Multi-modal analysis reveals tumor and immune features distinguishing EBV-positive and EBV-negative post-transplant lymphoproliferative disorders Authors: Jiaying Toh, Andrea J Reitsma, Tetsuya Tajima, Sheren F Younes, Chimere Ezeiruaku, Kayla C Jenkins, Josselyn K Peña, Shuchun Zhao, Xi Wang, Esmond YZ Lee, Marla C Glass, Laurynas Kalesinskas, Ananthakrishnan Ganesan, Irene Liang, Joy A Pai, James T Harden, Francesco Vallania, Edward A Vizcarra, Govind Bhagat, Fiona E Craig, Steven H Swerdlow, Julie Morscio, Daan Dierickx, Thomas Tousseyn, Ansuman T Satpathy, Sheri M Krams, Yasodha Natkunam, Purvesh Khatri, Olivia M Martinez Published: 2024 ABSTRACT SummaryThe oncogenic Epstein-Barr virus (EBV) can drive tumorigenesis with disrupted host immunity, causing malignancies including post-transplant lymphoproliferative disorders (PTLDs). PTLD can also arise in the absence of EBV, but the biological differences underlying EBV(+) and EBV(−) B cell PTLD and the associated host-EBV-tumor interactions remain poorly understood. Here, we reveal the core differences between EBV(+) and EBV(−) PTLD, characterized by increased expression of genes related to immune processes or DNA interactions, respectively, and the augmented ability of EBV(+) PTLD B cells to modulate the tumor microenvironment through elaboration of monocyte-attracting cytokines/chemokines. We create a reference resource of proteins distinguishing EBV(+) B lymphoma cells from EBV(−) B lymphoma including the immunomodulatory molecules CD300a and CD24, respectively. Moreover, we show that CD300a is essential for maximal survival of EBV(+) PTLD B lymphoma cells. Our comprehensive multi-modal analyses uncover the biological underpinnings of PTLD and offer opportunities for precision therapies.Keywords: Epstein-Barr virus, post-transplant lymphoproliferative disorder, B cell lymphoma, CD300a, tumor-immune interactions, multi-cohort analysis, multi-omics, tumor microenvironment Introduction Epstein-Barr virus (EBV) is a broadly disseminated gamma-1 herpesvirus estimated to infect 90%–95% of adults globally.1,2 EBV infection is mostly asymptomatic, with the virus proceeding through a complex life cycle including a productive lytic phase but ultimately establishing a latent infection in a subset of memory B cells.3 In this manner, EBV persists for the lifetime of the host, though periodic reactivation of EBV drives further lytic replication and viral transmission. Host immunity mediated by vigorous natural killer cell and antiviral T cell responses is vital in controlling EBV infection.4,5,6 EBV has potent oncogenic function and is linked to epithelial malignancies including nasopharyngeal carcinoma (NPC) as well as lymphoid malignancies such as Burkitt and Hodgkin lymphoma. Individuals with immune perturbations associated with aging, HIV infection, genetic immunodeficiency, or iatrogenic immunosuppression are at increased risk for the development of EBV(+) B cell lymphoma. In the setting of bone marrow or organ transplantation, the immunosuppression required to prevent graft rejection can impair the development of protective immunity following primary EBV infection and attenuate the immune response during reactivation of latent EBV infection, thereby predisposing the uncontrolled expansion of EBV-transformed B cells.7 Thus, EBV(+) B cell lymphomas constitute a life-threatening manifestation of atypical post-transplant lymphoproliferations,8 collectively termed post-transplant lymphoproliferative disorder (PTLD). The incidence of PTLD varies depending on the organ transplanted with reported incidences of 2%–20%8,9,10 and significant mortality rates.11 Although most B cell PTLDs are associated with EBV, a significant portion of cases are EBV(−) and similarly carry considerable risk.8,12,13 The underlying tumor-intrinsic features, as well as the characteristics of the tumor microenvironment (TME), of EBV(+) versus EBV(−) PTLD are poorly understood. Consequently, current therapeutic strategies are generally similar for EBV(+) and EBV(−) PTLD, including reduction of immunosuppression, chemotherapy, and the anti-CD20 monoclonal antibody rituximab. A better understanding of the molecular characteristics distinguishing EBV(+) and EBV(−) PTLD would advance more precise therapeutic strategies for the treatment of these B cell malignancies. Here, we integrated computational and multi-omics approaches to identify the molecular signatures distinguishing EBV(+) and EBV(−) B cell PTLD, as well as their respective TME. We identified tumor-derived chemokines that shape the TME of EBV(+) B cell PTLD characterized by an enrichment of CD163+ anti-inflammatory monocytes and a reduction in T cells. We also establish that CD300a expression on B lymphoma cells is specifically associated with EBV(+) B cell PTLD. We use gene editing approaches and pre-clinical models to demonstrate that CD300a is required for maximal survival of EBV(+) B cell lymphomas. Together, these data demonstrate the utility of implementing multi-modal systems analysis approaches to reveal molecular insights into human cancers and facilitate target discovery for the development of more effective therapies. Results Integrated multi-cohort analysis of PTLD transcriptomic profiles identifies differentially expressed genes between EBV(+) and EBV(−) PTLD We applied an integrated multi-modal approach in order to determine the distinguishing characteristics of the B lymphoma cells and the TME in EBV(+) and EBV(−) PTLD. To define the tumor and TME transcriptional landscape in B cell PTLD, we collected three existing datasets that contained whole transcriptome profiles from 60 fresh-frozen samples from patients with EBV(+) (n = 42) and EBV(−) (n = 18) PTLD lesions from three institutions14,15,16 (Table S1). While these samples reflect the heterogeneity of B cell lymphoproliferations in PTLD, the majority are monomorphic diffuse large B cell lymphoma (DLBCL), the most common clinical subtype of PTLD.17,18 An integrated multi-cohort analysis of these three datasets using MetaIntegrator19 identified 189 significantly differentially expressed genes (DEGs) (false discovery rate [FDR] < 20%, log2|effect size| (ES) > 0.59, measured in ≥2 out of 3 datasets) between EBV(+) PTLD and EBV(−) PTLD (Figure 1A), of which 113 were overexpressed and 76 underexpressed in EBV(+) PTLD patients compared to those with EBV(−) PTLD (Table S2). These 189 genes had concordant expression patterns across all three datasets analyzed for either EBV(+) or EBV(−) PTLD (Figure 1B). Collectively, we refer to these as the EBV-associated PTLD gene signature. Integrated multi-cohort analysis identifies a 189-gene signature distinguishing EBV(+) PTLD from EBV(−) PTLD (A) Overview of multi-cohort gene expression analysis comparing samples from EBV(+) and EBV(−) PTLD lesions. (B) Heatmap of log2-transformed effect size values for each gene in the EBV-associated PTLD signature. Gray cells denote genes not measured in the dataset. Red: overexpressed in EBV(+) PTLD; blue: overexpressed in EBV(−) PTLD. (C) Volcano plot of gene expression data from PTLD tumor microarray datasets. Red: overexpressed in EBV(+) PTLD; blue: overexpressed in EBV(−) PTLD. (D) Pathway overrepresentation analysis of genes in the EBV-associated PTLD signature. The top 15 Gene Ontology terms as ranked by adjusted p values are shown. See also Tables S1 and S2. Overexpressed DEGs in EBV(+) PTLD included cytokines and chemokines (CCL3, CCL4, CCL8, IL15, IFNG), immune cell surface markers (CD38, CD163, CD300A), interferon-stimulated genes (SIGLEC1, OAS1, OAS2, OASL, ISG15, IFI44L, IFIT3, IFIT5), cytotoxic mediators (PRF1, GNLY, GZMH), and the antiviral apolipoprotein B mRNA editing catalytic polypeptide-like (APOBEC) protein family member APOBEC3G, as well as CTC-338M12.4, an uncharacterized non-coding RNA gene, and PIK3R5, encoding the understudied p101 regulatory subunit of PI3Kγ, a crucial kinase in the PI3K/Akt/mTOR pathway that we established is constitutively active in EBV(+) PTLD20,21,22 (Figure 1C; Table S2). Conversely, among the DEGs overexpressed in EBV(−) PTLD were anti-apoptotic transcription factor genes (BCL6, BCL11A), a chemokine receptor gene (CXCR5), genes encoding proteins involved in ubiquitin-mediated protein degradation (UBE2A, CUL3, and CUL4B), oncogenes (BZW2, LMO2), and an uncharacterized gene (CXorf56, also known as STEEP1) (Figure 1C; Table S2). To gain insight into the biological processes associated with EBV(+) and EBV(−) PTLD, we performed Gene Ontology (GO) pathway overrepresentation analysis23,24,25 on our signature genes. DEGs that were upregulated or downregulated in EBV(+) PTLD (i.e., genes with higher expression in EBV(+) PTLD or EBV(−) PTLD, respectively) were analyzed separately. For the DEGs more highly expressed in EBV(+) PTLD, many of the top enriched pathways involve viral recognition and immune-related processes (Figure 1D). These findings imply that despite post-transplant immunosuppression, host immunity can detect and respond to EBV, and/or EBV induces a distinct set of immune processes, thereby contributing to the transcriptomic differences in EBV(+) and EBV(−) PTLD. In contrast, DEGs with higher expression in EBV(−) PTLD were associated with lipid metabolism and DNA interaction or DNA repair pathways (Figure 1D), indicative of the similarity of EBV(−) PTLD to B cell lymphomas arising in immunocompetent individuals, which have been characterized to have higher mutational burden and aberrant metabolism.26,27,28 Overall, these data provide further evidence that the disparate drivers of oncogenesis between EBV(+) and EBV(−) PTLD lead to distinct gene expression patterns and hence implicate different underlying biological processes in the development and persistence of EBV(+) and EBV(−) PTLD. EBV(+) PTLD B cells are transcriptomically distinct from EBV(−) B lymphoma cells Since the datasets in our multi-cohort analysis contained transcriptomic profiles of fresh-frozen PTLD tumor specimens, each sample included RNA transcripts from the transformed B cells, together with the background tissue, stromal cells, and immune infiltrate within the lesion. In order to elucidate the cell-intrinsic differences between B