 30 min. Samples were boiled at 99 °C with shaking at 500 rpm for 15 min and then sonicated (5 min, 4 °C, 30 s on and off, high frequency; Bioruptor; Diagenode). Supernatants were collected after centrifugation at 15,000g at 4 °C for 30 min. Total proteins (50 µg) were predigested with 1 µg LysC (Wako Chemicals) for 3 h at 37 °C, followed by a 1:5 dilution in 0.1 M Tris-HCl pH 8.0 and digestion for an additional 15 h at 30 °C with 1 µg trypsin (Sigma-Aldrich). The digest was stopped and the peptides solubilized by the addition of 0.6% TFA and 2% ACN. Samples were spin-centrifuged, and the cleared peptide supernatant was transferred into new tubes before peptide purification. Peptides were purified on StageTips with three layers of C18 Empore filter discs (3M). MS analysis was performed on an EASY-nLC 1200 system (Thermo Fisher Scientific), directly coupled online with the mass spectrometer (Q Exactive HF-X; Thermo Fisher Scientific) via a nano-electrospray source as previously described²⁵. Briefly, peptides were eluted on a 50 cm reverse-phase analytical column (75 µm diameter; ReproSil-Pur C18-AQ 1.9 µm resin; Dr. Maisch) using a gradient of ACN in 0.1% formic acid at a flow rate of 300 nl min⁻¹ (sequential linear gradients of 80% ACN: 5–30% for 150 min, 30–60% for 5 min, 60–95% for 5 min, followed by a stationary step at 95% for 5 min to elute the most hydrophobic peptides and re-equilibration of the column at 5%). To avoid carryover of remaining peptides across conditions, the column was washed with 95% of 80% ACN for 15 min between quadruplicates. The mass spectrometer was operated and MS spectra acquired using the XCalibur software (Thermo Fisher Scientific) with DDA mode. Full MS scans (300–1,650 m/z, R = 120,000) were acquired at an ion target of 3 × 10⁶. The top 15 most abundant precursor peptides were fragmented by HCD with a normalized collision energy of 27% and MS/MS scan (R = 15,000) acquired at an ion target of 1 × 10⁵ and a maximum injection time of 25 ms. Isolation and fragmentation of the same peptide precursor were eliminated by dynamic exclusion for 20 s.

Transcriptome changes induced by NPHP4 gene depletion in VZV-infected SK-N-BE2 cells

Sample preparation and sequencing

SK-N-BE2 cells, control or depleted for NPHP4, were infected by co-culture with MeWo cells infected with VZV(pOka)-63-RFP/70-RFP using a transwell system to prevent contamination with inoculum cells³¹, as described earlier, with a ratio of 3:1 of uninfected:infected MeWo. At 48 h post-infection, SK-N-BE2 cells were scraped from the transwell membrane, washed in ice-cold PBS, lysed in LBP buffer (Macherey-Nagel), flash frozen and kept at −80 °C. RNA was extracted according to the supplier’s recommendation (Macherey-Nagel). Bulk sequencing of poly(A)-RNA was done as previously described¹⁰¹. Barcoded cDNA of each sample was generated using a Maxima RT polymerase (Thermo Fisher) using oligo(dT) primer containing barcodes, unique molecular identifiers (UMIs) and an adaptor. 5′ ends of the cDNAs were extended by a template switch oligonucleotide, and full-length cDNA was amplified with primers binding to the template switch oligonucleotide site and the adaptor. The NEB UltraII FS kit was used to fragment cDNA. After end repair and A-tailing, a TruSeq adaptor was ligated and 3′ end fragments were amplified using primers with Illumina P5 and P7 overhangs. The library was sequenced on a NextSeq 500 (Illumina) with 61 cycles for the cDNA in read 1 and 19 cycles for the barcodes and UMIs in read 2.

AP of SK-N-BE2 cells expressing HA-tagged NPHP4 wild type or S862N variant

SK-N-BE2 cells transduced with NPHP4 wild type, the S862M mutant or GFP fused to the N-terminal HA tag were expanded in quadruplicates to reach 2 confluent 15 cm dishes per replicate. Cells were gently washed in ice-cold PBS, scraped, pooled per replicate and washed twice in ice-cold PBS by centrifugation at 600g at 10 °C for 10 min. All samples were flash frozen in liquid nitrogen before storage at −80 °C. Frozen cell pellets were thawed and lysed on ice for 30 min in 1 ml lysis buffer (0.2% NP-40, 100 mM NaCl, 5% glycine, 1.5 mM MgCl₂, 50 mM Tris-HCl pH 7.5) supplemented with 1% in-house benzonase and EDTA-free Complete Protease Inhibitor (Roche). Samples were sonicated (5 min, 4 °C, 30 s on and off, high frequency; Bioruptor; Diagenode) and centrifuged at 15,000g at 4 °C for 30 min. Total protein concentrations were measured by Pierce 660 nm Protein Assay (Thermo Fisher Scientific) and normalized to 2 mg in 1 ml lysis buffer supplemented with EDTA-free Complete Protease Inhibitor (Roche). Cleared lysates were mixed with 40 µl of anti-HA agarose bead slurry (A2095; Sigma-Aldrich), previously equilibrated in lysis buffer, and agitated for 3 h at 4 °C. Immune complexes attached to the beads were washed 5× with 1 ml lysis buffer and 4× in 1 ml wash buffer (lysis buffer without NP-40) to eliminate remaining detergents. Excess buffer was removed and beads were resuspended in 20 µl of 1:10 diluted guanidinium chloride buffer (0.6 M GdmCl, 1 mM TCEP, 4 mM CAA in 0.1 M Tris-HCl pH 8.0) to allow denaturation, reduction, and alkylation of the enriched proteins, and then frozen at −20 °C. Samples were thawed at room temperature and predigested with 0.5 µg LysC (Wako Chemicals) for 3 h at 37 °C, followed by a 1:4 dilution in 0.1 M Tris-HCl pH 8.0 and digestion for an additional 16 h at 30 °C with 0.5 µg trypsin (Sequencing Grade; Promega). The digest was stopped and peptides solubilized in 0.6% TFA and 2% ACN. Beads were sedimented by spin centrifugation for 5 min at 10,000 rpm, peptides were transferred into new tubes, processed by StageTip purification with 3 layers of C18 Empore filter discs (3M) and resuspended in 2% ACN, 0.1% TFA. MS analysis was performed on an EASY-nLC 1200 system (Thermo Fisher Scientific), coupled online with the mass spectrometer (Q Exactive HF-X; Thermo Fisher Scientific) via a nano-electrospray source as previously described²⁵. Briefly, peptides were eluted on a 20 cm reverse-phase analytical column (75 µm diameter; ReproSil-Pur C18-AQ 1.9 µm resin; Dr. Maisch) using a gradient of ACN in 0.1% formic acid at a flow rate of 300 nl min⁻¹ (sequential linear gradients of 80% ACN: 5–30% for 37 min, 30–60% for 6 min, 60–80% for 3 min, and 80–95% for 1 min, followed by a stationary step at 95% for 5 min to elute the most hydrophobic peptides and re-equilibration of the column at 5%). The column was washed with 95% of 80% ACN for 15 min between each quadruplicate. The mass spectrometer was operated and MS spectra acquired using the XCalibur software (Thermo Fisher Scientific) with DDA mode. Full MS scans (300–1,650 m/z, R = 60,000) were acquired at an ion target of 3 × 10⁶. The top 15 most abundant precursor peptides were fragmented by HCD with a normalized collision energy of 27% and MS/MS scan (R = 15,000) acquired at an ion target of 1 × 10⁵ and a maximum injection time of 25 ms. Isolation and fragmentation of the same peptide precursor were eliminated by dynamic exclusion for 20 s.

