High-Definition Analysis of Host Protein Stability during Human Cytomegalovirus Infection Reveals Antiviral Factors and Viral Evasion Mechanisms

Cells and Cell Culture

Primary human fetal foreskin fibroblast cells (HFFFs, male) and HFFFs immortalised with human telomerase (HFFF-TERTs) were grown in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with foetal bovine serum (FBC: 10% v/v), and penicillin/streptomycin at 37°C in 5% CO₂. HFFFs and HFFF-TERTs have been tested at regular intervals since isolation to confirm both that the HLA & MICA genotypes and the morphology and antibiotic resistances are consistent with the original cells. In addition, the HCMV Merlin strain used is only permissive in human fibroblasts (dermal or foreskin), further limiting the chances that the cells have been contaminated with another cell type.

For pulsed SILAC analysis, cells were grown for seven divisions in DMEM for SILAC, which was supplied without light arginine or lysine. This medium was supplemented with 10% dialysed FBS, penicillin/streptomycin, 84 mg/ml light arginine, 280 mg/l L-proline and either 50 mg/ml medium lysine (Lys 4) or 146 mg/ml heavy lysine (Lys 8). For SILAC immunoprecipitations, cells were grown identically but the medium was supplemented with 10% dialysed FBS, penicillin/streptomycin, 280 mg/l L-proline and either light (Arg 0, Lys 0) or medium (Arg 6, Lys 4) amino acids at 50 mg/l. Incorporation of heavy label was >98% for both arginine and lysine-containing peptides.

Viruses

We used virus (RCMV1111) derived by transfection of a BAC clone of HCMV strain Merlin, the genome of which is designated the reference HCMV sequence by the National Center for Biotechnology Information and was sequenced after 3 passages in vitro (Dolan et al., 2004) (Stanton et al., 2010). RCMV1111 contains point mutations in two genes (RL13 and UL128) that enhance replication in fibroblasts (Stanton et al., 2010). The 10 block HCMV deletion mutants were generated on a strain Merlin background (wt1), or wt1 that lacked UL16 and UL18 (wt2) by transfection of recombinant BACs (Stanton et al., 2010) as described in (Fielding et al., 2014). The wt2 background was originally employed to facilitate detection of NK evasion functions (deletion and backbone details shown in Table S5A) (Fielding et al., 2014). HCMV strain AD169 varUK/BK000394 was used for the U_L/b’ mutant, as a deletion in this region (plus additional defects) has been acquired during passage in culture. Single gene deletion mutants of all the canonical genes in the U_L/b’ region were generated by recombineering the strain Merlin BAC as described previously (Stanton et al., 2010). Whole-genome consensus sequences of passage 1 of each RCMV were derived using the Illumina platform as described previously (Fielding et al., 2014), and deposited in GenBank. HCMV expressing rGFP from a P2A self-cleaving peptide following the UL36 ORF, and UL145-V5 recombinants were generated as described in (Fielding et al., 2014). For part of the inhibitor screen, viruses were irradiated with a dose of 3500 Gy using a Gammacell 1000 Elite (Nordion International), and inactivation was verified by the absence of immunofluorescence for IE1 compared to control (data not shown) (Weekes et al., 2014).

Virus stocks were prepared from HFFF-TERTs as described previously (Stanton et al., 2007). Tissue culture supernatants were kept when a 100% cytopathic effect was observed, and were centrifuged to remove cell debris. Cell-free virus was pelleted from supernatant by centrifugation at 22,000 × g for 2 h and then resuspended in fresh DMEM. Residual debris was removed from the resulting virus stocks by centrifugation at 16,000 x g for 1 min.

Virus Infections and Inhibitors for Proteomic Experiments

1 x 10⁶ HFFFs (RNA/protein screen, block deletion mutant screen) or HFFF-TERTs (inhibitor, pSILAC, wt vs ΔUL145 virus screens) were plated in a 25cm² flask. We found that primary HFFFs were limited to a total of ∼25 passages and exhibited diminished rates of growth as passage number increased. HFFF-TERTs were therefore used in the pSILAC screen, due the need for seven cell divisions prior to infection, and in the inhibitor screen to ensure comparability of results. Cells were infected at multiplicity of infection 5 or 10 with HCMV strain Merlin as previously described (Weekes et al., 2014). Briefly, the requisite volume of viral stock was added to 1 ml DMEM and mixed gently prior to being applied to cells. Mock infections were performed identically but with additional DMEM instead of viral stock. Time 0 in any experiment was considered to be the initial point of infection with virus. In each experiment, cells were incubated with virus for 2 h prior to a change in medium with the exception of the 6 h pSILAC experiment, where cells were incubated with virus for 1.5 h. Inhibitors added at the indicated times were: 10 μM MG132 (Merck) or 200 μM Leupeptin (Merck).

