Characterization of target gene regulation by the two Epstein-Barr virus oncogene LMP1 domains essential for B-cell transformation

ABSTRACT The Epstein-Barr virus (EBV) oncogene latent membrane protein 1 (LMP1) mimics CD40 signaling and is expressed by multiple malignancies. Two LMP1 C-terminal cytoplasmic tail regions, termed transformation essential sites (TES) 1 and 2, are critical for EBV transformation of B lymphocytes into immortalized lymphoblastoid cell lines (LCL). However, TES1 versus TES2 B-cell target genes have remained incompletely characterized, and whether both are required for LCL survival has remained unknown. To define LCL LMP1 target genes, we profiled transcriptome-wide effects of acute LMP1 CRISPR knockout (KO) prior to cell death. To then characterize specific LCL TES1 and TES2 roles, we conditionally expressed wildtype, TES1 null, TES2 null, or double TES1/TES2 null LMP1 alleles upon endogenous LMP1 KO. Unexpectedly, TES1 but not TES2 signaling was critical for LCL survival. The LCL dependency factor cFLIP, which plays obligatory roles in blockade of LCL apoptosis, was highly downmodulated by loss of TES1 signaling. To further characterize TES1 vs TES2 roles, we conditionally expressed wildtype, TES1, and/or TES2 null LMP1 alleles in two Burkitt models. Systematic RNAseq analyses revealed gene clusters that responded more strongly to TES1 vs TES2, that respond strongly to both or that are oppositely regulated. Robust TES1 effects on cFLIP induction were again noted. TES1 and 2 effects on expression of additional LCL dependency factors, including BATF and IRF4, and on EBV super-enhancers were identified. Collectively, these studies suggest a model by which LMP1 TES1 and TES2 jointly remodel the B-cell transcriptome and highlight TES1 as a key therapeutic target. IMPORTANCE Epstein-Barr virus (EBV) causes multiple human cancers, including B-cell lymphomas. In cell culture, EBV converts healthy human B-cells into immortalized ones that grow continuously, which model post-transplant lymphomas. Constitutive signaling from two cytoplasmic tail domains of the EBV oncogene latent membrane protein 1 (LMP1) is required for this transformation, yet there has not been systematic analysis of their host gene targets. We identified that only signaling from the membrane proximal domain is required for survival of these EBV-immortalized cells and that its loss triggers apoptosis. We identified key LMP1 target genes, whose abundance changed significantly with loss of LMP1 signals, or that were instead upregulated in response to switching on signaling by one or both LMP1 domains in an EBV-uninfected human B-cell model. These included major anti-apoptotic factors necessary for EBV-infected B-cell survival. Bioinformatics analyses identified clusters of B-cell genes that respond differently to signaling by either or both domains.

(A)Relative mean + standard deviation (SD) live cell numbers from CellTitreGlo analysis of n=3 replicates of Cas9+ GM12878 LCLs, transduced with lentiviruses that expressed control or LMP1 sgRNAs and puromycin selected, for 0 versus 2 days.
(B) RT-PCR analysis of CCL22 mRNA abundance in Cas9+ GM12878 post-transduction with lentiviruses that expressed control or LMP1 sgRNA and puromycin selected for 2 days, as in Fig. 1B.Values from cells with control sgRNA were set to 1, and mean fold-change of CCL22 mRNA abundance + SD in cells with LMP1 sgRNA are shown.p-values were determined by one-sided Fisher's exact test from two independent experiments, each with two technical replicates.***p<0.001.The same RNA used for RNA-seq (Fig. 1B) was used for these qPCR experiments.
(C) RT-PCR analysis of EBI3 mRNA abundance in Cas9+ GM12878 post-transduction with lentiviruses that expressed control or LMP1 sgRNA and puromycin selected for 2 days, as in Fig. 1B.Values from cells with control sgRNA were set to 1, and mean fold-change of EBI3 mRNA abundance + SD in cells with LMP1 sgRNA are shown.p-values were determined by one-sided Fisher's exact test from two independent experiments, each with two technical replicates.***p<0.001.The same RNA used for RNA-seq (Fig. 1B) was used for these qPCR experiments.
(D) RT-PCR analysis of IRF4 mRNA abundance in Cas9+ GM12878 post-transduction with lentiviruses that expressed control or LMP1 sgRNA and puromycin selected for 2 days, as in Fig. 1B.Values from cells with control sgRNA were set to 1, and mean fold-change of IRF4 mRNA abundance + SD in cells with LMP1 sgRNA are shown.p-values were determined by one-sided Fisher's exact test from two independent experiments, each with two technical replicates.***p<0.001.The same RNA used for RNA-seq (Fig. 1B) was used for these qPCR experiments.
