Quantitative temporal viromics: an approach to investigate host-pathogen interaction

Abstract

A systematic quantitative analysis of temporal changes in host and viral proteins throughout the course of a productive infection could provide dynamic insights into virus-host interaction. We developed a proteomic technique called "quantitative temporal viromics" (QTV), which employs multiplexed tandem-mass-tag-based mass spectrometry. Human cytomegalovirus (HCMV) is not only an important pathogen but a paradigm of viral immune evasion. QTV detailed how HCMV orchestrates the expression of >8,000 cellular proteins, including 1,200 cell-surface proteins to manipulate signaling pathways and counterintrinsic, innate, and adaptive immune defenses. QTV predicted natural killer and T cell ligands, as well as 29 viral proteins present at the cell surface, potential therapeutic targets. Temporal profiles of >80% of HCMV canonical genes and 14 noncanonical HCMV open reading frames were defined. QTV is a powerful method that can yield important insights into viral infection and is applicable to any virus with a robust in vitro model.

Graphical abstract

Figure 1

Temporal Plasma Membrane Profiling of HCMV-Infected Fibroblasts (A) Workflow of experiments PM1 and WCL1. (B) Hierarchical cluster analysis of all proteins quantified in experiment PM1 and annotated “plasma membrane,” “cell surface,” “extracellular,” or “short GO” (Figure S1A). (C) All ABC transporters quantified. One-way ANOVA with multiple hypothesis correction: ^∗p < 0.005, ^∗∗p < 0.0001. (D) Quantitation of all HCMV proteins reported present at the surface of infected fibroblasts. gB, gH, gL, gO, and UL132 are virion envelope glycoproteins expressed late in infection. One-way ANOVA with multiple hypothesis correction: ^∗p < 0.005, ^∗∗p < 0.0001. (E) Principal component analysis of all quantified proteins from experiments PM1 and WCL1 confirmed that biological replicates were highly reproducible and suggested that the major source of variability within a given experiment was duration of infection. (F) Correlation between proteins quantified in experiments PM1 and PM2. See also Figure S1 and Table S1.

Figure 2

Temporal WCL Analysis of HCMV-Infected Fibroblasts Demonstrates Exquisite Regulation of ISGs (A) Hierarchical cluster analysis of all proteins quantified in experiment WCL2, and enlargement of the top cluster A that included multiple IFN-induced antiviral proteins. Right panels, example temporal profiles. The y axis shows relative abundance of each protein. Red diamonds, 12 hr after infection with irradiated HCMV. (B) Immunoblots of HFFF infected with HCMV confirm proteomic profiles. (C) Interferon-induced proteins were more potently upregulated by productive infection than infection with irradiated HCMV. For each protein, signal:noise from the 12 hr irradiated virus or productive infection samples, and both mock samples was normalized to 1. The order of samples mock 1 and mock 2 was randomized into mock (a) and mock (b). The difference in normalized signal:noise was then calculated as indicated, and histograms were plotted. Where <0.05% of proteins were upregulated by infection with irradiated virus, 2% of proteins were upregulated by infection with live virus. These included 69 viral proteins and 84 human proteins, of which 39% are known ISGs. See also Figure S2 and Table S3.

Figure 3

Modulation of Intracellular Signaling Pathways during HCMV Infection (A) Quantitation of interferon induction and response pathways. The temporal profile of each protein is shown over 96 hr of infection, and colored red (downregulation), blue (unchanged), or green (upregulation). Data were derived from experiment WCL2 apart from IFNAR1 and IFNAR2, from PM2. Expression of certain ISGs is known to occur in the absence of IFN, in an IRF3-dependent manner. PRR, pattern recognition receptors. (B) Average temporal profiles from 3-class k-means clustering of proteins quantified in experiments WCL2 and PM2. The three classes divided proteins into downregulated (red), unchanged (blue), or upregulated (green). (C) Enrichment of KEGG pathways within each class was determined using DAVID software, against a background of all quantified proteins. Benjamini-Hochberg adjusted p values are shown for each indicated bar (^∗p < 0.00001, ^∗∗p < 0.0001, ^∗∗∗p < 0.01, ^∗∗∗∗p < 0.05). Individual pathways are shown in Figures S3A–S3C, and pathway members are shown in Table S4.