Pulsed SILAC Analysis

For pulsed SILAC, cells were grown for seven divisions in medium-labelled SILAC DMEM as detailed above. Seeding at 1 x 10⁶ per T25 gave a confluent culture, resulting in the arrest of cell division by contact inhibition. This ensured that protein turnover was the sole determinant of labeling kinetics, as opposed to dilution of cellular proteins by cell division. At time 0, media was changed to heavy-labelled SILAC DMEM containing HCMV or equivalent virus-free medium for mock infection. Without frequent media changes, substantial recycling of medium-labelled amino acids occurred (data not shown). We therefore (a) increased the concentration of heavy lysine to that of standard DMEM, namely 146 mg/l. (b) changed media every 45 mins for the whole course of the experiment (18 h pSILAC experiment) or every 30 mins for the whole experiment (6 h pSILAC experiment). Cells were harvested at time points detailed in Figure 2A. For the 3 h time point in Experiment 2, HCMV-infected and not mock-infected cells were harvested, as only 11 TMT labelling reagents were available.

Whole Cell Lysate Protein Digestion

Cells were washed twice with PBS, and 250 μl lysis buffer added (6M Guanidine/50 mM HEPES pH 8.5). Cell lifters (Corning) were used to scrape cells in lysis buffer, which was removed to an eppendorf tube, vortexed extensively then sonicated. Cell debris was removed by centrifuging at 21,000 g for 10 min twice. Although this method prohibited cell counting immediately prior to lysis, it avoided the need for cellular detachment. Half of each sample was kept for subsequent analysis by immunoblot where required. For the other half, dithiothreitol (DTT) was added to a final concentration of 5 mM and samples were incubated for 20 mins. Cysteines were alkylated with 14 mM iodoacetamide and incubated 20 min at room temperature in the dark. Excess iodoacetamide was quenched with DTT for 15 mins. Samples were diluted with 200 mM HEPES pH 8.5 to 1.5 M Guanidine followed by digestion at room temperature for 3 h with LysC protease at a 1:100 protease-to-protein ratio. Samples were further diluted with 200 mM HEPES pH 8.5 to 0.5 M Guanidine. Trypsin was then added at a 1:100 protease-to-protein ratio followed by overnight incubation at 37°C. The reaction was quenched with 5% formic acid, then centrifuged at 21,000 g for 10 min to remove undigested protein. Peptides were subjected to C18 solid-phase extraction (SPE, Sep-Pak, Waters) and vacuum-centrifuged to near-dryness.

Peptide Labeling with Tandem Mass Tags

In preparation for TMT labeling, desalted peptides were dissolved in 200 mM HEPES pH 8.5. Peptide concentration was measured by microBCA (Pierce), and 25 μg of peptide labeled with TMT reagent. TMT reagents (0.8 mg) were dissolved in 43 μl anhydrous acetonitrile and 3 μl added to peptide at a final acetonitrile concentration of 30% (v/v). Sample labelling was as indicated in Table S7A. Following incubation at room temperature for 1 h, the reaction was quenched with hydroxylamine to a final concentration of 0.3% (v/v). TMT-labeled samples were combined at a 1:1:1:1:1:1:1:1:1:1:1 ratio. The sample was vacuum-centrifuged to near dryness and subjected to C18 SPE (Sep-Pak, Waters). An unfractionated singleshot was analysed initially to ensure similar peptide loading across each TMT channel, thus avoiding the need for excessive electronic normalization. As all normalisation factors were >0.5 and <2, data for each singleshot experiment was analysed with data for the corresponding fractions to increase the overall number of peptides quantified. Normalisation is discussed in ‘Data Analysis’, and high pH reversed-phase (HpRP) fractionation is discussed below.