(E) Scatter plot analysis cross-comparing the significance of changes in LCL dependency factor expression upon GM127878 LMP1 KO versus the CRISPR screen significance score for selection against sgRNAs in LCL vs Burkitt dependency factor analysis (25).Shown on the Y-axis are -log10 transformed P-values from RNAseq analysis of GM12878 LCLs transduced with lentiviruses expressing LMP1 versus control sgRNA (as in Fig. 1F), versus -log10 transformed P-values from CRISPR LCL vs Burkitt cell dependency factor analysis (25).Higher Y-axis scores indicate more significant differences in expression for the indicated genes in GM12878 with LMP1 vs control sgRNA.Higher X-axis scores indicate a stronger selection against sgRNA targeting the indicated genes in GM12878 LCLs versus P3HR1 Burkitt cells over 21 days of cell culture.Shown are genes with p<0.05 in both analyses.
(F) Volcano plot analysis visualizing KEGG Hodgkin lymphoma pathway gene -Log10 (P-value) on the yaxis versus Log2 transformed fold change in mRNA abundances on the x-axis of GM12878 genes in cells expressing LMP1 versus control sgRNA (as in Fig. 1F).P-value <0.05 and >2-fold change mRNA abundance cutoffs were used.Mean ± SD of fold change plasma membrane annexin V values from n=3 independent experiments, using GM12878 with the indicated control or LMP1 sgRNA and rescue cDNA expression.Values in GM12878 with control sgRNA and no LMP1 rescue cDNA were set to 1.  (A) Heatmap analysis of CRISPR defined LCL dependency factor gene relative row Z-scores from RNAseq of GM12878 expressing LMP1 sgRNA and the indicated rescue cDNA, as in Fig. 3.The Z-score scale is shown at bottom, where blue and red colors indicate lower versus higher relative expression, respectively.Two-way ANOVA P-value cutoff of <0.05 and >2-fold gene expression cutoffs were used.

LMP1
(B) Heatmap analysis of KEGG Hodgkin Lymphoma pathway gene relative row Z-scores from RNAseq of GM12878 expressing LMP1 sgRNA and the indicated rescue cDNA, as in Fig. 3. Two-way ANOVA P-value cutoff of <0.05 and >2-fold gene expression cutoffs were used.
(C) RT-PCR analysis of CCL22 mRNA abundance in GM12878 LCLs transduced with lentivirus expressing LMP1 sgRNA and induced for WT, TES1m or TES2m rescue cDNA expression for 6 days, as in Fig. 3A.The same RNA used for RNA-seq in Fig. 3A     (  (A) Volcano plot analysis of host transcriptome-wide genes differentially expressed in BL-41 cells conditionally induced for WT LMP1 expression for 24h by 250 ng/ml Dox versus in mock induced cells.Higher X-axis fold changes indicate genes more highly expressed in cells with WT LMP1 expression, whereas lower X-axis fold changes indicate higher expression in cells mock induced for LMP1.Data are from n=3 RNAseq datasets.
(B) Enrichr analysis of KEGG pathways most highly enriched in RNAseq data as in (A) amongst genes more highly expressed in BL-41 with WT LMP1 (red) vs amongst genes more highly expressed with mock LMP1 induction (blue).
(C) Volcano plot cross-comparison of Log2 transformed fold change of host mRNA levels in BL-41 cells (Xaxis) versus Akata cells (Y-axis) uninduced versus induced for WT LMP1 by 250 ng/ml Dox for 24 hours.Selected genes highly WT LMP1 induced in both Burkitt contexts are highlighted in red, whereas selected genes suppressed by LMP1 in both Burkitt contexts are highlighted in blue.
(D) Volcano plot analysis of host transcriptome-wide genes differentially expressed in BL-41 cells conditionally induced for DM versus WT LMP1 expression for 24h by 250 ng/ml Dox.Higher X-axis fold changes indicate genes more highly expressed in cells with WT LMP1 expression, whereas lower X-axis fold changes indicate higher expression in cells induced for DM LMP1.Data are from n=3 RNAseq datasets.
(B) Enrichr analysis of KEGG pathways most highly enriched in RNAseq data as in (D) amongst genes more highly expressed in BL-41 with WT LMP1 (red) vs amongst genes more highly expressed with DM LMP1 induction (blue).
(C) Volcano plot cross-comparison of Log2 transformed fold change of host mRNA levels in BL-41 cells (Xaxis) versus Akata cells (Y-axis) induced for DM versus WT LMP1 by 250 ng/ml Dox for 24 hours.Selected genes highly WT LMP1 induced in both Burkitt contexts relative to levels in cells with DM LMP1 expression are highlighted in red, whereas selected genes suppressed by WT LMP1 in both Burkitt contexts are highlighted in blue.A) Volcano plot analysis of host transcriptome-wide genes differentially expressed in BL-41 cells conditionally induced for TES1m vs WT LMP1 expression for 24h by 250 ng/ml Dox.Higher X-axis fold changes indicate genes more highly expressed in cells with WT LMP1 expression, whereas lower X-axis fold changes indicate higher expression induced for TES1m LMP1.Data are from n=3 RNAseq datasets.