cells transformed with, or in the absence of, EBV, we next investigated the transcriptome of EBV(+) B cell lymphoma lines derived from patients with EBV(+) PTLD with morphology corresponding to DLBCL, and EBV(−) DLBCL cell lines, by microarray gene expression analysis. Unsupervised clustering of all seven B lymphoma lines by their transcriptome showed that they segregate by EBV status (Figure 2A), again indicating that malignant EBV(+) and EBV(−) B cells are clearly distinct in their underlying biology. Independent validation of gene signature on EBV(+) PTLD patient-derived and EBV(−) B cell lines (A) tSNE plot and unsupervised affinity propagation clustering of EBV(+) and EBV(−) B cell lines by their transcriptomic profiles. (B) Volcano plot of gene expression data from EBV(+) and EBV(−) B cell lines. Red: overexpressed in EBV(+) cell lines; blue: overexpressed in EBV(−) cell lines. (C) Heatmap of log2-transformed fold change values for each of 1,407 DEGs between EBV(+) and EBV(−) B cell lines. Red: higher gene expression; blue: lower gene expression. (D) Pathway overrepresentation analysis on DEGs in EBV(+) B cell lines. The top 15 Gene Ontology terms as ranked by adjusted p values are shown. (E) Correlation of effect size from EBV-associated PTLD signature with fold change observed between EBV(+) and EBV(−) B cell lines. (F) Strategy overview to select genes for further in vitro and in vivo investigation using EBV(+) B cell lines. See also Table S2. We identified 1,407 DEGs (Figures 2B and 2C), of which 613 were overexpressed and 794 were underexpressed in the EBV(+) versus the EBV(−) B cell lymphoma lines. Specifically, the upregulated DEGs included genes encoding a multitude of chemokines, cluster of differentiation (CD) antigens, interferon-induced genes, and antigen presentation molecules, while the downregulated DEGs included genes encoding proteins associated with DNA organization or interactions, such as zinc fingers, transcription factors, and centromere-associated proteins (Table S2). GO pathway overrepresentation analysis revealed that the upregulated DEGs in the EBV(+) B cell lines were enriched for pathways involved in pathogenic infection and immune signaling, while the downregulated DEGs (i.e., DEGs with higher expression in EBV(−) B cell lines) were most associated with cell cycle and DNA-related pathways (Figure 2D). We further filtered the DEGs to identify the genes in the B cell lymphomas that overlapped with, and had expression patterns consistent with, those observed in the EBV-associated PTLD gene signature from our multi-cohort analysis (Figures 2E and 2F). The DEGs with concordantly higher transcriptomic expression in patients with EBV(+) PTLD and B cell lines were mainly associated with immune processes, including cytokine/chemokine genes (CCL3, CCL4, IL15), an antiviral gene (APOBEC3G), interferon-related genes (ISG15, IFI44L, IFIT3, OAS1, OAS2), and immune receptor genes (CD300A, BTN3A1). Meanwhile, the DEGs with similarly increased expression in EBV(−) PTLD and B cell lines instead encompassed genes with anti-apoptotic functions (BCL6, BCL11A, SYPL1, SYK) and genes linked to more aggressive tumor presentations (SWAP70, GDI2, RBBP7).29,30,31 These transcriptomic differences reflect the capability of EBV-transformed B cells to orchestrate changes in immune signaling pathways and within the TME, even in the absence of interactions with the immune infiltrate and background stromal tissue, and in the case of EBV(−) B cell malignancies, the potentiation of their development through dysregulation of homeostatic processes. EBV(+) PTLD B cell lines have increased expression and secretion of chemokines, which correlate with enhanced monocyte migration and enrichment of CD163+ monocytes in the TME of primary EBV(+) PTLD lesions. While little is known about the TME in EBV(+) PTLD, there is evidence of an immune inflammatory environment.32 To better understand how EBV(+) B cell PTLD may impact tumor-host cell interactions, we initially focused on EBV(+) PTLD B cell DEGs that could contribute to TME modulation. The chemokine genes CCL3, CCL3L1, CCL3L3, and CCL4 were upregulated in the EBV(+) B cell lymphoma lines and the EBV(+) PTLD signature genes (Figure 3A). We validated their concordant overexpression in the EBV(+) B lymphoma lines relative to the EBV(−) lymphoma lines at the transcript level (Figure 3B), and the protein level for CCL3 and CCL4 (Figure 3C). EBV(+) B cell lines have increased expression and secretion of chemokines compared to EBV(−) B cell lines, which correlate with enhanced monocyte migration (A) Forest plots of CCL3, CCL3L1, CCL3L3, and CCL4 expression from PTLD tumor microarray datasets. The x axes represent the log2-transformed effect size between groups. Rectangle size is proportional to the standard error of mean (SEM) difference in the study. Whiskers represent the 95% confidence interval. Diamonds represent the overall, combined effect size for the specific gene between groups. Width of the diamonds represents the 95% confidence interval of the summary effect size. (B) Relative gene expression of CCL3, CCL3L1, CCL3L3, and CCL4 in EBV(+) and EBV(−) B cell lines, quantified by quantitative reverse-transcription PCR (RT-qPCR). Red: EBV(+) cell lines; blue: EBV(−) cell lines. Data in quadruplicate are represented as mean ± SEM. (C) Secretion of CCL3 and CCL4 by EBV(+) and EBV(−) B cell lines, quantified by ELISA. Red: EBV(+) cell lines; blue: EBV(−) cell lines. Data in triplicate are represented as mean ± SEM. (D) Overview of in silico deconvolution and multi-cohort analysis of PTLD tumor microarray datasets. (E) Forest plots of estimated monocyte proportions from PTLD tumor microarray datasets. (F) Experimental schematic for transwell migration assay performed with cell culture supernatants from EBV(+) and EBV(−) B cell lines. (G) THP-1 migration indices for cell culture supernatants from EBV(+) and EBV(−) cell lines. ∗, adjusted p < 0.05; ∗∗, adjusted p < 0.01 by Wilcoxon rank-sum test. Pooled data from 3 independent experiments of n = 2 replicates each are represented as mean ± SEM. (H) Heatmap of log10-transformed mean fluorescence intensity (MFI) values from Luminex assay of cell culture supernatants. ND: cytokine MFI values below the lowest point on the standard curve (not detected). (I) Correlation of cytokine MFI values with THP-1 migration index computed as Pearson’s R. See also Figure S1, Table S7. Given the established roles of these chemokines in immune cell recruitment, we reasoned that their secretion by EBV(+) B cell PTLD could influence the composition of the immune infiltrate. Therefore, we estimated the proportion of various immune cell subsets present within each sample using in silico cell-mixture deconvolution with the basis matrix immunoStates33,34,35,36 on the original multi-cohort analysis datasets (Table S1). We then conducted another multi-cohort analysis on the estimated immune cell proportions to compare changes between EBV(+) PTLD and EBV(−) PTLD primary lesions (Figure 3D). We determined that the proportions of total monocytes were significantly higher in EBV(+) PTLD compared to EBV(−) PTLD lesions (ES = 0.707, FDR = 0.018), largely driven by the significant enrichment of CD14+ monocytes in EBV(+) PTLD (ES = 0.716, FDR = 0.0166) (Figure 3E). The increased monocyte proportions in EBV(+) PTLD dovetail with our previous finding that CCL3 and CCL4, known monocyte chemoattractants, are overexpressed in EBV(+) PTLD, suggesting that lymphoma-derived chemokines promote monocyte infiltration into EBV(+) PTLD lesions, thereby altering the TME. To address this possibility, we determined whether the CCL3 and CCL4 quantities secreted by the EBV(+) PTLD cell lines were sufficient to affect monocyte recruitment. We performed transwell migration assays with the monocytic cell line THP-1, using culture supernatants from EBV(+) and EBV(−) B lymphoma cell lines as chemoattractants (Figure 3F). Culture supernatants from three EBV(+) PTLD cell lines (AB5, JB7, and JC62) induced significantly higher THP-1 migration than supernatants from both the EBV(−) B lymphoma cell lines (Figure 3G). To determine if other secreted factors could also contribute to the differential monocyte migration observed, we next assessed EBV(+) and EBV(−) culture supernatants via Luminex assays. Unsupervised hierarchical clustering of all seven cell lines by their cytokine secretion profile showed clear segregation by their EBV status (Figure 3H). In addition to CCL3 and CCL4, multiple cytokines and chemokines (FLT3L, GM-CSF, interleukin [IL]-27, CCL2, M-CSF, CXCL9, and CCL5), many with known roles in monocyte recruitment, were also secreted by the EBV(+) cell lines at consistently higher levels than the EBV(−) cell lines (Figures 3H and S1). At the transcript level, CCL5 and CSF1 (M-CSF) were among the DEGs identified to also be upregulated in EBV(+) B cell lines (Figure 2B). Furthermore, the THP-1 migration index was positively correlated (Pearson R ≥ 0.4) with FLT3L, GM-CSF, IL-27, CCL2, and M-CSF concentrations (Figure 3I). In particular, secreted CCL2 in the EBV(+) B cells was strongly positively correlated (Pearson R = 0.96, p = 0.00062) with the extent of monocyte migration. These data indicate that chemokines produced by EBV(+) B cell lymphoma cells can promote monocyte infiltration into the TME of EBV(+) PTLD. To directly assess the monocyte infiltrate within primary EBV(+) and EBV(−) PTLD tumor lesions, we performed co-detection by indexing (CODEX), immunohistochemistry (IHC) and in situ hybridization (ISH) on formalin-fixed paraffin-embedded (FFPE) tissue microarrays from human DLBCL PTLD archival specimens (n = 33). We found that the median intensity of the monocyte/macrophage marker CD14 was significantly increased in non-B/non-T cells within EBV(+) compared to EBV(−) PTLD lesions (Figure 4A; p = 0.033, q = 0.066), in agreement with our multi-cohort analysis (Figure 3E). Although the median expression of CD163 as determined by CODEX was not significantly elevated (Figure 4A; p = 0.076, q = 0.101) in EBV(+) PTLD, the proportions of CD163+ cells tended to be higher in EBV(+) PTLD lesions (Figures 4B and 4C; p = 0.017, q = 