Data processing and bioinformatic analysis

Host full proteome changes induced by VZV infection of SK-N-BE2 cells

Data processing

The raw MS data files were analysed with MaxQuant (v.1.6.0.15) using the standard settings and label-free quantification (LFQ) with ‘match between runs’ enabled (‘LFQ min ratio count’ set to 2 and ‘stabilization of large LFQ ratios’ enabled). The MS2 spectra were searched against forward and reverse (used as decoy) sequences of the reviewed human proteome, including isoforms (UniprotKB, release 2018.02) and VZV proteins (pOka; UniprotKB, release 2017.12), using the built-in Andromeda search engine.

Statistical analysis

LFQ intensities of protein groups were imported from MaxQuant (proteinGroup file) into Perseus (v.1.6.5.0) for statistical analysis of infected proteome changes. Data from two repeats, each including independent quadruplicate mock and infected samples, were analysed. Contaminants, reverse and only-by-site identifications were removed, as were protein groups not quantified in at least three of four samples in at least one condition. Missing intensities were imputed by random sampling from a normal distribution N(I₀ − 1.8σ, (0.3σ)²), where I₀ and σ are the mean and s.d. of total measured log₂ intensities, respectively. Log-transformed fold changes were calculated as the log₂ difference of the median intensity within infected samples over mock samples for each repeat. The significance was assessed by the Student’s t-test with permutation-based multiple hypothesis testing correction. Host proteins with absolute log₂ difference ≥ 0.5 and adjusted P ≤ 0.05 in both repeats were reported as strong hits (the direction of the change in both sets also had to match). Gene set enrichment analysis was performed via one-sided Fisher’s exact tests on gene ontologies from the GO collections, adjusted by Benjamini–Hochberg false discovery rate (FDR). Terms enriched with an enrichment factor ≥4 and an adjusted P ≤ 0.05 were considered to be significant. The least-enriched terms that were redundant between GO collections were excluded.

AP of V5-tagged VZV proteins in SK-N-BE2 cells (interactomes)

Data processing

The raw MS data files of the AP–MS experiments, full proteome of SK-N-BE2 cells and six fractions from deep coverage of the SK-N-BE2 cell proteome were analysed with MaxQuant (v.1.6.0.15) using the standard settings and fast LFQ enabled (‘LFQ min ratio count’ set to 2, ‘stabilize large LFQ ratios’ disabled and normalization skipped). The AP–MS and fractionated SK-N-BE2 deep proteomic samples were assigned to different parameter groups with match between runs enabled. The MS2 spectra were searched against forward and reverse (used as decoy) sequences of the reviewed human proteome, including isoforms (UniprotKB, release 2018.02), VZV proteins (pOka; UniprotKB, release 2017.12), the GFP protein and the V5-tag sequence.

Statistical analysis

R (v.3.6), Julia (v.1.5) and Python (v.3.8) using a collection of in-house scripts¹⁰² were used in this analysis.

MaxQuant output files were imported into R using the in-house msimportr R package¹⁰³. A Bayesian linear mixed effects model, implemented in the msglm R package⁴³ was used to estimate the enrichment of protein groups in the AP–MS experiments with viral baits. In R generalized linear model (GLM) formula language, the model could be specified as

where intensity is the LFQ intensity of a given protein group in a given sample, the APMS effect corresponds to the average shift of protein group intensity (enrichment) in AP–MS data in comparison to the full SK-N-BE2 proteome, and the interaction effect APMS:bait_i models the enrichment of a protein group specific to the viral bait of the ith MS replicate; the effects MSbatch_i and BioChemBatch_i account for the protein intensity variations specific to the MS measurement and the AP batches of the ith sample, respectively. An additional continuous batch effect, PCAbatch_i, corresponds to the specific protein contamination pattern recurring across the AP samples with varying intensity. The pattern was associated with the second principal component of the protein × sample log-intensities matrix, and we used the following formula to define the intensity of the contamination in the ith sample:

if w_i > median(w) + max(w), and 0 otherwise, where w_i is the weight of the second principal component in the ith sample.

Due to the inherent sparsity of AP data, the effects of the model associated with experimental conditions had horseshoe priors¹⁰⁴. The MSGLM model assumes that the measurement error of the MS instrument follows a Laplace distribution, and the parameters of this distribution depend on the signal intensity (heteroscedastic intensities noise model). The MSGLM noise model was calibrated with technical replicates of our MS instrument. The same MS data were also used to calibrate the missing data model that defines the probability of protein identification as a logit-transformed expected abundance. The model was applied to unnormalized LFQ intensities with per-MS-run normalization multipliers, matching the predictions to the expected intensity of a given MS run. This scheme allows the model likelihood to automatically account for the quality of each individual sample. The MSGLM model was applied to each protein group separately. To infer the posterior distribution of model parameters, 4,000 iterations (2,000 warm-up and 2,000 sampling iterations) of the no-U-turn Markov chain Monte Carlo method, implemented using the rstan package¹⁰⁵ (v.2.19), were run in 8 independent chains, with every 4th collected sample.