Offline HpRP Fractionation

TMT-labelled tryptic peptides were subjected to HpRP fractionation using an Ultimate 3000 RSLC UHPLC system (Thermo Fisher Scientific) equipped with a 2.1 mm internal diameter (ID) x 25 cm long, 1.7 μm particle Kinetix Evo C18 column (Phenomenex). Mobile phase consisted of A: 3% acetonitrile (MeCN), B: MeCN and C: 200 mM ammonium formate pH 10. Isocratic conditions were 90% A/10% C, and C was maintained at 10% throughout the gradient elution. Separations were conducted at 45°C. Samples were loaded at 200 μl/minute for 5 minutes. The flow rate was then increased to 400 μl/minute over 5 minutes, after which the gradient elution proceed as follows: 0-19% B over 10 minutes, 19-34% B over 14.25 minutes, 34-50% B over 8.75 minutes, followed by a 10 minutes wash at 90% B. UV absorbance was monitored at 280 nm and 15 s fractions were collected into 96 well microplates using the integrated fraction collector. Fractions were recombined orthogonally in a checkerboard fashion, combining alternate wells from each column of the plate into a single fraction, and commencing combination of adjacent fractions in alternating rows. Wells were excluded prior to the start or after the cessation of elution of peptide-rich fractions, as identified from the UV trace. This yielded two sets of 12 combined fractions, A and B, which were dried in a vacuum centrifuge and resuspended in 10 μl MS solvent (4% MeCN/5% formic acid) prior to LC-MS3. 12 set ‘A’ fractions were used for MS analysis of all experiments. For the 18 h pSILAC (pSILAC_18) experiment, an additional 6 set ‘B’ fractions were used, as 6/12 original fractions were suboptimally analysed on the Orbitrap Lumos. For the 12 h inhibitor experiment (Deg_12), a single fraction failed to run optimally and a further set ‘B’ fraction was analysed. Both set ‘A’ and set ‘B’ runs were included in the final analysis in each case (Table S7B).

LC-MS3 for TMT and TMT/SILAC Experiments

Mass spectrometry data was acquired using an Orbitrap Lumos for all experiments apart from 6 fractions from the pSILAC_18 experiment, where an Orbitrap Fusion mass spectrometer was used instead (Thermo Fisher Scientific, San Jose, CA). In both cases, an Ultimate 3000 RSLC nano UHPLC equipped with a 300 μm ID x 5 mm Acclaim PepMap μ-Precolumn (Thermo Fisher Scientific) and a 75 μm ID x 50 cm 2.1 μm particle Acclaim PepMap RSLC analytical column was used.

Immunoprecipitation and Protein Digestion

Cells were harvested in lysis buffer (50 mM Tris pH 7.5, 300 mM NaCl, 0.5% (v/v) NP40, 1 mM DTT and Roche protease inhibitor cocktail), tumbled for 15 minutes at 4°C and then centrifuged at 16,100 g for 20 minutes at 4°C. Lysates were then clarified by filtration through a 0.7 μm filter and incubated for 3 h with immobilised mouse monoclonal anti-V5 agarose resin. Samples were washed multiple times with lysis buffer, followed by multiple PBS pH 7.4 washes. Subsequently, proteins bound to the anti-V5 resin were eluted twice by adding 200 μl of 250 μg/ml V5 peptide (Alpha Diagnostic International) in PBS at 37°C for 30 minutes with agitation. Finally, proteins were precipitated with 20% TCA, washed once with 10% TCA, washed three times with cold acetone and dried to completion using a centrifugal evaporator. Samples were re-suspended in protein loading dye, electrophoresed approximately 2 cm into a precast SDS-Polyacrylamide gel and stained with SimplyBlue Safe Stain (Novex). The lane was excised, and the proteins digested in-gel for mass spectrometry on the Orbitrap Lumos.

LC-MS/MS for Immunoprecipitation Experiments

Loading solvent was 3% MeCN, 0.1% FA, analytical solvent A: 0.1% FA and B: MeCN + 0.1% FA. All separations were carried out at 55°C. Samples were loaded at 5 μl/minute for 5 minutes in loading solvent before beginning the analytical gradient. The following gradient was used: 3-40% B over 29 minutes followed by a 3 minute wash at 95% B and equilibration at 3% B for 10 minutes. The following settings were used: MS1: 300-1500 Th, 120,000 resolution, 4x10⁵ AGC target, 50 ms maximum injection time. MS2: Quadrupole isolation at an isolation width of m/z 1.6, HCD fragmentation (NCE 35) with fragment ions scanning in the Orbitrap from m/z 110, 5x10⁴ AGC target, 60 ms maximum injection time, ions accumulated for all parallelisable time. Dynamic exclusion was set to +/- 10 ppm for 60 s. MS2 fragmentation was trigged on precursors 5x10⁴ counts and above.

RNAseq Analysis

RNAseq analysis was performed in biological triplicate at three time points of infection: 0h (mock), 24 h and 72 h. For each sample, RNA was extracted from a 75 cm² flask of HFFFs infected at moi 10 or mock-infected using an RNeasy Plus kit (Qiagen). Infections and harvests were performed simultaneously with protein samples for experiment WCL2 (Weekes et al., 2014). Poly(A) RNA was enriched using a Poly(A)Purist MAG kit (Thermo). 250 ng of poly(A) RNA from each sample was used to prepare a cDNA library using a PrepX RNA-Seq Library Kit (Wafergen biosystems) on an Apollo 324 (WaferGen biosystems), according to the manufacturer's protocol. The following barcode sequences were used: Mock1 (ATCACGAT); Mock2 (CGATGTAT); Mock3 (TTAGGCAT); 24h_1 (TGACCAAT); 24h_2 (ACAGTGAT); 24h_3 (GCCAATAT); 72h_1 (CAGATCAT); 72h_2 (ACTTGAAT); 72h_3 (GATCAGAT). The resulting libraries were quantified using an Agilent Bioanalyser 2100 (Agilent Technologies, Santa Clara, CA), then were pooled for sequencing on a single lane of Illumina HiSeq2500 (1 × 50 bp reads).