0.085). IHC analysis further corroborated the enrichment of CD163+ cells in the TME of EBV-encoded RNA (EBER)+, i.e., EBV(+), PTLD (Figure 4D; Table S3; q = 0.024). Latent membrane protein (LMP)1+ and Epstein-Barr nuclear antigen (EBNA)+ PTLD cases also had increased proportions of CD163+ cells (Table S3; q = 0.024 and 0.027, respectively). Moreover, in our original multi-cohort analysis datasets, CD163 was also upregulated within the EBV-associated PTLD signature genes (Table S2; ES = 0.927, FDR = 2.79 × 10−4). Taken together, our data suggest that the predominance of M2-like monocytes in EBV(+) PTLD could create an immunosuppressive environment. Consistent with this, we observed in LMP1+ or EBNA+ PTLD cases significantly decreased frequencies of both CD3+ T cells (LMP1: q = 0.046; EBNA: q = 0.024) and CD4+ T cells (LMP1: q = 0.024; EBNA: q = 0.046)) (Figures 4D; Table S3). Furthermore, the median expression of the myeloid lineage-restricted myeloperoxidase (MPO), whose enzymatic activity has been linked to poorer outcomes in cancer,37,38 was also significantly increased in EBV(+) PTLD (Figure 4A; p = 0.008, q = 0.030). Within the EBV(+) PTLD cases, latency III cases were more likely to have lower CD3+ T cell proportions (Table S3; q = 0.024). CD163+ monocytes are increased in the tumor microenvironment of primary EBV(+) PTLD lesions (A) Expression of CD14, CD163, and MPO in non-B/non-T cells within primary EBV(+) or EBV(−) PTLD lesions. MFIs for each marker within the membrane compartment are quantified at the single-cell level and aggregated by computing the median marker MFI from all cells within each sample. Orange: EBV(+) (n = 18); blue: EBV(−) (n = 7); points represent individual cases. Data are represented as mean ± SEM. q = adjusted p values by Wilcoxon rank-sum test. (B) Percentage of CD163+ cells within primary EBV(+) or EBV(−) PTLD lesions, quantified by CODEX. Orange: EBV(+) (n = 11); blue: EBV(−) PTLD (n = 5); points represent individual cases. Data are represented as mean ± SEM. (C) CODEX images of FFPE tissue from diagnostic biopsies of EBV(+) and EBV(−) monomorphic PTLD (DLBCL subtype). Yellow: EBNA2; red: CD3/CD4; blue: CD68/CD163. A representative case in each group is shown in parallel for comparison. Scale bar, 50 μm. (D) Histology of FFPE tissue from diagnostic biopsies of EBV(+) and EBV(−) monomorphic PTLD (DLBCL subtype). EBER in situ hybridization and IHC staining for CD163, CD3, and CD4 from a representative case per group are shown in parallel for comparison. Scale bar, 50 μm. See also Tables S3 and S8. Collectively, our data suggest that EBV(+) B lymphoma cells alter the TME through producing monocyte-attracting chemokines, leading to an immunosuppressive environment in EBV(+) PTLD driven by tumor-associated M2 monocytes and diminished T cell proportions. Single-cell surface proteomic analysis distinguishes EBV(+) PTLD B lymphoma cells from EBV(−) B lymphoma cells and healthy B cells We next focused on deciphering the differences in the B lymphoma cells in EBV(+) and EBV(−) PTLD by performing high-dimensional single-cell analysis using cytometry by time of flight (CyTOF) and a custom antibody panel recognizing B cell surface molecules (Table S4). Both stimulated and unstimulated B cells from PBMCs of a healthy blood donor (control B cells) were included (Figure S2A). Principal-component analysis (PCA) performed on the median expression of individual markers for each sample demonstrates that EBV(+) and EBV(−) B lymphoma cell lines clearly segregate from each other and from control B cells using the first two principal components (Figure 5A). Moreover, uniform manifold approximation and projection (UMAP) plots mapping the EBV(+) and EBV(−) lymphoma lines to the same high-dimensional space also show similarly distinct segregation between both groups (Figures 5B and S2B). Within the UMAP space, four clusters could be identified. EBV(+) B cell lines constituted almost the entirety of clusters A and B, while clusters C and D were similarly overrepresented by the EBV(−) B cell lines (Figure 5C). EBV(+) B cell lines have a distinct surface phenotype from EBV(−) B cell lines and healthy B cells (A) PCA of EBV(+) and EBV(−) B cell lines and stimulated and unstimulated B cells from a healthy donor (control B cells), using 27 markers measured by CyTOF. Points represent individual samples, color-coded by group. Eigenvectors of individual markers are indicated by arrows. (B) UMAP plots using 3,000 cells per cell line, using all 27 chosen markers. Cells are color-coded by sample and aggregated by EBV status (top) or colored by relative signal intensity of indicated surface markers (bottom). (C) Clustering of EBV(+) and EBV(−) B cell lines. Based on the UMAP in (B), 4 distinct clusters were identified (left) and the proportion of each cluster by EBV status was calculated (right). (D) Histograms for the expression of indicated phenotypic markers, represented as arcsinh-transformed signal intensity values. Each histogram represents marker expression in individual samples for EBV(+) and EBV(−) B cell lines, as well as stimulated and unstimulated control B cells. (E) Comparison of CD300a median signal intensity for indicated phenotypic markers between EBV(+) cell lines (n = 5), EBV(−) cell lines (n = 2), and control B cells (n = 2). q = adjusted p values by Kruskal-Wallis test. Data are shown as mean ± SEM. (F) Heatmap of globally scaled B cell phenotypic marker expression in B cell lines and control B cell subsets. Median arcsinh-transformed signal intensity values of each marker were scaled between 0 and 1 across all samples. Unsupervised clustering was carried out on both individual cell lines/B cell subsets (rows) and phenotypic markers (columns). (G) Heatmap of scaled B cell phenotypic marker expression in B cell lines and control B cell subsets, with cell lines and control B cells scaled separately. Median arcsinh-transformed signal intensity values of each marker were scaled between 0 and 1 on all B cell lines, and separately on all control B cell subsets, to map B cell lines to the closest control B cell subset. Unsupervised clustering was performed on both individual cell lines/B cell subsets (rows) and markers (columns). See also Figure S2, Table S4. Notably, the EBV(+) B cell lines are characterized by significantly higher CD23, CD95 (Fas receptor), and CD300a surface expression when compared to EBV(−) B cell lines and control B cells (Figures 5B, 5D, and 5E), as well as increased expression of the activation marker CD69 (Figures S2C–S2E), congruent with their activated phenotype.39 In contrast, both EBV(−) cell lines exhibit lower CD22 expression and significantly higher CD24 expression, compared to all five EBV(+) cell lines and control B cells (Figures 5B, 5D, and 5E). Additionally, CD95, CD300A, CD22, and CD24 are among the DEGs distinguishing the EBV(+) and EBV(−) B cell lines at the transcript level (Figures 2B; Table S2), concurring with their protein expression. Since B cells from healthy individuals consist of multiple functional subsets, we then classified the control B cell samples into six subpopulations (Figures S2F–S2H) based on previously described B cell phenotypic characterization schemes.40,41 Unsupervised hierarchical clustering of all cell lines and control B cell subsets by their median marker expression separated the samples into three clusters corresponding to their EBV and transformation status, with notably distinct patterns of phenotypic marker expression between all experimental groups (Figure 5F). We accounted for the marked separation of the EBV(+) and EBV(−) cell lines from the untransformed B cells by scaling the median marker expression of all cell lines separately from that of the control B cell subsets. With the median expression values scaled to the same relative range, we then sought to relate each B cell line to its closest control B cell subset. Subsequently, we performed unsupervised hierarchical clustering. This analysis revealed that the EBV(+) cell line JB7 was most similar to naive B cells while all other EBV(+) cell lines more closely resembled various CD45RB + CD27+ memory B cell subsets. Among the EBV(−) cell lines, Pfeiffer was most associated with the CD45RB + CD27−memory B cell subset, while Toledo did not correlate with any of the control B cell subsets (Figure 5G). CD300a is more highly expressed and is required for maximal proliferation of EBV(+) PTLD B lymphoma cells Having applied an integrated multi-omics approach to resolve the differences between EBV(+) and EBV(−) PTLD, we next evaluated whether this approach could pinpoint molecules with a functional impact. We noted CD300A transcripts were concordantly upregulated in both primary EBV(+) PTLD lesions and EBV(+) B lymphoma lines (Figures 2B and S3; Table S2). At the protein level, CD300a is similarly overexpressed on EBV(+) versus EBV(−) cell lines, while healthy B cells have negligible CD300a surface expression (Figures 5D–5F and S3C). Since CD300A exhibited increased expression in primary EBV(+) PTLD lesions and EBV(+) B lymphoma cell lines, at both the transcript and protein levels, we chose CD300A for further investigation (Figure 2F). We first evaluated whether EBV can elicit expression of CD300a in B cells, by assessing the levels of surface CD300a on peripheral blood B cells from healthy donors and autologous B cells following EBV-mediated transformation. Three of three EBV(+) lymphoblastoid cell lines (LCLs), each generated by EBV infection of peripheral B cells from a healthy donor, show increased CD300a expression relative to autologous uninfected B cells (Figures 6A and 6B). These data indicate that EBV infection and transformation of human B cells induce CD300a expression. CD300a is more highly expressed on, and is required for, optimal growth of EBV(+) PTLD B lymphoma cells (A and B) Surface expression of CD300a in peripheral blood B cells from healthy donors and autologous LCL transformed in vitro by EBV (n = 3). Biaxial plots (A, left), histograms (A, right), and comparison of CD300a median signal intensity (B) are shown. (C) Schematic for CRISPR-Cas9-mediated gene editing for CD300A in EBV(+) PTLD B cell lines. (D) In vitro