Figure 4

Temporal Changes in Known and Putative Cell-Surface Immunomodulators (A) Temporal profiles of known NK ligands whose modulation by HCMV had not previously been recognized. (B) Temporal profiles of T cell ligands not previously known to be modulated during infection. (C) Temporal profiles of all quantified protocadherins. (D) Validation of the temporal profile of PDCHγC3 by flow cytometry. Red diamonds, 12 hr after infection with irradiated HCMV. See also Figure S4, Table S5, and Data S1.

Figure 5

Definition of Temporal Classes of HCMV Gene Expression (A) The k-means method was used to cluster all quantified HCMV proteins (experiment WCL2) into four or five classes. Shown are the average temporal profiles of each class. With four classes, proteins grouped into classes similar to the classical IE/E/E-L/L cascades. With five classes, a distinct temporal profile appeared (blue). (B) Number of temporal classes of HCMV gene expression. The summed distance of each protein from its cluster centroid was calculated for one to 14 classes and plotted. The point of inflexion fell between five and seven classes. (C) Top: temporal profiles of proteins in each k-means class (experiment WCL2) were subjected to hierarchical clustering by Euclidian distance. Both UL112 and isoform p50 of UL112 (UL112-2) were quantified. UL122 was excluded from clustering due to uncertainty in peptide assignment. Bottom: experiment WCL3. Viral protein profiles were further assessed in the presence or absence of the viral DNA replication inhibitor PFA. Proteins are displayed in the same order as the clusters defined in the upper panels. (D) Temporal profiles of typical proteins from each cluster (upper panel), and the corresponding profiles in the presence or absence of PFA (lower panel). See also Figure S5, Table S6, and Data S1.

Figure 6

HCMV Proteins Quantified at the Surface of Infected Fibroblasts (A) Histogram of peptide ratios for all GO-annotated proteins quantified in experiments PM1 or PM2. “PM only,” not detected in experiments WCL1 or WCL2. “PM annotation”: “plasma membrane”, “cell surface,” “extracellular,” or “short GO.” (B) Temporal profiles of all high-confidence PM proteins (Table S6). Virion envelope glycoproteins were generally detected significantly earlier in whole-cell lysates than in plasma membrane samples. Arrows, quantitation of fusion or binding of the virion envelope and the plasma membrane. See also Figure S6 and Table S7.

Figure 7

QTV Provides Mechanistic Insights into Downregulated Cell-Surface Targets (A) Proteins that are known to be sequestered within the cell accumulated in WCL samples during infection. (B) Proteins targeted for lysosomal or proteasomal degradation declined during infection. Red diamonds, 12 hr after infection with irradiated HCMV. See also Figures S7A and S7B.

Figure S1

Temporal Plasma Membrane Profiling of HCMV-Infected Fibroblasts, Related to Figure 1 (A) Gene ontology annotation of proteins quantified in experiment PM1. Peptides were analyzed by mass spectrometry either unfractionated, or after division into 6 fractions using offline strong cation exchange. ‘'Short GO” refers to a subset of proteins annotated by GO as integral to the membrane, but with no subcellular assignment and a short 4- or 5-part GO cellular compartment term (Weekes et al., 2012). (B) Quantitation of all cell surface proteins that exhibit previously reported changes during productive HCMV infection in HFFFs. Protein temporal profiles correlated well between repeat time courses PM1 (upper panels) and PM2 (lower panels). MICB and ULBP1 were not quantified in experiment PM1. Because of sequence similarities it was not possible to distinguish individual isoforms of HLA-B, and these data are not included. Red diamonds – irradiated HCMV infection at 12h. One-way ANOVA with multiple hypothesis correction: ^∗p < 0.005, ^∗∗p < 0.0001. Red diamonds – 12h after infection with irradiated HCMV. (C) Hierarchical clustering of proteins quantified in whole-cell lysate experiment WCL1. Fold change was limited to a maximum of 50. (D) Workflow of 10-plex TMT experiments. ^∗12h after infection with irradiated HCMV. (E) Hierarchical clustering of proteins quantified in plasma membrane experiment PM2. Fold change was limited to a maximum of 50. (F) Correlation between proteins quantified in both experiments WCL1 and WCL2.