Plasmid Construction

For exogenous gene expression, a V5-tagged UL145 construct was amplified from an adenoviral template, which expressed UL145-V5 under the control of a CMV promoter. Primers were designed to recognise the 3’ end of the CMV promoter (forward) and the V5 tag (reverse) (Key Resources Table). A control construct was prepared by annealing two oligonucleotides. Both primers and oligonucleotides had flanking Gateway attB sequences (Key Resources Table). PCR employed the PfuUltra II Fusion HS DNA Polymerase (Agilent). Constructs were subsequently cloned into lentiviral destination vector pHAGE-pSFFV using the Gateway system (Thermo Scientific). pHAGE-pSFFV has a spleen focus-forming virus (SFFV) promoter replacing the CMV promoter in pHAGE-pCMV to prevent promoter inactivation during HCMV infection. For shRNA, two partially complementary oligonucleotides (Table S7C) were annealed. The resulting product was ligated into the pHR-SIREN vector (gift from Prof. Paul Lehner, University of Cambridge) as a BamHI–EcoRI fragment using T4 ligase (Thermo Scientific). All constructed plasmids were transformed into Alpha-Select Silver Efficiency Competent E. coli cells (Bioline) at 42°C for 1 min and selected on antibiotic-containing LB agar plates.

Stable Cell Line Production

Lentiviral particles were generated through transfection of HEK293T cells with the lentiviral transfer vector plus four helper plasmids (VSVG, TAT1B, MGPM2, CMV-Rev1B), using TransIT-293 transfection reagent (Mirus) according to the manufacturer's recommendations. Viral supernatant was typically harvested 48 h after transfection, cell debris was removed with a 0.22 μm filter, and target cells were transduced for 48 h then subjected to antibiotic selection for two weeks.

CRISPR/Cas9-Mediated Gene Disruption

HFFF-TERT cells stably expressing pHRSIN-P_SFFV-Cas9-P_PGK-Hygro (gift from Professor Paul Lehner, University of Cambridge) were transduced with lentivirus employing the pKLV-U6gRNA(BbsI)-PGKpuro2ABFP plasmid (Addgene Plasmid #50946), that constitutively expressed a given gRNA (Table S7C). Confirmation of protein level reduction in low passage polyclonal populations of cells expressing both Cas9 and the gRNA of choice was then achieved by immunoblot. Polyclonal selected cell populations were used in this study.

siRNA Knockdown

24 h prior to transfection, 3x10⁵ 293Ts constitutively expressing UL145-V5 or control were plated in 6 well plates. Cells were transfected with a pool of CUL4A siRNAs (L-012610-00, Dharmafect) or a pool of non-targeting siRNAs (D-001810-10, Dharmafect) with DHARMAfect 1 Transfection Reagent (T-2001, Dharmafect) giving a final siRNA concentration of 25 nM. Cellular lysates were harvested 48 h post transfection for immunoblot.

Immunoblotting

HFFF-TERTs were used for all experiments apart from Figure 6C, where 293T cells were used. For most immunoblots, cells were lysed with RIPA buffer (Cell Signaling) containing Complete Protease Inhibitor Cocktail (Roche) and then lysates were sonicated. For cells infected by single gene deletion viruses, 6 M Guanidine whole cell lysates were precipitated using a ProteoExtract protein precipitation kit (Calbiochem) and re-dissolved in 2% SDS/Tris 200 mM pH 8.5 with sonication. Protein concentration was measured by BCA (Pierce). Lysates were reduced with 6X Protein Loading Dye (Tris 375 mM pH 6.8, 12% SDS, 30% glycerol, 0.6 M DTT, 0.06% bromophenol blue) for 5 min at 95°C. 50 μg of protein for each sample was separated by PAGE using 4-15% TGX Precast Protein Gels (Bio-rad), then transferred to PVDF membranes using Trans-Blot Systems (Bio-rad). The following primary antibodies were used: anti-HLTF (ab17984, Abcam), anti-HCMV IE1/2 (ab53495, Abcam), anti-GAPDH (MAB5718, R&D Systems), anti-V5 (MA5-15253, Thermo), anti-Sp100 (GTX131570, GeneTex). Secondary antibodies were IRDye 680RD goat anti-mouse (925-68070, LI-COR) and IRDye 800CW goat anti-rabbit (925-32211, LI-COR). Fluorescent signals were detected using a LI-COR Odyssey, and images were processed using Image Studio Lite (LI-COR).