growth of CD300A KO cells relative to mock-edited controls of EBV(+) PTLD B cell lines. Cell quantities in mock-edited controls after 48 h of culture are used as the reference for normalization. ∗, p < 0.05 by Wilcoxon rank-sum test. Data shown in quadruplicate with mean ± SEM. (E) Apoptosis assay of CD300A KO or mock-edited EBV(+) PTLD B cell lines. Apoptosis stages were determined by DAPI and annexin V staining. ∗, p < 0.05; ∗∗, p < 0.01; ∗∗∗, p < 0.001; ∗∗∗∗, p < 0.0001 by unpaired t test. Data shown in triplicate with mean ± SEM. (F) Pathway overrepresentation analysis of genes downregulated in CD300A KO compared to mock control EBV(+) B cell lines using clusterProfiler. See also Figures S3–S5, Tables S4 and S5. CD300a is a member of the CD300 multi-gene family of receptors, mainly expressed externally on immune cells. Earlier studies show that CD300a has inhibitory function,42,43,44,45,46,47,48,49,50,51 mediated by the four immunoreceptor tyrosine inhibitory motifs (ITIMs) within its cytoplasmic tail. To elucidate CD300a’s role in B cell lymphoma, we generated CD300A knockout (KO) cell lines from three EBV(+) PTLD-derived B cell lines (AB5, JB7, and VB5) using CRISPR-Cas9 gene editing (Figure 6C). KO efficiencies exceeding 85% were achieved with two CD300A KO cell lines (AB5 and JB7), while the third (VB5) exhibited a KO efficiency of 40%–50% (Figure S4). We first evaluated the effect of CD300a deficiency on the spontaneous in vitro growth of these cells. All three CD300A KO EBV(+) B lymphoma cell lines had significantly decreased expansion relative to their corresponding mock-edited parental cell controls (Figure 6D). Furthermore, all three CD300A KO EBV(+) cell lines had higher proportions of cells in late apoptosis (DAPI+, annexin V+ cells) vis-à-vis the mock-edited parental cells (Figures 6E and S5A). We did not observe any evidence of cell-cycle arrest, nor significant differences in the cellular distribution throughout the cell cycle (Figure S5B). Given that annexin V binds to phosphatidylserine, one of the only two known ligands of CD300a,52 our observation of annexin V+ cells in the parental EBV(+) lymphoma lines (Figure S5A) provides evidence for the presence of a physiological ligand capable of engaging CD300a. Together, our data suggest that CD300a may promote the growth and/or provide a signal to impede apoptosis of EBV-transformed B cells. To directly evaluate the impact of CD300A depletion on cytokine/chemokine secretion in EBV-transformed B cell lymphomas, we conducted Luminex assays of cell culture supernatants from the CD300A KO EBV(+) cell lines and their corresponding mock-edited controls. Unsupervised hierarchical clustering of all samples by their cytokine secretion profile showed clear segregation only by their parental cell line, but not their CD300A expression nor the culture duration post-gene KO (Figure S5C). To further understand how CD300a expression can contribute toward tumor growth, we characterized the gene expression changes associated with CD300A depletion using single-cell RNA sequencing (scRNA-seq) of the CD300A KO EBV(+) B lymphoma lines and their corresponding mock-edited controls. DEGs were identified between CD300A KO and mock-edited cells as those with fold change >1.15 and adjusted p value <0.05. GO pathway overrepresentation analysis using all DEGs between CD300A KO and mock-edited cells from all three EBV(+) cell lines showed that the DEGs downregulated in CD300A KO cells enriched for pathways involving respiration and homeostatic metabolic processes (Figures 6F; Table S5). A second, independent evaluation of DEG and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis using the GSEApy package also implicated alterations in metabolic processes in CD300A KO cells (Figures S5D and S5E), further indicating that CD300a may participate in regulating homeostatic signaling pertaining to general cellular respiration, which could contribute to the reduced growth observed with CD300A deletion. Loss of CD300A expression leads to reduced growth of EBV(+) B cell PTLD in vivo To evaluate if CD300a deficiency affected lymphoma growth in vivo, we utilized a xenograft murine model of EBV(+) PTLD.22,53,54 Immunodeficient non-obese diabetic-severe combined immunodeficiency (NOD-SCID) mice were injected subcutaneously with either CD300A KO or mock-edited cells from the EBV(+) B cell lines AB5 or JB7. Mice receiving CD300A KO cells from both EBV(+) cell lines had significantly inhibited tumor growth compared to those receiving parental mock-edited cells within 5–7 days after tumor detection (Figures 7A and 7B). At sacrifice, tumor tissue was excised and harvested. The average tumor volume was significantly smaller in the CD300A KO group than the mock group for both AB5 (755 ± 217 mm3 vs. 1,874 ± 793 mm3; mean ± standard deviation, p < 0.01) and JB7 (1,407 ± 387 mm3 vs. 2,176 ± 413 mm3, p < 0.05) (Figures 7C and 7D). Loss of CD300a leads to reduced growth of EBV(+) B cell lymphomas in a xenograft mouse model of PTLD (A and B) Average tumor volume (mm3) calculated from mice injected with CD300A KO or mock-edited AB5 (A) or JB7 (B) cells (n = 5 mice per group). Data are shown as mean tumor volume ± SD for each group. Differences in tumor volume across all time points compared using a 2-way repeated measures ANOVA. (AB5: F(6, 48) = 7.987; JB7: F(7, 56) = 4.559). ∗, p < 0.05; ∗∗, p < 0.01; ∗∗∗, p < 0.001 with Benjamini-Hochberg FDR correction. (C and D) Tumor volume (mm3) at the time of sacrifice from mice injected with CD300A KO or mock-edited AB5 (C) or JB7 (D) cells. Individual tumor volumes (n = 5) are shown with mean ± SD. ∗, p < 0.05; ∗∗, p < 0.01 by Wilcoxon rank-sum test. (E) Histology of tumors harvested at sacrifice from mice injected with CD300A KO or mock-edited AB5 or JB7 cells. H&E staining, EBER in situ hybridization, and IHC staining for CD20 and Ki67 of tumors from a representative mouse per group are shown in parallel for comparison. Scale bar, 50 μm. (F) Necrosis scores of tumors harvested at sacrifice from mice injected with CD300A KO or mock-edited AB5 (top) or JB7 (bottom) cells. Scores from representative tumors (n = 3) are shown with mean ± SEM. See also Figure S6, Table S6. Histological and immunohistochemical analyses were also performed on three representative tumors per group. Notably, the CD300A KO group had lower Ki67 labeling indices (Figure 7E) and increased necrosis than the mock group for both AB5 (30% ± 11.55% vs. 20% ± 5.77%; mean ± standard error of mean) and JB7 (8.33% ± 2.89% vs. 5.00% ± 0.00%) (Figure 7F; Table S6). Strong membranous CD20 staining and nuclear EBER ISH signals were observed in all tumors, confirming their human B cell lineage and EBV positivity (Figure 7E). While mouse monocytes could be detected in the human tumor xenografts (Figure S6), there were no significant differences in the frequency of monocytes infiltrating tumors derived from CD300A KO or mock-edited AB5 or JB7 cells (Table S6), consistent with the highly similar cytokine/chemokine secretion profiles observed between CD300A KO and mock-edited cells in vitro (Figure S5C). Together, these data indicate that CD300A deletion impairs the in vivo development of solid tumors derived from EBV(+) B cell PTLD lines, thus providing clear evidence that CD300a mediates pro-tumor effects in EBV(+) B cell PTLD. Discussion The interplay between EBV and host immunity in EBV-driven malignancies is presently poorly characterized, as are the underlying molecular and immunological differences between EBV(+) and EBV(−) cancers of the same cell type.2 In the post-transplant context, this translates to similar therapeutic approaches for managing EBV(+) and EBV(−) B cell PTLD. These strategies generally fail to consider EBV-driven disease pathogenesis and only broadly modulate host immunity without specifically targeting the virus. They may also heighten the risk of organ rejection55 and/or opportunistic infection.56,57 While adoptive transfer of autologous or allogeneic EBV-specific cytotoxic T cells has shown promise, this approach has not been broadly applicable.58 Consequently, there is an unmet need to understand the divergent biological mechanisms involved when EBV is present or absent as an oncogenic driver to then facilitate personalized treatment options. In this study, we employed an integrated multi-modal strategy combining computational analyses of gene expression data from EBV(+) and EBV(−) PTLD with high-dimensional multi-omics to identify the underlying tumor and immune characteristics of EBV(+) and EBV(−) B cell PTLD. We establish that viral-immune dominance is a tumor-intrinsic feature of EBV(+) B cell PTLD in marked contrast to the DNA-interacting proteins, DNA repair, and altered lipid metabolism landscape that characterize EBV(−) B cell lymphomas. Second, we demonstrate that EBV(+) B cell lymphomas directly shape the TME through the elaboration of an array of monocyte-attracting chemokines, and an enrichment of M2-linked monocytes in EBV(+) PTLD, which may impact host anti-tumor responses via reducing T cells in the TME. Finally, we show that EBV transformation induces CD300a upregulation, required for maximal cell survival. CD300a-deficient EBV(+) B lymphoma lines have significantly diminished in vitro/in vivo growth vis-à-vis parental EBV(+) CD300a-expressing lines. The concordance of our EBV(+) versus EBV(−) B lymphoma transcriptomic analyses (across previously published datasets and B cell lines) was robust and revealed associations of EBV(+) lymphomas with immune response genes, and EBV(−) lymphomas with genes encoding DNA-interacting proteins. These findings support the hypotheses that EBV(+) B cell PTLD is either sufficiently immunogenic to elicit antiviral and/or anti-tumor responses or driven by EBV subverting immune pathways as it transforms the host cell, or both, whereas EBV(−) PTLD exhibits closer similarity to immunocompetent B cell lymphomas, with altered lipid metabolism and DNA interaction/repair processes. Previous work from our group has demonstrated that EBV induces constitutive IL-10 production.59 IL-10 has dual roles as an autocrine growth factor for EBV(+) B cell PTLD, and an immunosuppressive