Figure S2

Modulation of Viral Restriction Factors during HCMV Infection, Related to Figure 2 (A) Sp100 is known to be targeted for degradation by HCMV IE1 (Kim et al., 2011a). Modulation of SAMHD1, ZAP and SLFN11 had not previously been recognized. (B) Quantitation of 21 TRIMs during infection. Results are shown from experiments WCL2 (top sets of panels) and WCL1 (bottom sets of panels). 8 TRIMs were not quantified in experiment WCL1. One-way ANOVA with multiple hypothesis correction (experiment WCL1): ^∗p < 0.005. Red diamonds – 12h after infection with irradiated HCMV.

Figure S3

QTV Provides Insights into Modulated Signaling Pathways, Related to Figure 3 (A) The Toll-like receptor signaling pathway and its modulation during HCMV infection, based on the KEGG pathway (Kanehisa and Goto, 2000). Proteins (ovals) are shaded to correspond to their clusters (Figure 3B): red (downregulated), blue (unchanged), green (upregulated), gray (not quantified). Edges (from KEGG) are shaded green with arrows, to indicate that protein A activates protein B and pink to indicate that protein A inhibits protein B. A single gray/black edge with arrow indicates that protein A affects protein B - not labeled as activation or inhibition. A paired gray/black arrow between proteins A and B: protein A and B bind to or are associated with one another. Phosphorylation events are labeled with p+. (B) Immunoblots of HFFF infected with HCMV confirm proteomic profiles for three TLR-pathway members. (C) The gap junction pathway, shaded as for Figure S3A. (D) Modulation of cell surface wnt receptors. 11/13 quantified canonical and non-canonical wnt receptors were downregulated. The WCL2 β-catenin profile demonstrated a modest decline over time. Red diamonds – 12h after infection with irradiated HCMV.

Figure S4

Validation of Temporal Profiles by Flow Cytometry and Immunoblot and Investigation of Novel Immune Ligands, Related to Figure 4 (A) PM and WCL profiles of MHC-I (HLA-A11) and CD99, validated by flow cytometry and immunoblot. ≥ 95% of HFFF were routinely infected by HCMV, assessed by flow cytometry for cell surface MHC-I. (B) Representative intracellular flow cytometry of 24h-infected HFFF with anti-IE1 confirms > 95% infection efficiency. (C) Flow cytometry of HFFF infected with HCMV confirms proteomic profiles for five additional cell surface proteins. (D and E) NK degranulation assays suggest that CLEC1A and FAT1 are novel activating NK ligands. Top panels – validation of temporal PM and WCL profiles by flow cytometry and immunoblot. CLEC1A was not quantified in any WCL QTV experiments but accumulated by immunoblot of whole-cell lysates, while depleting from the PM. Bottom panels – target cells underwent siRNA knockdown of CLEC1A (D) or FAT1 (E) and were then incubated with stimulated polyclonal NK cells from each of three donors. Degranulation of NK cells in response to both CLEC1A and FAT1 knockdown targets was significantly reduced compared to control. Error bars: +/− SEM (donors A, B), +/− range (donor C). ^∗ two-tailed p-value < 0.05, ^∗∗ two-tailed p-value < 0.005. Cell surface MHC-I was unaffected by siRNA (right bottom panels, staining with W6-32 antibody or control Ig). (F) CD8+ T cell degranulation assay suggests that CEACAM1 is a novel inhibitory ligand for CMV-specific cytotoxic T cells. Top panel – validation of temporal PM profile by flow cytometry. Bottom panels (left) – flow cytometry of a CD8+ T cell line specific to the HCMV HLA-A2 restricted IE1 peptide VLEETSVML confirmed CEACAM1 surface expression. (right) HCMV peptide-specific CD8+ effector cells were incubated with autologous fibroblasts that had been infected with HCMV for 72h then pulsed with peptide or left unpulsed. Effectors and targets were treated with control Ig or anti-CEACAM1. HCMV-specific T cell degranulation was significantly increased with CEACAM1 block. Error bars ± SEM. ^∗ two-tailed p-value < 0.0001. y-axes of QTV plots represent relative abundance, and y-axes of flow cytometry plots represent % of max.