Restriction Assay and Flow Cytometry

24 h prior to infection, 1.5x10⁵ HFFF-TERTs stably expressing shRNA constructs targeted against Sp100, HLTF or control were plated in 24 well plates. Cells were infected with HCMV UL36-GFP at a range of low moi (0.003 – 0.3). The requisite volume of viral stock was added to 150 μl DMEM, and mixed gently prior to being applied to cells. Cells were incubated with virus for 2 h prior to replacing the medium. 24 h after infection, cells were harvested and fixed in 4% paraformaldehyde. 30,000 events were acquired with a FACSCalibur flow cytometer and analysed with FlowJo vX software. A similar approach was performed for polyclonal selected CRISPR cell populations.

Immunofluorescence Microscopy

HFFF-TERTs were infected on coverslips with a recombinant Merlin strain with a C-terminal UL145 V5 tag, at moi 0.1 for 24 h. Cells were then cross-linked with fixation buffer (Biolegend), permeabilised with ice-cold methanol, and blocked with Human TruStain FcX (Biolegend). Two primary antibodies were used: rabbit anti-HLTF (ab17984, Abcam) and mouse anti-V5 (MA5-15253, Thermo). Secondary antibodies were anti-mouse Alexa Fluor 488 (4408S, Cell Signaling) and anti-rabbit Alexa Fluor 647 (A31573, Thermo). Cell nuclei were stained with DAPI (Cell Signaling). Fluorescence were observed using a confocal microscope (Zeiss LSM 710).

RT-qPCR

Total RNA from mock- or HCMV-infected HFFF-TERTs was extracted using an RNeasy Mini Kit (Qiagen). cDNA was synthesized using GoScript Reverse Transcriptase (Promega), followed by RT-qPCR using Fast SYBR Green Master Mix (Applied Biosystems) and 7500 Fast & 7500 Real-Time PCR Systems (Applied Biosystems). Primers targeting HCMV UL145 or GAPDH (as an internal control) are shown in the Key Resources Table. The PCR program started with activation at 95°C for 2 min, followed by 40 cycles of denaturation at 95°C for 5 s and annealing/extension at 60°C for 30 s. Melting curve analyses were performed to verify the amplification specificity. All mock-infected samples exhibited non-singular melting curves, indicating non-specific amplification; values for these samples were set to zero.

Data Analysis

Mass spectra were processed using a Sequest-based software pipeline for quantitative proteomics, “MassPike”, through a collaborative arrangement with Professor Steven Gygi’s laboratory at Harvard Medical School. MS spectra were converted to mzXML using an extractor built upon Thermo Fisher’s RAW File Reader library (version 4.0.26). In this extractor, the standard mzxml format has been augmented with additional custom fields that are specific to ion trap and Orbitrap mass spectrometry and essential for TMT quantitation. These additional fields include ion injection times for each scan, Fourier Transform-derived baseline and noise values calculated for every Orbitrap scan, isolation widths for each scan type, scan event numbers, and elapsed scan times. This software is a component of the MassPike software platform and is licensed by Harvard Medical School.

A combined database was constructed from (a) the human Uniprot database (26^th January, 2017), (b) the HCMV strain Merlin Uniprot database, (c) all additional non-canonical human cytomegalovirus ORFs described by Stern-Ginossar et al. (Stern-Ginossar et al., 2012), (d) a six-frame translation of HCMV strain Merlin filtered to include all potential ORFs of ≥8 amino acids (delimited by stop-stop rather than requiring ATG-stop) and (e) common contaminants such as porcine trypsin and endoproteinase LysC. ORFs from the six-frame translation (6FT-ORFs) were named as follows: 6FT_Frame_ORFnumber_length, where Frame is numbered 1-6, and length is the length in amino acids. The combined database was concatenated with a reverse database composed of all protein sequences in reversed order. Searches were performed using a 20 ppm precursor ion tolerance. Fragment ion tolerance was set to 1.0 Th. TMT tags on lysine residues and peptide N termini (229.162932 Da) and carbamidomethylation of cysteine residues (57.02146 Da) were set as static modifications, while oxidation of methionine residues (15.99492 Da) was set as a variable modification. For SILAC analysis, the following variable modifications were used: heavy lysine (8.01420 Da), heavy arginine (10.00827 Da), medium lysine (4.02511 Da), medium arginine (6.02013 Da). SILAC-only searches were performed in the same manner, omitting the TMT static modification.