cytokine dampening anti-tumor responses. Here, we found that EBV(+) B cell lines have enhanced secretion of several monocyte-chemotactic cytokines/chemokines including CCL2, CCL3, and CCL4. EBV’s main oncogene, LMP1, induces CCL3 and CCL4 expression in NPC cell lines via the c-Jun N-terminal kinase pathway,60 as well as CCL2 expression in an epithelial cell line via nuclear factor κB (NF-κB).61 CCL3 and CCL4 have also been described as autocrine growth factors within multiple myeloma cell lines62 and in vitro-generated B LCLs.60 These chemokines may similarly function in spontaneously arising EBV(+) B cell lymphomas as well, given the expression of their cognate receptor CCR5 on all EBV(+) lymphoma lines in this study. Monocytes are an important immune population in the TME, given their potential to differentiate into macrophages and the role for tumor-associated macrophages (TAMs) with their spectrum of polarization states as crucial orchestrators of many hallmarks of cancer, including angiogenesis, invasion, and creation of an immunosuppressive TME.63 More recently, the role of monocytes themselves within the TME has become increasingly recognized, notably their acquisition of immunosuppressive function and development into myeloid-derived suppressor cells.64,65 Studies investigating the TME in PTLD remain uncommon, although macrophage enrichment has been reported to correlate with EBV presence.32,66,67 Altogether, our findings further establish EBV’s capacity to alter its transformed host cell’s secretome, thereby manipulating the TME via monocyte recruitment. Moreover, we demonstrate that the TME in primary EBV(+) PTLD specimens is characterized by an enrichment of M2-like monocytes and a corresponding decrease in T cells, which may, in tandem with the aforementioned production of IL-10 by EBV+ B lymphomas,20,59 create an immunosuppressive environment that can dampen anti-EBV T effector function. While PTLD includes a spectrum of abnormal lymphoproliferations, many are of the monomorphic DLBCL type.17,18 Whereas monomorphic DLBCL was the predominant subtype in our multi-cohort analysis, our EBV(+)/EBV(−) B cell lines were also of the DLBCL subtype. Thus, our combined transcriptomic analyses and proteomic characterization of EBV(+) and EBV(−) B cell lymphomas provide a vast resource for further studies to interrogate biologic functions of specific differentially expressed molecules and their potential as therapeutic targets. Along these lines, CyTOF analysis of EBV(+) and EBV(−) B cell lines revealed further surface phenotypic differences, with EBV(+) B cell lines characterized by CD23, CD95, and CD300a overexpression and EBV(−) B cell lines by their lower CD22 expression and higher CD24 expression. CD24 is an anti-phagocytosis “don’t-eat-me” signal; when expressed on tumor cells, it impairs macrophage phagocytosis by binding the inhibitory receptor Siglec-10 on macrophages.68 CD22 (Siglec-2) is predominantly expressed on B cells and is canonically considered an inhibitory receptor due to its four cytoplasmic ITIMs. CD22 negatively regulates B cell receptor signaling69; CD22-deficient mice exhibit augmented responses to B cell receptor (BCR) activation70 and autoantibody titers.71 Hence, EBV(−) B cell lymphomas could be co-opting CD24 as an immune escape mechanism, while downregulating CD22 to maintain a heightened activation state. CD95 is a death receptor capable of triggering apoptosis upon activation. Its upregulation by the host cell could represent a defense mechanism attempting to induce cell death. Our lab has previously reported high CD95 expression on the same EBV(+) PTLD B cell lines and demonstrated that LMP1 can protect EBV(+) B cell lymphomas from apoptosis by NF-κB-dependent induction of cellular FLICE (FADD-like IL-1β-converting enzyme)-inhibitory protein (c-FLIP), blocking signal propagation downstream of CD95.72,73 CD23 is an immunoglobulin E (IgE) receptor (FcεRII) whose cleavage releases soluble CD23, which maintains B cell survival and growth74,75 by inhibiting apoptosis.76 We show that infection and transformation of B cells by EBV also resulted in higher CD300a surface expression. Interestingly, the EBV gene EBNA2 has been described to enhance FCER2 (CD23) and CD300A transcript expression,77 raising the possibility of their upregulation being another mechanism utilized by EBV to drive tumor growth. In this report, we show that CD300a expression is essential to maximal survival of EBV(+) B cell lymphomas. Paradoxically, CD300a is a receptor containing three classical and one non-classical cytoplasmic ITIM motifs.78 Early studies in B cells described CD300a as an inhibitory molecule because CD300a engagement suppressed calcium mobilization and NFAT activity elicited by concurrent BCR ligation.45,46 However, subsequent reports on CD300a′s function have been mixed: a multitude of studies, predominantly in primary cells, showed that CD300a exhibits inhibitory function,42,43,44,47,48,49,50,51 while others described its pro-tumor functions.79,80,81 The latter implicate downstream activation of the PI3K/Akt pathway, which our group has demonstrated to be constitutively active in EBV(+) PTLD.20,21,22 CD300a mediates its inhibitory signaling via SHP-148 after engaging its ligands phosphatidylserine or phosphatidylethanolamine,52 exposed on the outer membrane leaflet of stressed or apoptotic cells. These cells commonly arise in cell culture systems and in vivo tumor environments; here, we observe phosphatidylserine presence on our EBV(+) B cell lines. More intriguingly, it is unknown if CD300a has other yet-undiscovered ligands that can potentiate its pro-tumor effects. Additionally, the studies demonstrating CD300a inhibitory function generally used immune cells isolated from healthy individuals or cell lines provided with strong exogenous stimuli, whereas those implicating CD300a in pro-tumor roles were performed using cancer cell lines without external stimulation. Therefore, it is plausible that CD300a has pleiotropic functions, depending on the biological context. Using CRISPR-Cas9 to deplete CD300A expression in EBV(+) B cell lines, we demonstrate that CD300a supports both in vitro and in vivo growth of EBV(+) B cell lymphomas. Moreover, scRNA-seq reveals an association of CD300a depletion with decreased expression of genes enriched in energy, aerobic, and cellular respiration. Interestingly, these include cyclooxygenase (COX) family members and genes linked to the PI3K/Akt/mTOR pathway, known to be critically involved in cellular metabolism and, as we have previously shown, essential to EBV(+) PTLD B cell survival.20,21,22,54 Coupled with our finding that EBV infection and transformation elicit CD300a expression in B cells, these data collectively imply that EBV may maintain the survival and proliferative potential of its transformed host B cells through upregulating CD300a expression. Notably, our findings implicate CD300a in transformed cells with a viral driver of oncogenesis. Prior studies examining CD300a’s role in tumors described its overexpression in malignant tissues versus normal tissues, without sufficient resolution on tumor EBV status and/or etiology. How exactly CD300a, with its inhibitory properties, potentiates EBV(+) B cell lymphoma growth remains to be more thoroughly determined mechanistically. However, it is intriguing to consider that CD300a may tune constitutive pro-activation and proliferation signals provided by EBV proteins, to avoid surpassing signaling thresholds that may elicit cell death in transformed cells.82 In support of this hypothesized role for CD300a in EBV-transformed cells, we show that higher frequencies of apoptotic cells accompany CD300A depletion without affecting cell-cycle arrest/progression, indicating that CD300a could provide anti-apoptotic signaling to EBV(+) B cell lymphomas in PTLD, thereby offering mechanistic insight into the observed reduced expansion of CD300A KO cells. Collectively, our results highlight that EBV(+) and EBV(−) PTLD and B cell lymphomas have distinct molecular pathways and immune interactions, providing further evidence supporting the predominance of immune pathways in EBV(+) PTLD versus cytogenetic abnormalities in EBV(−) PTLD. We demonstrate the TME-altering potential of EBV-driven malignancies: EBV-transformed B cells secrete larger quantities of monocyte-chemotactic cytokines/chemokines, possibly contributing to the increased monocyte presence in EBV(+) PTLD lesions. Additionally, our integrated multi-modal approach enabled our identification of CD300a overexpression in EBV(+) PTLD, crucial for optimal expansion of EBV-transformed cells. Overall, these data establish the differential involvement of specific processes in the pathogenesis of EBV(+) and EBV(−) PTLD, thereby providing a basis for exploiting these biological targets to further refine the therapeutic management of PTLD based on tumor’s association with EBV or lack thereof. Limitations of the study EBV(−) sample sizes in our analyses were limited. As with cell lines, the B cell lines we used to model PTLD in vitro may not fully capture the variety of physiological processes at play in PTLD development and persistence. To mitigate this, we focused on genes with concordantly differential expression between EBV(+) and EBV(−) samples from both primary PTLD tumors and cell lines. While we have addressed CD300a’s functional impact on the tumor, its exact mechanistic contributions toward tumor growth have yet to be comprehensively explored. Unfortunately, we could not experimentally demonstrate the therapeutic effects of targeting CD300a in PTLD due to the lack of commercially available small molecules/drugs or reliable agonistic/antagonistic monoclonal antibodies capable of specifically modulating CD300a’s downstream activity. Future studies should focus on validating the therapeutic potential of targeting CD300a in PTLD. Lastly, the molecular mechanism(s) behind the induction of these differential gene/protein expression patterns upon EBV infection and transformation remain to be determined. Resource availability Lead contact Further information and requests for resources, software, and data should be directed