Figure S5

Further Details of Specific HCMV Proteins and Protein/mRNA Profiles, Related to Figure 5 (A) All peptides quantified from the major immediate early region spanning UL122 (IE2) – UL123 (IE1). Peptides from exon 4 are unique to UL123. Peptides from exons 1-3 were assigned to IE1 protein by our data processing software, according to the principles of parsimony. Expression of ten exon 5 peptides corresponding to ORFL265C.iORF1 (lower panel) peaked late at 96h, in comparison to a single peptide N-terminal to this region, which is likely to have derived from UL122 itself. The lower panel shows a map of internal ORFs detected by exon 5 ribosomal footprinting (Stern-Ginossar et al., 2012). ^∗ in peptide sequence: methionine oxidation. (B) Relationship between four novel ORFs and their canonical HCMV counterparts, with temporal profiles. Each of the novel ORFs were quantified based only on unique peptides that could only have originated from that ORF. Peptides that could either have originated from the canonical protein or the novel ORF were assigned to the canonical protein. (C) k-means clusters of (i) all quantified canonical HCMV proteins with 9 novel ORFs (experiment WCL2, also shown in Figure 5A) and (ii) all quantified canonical HCMV mRNAs and the same 9 novel ORFs (Stern-Ginossar et al., 2012). 5 clusters were selected in each case. The plots shown represent the average temporal profile for each cluster. As there were no intermediate mRNA time points between 5 and 24 hr, or between 24 and 72 hr, there was insufficient information to make an accurate comparison between the central three mRNA clusters and our Tp2, Tp3, or Tp4 class proteins. We therefore used mRNA data to define 3 classes: Tr1, Tr2-4, and Tr5. See Table S6C for details of the class of each protein. (D) Comparison between temporal protein profiles (this study) and mRNA expression profiles (Stern-Ginossar et al., 2012), grouped according to protein class. The 134 viral genes quantified in both studies are shown. For each protein or transcript, expression was normalized to the maximum across the measured time points. There was extremely good correspondence between protein and mRNA temporal profiles for Tp1 and Tp5 protein classes. Correspondence was less good for Tp2-4 protein classes as there were insufficient intermediate RNA time points to determine when maximal mRNA expression occurred. Red diamonds – 12h after infection with irradiated HCMV.

Figure S6

Temporal Profiles of “High-Confidence” Viral PM Proteins that Were Quantified in Experiment PM1, Related to Figure 6 Known virion envelope glycoproteins (starred) were generally detected significantly earlier in whole-cell lysates than in plasma membrane samples (Figure 6). Values shown are averages of two biological replicates, +/− range. See also Table S7.

Figure S7

QTV Informs about Mechanism of Modulation of Cell-Surface Targets, Related to Figure 7 (A) Further examples of mechanistic insights into downregulated cell surface immunomodulators. TRAILR2 is retained in the ER by UL141 (Smith et al., 2013). MICA is retained in the cis-Golgi by UL142 (Ashiru et al., 2009). UL142 was only quantified at the plasma membrane (Figure S5A) and is not shown here. (B) Examples of mechanistic insights into upregulated cell surface ligands, or receptors with complex patterns of expression. Signal peptide from HCMV UL40 acts to promote cell surface expression of HLA-E (Prod’homme et al., 2012). The WCL pattern of expression of UL40 was extremely similar to cell surface HLA-E, and is shown overlying both PM and WCL data to illustrate this. TNFR1 cell surface expression is upregulated by UL138, which has a dominant effect. Other virally encoded functions may downregulate TNFR1 expression (Montag et al., 2011). The WCL pattern of expression of UL138 is shown overlying both PM and WCL data. Red diamonds – 12h after infection with irradiated HCMV.