To control the fraction of erroneous protein identifications, a target-decoy strategy was employed (Huttlin et al., 2010). Peptide spectral matches (PSMs) were filtered to an initial peptide-level false discovery rate (FDR) of 1% with subsequent filtering to attain a final protein-level FDR of 1%. PSM filtering was performed using a linear discriminant analysis, as described previously (Huttlin et al., 2010). This distinguishes correct from incorrect peptide IDs in a manner analogous to the widely used Percolator algorithm (https://noble.gs.washington.edu/proj/percolator/), though employing a distinct machine learning algorithm. The following parameters were considered: XCorr, ΔCn, missed cleavages, peptide length, charge state, and precursor mass accuracy.

Protein assembly was guided by principles of parsimony to produce the smallest set of proteins necessary to account for all observed peptides (algorithm described in Huttlin et al., 2010). Where all PSMs from a given HCMV protein could be explained either by a canonical gene or non-canonical ORF, the canonical gene was picked in preference.

In eleven cases, PSMs assigned to a non-canonical or 6FT-ORF were a mixture of peptides from the canonical protein and the ORF. This most commonly occurred where the ORF was a 5’-terminal extension of the canonical protein (thus meaning that the smallest set of proteins necessary to account for all observed peptides included the ORFs alone). In these cases, the peptides corresponding to the canonical protein were separated from those unique to the ORF, generating two separate entries. In a single case, PSM were assigned to the 6FT-ORF 6FT_6_ORF1202_676aa, which is a 5’-terminal extension of the non-canonical ORF ORFL147C. The principles described above were used to separate these two ORFs.

Proteins were quantified by summing TMT reporter ion counts across all matching peptide-spectral matches using ”MassPike”, as described previously (McAlister et al., 2014). Briefly, a 0.003 Th window around the theoretical m/z of each reporter ion (126, 127n, 127c, 128n, 128c, 129n, 129c, 130n, 130c, 131n, 131c) was scanned for ions, and the maximum intensity nearest to the theoretical m/z was used. The primary determinant of quantitation quality is the number of TMT reporter ions detected in each MS3 spectrum, which is directly proportional to the signal-to-noise (S:N) ratio observed for each ion. Conservatively, every individual peptide used for quantitation was required to contribute sufficient TMT reporter ions (minimum of ∼500 per spectrum) so that each on its own could be expected to provide a representative picture of relative protein abundance (McAlister et al., 2014). An isolation specificity filter with a cutoff of 50% was additionally employed to minimise peptide co-isolation (McAlister et al., 2014). Peptide-spectral matches with poor quality MS3 spectra (more than 9 TMT channels missing and/or a combined S:N ratio of less than 100 across all TMT reporter ions) or no MS3 spectra at all were excluded from quantitation. Peptides meeting the stated criteria for reliable quantitation were then summed by parent protein, in effect weighting the contributions of individual peptides to the total protein signal based on their individual TMT reporter ion yields. Protein quantitation values were exported for further analysis in Excel.

For protein quantitation, reverse and contaminant proteins were removed, then each reporter ion channel was summed across all quantified proteins and normalised assuming equal protein loading across all channels. For further analysis and display in Figures, fractional TMT signals were used (i.e. reporting the fraction of maximal signal observed for each protein in each TMT channel, rather than the absolute normalized signal intensity). This effectively corrected for differences in the numbers of peptides observed per protein.

For pulsed SILAC experiments, after protein assembly, medium-labelled peptides (measuring protein degradation), and heavy-labelled peptides (measuring protein synthesis) were extracted then re-assembled into medium- and heavy-labelled proteins using an in-house script written in Python (version 2.7). Protein normalisation across reporter ion channels again assumed equal protein loading (i.e. for each TMT channel, the summed protein S:N including all medium- and heavy-labelled proteins was the same). For each protein, values were further normalised to the time 0 sample for display in Figures. For all TMT or pSILAC experiments, normalised S:N values are presented in Table S1 (‘Data’ worksheet). For SILAC immunoprecipitations, normalisation assumed equal protein loading across all samples.

Although peptides were assigned appropriately to HLA-A alleles, it was not possible confidently to assign peptides to only two HLA-B or HLA-C alleles. For the 5 HLA-B or HLA-C alleles that had the greatest summed number of peptides across all experiments, signal:noise values were further summed to give a single combined result for HLA-B or HLA-C.

Hierarchical centroid clustering based on uncentered Pearson correlation, and k-means clustering were performed using Cluster 3.0 (Stanford University) and visualised using Java Treeview (http://jtreeview.sourceforge.net) unless otherwise noted. Multiple sequence alignment was performed using Clustal Omega (http://www.ebi.ac.uk/Tools/msa/clustalo/) provided by EMBL-EBI.