The modelling of MS data provided the enrichment estimates for each protein in each AP experiment cleared from the batch effects. The specific AP–MS interactions between a viral bait and a protein had to be significantly enriched relative to the background distribution of that protein, calculated from its abundance in all the other baits, with the 25% least abundant and 10% most abundant viral baits removed. The difference between the medians of the bait-specific and background posterior distribution of protein log₂ intensity had to be ≥1 and with P ≤ 0.01. The P value was calculated as the probability that a random sample from the posterior distribution of the protein abundance in the viral bait AP–MS experiment would be smaller than a random sample drawn from the background distribution. No P value adjustment was done as multiple-hypothesis correction is handled by the choice of Bayesian priors. Proteins with the median estimated effect size of the PCAbatch term above 2.5 and P ≤ 10⁻⁵ were annotated as ‘putative contaminants’. Proteins identified by only one peptide needed to be detected in at least three of the four independent experiments, and identified in at least two experiments by MS/MS. Multiple isoforms identified within a single VZV bait interactome were combined under the canonical prey name, keeping the enrichment and P value of the interaction quantified with the highest number of peptides. To exclude possible contamination due to bait overexpression effects on the background proteome, host preys detected as upregulated in the effectome data for the given bait (see below), with a median log₂ difference smaller than four times the AP–MS enrichment, were filtered out.

Host full proteome changes induced the expression of V5-tagged VZV proteins in SK-N-BE2 cells (effectomes)

Data processing

The raw MS data files of the effectome experiments were analysed with MaxQuant (v.1.6.14.0) using the standard settings and fast LFQ with match between runs (LFQ min ratio count 2, classic normalization and stabilize large LFQ ratios enabled). The MS2 spectra were searched against forward and reverse (used as decoy) sequences of the reviewed human proteome (UniprotKB, release 2019.12), VZV proteins (pOka; UniprotKB, release 2017.12), the GFP protein and the V5 tag independently using the built-in Andromeda search engine.

Statistical analysis

Protein groups and LFQ intensities generated by MaxQuant (proteinGroups file) were imported into R Studio (v.2022.07.2+576). MS runs that did not reach a satisfactory performance were excluded. This led to the removal of the data for 18 VZV ORFs. Runs were grouped per batch corresponding to MS measurement sessions over the course of time. VZV ORF33.5 runs, being the only satisfactory runs in their batch, were associated with another batch with a similar background. Protein groups were filtered for contaminants and reverse identifications. Missing intensities were imputed from a log normal distribution of log-transformed values with a width of 0.3× the s.d. of the measured intensities and downshifted by 1.8 s.d. to simulate the detection limit of the mass spectrometer. VZV ORF-specific effects were calculated as the log₂ difference of the median intensity within the proteome of the given bait and its background. The background was defined as the median intensity within all other ORFs of the batch with the 10% least abundant and 10% most abundant ORFs removed. Significance was assessed using the Wilcoxon rank-sum test, adjusted by the Benjamini–Hochberg approach. Absolute log₂ difference ≥ 0.5 and adjusted P ≤ 0.05 were used to define a significant effect. An absolute log₂ difference ≥ 1 was used to define ‘strong’ significant effects, which were used as input for the network diffusion analysis. Finally, ‘high confident’ significant effects are defined as an adjusted P ≤ 0.01 and an identification by MS/MS in 3 of 4 of the replicates within the group showing higher abundance (that is, within the VZV ORF replicates if upregulated and within the background if downregulated) on proteins identified by at least 3 peptides.

Systematic subcellular localization analysis of the interactome preys

Host preys identified in the VZV–host interactome datasets were annotated using the GO Cellular Component terms, curated UniProt subcellular locations and both main and additional locations from the Protein Atlas. Each prey’s localizations were summarized within the following categories using a keyword search: nucleus (‘nucl’), cytosol (‘cytosol’), Golgi apparatus (‘golgi’), endoplasmic reticulum (‘reticulum’), mitochondria (‘mitochon’), cytoskeleton (‘actin’, ‘microtubule’ and ‘cytoskelet’) and cell membrane (‘plasma’ and ‘cell membrane’).

Gene set enrichment analysis of the V5-tagged VZV proteins interactome and effectome

We used EnrichmentMap gene sets of human proteins (v.2020.10)¹⁰⁶. For interactomes analysis, we used GO (v.2020.10). For effectome analysis, we used GO (v.2021.12) and protein complex annotations from IntAct Complex Portal (v.2019.12)¹⁰⁷ and CORUM (v.2019)¹⁰⁸ and transcriptional interactions from OmniPath (v.2021.06)¹⁰⁹.

To find the non-redundant collection of annotations describing the unique and shared features of multiple experiments in a dataset, we used the in-house Julia package OptEnrichedSetCover¹¹⁰, which uses an evolutionary multi-objective optimization technique to find a collection of annotation terms that have both significant enrichments in the individual experiments and minimal pairwise overlaps. The resulting set of terms was further filtered by requiring that the annotation term has to be significant with the specified unadjusted Fisher’s exact test P value cut-off in at least one of the experiments or comparisons.

For interactomes, GO Cellular Component terms with P ≤ 0.0001 were retained and only 1 of the redundant terms (same intersect_genes within ORF) was retained, with priority given to that with the lowest P value (Fig. 2b). The analysis of the effectomes was performed on all significant effects (absolute log₂ difference ≥ 0.5 and adjusted P ≤ 0.05). Terms from GO Biological Processes, protein complexes and transcriptional interactions enriched with P ≤ 0.001 were kept. The terms enriched within the two ORF66 batch controls were merged. For clarity of the representative figure, only one of the redundant terms (same intersect_genes within ORF) was kept, with priority given to the term enriched in other ORF (Fig. 3b).

Integration of the interactome and effectome data by network diffusion analysis

To integrate the AP–MS viral–host protein interactions and the virus protein-induced protein changes, we used the HierarchicalHotNet-based method⁷¹, as we have previously done for the severe acute respiratory syndrome coronavirus 2 (ref. ²⁵), with a few modifications as described later. The HierarchicalHotNet.jl v.1.1 Julia Package¹¹¹ and in-house script¹⁰² were used for network diffusion and the following analysis of predictions of statistical significance. The analysis was based on the ReactomeFI network of gene functional interactions (v.2019)¹¹². Instead of directly using the weighted graph of gene–gene random walk transitions, we used the results of this random walk to weight the edges of the ReactomeFI network:

where r is the random walk restart probability, w_i is the probability to restart the random walk from the ith gene, S is the weighted adjacency matrix of the gene functional interactions (ReactomeFI) and W is the matrix of random walk transition probabilities. The proteins with significant abundance changes upon bait overexpression (|median(log₂(fold change))| ≥ 1.0, Benjamini–Hochberg adjusted P ≤ 0.05) (strong significant effects) were used as the sources of signal diffusion, with node weights set to \({w}_{i}=w({n}_{i})\)\(=\sqrt{{|{\rm{median}}}({\log }_{2}({\rm{fold}}\; {\rm{change}}))|\times |{\log }_{10}{P|}}\), otherwise the node weight was set to zero. The weight of the edge g_i → g_j, between a pair of genes g, was set to w_i,j = 1 + w(n_j) if |median(log₂(fold change))| ≥ 0.25, unadjusted P ≤ 0.001, otherwise w_i,j = 1. The random walk restart probability was set to 0.4 as previously described²⁵. To define the optimal subnetwork, we applied the same random permutation-based procedure as in ref. ²⁵, which searches for the optimal edge weight threshold t_*, which maximizes the weighted difference in the average (avg) inverse path length (\({L}_{\rm{avg}}^{-1}(t)\)) between the real data and permuted (perm) data-based analysis:

where N₅(t) is the number of nodes in the 5 largest components of the subnetwork with the minimal edge weight threshold t.

To assess the significance of the edge presence in the resulting network, we calculated the edge P value as the probability that its weight in the permuted-data-based analysis would be higher than in the one based on real data:

This P value was stored as the ‘prob_perm_walkweight_greater’ edge attribute. The specific subnetworks predicted by the network diffusion were filtered for edges with P ≤ 0.05. Effects that are part of the functional subnetwork are shown as big squares, interactors that are connected to these effects are shown as circles, and intermediate proteins that are functionally connected but not affected at their expression level or identified as an interactor are shown as small squares. The networks were exported to the GraphML format and prepared for publication using yEd software (v.3.20; https://www.yworks.com). The catalogue of the networks for each viral bait is available as Supplementary Data 1.

Assembling public interactions

To compare the VZV AP–MS interactome with the known interactomes of VZV and HSV-1, we assembled the known viral interactions from BioGRID, IntAct, VirHostNet (v.2020.12; see Supplementary Table 7 for the taxonomy and molecular interactions identifiers considered) and interactions manually extracted from publications^15,113. We mapped the herpesvirus proteins to the homologous VZV proteins according to gene and locus names (Supplementary Table 2).

Full proteome changes induced by the depletion of the MPP8 gene in SK-N-BE2 cells

Data processing

The raw MS data files were analysed with MaxQuant (v.1.6.14.0) using the standard settings and LFQ with match between runs enabled (LFQ min ratio count 2 and stabilize large LFQ ratios enabled). The MS2 spectra were searched against forward and reverse (used as decoy) sequences of the reviewed human proteome (UniProtKB, release 2019_12) by the built-in Andromeda search engine.

Statistical analysis

R (v.4.1) and Julia (v.1.6) were used.

The MaxQuant output files were imported into R using the in-house msimportr R package¹⁰³. A Bayesian linear random effects model, implemented in the msglm R package¹¹⁴, was used to estimate the change in abundance of protein groups in the knockout experiment. In R GLM formula language, the model could be specified as:

where intensity is the intensity of a given protein group and KO models the average shift of intensity in the knockout samples in comparison to the control samples.

The peptide-based model was applied to unnormalized MS1 peak intensities of each protein group (evidence.txt table of MaxQuant output), using MS-run-specific normalization multipliers to scale the inferred abundance. The KO effect has regularized horseshoe+ priors¹¹⁵. As described in ‘AP of V5-tagged VZV proteins in SK-N-BE2 cells (interactomes)’—‘Statistical analysis’, we assumed a measuring error of the instrument and a logit-based probability of missing MS data. To infer the posterior distribution of the model parameters, 4,000 iterations (2,000 warm-up and 2,000 sampling iterations) of the no-U-turn Markov chain Monte Carlo method, implemented in the cmstanr package (v.0.4.0), were run in 8 independent chains, with every 4th sample collected. The P value was calculated as the probability that on average replicates from the posterior distribution of the knockout condition are different from the control condition. There was no correction for multiple hypothesis testing as this was resolved via the choice of model priors.

A protein group was detected as significantly regulated if (1) the difference between the medians of the posterior distributions of the log₂ intensity in the knockout condition and the control condition was ≥0.5 with P ≤ 0.001, and (2) the protein group was identified with a minimum of 2 peptides detected with MS/MS and at least with 1 peptide identified in 2 of the 3 replicates.

Host transcriptome changes induced by NPHP4 gene depletion in VZV-infected SK-N-BE2 cells

Data processing

Data were processed using the published Drop-seq pipeline (v.1.0) to generate sample- and gene-wise UMI tables¹¹⁶. The reference genome (GRCh38) was used for alignment. Transcript and gene definitions were used according to GENCODE v.38. The VZV genome was derived from GenBank: NC_001348.1. Drop-seq tools v.1.12 was used for mapping raw sequencing data to the reference genome.

Statistical analysis

Data normalization, differential expression analysis and P value adjustment were performed with the DESeq2 package (v.1.38.1)¹¹⁷ using GLM linear modelling with the standard settings. Independent filtering was disabled. For each condition, one of the quadruplicates was excluded following PCA. Transcripts with less than ten counts in at least three samples were excluded. The best sensitivity to identify differentially expressed genes (DEGs) was achieved using the following designs:

• VZV versus mock in NTC cells: defined DEGs following VZV infection of NTC cells (1).

• VZV versus mock in KO cells: defined DEGs following VZV infection of NPHP4-depleted cells (2).

  The log₂-transformed fold changes were shrunken using apeglm. For visualization, we used a scatter plot of shrunken log₂-transformed fold changes between comparisons (1) and (2). Shared DEGs following VZV infection between NTC and KO cells were defined by adjusted P ≤ 10⁻⁴ and |shrunken log₂(fold change)| ≥ 1 in each comparison.

• Statistical evaluation of DEGs between KO and NTC cells following infection was performed as follows: KO versus NTC in VZV samples: log₂-transformed fold changes were left unshrunken. DEGs were defined by |log₂(fold change)| ≥ 0.5 and (i) unadjusted P ≤ 0.02 or (ii) unadjusted P ≤ 0.1 and defined as DEGs in (1) or in (2).

• DEGs following NPHP4 gene depletion in mock cells were defined as follows: KO versus NTC in mock samples: log₂-transformed fold changes were left unshrunken. DEGs were defined by |log₂(fold change)| ≥ 0.25 and adjusted P ≤ 0.1.