to and will be fulfilled by the lead contact, Olivia M. Martinez (omm@stanford.edu). Materials availability There are restrictions to the availability of the EBV(+) PTLD patient-derived spontaneous lymphoblastoid cell lines (SLCLs), AB5, JB7, JC62, MF4, VB5, and their gene-edited derivatives. The informed consent utilized for the establishment of these lines, or their derivatives, precluded transfer to other parties; therefore, they cannot be shared with other investigators. This study did not generate any other new unique reagents. Further information and requests for resources, software, and data should be directed to and will be fulfilled by the lead contact, Olivia M. Martinez (omm@stanford.edu). Data and code availability B cell line microarray data and CD300A KO scRNA-seq data have been deposited at National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) as GEO: GSE279598 and GEO: GSE279989, respectively, within superseries GEO: GSE279991 and are publicly available as of the date of publication. This study does not report original code. All software and algorithms used in this study are publicly available. Any datasets or additional information required to reanalyze the data reported in this paper is available from the lead contact/corresponding authors upon request. Acknowledgments This project was supported by the National Institutes of Health (NIH) National Institute of Allergy and Infectious Diseases (NIAID) grant RO1AI113130 (O.M.M.). P.K. is funded in part by the Bill and Melinda Gates Foundation (OPP1113682), NIH NIAID grants U19AI057229 and U19AI67903, NIH NIAID contract 75N93022C00052, and DOD contracts W81XWH1910235 and W911NF2320019. J.T. and E.Y.Z.L. received National Science Scholarships from the Agency for Science, Technology, and Research, Singapore. J.K.P. was supported by NIH T32AI1007290 (O.M.M.) and the Stanford Diversifying Academia, Recruiting Excellence Doctoral Fellowship Program. Flow cytometry was performed using BD LSRII instruments in the Stanford Shared FACS Facility. Primer synthesis and microarray gene expression profiling were conducted by Stanford University’s Protein and Nucleic Acid facility. Luminex assays were carried out at Stanford University’s Human Immune Monitoring Center. CODEX analysis was performed with the support of the Stanford C-ShaRP Instrumentation and Enhancement Program. Single-cell RNA sequencing was performed at the Stanford Functional Genomics Facility with instrumentation purchased with NIH funding (S10OD025212 and 1S10OD021763). We thank the following individuals for providing technical advice: Sean C. Bendall (CyTOF data analysis); Liang Wei Wang, Sizun Jiang, and Borja Ocón Moreno (experimental protocols); Jian Yang “Lumpy” Lum (statistical analyses); and members of the Martinez, Krams, and Khatri labs, particularly Wenming Zhang and Mahil Rao (CyTOF data analysis). Author contributions J.T., P.K., and O.M.M. conceptualized and designed the study. J.T., A.J.R., T. Tajima, C.E., K.C.J., J.K.P., X.W., E.Y.Z.L., M.C.G., and E.A.V. designed and conducted experiments. J.T., A.J.R., T. Tajima, C.E., K.C.J., and J.K.P. contributed to experimental data collection, analysis, and interpretation. J.T., K.C.J., M.C.G., and J.T.H. performed CyTOF data pre-processing and analysis. J.T., L.K., and A.G. performed the comparison of B cell lines to healthy control B cell subsets. J.T. and F.V. performed in silico deconvolution. S.F.Y., S.Z., and Y.N. performed IHC staining, CODEX, and histological analyses. J.T. and I.L. analyzed IHC and CODEX staining data. I.L., J.A.P., and A.T.S. performed scRNA-seq profiling and comparison of CD300A knockout to mock-edited EBV(+) B cell lines. C.E. and E.A.V. performed proliferation and cell-cycle analysis of CD300A knockout and mock-edited EBV(+) B cell lines. G.B., F.E.C., S.H.S., J.M., D.D., and T. Tousseyn provided transcriptomic data from EBV(+) and EBV(−) PTLD tumor biopsies. S.M.K. contributed to study design and data interpretation and gave conceptual advice. J.T., O.M.M., and P.K. wrote the manuscript. O.M.M. and P.K. supervised the study. All authors read the manuscript and approved the submission. Declaration of interests The authors declare no competing interests. STAR★Methods Key resources table REAGENT or RESOURCESOURCEIDENTIFIERAntibodiesanti-human CD14 (clone HCD14) - APCBioLegendCat#325607; RRID:AB_830680anti-CD45 (clone H130) - 89 YFluidigm (now Standard BioTools)Cat#3089003B; RRID:AB_2938863anti-CD19 (clone HIB19) - 165 HoFluidigm (now Standard BioTools)Cat#3165025Banti-CD20 (clone 2H7) - 147 SmFluidigm (now Standard BioTools)Cat#3147007Banti-CD3 (clone OKT3) - biotinBioLegendCat#317320; RRID:AB_10916519anti-CD7 (clone CD7-6B7) - biotinMiltenyi BiotechCat#130-123-562; RRID:AB_2802048anti-CD15 (clone HI98) - biotinBioLegendCat#301194anti-CD33 (clone WM53) - biotinBioLegendCat#303426; RRID:AB_2650842anti-CD56 (clone 5.1H11) - biotinBioLegendCat#362536; RRID:AB_2565653anti-CD61 (clone Y2/51) - biotinMiltenyi BiotecCat#130-098-522; RRID:AB_2660502anti-CD235ab (clone HIR2) - biotinBioLegendCat#306618; RRID:AB_2565773anti-Biotin (clone 1D4-C5) - 113 lnBioLegendCat#409002; RRID:AB_10642032anti-CD9 (clone HI9a) - 141 PrBioLegendCat#312102; RRID:AB_314907anti-CD39/ENTPD1 (clone A1) - 142 NdBioLegendCat#328221; RRID:AB_2563747anti-CD5 (clone UCHT2) - 143 NdBioLegendCat#300627; RRID:AB_2563756anti-CD69 (clone FN50) - 144 NdFluidigm (now Standard BioTools)Cat#3144018Banti-CD45RB (clone MEM-55) - 145 NdFluidigm (now Standard BioTools)Cat#3145009Banti-CD25 (clone 2A3) - 149 SmFluidigm (now Standard BioTools)Cat#3149010B; RRID:AB_2756416anti-CD81 (clone 5A6) - 150 NdBioLegendCat#349502; RRID:AB_10643417anti-CD21 (clone BL13) - 152 SmFluidigm (now Standard BioTools)Cat#3152010B; RRID:AB_3661727anti-CCR5 (clone J418F1) - 153 EuBioLegendCat#359102; RRID:AB_2562457anti-TGFbRII (clone W17055E) - 154 SmBioLegendCat#399702; RRID:AB_2861013anti-CD27 (clone L128) - 155 GdFluidigm (now Standard BioTools)Cat#3155001B; RRID:AB_2687645anti-CD183 (clone G025H7) - 156 GdFluidigim (now Standard BioTools)Cat#3156004B; RRID:AB_2687646anti-CD72 (clone 3F3) - 157 GdBiolegendCat#316202; RRID:AB_2260104anti-CD300a (clone MEM-260) - 159 TbThermoFisher ScientificCat#MA1-10190; RRID:AB_11152437anti-CD23 (clone EBVCS5) - 160 GdFluidigm (now Standard BioTools)Cat#3160020Banti-CD22 (clone HIB22) - 161 DyBioLegendCat#302511; RRID:AB_2563730anti-LAP/TGFβ latent complex (clone TW4-6H10) - 163 DyFluidigm (now Standard BioTools)Cat#3163010Banti-CD95/FasR (clone DX2) - 164 DyBioLegendCat#305631; RRID:AB_2563766anti-CD197/CCR7 (clone G043H7) - 167 ErFluidigm (now Standard BioTools)Cat#3167009A; RRID:AB_2858236anti-CD73/NT5E (clone AD2) - 168 ErFluidigm (now Standard BioTools)Cat#3168015B; RRID:AB_2810249anti-CD24 (clone ML5) - 169 TmBioLegendCat#311127; RRID:AB_2563733anti-HLA-DR (clone L243) - 170 ErFluidigm (now Standard BioTools)Cat#3170013B; RRID:AB_2888929anti-CD1d (clone 51.1) - 171 YbBioLegendCat#350302; RRID:AB_10641848anti-CD38 (clone HIT2) - 172 Yb and 154 SmBioLegendCat#303535; RRID:AB_2562819anti-CD1c (clone L161) - 173 YbBioLegendCat#331502; RRID:AB_1088995anti-TGFβRI (polyclonal) - 174 YbSigma-AldrichCat#ABF17-1anti-CD71 (clone OKT-9) - 175 LuFluidigm (now Standard BioTools)Cat#3175011B; RRID:AB_2756287anti-CD305/LAIR1 (clone NKTA255) - 176 YbGenetexCat#GTX75993; RRID:AB_377148anti-IgD (clone IA6-2) - 146 NdBioLegendCat#348235; RRID:AB_2563775anti-IgA (polyclonal) - 148 NdFluidigm (now Standard BioTools)Cat#3148007B; RRID:AB_2810852anti-IgG (clone M1310G05) - 151 EuBioLegendCat#410701; RRID:AB_2565624anti-IgM (clone MHM-88) - 158 GdBioLegendCat#314527; RRID:AB_2563776anti-PD-L1 (Clone 29E.2A3) - 162 DyBioLegendCat#329719; RRID:AB_2565429anti-HLA-E (clone 3D12) - 163 DyBioLegendCat#342602; RRID:AB_1659247anti-HLA-F (clone 3D11/HLA-F) - 166 DyBioLegendCat#373202; RRID:AB_2650870anti-B2 microglobulin (clone 2M2) - 172 YbBioLegendCat#316302; RRID:AB_492835anti-HLA-A,B,C (clone W6/32) – 174 YbBioLegendCat#311402; RRID:AB_314871anti-Ki67 (clone Ki-67) - Alexa Fluor® 647BioLegendCat#350510; RRID:AB_10900821anti-LMP1 (clones CS.1/CS.2/CS.3/CS.4)DakoCat#IR753anti-EBNA2 (clone PE2)AbcamCat#ab49498; RRID:AB_2049594anti-CD3 (clone 2GV6)VentanaCat#790–4341; RRID:AB_2335978anti-FOXP3 (clone 236A/E7)AbcamCat#ab96048; RRID:AB_10861686anti-CD25 (clone 4C9)LeicaCat#PA0306; RRID:AB_10556556anti-CD163 (clone MRQ-26)VentanaCat#760–4437; RRID:AB_2335969anti-CD20 (clone SP32) - BX035AbcamCat#ab64089; RRID:AB_1139887anti-PAX5 (clone D7H5X) - BX021Cell Signaling TechnologyCat#93009; RRID:AB_3080867anti-CD79a (clone EP3618) - BX004AbcamCat#ab79414; RRID:AB_2260147anti-CD138/Syndecan-1 (clone SP152) - BX037AbcamCat#ab130465, RRID:AB_11157586anti-CD38 (clone EPR4106) - BX050AbcamCat#ab176886, RRID:AB_2864383anti-CD30 (clone BLR055F) - BX005AbcamCat#ab272089anti-CD27 (clone EPR8569) - BX001AbcamCat#ab131254, RRID:AB_11155136anti-CD3e (clone EP449E) - BX045Akoya BiosciencesCat#240006, RRID:AB_2928989anti-CD4 (clone EPR6855) - BX003Akoya BiosciencesCat#4350018, RRID:AB_2915936anti-CD8 (clone C8/144B) - BX026Akoya BiosciencesCat#232151, RRID:AB_2928985anti- PD1 (clone D4W2J) - BX014Cell Signaling TechnologyCat#86163, RRID:AB_2728833anti-CD39 (clone EPR20627) - BX099Akoya BiosciencesCat#4250076anti-ICOS (clone D1K2T) - BX042Cell Signaling TechnologyCat#89601, RRID:AB_2800142anti-IDO1 (clone D5J4E) - BX025Cell Signaling TechnologyCat#86630, RRID:AB_2636818anti-LAG3 (clone EPR20261) - BX027AbcamCat#ab209236, RRID:AB_2883982anti-CD56/NCAM1 (clone EP2567Y) - BX052AbcamCat#ab215981, RRID:AB_3096036anti-CD68 (clone KP1) - BX015Akoya BiosciencesCat#4550113, RRID:AB_2935894anti-CD69 (clone EPR21814) - BX017AbcamCat#ab234512, RRID:AB_2891140anti-CD163 (clone EPR19518) - BX069Akoya BiosciencesCat#4250079, RRID:AB_2935895anti-CD14 (clone EPR3653) - BX030AbcamCat#ab133335, RRID:AB_2889158anti-MPO (clone EPR20257) - BX006AbcamCat#ab221847, RRID:AB_3086778Biological samplesLeukocyte reduction chamber from adult human bloodStanford Blood Centerhttps://stanfordbloodcenter.org/Whole blood from healthy adult donorsObtained under informed consent with IRB approvalN/AChemicals, peptides, and recombinant proteinsHeat Inactivated Premium