Comparative Data Analysis Using Proteome Discoverer

To compare data generated using “MassPike” with another platform, we re-analysed raw MS files for the 12, 18 and 24 h inhibitor experiments using Proteome Discoverer 2.2 (Thermo Fisher Scientific) (Figure S7). Data were searched in Sequest against an identical combined database to that described in the “Data analysis” section. Searches were performed using a 20 ppm precursor ion tolerance and fragment mass tolerance of 0.5 Da. TMT tags on lysine residues and peptide N termini and carbamidomethylation of cysteine residues were set as static modifications, and oxidation of methionine residues was set as a variable modification. Percolator (https://noble.gs.washington.edu/proj/percolator/) was used to control the fraction of erroneous protein identifications, with a peptide false discovery rate of 1% for ‘high’ confidence PSMs and 5% for ‘medium’ confidence PSMs. Proteins were subsequently filtered to attain a final protein-level FDR of 1% (‘strict’ criteria) or 5% (‘relaxed’ criteria). Protein assembly was guided by principles of parsimony. MS3 spectra were used for reporter ion based quantitation with a most intense centroid tolerance of 20 ppm. An average reporter S:N value of 10 was used for quantitation and an isolation specificity filter with a cutoff of 50% was employed to minimize peptide co-isolation. Peptides meeting these criteria were summed by parent protein, and quantitation values were exported for further analysis in Excel. For protein quantitation, reverse and contaminant proteins were removed, and then each reporter ion channel was summed across all quantified proteins and normalised assuming equal protein loading across all channels.

Histone Proteomic Ruler

The cellular concentrations of viral proteins in whole cell lysates from the 6 h pSILAC experiment (Figure 7A) were calculated using a ‘proteomic ruler’ approach implemented in the Perseus plugin (http://www.coxdocs.org/doku.php?id=perseus:user:plugins:proteomicruler:estimatecopynumbers) (Wisniewski et al., 2014). This used the mass spectrometry signal of histones from the 6 h HCMV-infected sample to scale other proteins of unknown concentration from the same sample. Briefly, intensity values that had been normalised assuming equal protein loading across all samples were imported into Perseus. Molecular weights of all human and canonical HCMV proteins were obtained from Uniprot. Predicted masses of non-canonical HCMV ORFs and six-frame translations were obtained using an online molecular weight calculator (https://www.bioinformatics.org/sms/prot_mw.html). Scaling was using the histone proteomic ruler, assuming a ploidy of 2.

RNAseq Data Analysis

RNA samples were collected in biological triplicate. Reads were aligned to the Human genome (hg19) and HCMV Merlin strain genome (NC_006273.2) using the aligner STAR version 2.5.2b (http://code.google.com/p/rna-star/). Read counts for each gene/transcript were determined using HTSeq version 0.6.1 (https://pypi.org/project/HTSeq/) with the following optional parameters: --m union,-r pos,-i transcript_id,-a 10,--stranded=no. For human genes, normalisation assumed equal human reads per run; HCMV reads were not included in this normalisation since by 72 h HCMV accounted for a significant proportion of all reads (0h: <0.01%, 24 h: 4.7-5.8%; 72 h: 41.8-42.3% HCMV reads). For HCMV, reads for each barcode were normalised assuming equal total reads (human plus HCMV) per sequencing run. Reads per Kilobase per Million (RPKM) values were calculated in Excel and further normalised to 1. Mean and standard error of the mean (SEM) values are shown in Table S1 and Figures 3, 5, S1–S3, and S5. Although data for HCMV reads is included in Table S1, this analysis may be confounded by overlapping viral transcripts. For Figure 5F, as UL145 transcripts do not overlap with transcripts from the neighboring UL144 and UL146 genes, the sequences detected by RNAseq could be reliably ascribed to UL145 (Sun et al., 2007).

Statistical Analysis

The exact value of n within figures is indicated in the respective figure legends, and refers to the number of biological replicates. Blinding or sample-size estimation was not appropriate for this study. There were no inclusion criteria and no data was excluded.

Figures 1, S2, S3, and S5. The inhibitor experiments were performed in single replicates at each of 12, 18 and 24 h after infection. Protein ‘rescue ratios’ were approximately normally distributed (Figure S1B). The method of significance A was used to estimate the p-value that each ratio was significantly different to 1 (Cox and Mann, 2008). Values were calculated and corrected for multiple hypothesis testing using the method of Benjamini-Hochberg in Perseus version 1.5.1.6 (Cox and Mann, 2008). A corrected p-value <0.01 was considered statistically significant.