AP of SK-N-BE2 cells expressing HA-tagged NPHP4 wild type or S862N mutant

Data processing

The raw MS data files were analysed with MaxQuant (v.1.6.14.0) using the standard settings, with match between runs and LFQ enabled (LFQ min ratio count set to 2, stabilize large LFQ ratios disabled and normalization skipped). The MS2 spectra were searched against forward and reverse (used as decoy) sequences of the reviewed human proteome, including isoforms (UniProtKB, release 2019) and GFP using the built-in Andromeda search engine.

Statistical analysis

Protein groups and LFQ intensities generated by MaxQuant (proteinGroups file) were imported into Perseus (2.0.7.0). Protein groups were filtered for contaminants, reverse identification, if only identified by a modified peptide, and if identified in less than two replicates in at least one bait. A background was defined by protein groups identified in all samples (in four of four replicates of all baits). log₂-transformed LFQ intensities were normalized by the median of the background of the given sample. Missing intensities were imputed from a log normal distribution of log-transformed normalized intensities with a width of 0.3× the s.d. of the measured intensities and downshifted by 1.8 s.d. to simulate the detection limit of the mass spectrometer. Enrichments were calculated as the log₂ difference of the median intensity of each NPHP4 variant with GFP to define interactors, or between the two NPHP4 variants to define specific interactors. Significance was assessed by the Welch t-test, adjusted by permutation-based FDR.

FACS analysis of VZV infection

SK-N-BE2 cells were infected by co-culture with VZV rOka as described earlier in three independent experiments. At 48 h post-infection, cells were detached from the transwell by trypsin-EDTA (Sigma-Aldrich), washed in PBS (Sigma-Aldrich) and fixed with 3.7% formaldehyde (Sigma-Aldrich) for 15 min at room temperature. After a wash in FACS buffer (1% FCS, 2 mM EDTA in PBS), cells were labelled with the VZV DFA Kit reagent (Light Diagnostics; Merck) containing two FITC-conjugated antibodies against VZV immediate-early ORF62 protein and late glycoprotein E for 15 min at room temperature. Labelled cells were analysed on an Attune NxT Acoustic Focusing Cytometer (Thermo Fisher Scientific) and data were processed with FlowJo (v.10). Forward and side scatters were used to exclude cell debris and gate single cells and FITC intensity was analysed.

WB analysis

Expression of the V5-tagged VZV proteins

A total of 5 × 10⁵ SK-N-BE2 cells expressing individual V5-tagged VZV ORF or GFP were generated as described earlier. Frozen cell pellets were thawed and lysed on ice for 30 min in lysis buffer (0.2% NP-40, 100 mM NaCl, 5% glycine, 1.5 mM MgCl₂, 50 mM Tris-HCl pH 7.5) supplemented with 1% in-house benzonase. Samples were sonicated (5 min, 4 °C, 30 s on and off, high frequency; Bioruptor; Diagenode) and centrifuged at 15,000g at 4 °C for 30 min. Supernatants were collected and total proteins were precipitated overnight at −20 °C in 80% acetone before being resuspended in SDS sample buffer (62.5 mM Tris-HCl pH 6.8, 2% SDS, 10% glycerol, 50 mM DTT, 0.01% bromophenol blue). After boiling for 5 min at 95 °C, samples were loaded on NuPAGE Novex 10% Bis-Tris (Invitrogen) and further submitted to WB using 0.45 µm nitrocellulose membranes (Amersham Protran). Imaging was performed by HRP luminescence using the SuperSignal West Femto kit (Thermo Fisher Scientific).

Expression levels were quantified from the V5-tag blot signal from VZV ORF-expressing SK-N-BE2 cell lines (Extended Data Fig. 2a) using ImageLab (v.6.0.1). The volume of the bait-specific bands was normalized by the volume of the lane-corresponding actin band (loading control). Intensities that required longer exposure detection were normalized by the fold change between short and long exposures, obtained from other bands on the same blot that were detected without saturation in both images. Bands that overlapped the actin signal were excluded from this analysis (ORF36, ORF9 and ORF63). The Pearson correlation coefficient (Pearson’s r) between the expression level and the numbers of targets and effects were calculated from non-transformed values.

Degradation of IFI16 in co-transfected HEK293T cells

A total of 1 × 10⁶ HEK293T cells were transfected with 500 ng of HA-tagged VZV ORF61, HSV-1 ICP0 or GFP and 100 ng of IFI16 for 24 h. Cells were washed in PBS and pellets were frozen at −20 °C. Pellets were lysed in 2% SDS, 1% DNAse lysis buffer, sonicated (10 min, 4 °C, 30 s on and off, high frequency; Bioruptor; Diagenode) and boiled for 5 min at 95 °C. Protein concentration was measured by Pierce 660 nm Protein Assay (Thermo Fisher Scientific) and normalized in SDS sample buffer. Protein (20 µg) was loaded on 10% Tris-glycine SDS–PAGE and further submitted to western blotting using 0.22 µm nitrocellulose membrane (Amersham Protran). Imaging was performed by HRP luminescence using the Western Lightning Plus-ECL kit (Perkin Elmer).

IP followed by western blot analysis

Sample preparation

VZV ORF32-V5 IP in SK-N-BE2 cells

A total of 4 × 10⁶ SK-N-BE2 cells expressing V5-tagged VZV ORF32 and control cells expressing V5-tagged VZV ORF23 were scraped, washed in ice-cold PBS and flash frozen in liquid nitrogen.

Reverse IP of VZV ORF32 by GFP-GTF2B in HeLa Kyoto cells

pLenti6.3-TO-V5 expressing VZV ORF32 or ORF23 (10 µg) was transfected as described above in 1 × 10⁶ HeLa Kyoto GFP-GTF2B. At 24 h post-transfection, cells were scraped, washed in ice-cold PBS and flash frozen in liquid nitrogen.

IP and WB

Cell lysis was performed using lysis buffer (0.2% NP-40, 100 mM NaCl, 5% glycine, 1.5 mM MgCl₂, 50 mM Tris-HCl pH 7.5) supplemented with 1% in-house benzonase as described for the AP–MS. Of the total lysate, 5% was retained as a total lysate control for the WB analysis. IP and washes were performed similarly to the AP–MS, using anti-V5 magnetic beads (MBL M215-11) for V5-ORF32 IP and GFP-Trap Magnetic Agarose (Chromotek) for GFP-GTF2B IP. Washed beads were resuspended in the wash buffer. IP samples and total lysate controls were mixed in SDS sample buffer (62.5 mM Tris-HCl pH 6.8, 2% SDS, 10% glycerol, 50 mM DTT, 0.01% bromophenol blue), boiled at 95 °C for 5 min and loaded on homemade 10% Tris-HCl gels. Proteins were blotted on a 0.45 µm nitrocellulose membrane (Amersham Protran). Imaging was performed by HRP luminescence using the Western Lightning Plus-ECL kit (Perkin Elmer).