Fetal Bovine SerumCorningCat#33-016-CVGlutaMAX SupplementGibcoCat#35050-0612-mercaptoethanol 1000X in DPBSGibcoCat#21985023RNeasy Mini KitQiagenCat#74104iScript™ Advance cDNA Synthesis Kit for RT-qPCRBioRadCat#172-5038PowerUP™ SYBR™ Green Master MixApplied BiosystemsCat#A25742Recombinant Human CCL3 (MIP-1a)BioLegendCat#759502Precision Count BeadsBioLegendCat#424902Zombie Aqua Fixable Viability KitBioLegendCat#423102CellTrace VioletInvitrogenCat#C34557Human TruStain FcX™ (Fc Receptor Blocking Solution)BioLegendCat#422302Ficoll-Paque PLUSCytivaCat#17-1440-02Benzonase Nuclease, 2500 unitsMilliporeCat#E1014-25KUeBioscience™ Cell Stimulation CocktailInvitrogenCat#00-4970-93Maxpar® PBSFluidigm (now Standard BioTools)Cat#201058Maxpar® Cell Staining BufferFluidigm (now Standard BioTools)Cat#201068Maxpar® Fix I BufferFluidigm (now Standard BioTools)Cat#201065Maxpar® 10X Barcode Perm BufferFluidigm (now Standard BioTools)Cat#201057Cell-ID 20-Plex Palladium Barcoding KitFluidigm (now Standard BioTools)Cat#201060Electron Microscopy Sciences 16% Paraformaldehyde Aqueous Solution, EM Grade, Ampoule 10 ML - pack of 10Fisher ScientificCat#50-980-487Cell-ID™ Intercalator-Ir—125 μMFluidigm (now Standard BioTools)Cat#201192AMaxpar® Fix and Perm BufferFluidigm (now Standard BioTools)Cat#201067EQ Four Element Calibration BeadsFluidigm (now Standard BioTools)Cat#201078Maxpar® Cell Acquision SolutionFluidigm (now Standard BioTools)Cat#201240Cell-ID™ CisplatinFluidigm (now Standard BioTools)Cat#201064MIBItag Conjugation Kit_113inIonPathCat#600113MIBItag Conjugation Kit_157GdIonPathCat#600157Maxpar® X8 Antibody Labeling Kit, 141Pr—4 RxnFluidigm (now Standard BioTools)Cat#201141AMaxpar® X8 Antibody Labeling Kit, 142Nd—4 RxnFluidigm (now Standard BioTools)Cat#201142AMaxpar® X8 Antibody Labeling Kit, 143Nd—4 RxnFluidigm (now Standard BioTools)Cat#201143AMaxpar® X8 Antibody Labeling Kit, 144Nd—4 RxnFluidigm (now Standard BioTools)Cat#201144AMaxpar® X8 Antibody Labeling Kit, 145Nd—4 RxnFluidigm (now Standard BioTools)Cat#201145AMaxpar® X8 Antibody Labeling Kit, 146Nd—4 RxnFluidigm (now Standard BioTools)Cat#201146AMaxpar® X8 Antibody Labeling Kit, 147Sm—4 RxnFluidigm (now Standard BioTools)Cat#201147AMaxpar® X8 Antibody Labeling Kit, 148Nd—4 RxnFluidigm (now Standard BioTools)Cat#201148AMaxpar® X8 Antibody Labeling Kit, 149Sm—4 RxnFluidigm (now Standard BioTools)Cat#201149AMaxpar® X8 Antibody Labeling Kit, 150Nd—4 RxnFluidigm (now Standard BioTools)Cat#201150AMaxpar® X8 Antibody Labeling Kit, 151Eu—4 RxnFluidigm (now Standard BioTools)Cat#201151AMaxpar® X8 Antibody Labeling Kit, 153Eu—4 RxnFluidigm (now Standard BioTools)Cat#201153AMaxpar® X8 Antibody Labeling Kit, 154Sm—4 RxnFluidigm (now Standard BioTools)Cat#201154AMaxpar® X8 Antibody Labeling Kit, 155Gd—4 RxnFluidigm (now Standard BioTools)Cat#201155AMaxpar® X8 Antibody Labeling Kit, 156Gd—4 RxnFluidigm (now Standard BioTools)Cat#201156AMaxpar® X8 Antibody Labeling Kit, 158Gd—4 RxnFluidigm (now Standard BioTools)Cat#201158AMaxpar® X8 Antibody Labeling Kit, 159Tb—4 RxnFluidigm (now Standard BioTools)Cat#201159AMaxpar® X8 Antibody Labeling Kit, 160Gd—4 RxnFluidigm (now Standard BioTools)Cat#201160AMaxpar® X8 Antibody Labeling Kit, 161Dy—4 RxnFluidigm (now Standard BioTools)Cat#201161AMaxpar® X8 Antibody Labeling Kit, 162Dy—4 RxnFluidigm (now Standard BioTools)Cat#201162AMaxpar® X8 Antibody Labeling Kit, 163Dy—4 RxnFluidigm (now Standard BioTools)Cat#201163AMaxpar® X8 Antibody Labeling Kit, 164Dy—4 RxnFluidigm (now Standard BioTools)Cat#201164AMaxpar® X8 Antibody Labeling Kit, 165Ho—4 RxnFluidigm (now Standard BioTools)Cat#201165AMaxpar® X8 Antibody Labeling Kit, 167Er—4 RxnFluidigm (now Standard BioTools)Cat#201167AMaxpar® X8 Antibody Labeling Kit, 168Er—4 RxnFluidigm (now Standard BioTools)Cat#201168AMaxpar® X8 Antibody Labeling Kit, 169Tm—4 RxnFluidigm (now Standard BioTools)Cat#201169AMaxpar® X8 Antibody Labeling Kit, 170Er—4 RxnFluidigm (now Standard BioTools)Cat#201170AMaxpar® X8 Antibody Labeling Kit, 171Yb—4 RxnFluidigm (now Standard BioTools)Cat#201171AMaxpar® X8 Antibody Labeling Kit, 172Yb—4 RxnFluidigm (now Standard BioTools)Cat#201172AMaxpar® X8 Antibody Labeling Kit, 173Yb—4 RxnFluidigm (now Standard BioTools)Cat#201173AMaxpar® X8 Antibody Labeling Kit, 174Yb—4 RxnFluidigm (now Standard BioTools)Cat#201174AMaxpar® X8 Antibody Labeling Kit, 176Yb—4 RxnFluidigm (now Standard BioTools)Cat#201176AGene Knockout Kit v2Synthegohttps://store.editco.bio/products/gene-knockout-kitsCas9 nuclease 2NLS, S. pyogenesSynthegoN/A (add-on item to Gene Knockout Kit v2); https://store.editco.bio/products/add-onsSF Cell Line 4D-Niucleofector X Kit LLonzaCat#V4XC-2024QuickExtract DNA Extraction SolutionLucigenCat#QE0905TPhusion Green Not Start II High-Fidelity PCR Master MixThermo ScientificCat#F-566SBenchTop 100bp DNA LadderPromegaCat#G8291Annexin-binding bufferThermo Fisher ScientificCat#422201Annexin V-Alexa Fluor™ 488Thermo Fisher ScientificCat#V13245DAPIBioLegendCat#422801BD Cytofix/Cytoperm™ Fixation/Permeabilization KitBD BiosciencesCat#554714Critical commercial assaysHuman CCL3/MIP-1alpha DuoSet ELISAR&D SystemsCat#DY270-05Human CCL4/MIP-1beta DuoSet ELISAR&D SystemsCat#DY271-05Human IL-15 DuoSet ELISAR&D SystemsCat#DY-247-05DuoSet ELISA Ancillar Reagent Kit 2R&D SystemsCat#DY008CyQUANT™ Direct Cell Proliferation Assay KitThermoFisher ScientificCat#C35011Deposited datascRNAseq data from CD300A KO vs. mock-edited EBV(+) PTLD B cell linesThis paperGEO: GSE279989 (superseries GEO: GSE279991)Gene expression microarray data from EBV(+) PTLD B cell lines and EBV(−) DLBCL linesThis paperGEO: GSE279598 (superseries GEO: GSE279991)Morscio et al., microarray gene expression dataset from EBV(+) and EBV(−) PTLD tumor biopsiesMorscio et al.14GEO: GSE38885Craig et al., microarray gene expression dataset from EBV(+) and EBV(−) PTLD tumor biopsiesCraig et al.16https://doi.org/10.1097/pdm.0b013e31804f54a9Vakiani et al., microarray gene expression dataset from EBV(+) and EBV(−) PTLD tumor biopsiesVakiani et al.15https://doi.org/10.1002/hon.859Experimental models: Cell linesHuman: EBV(+) PTLD B cell lines established from patients with EBV(+) B cell PTLDObtained under informed consent with IRB approvalN/AHuman: PfeifferAmerican Type Culture Collection (ATCC)Cat#CRL-2632; RRID:CVCL_3326Human: ToledoLaboratory of Dr. Ronald Levy (Stanford University, USA)N/AHuman: THP-1Laboratory of Dr. Maria-Grazia Roncarolo (Stanford University, USA)N/AExperimental models: Organisms/strainsMouse: NOD.CB17-Prkdcscid/JJackson LaboratoryRRID:IMSR_RJ:NOD-SCIDOligonucleotidesSee Table S7 for primers used in PCR and qPCR analysesN/ASoftware and algorithmsR programming languageThe R Foundationhttps://www.r-project.orgRStudioRStudiohttp://rstudio.comCellEnginePrimity Biohttps://cellengine.com/CellRanger software10X Genomicshttps://www.10xgenomics.com/supportGraphPad PrismGraphPad Softwarehttp://www.graphpad.com/Python 3.11Pythonhttps://www.python.org/QuPathBankhead, P. et al.83https://qupath.readthedocs.io/en/0.5/ Experimental model and study participant details Cell lines and culture All cell lines used in this study are of the DLBCL subtype of B cell lymphoma and were established prior to this study.59 Institutional Review Board (IRB) approval was obtained for collection of blood from transplant recipients and written informed consent obtained from subjects for establishment of in vitro cell lines. Five EBV(+) PTLD B cell lines were established from peripheral blood (JB7, JC62, MF4, VB5) or lymph nodes (AB5) of individual patients with EBV(+) B cell PTLD, as previously described and characterized.59 We previously reported that these EBV(+) PTLD lines each express the latency II/III gene LMP184,85 and three of three analyzed (JB7, JC62, and AB5) express the latency III gene LMP2a to varying degrees.84,85 The EBV(−) DLBCL line Pfeiffer was obtained from the American Type Culture Collection (ATCC; # CRL-2632). The EBV(−) DLBCL line Toledo and the monocytic cell line THP-1 were generous gifts from Dr. Ronald Levy (Stanford, USA) and Dr. Maria-Grazia Roncarolo (Stanford, USA) respectively. Cell lines were cultured in RPMI-1640 (Cytiva, # SH30096.FS) supplemented with 10% (v/v) heat-inactivated fetal bovine serum (FBS; Corning, # 35-016-CV) and 1% (v/v) Glutamax (Gibco, # 35050-061) at 37°C with 5% CO2. 0.05 mM β-mercaptoethanol (Gibco, # 21985023) was added to the cell culture medium for THP-1 cells. In vivo xenograft model of PTLD All animal studies were performed in accordance with the animal care guidelines set by Stanford University’s Administrative Panel on Laboratory Animal Care. Throughout experimentation, mice were co-housed with food and water provided ad libitum, and monitored every 1–2 days. Six-week-old male NOD-SCID (NOD.CB17-Prkdcscid/J) mice (Jackson Laboratory, Bar Harbor, ME) were injected subcutaneously over the right flank with 7.5 × 106 EBV(+) PTLD cells suspended in PBS. Mice were randomly allocated into four groups (n = 5): (a) AB5, CD300A KO; (b) AB5, mock control; (c) JB7, CD300A KO; (d) JB7, mock control. Tumors were measured every 1–2 days in two perpendicular dimensions along the plane of the body using Vernier calipers, and tumor volume calculated as previously described22,54 using the following modified ellipsoidal formula: Method details Integrated multi-cohort analysis Three microarray gene expression datasets generated previously from EBV(+) and EBV(−) PTLD primary tumor biopsies by three separate studies were obtained from the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus or kindly provided by the corresponding study authors.14,15,16 Altogether, these consist of 60 PTLD primary tumor samples, encompassing 42 EBV(+) PTLD and 18 EBV(−) PTLD samples, collected and profiled at three independent institutions in two countries. Cohort details and patient clinical characteristics are provided as reported by their primary studies in Table S1. A majority of these samples are of the monomorphic diffuse large B cell lymphoma (DLBCL) subtype of PTLD. To identify differentially expressed genes (DEGs) between EBV(+) and EBV(−) PTLD samples, a multi-cohort analysis was performed using the R package MetaIntegrator (v2.1.3) as previously described.19,86 Briefly, the