Figures 2, S2, S3, and S5. The pSILAC experiments were performed in single replicates at each of 6 and 18 h after infection. With respect to protein concentration, protein degradation typically follows first order kinetics whereas protein synthesis is a zero-order process. For pSILAC data, the rate of protein decline in mock- and HCMV-infected samples was therefore estimated using exponential regression in Excel and the formula [protein] (t) = e^{Kdeg x t} where Kdeg is the rate constant for degradation, and should be negative for degraded proteins. A degradation ratio was calculated by rdeg = Kdeg_HCMV/Kdeg_mock. In cases where this ratio could not be calculated because Kdeg_mock was greater than 0, a fold change (FC_HCMV) in protein abundance in the HCMV-infected sample at 6 h (6 h pSILAC experiment) or 18 h (18 h pSILAC experiment) was instead used, defined by FC_HCMV = 1/e^{Kdeg(HCMV) x t}. Protein half-life was estimated by t_1/2 = ln(0.5)/Kdeg. The corresponding rates of protein synthesis were estimated using linear regression in Excel and the formula [protein] (t) = Ksyn x t where Ksyn is the rate constant for synthesis. We determined if Kdeg_HCMV was significantly different to Kdeg_mock, and if Ksyn_HCMV was significantly different to Ksyn_mock using an in-house script written in R (version 3.4.2). For each peptide, the difference in paired normalised signal:noise values at each of the five measured time points was calculated. A simple linear regression model without an intercept was fitted to paired difference data from all peptides for each protein, and a p-value calculated for the null hypothesis of the slope being zero. All p-values were corrected for multiple hypothesis testing using the method of Benjamini-Hochberg (Benjamini and Hochberg, 1995). A Benjamini-Hochberg-corrected p-value <0.05 was considered statistically significant.

For Figure 2E, all viral proteins were included where Ksyn_HCMV was significantly higher than Ksyn_mock at p<0.05. As the measurement of viral proteins in mock-infected samples was at the level of noise, the value of the ratio Ksyn_HCMV/Ksyn_mock was not considered. All human proteins were included with Ksyn_HCMV/Ksyn_mock >3 and p<0.05.

Figures 3, S2, S3, and S5. The WCL2 experiment was performed using single replicates collected at multiple time points as detailed in the figure, apart from the mock sample which was collected in biological duplicate (Weekes et al., 2014). The RNAseq experiment was performed in biological triplicate at 0, 24 and 72 h after infection. Mean and SEM were calculated for normalised RPKM values for each time point 0, 24, 72 h (n=3). Fold change at 24h was calculated from mean RPKM(24 h)/mean RPKM(0h). A similar value was calculated for 72 h data. A Benjamini-Hochberg corrected student’s t-test was used to estimate the p-value for the hypothesis that a given transcript was expressed significantly differently at 24 or 72 h compared to mock infection. For protein expression from experiment WCL2, fold change at time t was calculated from S:N (t) / S:N (0h). p-values that a given protein was expressed significantly differently at 24 or 72 h compared to mock infection were estimated using Benjamini-Hochberg-corrected significance A values (Cox and Mann, 2008). XLStat (Addinsoft) was used to calculate the summed distance of each protein from its cluster centroid (Figure S4), and k-means clustering was performed in Cluster 3.0 (Stanford University).

Figures 5 and S6. The block deletion screens were conducted in partial biological duplicate as detailed below. Two block viral gene-deletion screens Block1 and Block2 were conducted. For each protein in each screen, a mean (μ) and standard deviation (σ) of all normalised S:N values was calculated. In each case, the maximum (x) value was omitted. For example, for HLTF in Figure 5A, μ and σ were calculated using values for wt1, wt2, RL10-UL1, RL11-UL11, UL2-UL11, US1-US11, US18-US22, US29-US34A but not the maximum U_L/b’. The formula z = (x – μ)/σ was then applied to calculate a z-score. Fold change (FC) compared to wild-type (wt) infection was calculated from normalised S:N values using FC = x/wt1. Because of the limits of multiplexing with TMT, block deletion viruses AD169 (U_L/b’), ΔUS27-28, ΔUL13-20, ΔUS12-17 were only examined in a single screen. All other block deletion viruses were examined in both screens. For each experiment, a given protein was initially assigned to the block corresponding to the TMT channel with the maximum S:N. To combine results to assign an overall block to each protein:

Pathway Analysis

The Database for Annotation, Visualisation and Integrated Discovery (DAVID) was used to determine pathway enrichment (Huang da et al., 2009). A given cluster was always searched against a background of all proteins quantified within the relevant experiment. For Figure 4A, DAVID analysis examined all proteins identified as degraded at any of the three time points for the inhibitor screen, proteins degraded in either of the two time courses for the pSILAC screen, or proteins degraded at either time point for the RNA/protein screen.