(B) Enrichr analysis of KEGG pathways most highly enriched in RNAseq data as in (A) amongst genes more highly expressed in BL-41 with WT LMP1 (red) vs amongst genes more highly expressed with TES1m LMP1 induction (blue).
(C) Volcano plot analysis of host transcriptome-wide genes differentially expressed in BL-41 cells conditionally induced for TES2m vs WT LMP1 expression for 24h by 250 ng/ml Dox.Higher X-axis fold changes indicate genes more highly expressed in cells with WT LMP1 expression, whereas lower X-axis fold changes indicate higher expression induced for TES2m LMP1.Data are from n=3 RNAseq datasets.
(D) Enrichr analysis of KEGG pathways most highly enriched in RNAseq data as in (A) amongst genes more highly expressed in BL-41 with WT LMP1 (red) vs amongst genes more highly expressed with TES2m LMP1 induction (blue).
(E) Volcano plot analysis of host transcriptome-wide genes differentially expressed in Akata cells conditionally induced for TES1m vs TES2m LMP1 expression for 24h by 250 ng/ml Dox.Higher X-axis fold changes indicate genes more highly expressed in cells with TES1m LMP1 expression, whereas lower Xaxis fold changes indicate higher expression induced for TES2m LMP1.Data are from n=3 RNAseq datasets.
(F) Enrichr analysis of KEGG pathways most highly enriched in RNAseq data as in (E) amongst genes more highly expressed in Akata with TES1m LMP1 (red) vs amongst genes more highly expressed with TES2m LMP1 induction (blue).
(G) Volcano plot analysis of host transcriptome-wide genes differentially expressed in BL-41 cells conditionally induced for TES1m vs TES2m LMP1 expression for 24h by 250 ng/ml Dox.Higher X-axis fold changes indicate genes more highly expressed in cells with TES1m LMP1 expression, whereas lower Xaxis fold changes indicate higher expression induced for TES2m LMP1.Data are from n=3 RNAseq datasets.
(B) Enrichr analysis of KEGG pathways most highly enriched in RNAseq data as in (G) amongst genes more highly expressed in BL-41 with TES1m LMP1 (red) vs amongst genes more highly expressed with TES2m LMP1 induction (blue).(A) Volcano plot analysis of host genes differentially expressed upon WT LMP1 induction in Akata (X-axis) versus upon LMP1 KO in GM12878 (Y-axis).Shown are Log2 transformed mRNA fold change values for Akata cells mock induced versus induced for LMP1 WT expression for 24 hours (X-axis) versus upon expression of LMP1 vs control sgRNA in GM18278 for 48 hours.Genes more highly expressed in mockinduced Akata have higher x-axis values, whereas genes more highly expressed in Akata induced for WT LMP1 have lower x-axis values.Likewise, genes with higher expression with control sgRNA expression have higher y-axis values, whereas genes with lower expression in GM12878 Figure S1 Figure S2 A 4 Figure S3A LMP1 cDNA: Figure S3.Characterization of TES1 vs TES2 LCL dependency factor and Hodgkin lymphoma pathway targets.

Figure S9 .
FigureS9 GM LMP1 KO and Akata Uniduced vs WT Figure S10 A) K-means heatmap analysis of RNAseq datasets from n=3 replicates generated in EBV-BL-41 Burkitt cells with conditional LMP1 WT, TES1m, TES2m or DM expression induced by 250 ng/ml doxycycline for 24 hours.The heatmap visualizes host gene Log2 Fold change across the four conditions, divided into six clusters.A two-way ANOVA P value cutoff of <0.01 and >2-fold gene expression were used.# of genes in each cluster is indicated at right.
Genes more highly expressed in Akata with TES1m than WT LMP1 expression have higher x-axis values, whereas genes more highly expressed in Akata induced for WT LMP1 than TES1m have lower x-axis values.Likewise, GM12878 genes with higher expression with TES1m rescue have higher y-axis values, whereas genes with lower expression in GM12878 with TES1m than WT rescue have lower Y-axis values.P value <0.05 and >2-fold gene expression cutoffs were used.(C)Volcanoplotanalysis of host genes differentially expressed upon TES2m vs WT LMP1 induction in Akata (X-axis) versus upon rescue of LMP1 KO GM12878 with TES2m versus WT LMP1 (Y-axis).Shown are Log2 transformed mRNA fold change values for Akata cells induced for TES2m versus WT LMP1 expression for 24 hours (X-axis) versus upon rescue of GM18278 LMP1 KO with TES2m vs WT LMP1 cDNA, as in Fig 3.Genes more highly expressed in Akata with TES2m than WT LMP1 expression have higher x-axis values, whereas genes more highly expressed in Akata induced for WT LMP1 than TES2m have lower x-axis values.Likewise, GM12878 genes with higher expression with TES2m rescue have higher y-axis values, whereas genes with lower expression in GM12878 with TES2m than WT rescue have lower Y-axis values.P value <0.05 and >2-fold gene expression cutoffs were used.