IF analysis

Fixation

If not specified otherwise, cells were fixed with 4% formaldehyde for 15 min at room temperature, permeabilized (0.1% Triton X-100 in PBS) for 15 min at room temperature and blocked (0.1% FCS, 1% donkey serum, 0.1% Triton X-100 in PBS) for 1 h at room temperature. Staining was performed by incubation with the indicated primary antibodies at 4 °C overnight, followed by incubation with secondary antibodies for 2 h at room temperature in blocking buffer. DAPI was incubated for 1 min in blocking buffer or included in the secondary antibody incubation.

Cellular localization of VZV ORFs in SK-N-BE2 cells

A total of 0.8–2 × 10⁵ SK-N-BE2 cells expressing individual V5-tagged VZV ORFs were grown on coverslips precoated with Attachment Factor (Gibco Fisher Scientific) for 48 h before fixation. Three samples were generated and stained independently. Confocal imaging was performed using an LSM900 confocal laser scanning microscope (ZEISS) equipped with a ×63/1.40 oil DIC M27 objective (ZEISS), with Airyscan z2. Airyscan images were processed in Zen 3.5 (blue edition, ZEISS), and brightness and contrast adjustments were made using Zen 3.8 (blue edition, ZEISS).

Cellular localization of VZV ORF32 and GTF2B

Cellular localization of VZV ORF32 and GTF2B was analysed by IF analysis in HeLa Kyoto cells stably expressing GFP-tagged GTF2B as previously described¹¹⁸. Briefly, HeLa Kyoto GFP-GTF2B cells were grown on coverslips and 2 µg of pLenti6.3-TO-V5 expressing VZV ORF32 or ORF23 were transfected as described earlier. After 36 h, cells were fixed with 4% paraformaldehyde (PFA) for 15 min at room temperature and blocked in blocking buffer (1× PBS containing 0.1% FCS and 0.1% Triton X-100) for 1 h at room temperature. Staining was performed with the indicated antibodies for 1 h at room temperature in blocking buffer. Confocal imaging was performed using an LSM780 confocal laser scanning microscope (ZEISS) equipped with a Plan-APO ×63/NA1.46 immersion oil objective (ZEISS).

IFI16 degradation and ORF61-UBXN7 co-localization study in VZV-infected cells

HFF cells were grown on coverslips precoated with Attachment Factor (Gibco Fisher Scientific) and left as mock or infected by co-culture with MeWo cells infected with VZV(pOka)-61-HA. At 2 h post-infection, cells were left untreated or treated with 1 µM MG-132. Cells were fixed 8 h post-infection. Samples from three experiments were generated and stained independently. Confocal imaging was performed using an LSM900 confocal laser scanning microscope (ZEISS) equipped with a Plan-APO ×20/0.8 M27 objective for the analysis of IFI16 degradation or a ×63/1.40 oil DIC M27 objective for the co-localization analysis of VZV ORF61 and UBXN7 (ZEISS), with Airyscan 2. Airyscan images were processed in Zen 3.5 (blue edition, ZEISS). IFI16 intensity was measured from 10 z-stack images of 0.5-µm intervals centred on IFI16 maximum intensity, processed by maximum intensity projection in Zen 3.5 (blue edition, ZEISS). For VZV ORF61-UBXN7 co-localization analysis, single images at defined z were acquired. Brightness and contrast adjustments and intensity line profile analysis were performed using Zen 3.8 (blue edition, ZEISS).

VZV and IFI16 quantification were analysed with FIJI (v.1.54f; ImageJ). Cell nuclei were masked from the DAPI signal and mean intensities of IFI16 (Alexa Fluor 488) and VZV (Alexa Fluor 647) were analysed. Background VZV signal was determined by its maximum mean intensity measured in mock conditions. Non-infected cells, as defined by the VZV background, were excluded from the infected conditions in the summarized IFI16 plot (Fig. 4c, left).

IFI16 degradation study in ORF61-expressing cells

HFF cells were transduced with shRNA-expressing lentivirus for scramble or UBXN7-targeting sequence. Five days later, cells were transduced with V5-tagged VZV ORF61 or GFP as described earlier. Cells were grown for 24 h in 8-well cell culture chambers on a glass slide (Sarstedt) and fixed 7–20 days post-transduction. Samples from two experiments were generated and stained independently as described above with rabbit anti-IFI16 and mouse anti-V5 primary antibodies, anti-rabbit Alexa Fluor 647, anti-mouse Alexa Fluor 488 and DAPI. Confocal imaging was performed using an LSM900 confocal laser scanning microscope (ZEISS) equipped with a Plan-APO ×20/0.8 M27 or ×10/0.45 M27 objective (ZEISS), with Airyscan 2. A z-stack of 1-µm intervals centred on IFI16 maximum intensity and covering the whole signal was acquired. Airyscan images were processed and maximum intensity projections were performed in Zen 3.5 (blue edition, ZEISS).

Data analysis was performed using ICY software (v.2.5.2.0). Nuclei were masked and sorted for positive V5 (Alexa Fluor 488 or Alexa Fluor 647) signal. The nucleoli were isolated using the wavelet spot detector plug-in and subtracted from the region of interest mask for the nuclei. Within these resulting subtraction regions of interest, integrated intensity values of IFI16 (Alexa Fluor 647) were calculated for each nucleus. Outliers were detected with the ROUT method and a maximum desired FDR Q = 0.5% using GraphPad Prism.

Host gene knockout screen on VZV growth

Host gene selection

Data-driven selection of host proteins upregulated in VZV-infected SK-N-BE2 cells (Supplementary Table 1) or enriched in at least 1 VZV ORF interactome with an enrichment factor above 13 (Supplementary Table 3) was applied. FAM208A, MPP8, APOBEC3B, CNOT6 and USP37 were selected based on biological interest. HCF1 was used as a dependency factor control. DHX9 and MYD88 were additionally included as restriction factor controls. In total, 116 genes from our data and 2 control genes were tested in the knockout screen.

Generation of knockout cells, VZV infection and flow cytometry analysis

As VZV propagates mostly via cell-to-cell contact, we set up a fluorescence-based co-culture infection assay to measure the replication of a recombinant VZV using flow cytometry analysis. SK-N-BE2-GFP(blastR) and the SK-N-BE2-spCas9(ZeoR)-BFP(blastR) cells were seeded in a 1:1 ratio in a 96-well plate. The next day, each well was transduced with individual sgRNAs from the host gene library or with an empty vector as NTC, using polybrene at 8 µg ml⁻¹. At 48 h post-transduction, transduced cells were selected with 2 µg ml⁻¹ puromycin and kept in culture for another 12 days to allow clearance of remaining expressed proteins in the target cells. Each well was split into four wells the day before infection. Cell density at infection was monitored by the non-lytic Real-Time Glo luminescence assay (Promega) acquired on an Infinite 200 plate reader (Tecan). Two wells were mock infected by exchanging the culture medium. Two wells were infected at a low multiplicity of infection (MOI; expected between 0.002 and 0.005, depending on individual knockout cell density) by exchanging the medium with a suspension of MeWo cells infected with VZV(pOka)-63-RFP/70-RFP. After 48 h (16 days post-selection), cells were fixed in 1% PFA for 15 min, resuspended in FACS buffer (5 mM EDTA pH 8, 25 mM HEPES, 1% FCS in PBS), and analysed on a CytoFLEX flow cytometer operated by CytExpert (Beckman Coulter). The whole gRNA library was screened within five batches, each batch consisting of independent transduction, selection, infection and measurement steps. NTC was included within each batch.

Data analysis

Flow cytometry data were analysed by FlowJo (v.10). Forward and side scatters were used to exclude cell debris and gate single cells. Knockout and control cells were gated from BFP+ and GFP+ cells, respectively. A low percentage of BFP cells in mock wells at 15 days post-selection of transduced cells was associated with a cell growth disadvantage and was used as a proxy for knockout toxicity. Thus, individual gRNA knockouts with less than 35% of BFP cells in mock wells were excluded from subsequent analysis. A total of 218 knockouts covering 108 host genes and the NTC remained. The MRI of the combined population of BFP (knockout) and GFP (control) cells (excluding unmarked inoculum MeWo cells) was used to assess the spread of VZV in each well. The MRIs of each BFP+ and GFP+ population were separately extracted to evaluate the spread of the virus within knockout or control cells, respectively. The normalized target-to-control MRI was used to evaluate each knockout’s effect on VZV replication. Pearson’s correlation coefficient r between the MRI, average or ratio across BFP and GFP populations, and the absolute difference to the median luminescence intensity were calculated to assess the bias from variability of cell density at infection (Supplementary Discussion). Infection batch effects were normalized by subtracting the batch median MRI ratio from each individual MRI ratio, and then dividing the result by the batch median MRI ratio. Normalized MRI ratios were averaged across replicates and represented as robust z-scores to account for the absolute median deviation within the screen. Host factors for which at least two sgRNAs showed the same effect on VZV growth (increase or decrease), and for which at least one had an absolute robust z-scored ratio above two, were identified as strong restriction or dependency genes for VZV, respectively.

Validation of the effect of NPHP4, MPP8 and ZNF280D knockouts on VZV infection

Sample generation, VZV infection and flow cytometry analysis

Stable knockout SK-N-BE2 cells expressing BFP were generated for host genes of interest as described earlier. Knockout and NTC cells were seeded in 24-well plates in technical duplicates. Just before infection, cell density was monitored by imaging (IncuCyte S3 Live-Cell Analysis System) and masking of the phase contrast (as phase percentage; IncuCyte Software v.2019B Rev2; Sartorius). Cells were infected at an MOI of 0.01 by co-culture with MeWo cells infected with VZV(pOka)-63-RFP/70-RFP. At 48 h post-infection, cells were fixed in 1% PFA for 15 min, washed 3× in PBS, resuspended in FACS buffer (5 mM EDTA pH8, 25 mM HEPES, 1% FCS in PBS) and analysed on a CytoFLEX flow cytometer operated by CytExpert (Beckman Coulter). The assay was performed in three independent biological replicates. Data were analysed using FlowJo (v.10). SK-N-BE2 cells were distinguished from the inoculum MeWo cells by gating the BFP+ population. VZV spread was then assessed by the MRI of the total BFP+ population. To account for technical variability of cell density at infection, which affects the spread of VZV, an estimated MRI based on the measured cell density at infection was calculated per well. Polynomial regression was performed to assess the correlation between measured cell density at infection and measured MRI in NTC wells. The model was then applied to calculate an ‘estimate’ MRI per well based on the measured cell density at infection. The effect of the knockout was then reported as the log₂ difference of the measured MRI to the estimate MRI. The log₂ differences were averaged across technical duplicates and plotted as biological replicates. P values were generated using a one-tailed unpaired t-test on biological triplicates of individual knockouts against the NTC.

Viral growth kinetics

A total of 2.5 × 10⁵ SK-N-BE2 cells knocked out for MPP8, ZNF280D, NPHP4 or NTC were seeded in 24-well plates in technical duplicates. Cells were infected at an MOI of 0.0002 by co-culture with MeWo cells infected with VZV(pOka)-63-RFP/70-RFP (10⁴ MeWo cells). Medium was exchanged 3 h post-infection and cells were imaged every 6 h for 5 days (IncuCyte S5 Live-Cell Analysis System). Red and phase signals were processed and masked using the IncuCyte Software (v.2021; Sartorius). Red areas were normalized to the phase and to the first acquisition time (3 h post-infection) to account for viral load (MeWo cell count) variability. The assay was performed in three independent biological replicates. Areas under the curve were calculated using GraphPad Prism.

Validation of gene knockout efficiency

NTC or knockout cells’ gDNA was extracted as recommended (69504; Qiagen). Each targeted region was amplified by PCR using the GoTaq G2 DNA Polymerase (Promega) and flanking primers, before Sanger sequencing (Eurofins Genomics). Data were analysed using an ICE tool by Synthego⁹⁸. The R² value indicates how well the calculation explains the sequencing trace.

MS analysis of protein depletion

For each knockout cell line, 1 × 10⁶ cells were prepared in triplicate and analysed by MS as described earlier (see ‘Proteome changes induced by MPP8 gene depletion in SK-N-BE2 cells’). To allow for the correct identification of the proteins of interest, matching samples were generated by transfection of 0.7 µg expression plasmid for the given protein (pWPI-nHA-PuroR backbone) in 5 × 10⁵ HEK293T cells. The next day, cells were prepared as described earlier, along with the knockout samples. The raw MS data files were processed with MaxQuant (v.1.6.10.0), using the standard settings with ‘iBAQ’ quantification and match between runs enabled. The MS2 spectra were searched against forward and reverse (used as decoy) sequences of the reviewed human proteome, including isoforms (UniProtKB, release 2018.02), by the built-in Andromeda search engine. iBAQ (intensity-based absolute quantification) intensities were normalized by the median of all intensities per sample and averaged across knockout replicates.

WB analysis of protein depletion

Knockout and NTC SK-N-BE2 cells expressing BFP were flash frozen in liquid nitrogen as cell pellet and stored at −80 °C. Frozen cell pellets were thawed and lysed on ice for 45 min in lysis buffer (RIPA with 1% SDS supplemented with Complete EDTA-free Protease Inhibitor; Roche). Samples were sonicated (10 min, 4 °C, 30 s on and off, high frequency; Bioruptor; Diagenode) and centrifuged at 15,000g at 4 °C for 30 min. Supernatants were collected and total proteins were precipitated overnight at −20 °C in 80% acetone before being resuspended in SDS sample buffer (62.5 mM Tris-HCl pH 6.8, 2% SDS, 10% glycerol, 50 mM DTT, 0.01% bromophenol blue). Total protein concentrations were measured by Pierce 660 nm Protein Assay (Thermo Fisher Scientific) and normalized to 25 µg in 10 µl SDS sample buffer. After boiling for 5 min at 95 °C, samples were loaded on NuPAGE Novex 4–12% Bis-Tris (Invitrogen) and further submitted to western blotting using 0.45 µm nitrocellulose membrane (Amersham Protran). Imaging was performed by HRP luminescence using the SuperSignal West Femto kit (Thermo Fisher Scientific).

Viability assay

Knockout and NTC SK-N-BE2 cell lines were seeded in a 96-well plate. At the given time point, 10 µg per well of resazurin was added in triplicate, the cells were incubated at 37 °C for 90 min, and the fluorescence was measured on the plate reader Infinite 200 (Tecan). Technical triplicate values were averaged. The assay was conducted in biological triplicate.

Rescue of NPHP4 knockout cell lines and VZV infection

Sample generation

NPHP4 knockout or NTC SK-N-BE2 cells, expressing BFP, were transduced with a control empty vector or a vector encoding for N-terminal HA-tagged NPHP4 or N-terminal HA-tagged NPHP4-S862N variant (pWPI-nHA-PuroR backbone) using 8 µg ml⁻¹ polybrene. Culture medium was exchanged 6 h post-transduction. The day before the assay, cells were counted and seeded for VZV infection or WB analysis. At 72 h post-transduction, 1 well was collected in SDS sample buffer (62.5 mM Tris-HCl pH 6.8, 2% SDS, 10% glycerol, 50 mM DTT, 0.01% bromophenol blue) and frozen at −20 °C for later WB analysis, 1 well was mock infected, and 3 wells were infected at an MOI of 0.02 by co-culture with MeWo cells infected with VZV(pOka)-63-RFP/70-RFP for 48 h.

Flow cytometry analysis

At 48 h, mock- and VZV-infected cells were fixed in 1% PFA for 15 min, washed 3× in PBS, resuspended in FACS buffer (5 mM EDTA pH8, 25 mM HEPES, 1% FCS in PBS) and analysed on a CytoFLEX flow cytometer operated by CytExpert (Beckman Coulter). The assay was performed in three independent biological replicates. Data were analysed using FlowJo (v.10). SK-N-BE2 cells were isolated from the inoculum MeWo cells by sorting the single BFP+ cells. VZV spread was assessed by the MRI. MRI was averaged across technical replicates and normalized across biological replicates by calculating the fold change to the given NTC empty vector sample. P values were generated by unpaired one-way analysis of variance (ANOVA) and adjusted using the Bonferroni method.

WB analysis of ectopic HA-tagged NPHP4 expression

Cells resuspended in SDS sample buffer were thawed at room temperature, sonicated (5 min, 4 °C, 30 s on and off, high frequency; Bioruptor; Diagenode) and boiled for 10 min at 95 °C. Samples were loaded on NuPAGE Novex 4–12% Bis-Tris (Invitrogen), and then submitted to western blotting using 0.22 µm nitrocellulose membrane (Amersham Protran). Imaging was performed by HRP luminescence using the Western Lightning Plus-ECL kit (Perkin Elmer).

VZV patient cohort and WES

Patients

The patient cohort consisted of adults (>18 years) admitted to the Department of Infectious Diseases at Aarhus University Hospital with a final diagnosis of VZV-associated encephalitis, meningoencephalitis or cerebral vasculitis. A total of 13 adult patients were included on the basis of pleocytosis and a positive PCR for VZV in the cerebrospinal fluid. PCR on cerebrospinal fluid for HSV-1, HSV-2, and enterovirus, along with bacterial cultures, were also performed. All results were negative for all included patients. Exclusion criteria were immunosuppressive therapy, known malignant disease, pregnancy and HIV positivity. All patients had experienced chickenpox in childhood and none had received VZV vaccination (Supplementary Table 6-1).

Blood sampling

A total of 50 ml of blood was drawn for isolation of PBMCs and DNA for further analysis. DNA was isolated from EDTA-stabilized blood using the EZ1 DNA Blood 350 µl Kit and an EZ1 Advanced XL instrument (Qiagen) according to the manufacturer’s instructions¹¹⁹.

WES, bioinformatics and integration of VZV–host interactions

WES was performed using Kapa HTP Library Preparation Kit and the Nimblegen SeqCap EZ MedExome Plus kit. Sequencing was performed on an Illumina Nextseq 550 system. Single nucleotide polymorphism calling was performed relative to hg19 using BWA. PCR and optical duplicates were identified and marked. The alignment was recalibrated using GATK. Single nucleotide polymorphisms were called using HaplotypeCaller from the GATK package. VCF files were submitted to Ingenuity Variant Analysis. Unconfident (call quality below 30.0, read depth below 5.0 and allele fraction below 25.0) and common (ExAC database, 1000 Genomes, and GnomAD minor allele frequency > 0.01%) variants were excluded. Known variants classified as ‘pathogenic’ or unknown variants with a SIFT/PolyPhen prediction as ‘damaging’ were considered as predicted deleterious. Further filtering was applied to keep only variants with a Combined Annotation Dependent Depletion score > 15, a mutation significance cut-off below the Combined Annotation Dependent Depletion score, localized in exons and non-synonymous. A list of genes assembled from the host proteins with differential protein abundance following infection (Supplementary Table 1) and VZV proteins’ host targets as identified by AP–MS (Supplementary Table 3) was submitted as a biological context filter. Finally, identified variants were manually checked by inspecting BAM files using IGV.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.