effect size (ES) was estimated for each gene in each dataset as Hedges’ adjusted g and combined across studies using a random-effects inverse variance model. p values were corrected for multiple hypotheses testing using the Benjamini-Hochberg correction to obtain the false discovery rate (FDR) and combined using Fisher’s sum of logs method to obtain the Fisher FDR. DEGs between EBV(+) PTLD and EBV(−) PTLD were identified by selecting genes with |log2(ES)| > 0.59 (equivalent to |ES| > 1.5), and Fisher and ES FDR <20%.87 An additional filtering step for only genes that were represented in at least two out of three datasets used was implemented, to ensure that chosen genes showed a consistent effect between independent datasets. We refer to this gene set as the EBV-associated PTLD gene signature and list these genes in Table S2. In silico cellular deconvolution We performed in silico cellular deconvolution using immunoStates as a basis matrix with support vector regression to estimate proportions of 25 immune cell subsets in each sample.33,34,35,36 To investigate changes in the immune cell proportions between EBV(+) and EBV(−) PTLD samples, we performed a multi-cohort analysis on the estimated immune cell proportions using the R package MetaIntegrator.19,86 Thresholds of |log2(ES)| > 0.59 and FDR <20%87 were applied to identify immune cell types with increased or decreased proportions between EBV(+) PTLD and EBV(−) PTLD. RNA extraction Cell pellets were obtained from healthy cell cultures in the log-phase of growth and immediately stored at −80°C for later use. Total RNA was isolated using the RNeasy Mini Kit (Qiagen, Santa Clarita, CA, # 74104) according to the manufacturer’s instructions. RNA was eluted in the supplied RNase-free water and immediately stored at −80°C for later use. Gene expression profiling (microarray) analysis Gene expression profiling using total RNA isolated from EBV(+) and EBV(−) B cell lines was carried out by the Protein And Nucleic Acid (PAN) Facility at Stanford University. Briefly, total RNA was checked for quality and integrity using a BioAnalyzer (Agilent). Then, in vitro transcription, fragmentation, and labeling steps were carried out, followed by hybridization of samples to human Clariom S arrays (Applied Biosystems) and scanning of the microarrays. Transcriptome Analysis Console software (Applied Biosystems) was utilized for downstream analysis of microarray gene expression data. Specifically, background reduction and quantile normalization were first carried out via the signal-space transformation and robust multi-chip analysis (SST-RMA) algorithms. Next, the linear models for microarray data (LIMMA) method was implemented to compare SST-RMA-normalized signal intensities between EBV(+) and EBV(−) B cell lines. Only genes with |log2(fold change)| ≥ 1 and p < 0.05 were considered to be significantly differentially expressed. A list of these DEGs is provided in Table S2. Pathway analysis Functional pathway analysis was carried out on the 189 EBV-associated PTLD signature genes as well as the 1407 DEGs between EBV(+) and EBV(−) B cell lines using the R package EnrichR.23,24,25,88 The Gene Ontology Biological Process 2018 repository was chosen as the gene set library. The EnrichR combined scores were calculated and the top 15 gene sets as ranked by p-values were visualized. Reverse transcription quantitative polymerase chain reaction (RT-qPCR) RNA quality and concentration were measured using a Nanodrop spectrophotometer (Nanodrop Technologies, Wilmington, DE, USA). Reverse transcription was carried out on 3μg of total RNA using the iScript cDNA Synthesis Kit (Bio-Rad, Hercules, CA, USA, # 1708890) according to the manufacturer’s instructions. cDNA was either immediately used for qPCR or stored at −20°C for later use. qPCR was carried out using the PowerUp SYBR Green Master Mix (Applied Biosystems, Waltham, MA, USA, # A25742). Gene-specific primers were either obtained as commercially available 20x PrimePCR SYBR Green Assay primer mixes (Bio-Rad) or synthesized at the Stanford University PAN Facility and used at a final concentration of 0.5μM. Ten 10μL reactions were set up in quadruplicate and fluorescent signals read on a CFX384 Real-Time PCR Detection System (Bio-Rad, Hercules, CA, USA). ACTB and GAPDH (2 primer sets each) were included as reference genes for normalization. Where applicable, primer sequences are included in Table S7. For each sample/gene combination, the arithmetic mean Ct was calculated across the quadruplicates. The mean Ct values of the four reference gene qPCR reactions were combined into an overall reference gene Ct value by computing their geometric mean. ΔCt values were computed by subtracting this overall reference gene Ct value from the Ct values of the gene of interest. Results are shown as relative gene expression, calculated as 2−ΔCt. Enzyme-linked immunosorbent assay (ELISA) Cell lines were cultured for 72h in 6-well plates before supernatants were collected and stored at −80°C for later use. CCL3 and CCL4 concentrations were assayed using DuoSet ELISA kits (R&D Biosystems, # DY270-05 and # DY271-05) according to the manufacturer’s instructions. Where necessary, supernatants were diluted in a 1x solution of the supplied reagent diluent to ensure that absorbance values fell within standard curve ranges. Supernatants were plated in triplicate. Absorbance values were measured at 450nm and 540nm (for wavelength correction) on a Spectramax i3x microplate reader (Molecular Devices, San Jose, CA, USA). Wavelength-corrected values were interpolated on the standard curve and the results were multiplied by the dilution factor (if applicable) and normalized to cell concentration at the end of the 72h culture. Transwell migration assay EBV(+) and EBV(−) cell lines were cultured for 72h in 6-well plates before supernatants were collected and stored at −80°C for later use as chemoattractants for transwell migration assays. Transwell migration assays were performed using polycarbonate membrane 8.0 μm pore Transwell inserts (Corning, # 3422). THP-1 cells were labeled with 5μM CellTrace Violet (Invitrogen, #C34557) for 20 min and resuspended in fresh culture medium at 1×106 cells/ml. Labeled THP-1 cells (100μL) were added to the top chamber. Culture supernatants from EBV(+) and EBV(−) cell lines (600μL) were placed in the bottom chamber. A positive control (10 ng/ml CCL3 in fresh culture medium) and negative control (fresh culture medium alone) were included in each assay. Following incubation at 37°C with 5% CO2 for 3 h, the transwells were removed and the contents of the bottom chamber were collected. Migrated cells were washed in cold phosphate-buffered saline (PBS) before staining with Zombie Aqua viability dye (BioLegend, #423102) according to the manufacturer’s instructions. Subsequently, the cells were stained with anti-human CD14-APC antibody (BioLegend, # 325607), fixed in 1% paraformaldehyde (PFA), and enumerated via flow cytometry on an LSR II flow cytometer (BD Biosciences) using Precision Count Beads (BioLegend, # 424902). Absolute counts of migrated cells for each sample were calculated using the following equation: Specific migration indices for each sample were calculated by normalizing the number of migrated cells in the sample to that from the negative control, as follows: p values were corrected for multiple hypothesis testing using the original Benjamini-Hochberg FDR method. Luminex assay and analysis Cell lines were cultured for 72h in 6-well plates before supernatants were collected and stored at −80°C for later use. Supernatants were profiled via Luminex using human 48-plex cytokine kits at the Human Immune Monitoring Center at Stanford University as previously described,89 with samples run in technical duplicates. Data points with a bead count <20 for the corresponding cytokine were excluded. For each cytokine, mean values were calculated from the median fluorescent intensity (MFI) of the technical duplicates. Samples with >20% difference between MFI values from both technical duplicates were flagged. Results are expressed as log10-transformed background-corrected MFI values and are normalized to the cell concentration at time of supernatant harvest where applicable. The Wilcoxon rank-sum test (Mann-Whitney U test) was used to compare cytokine secretion between experimental groups. Where appropriate, multiple hypothesis correction was performed using the Holm method. Tissue microarray construction All human tissue samples containing post-transplant lymphoproliferative disorders were archival specimens obtained from the Department of Pathology with Institutional Review Board approval. Tissue microarrays (TMAs) were constructed with duplicate 0.6 mm cores of formalin-fixed paraffin-embedded (FFPE) tissue from diagnostic biopsies of EBV(+) and EBV(−) diffuse large B cell lymphomas (monomorphic PTLD). TMA construction method, evaluation, and scoring were as previously described.90 The percentage of positive cells for CD3, CD4, CD25 and CD163 were evaluated by two pathologists (S.F.Y. and Y.N.) and categorized into: 1, 0–5% positive cells; 2, 6–29% positive cells; and 3, ≥30% positive cells. The percentage of FOXP3 positive cells was similarly evaluated and further categorized into: 1 < 30% and 2, ≥30% positive nuclei. Immunohistochemical staining and in situ hybridization Standardized automated immunohistochemistry (IHC) and in situ hybridization were performed on Ventana Benchmark Ultra (Roche Tissue Diagnostics, Tucson, AZ) or Leica BOND-III (Leica Biosystems, Buffalo Grove, IL). Commercially available antibodies were used as shown in Table S8. Co-Detection by Indexing (CODEX) Antibody conjugation and validation Antibodies were screened and validated by individual staining on FFPE human tonsil tissue samples. Immunohistochemistry was used to cross-validate antibodies with the same, non-conjugated antibody clones listed in Table S8. Five-micron thick sections of FFPE tissue were immobilized on charged slides (Superfrost PlusTM) (Fisher Scientific, Hampton, NH) according to the manufacturer’s instructions. Staining, multicycle setup and imaging FFPE tissue sections were incubated with a cocktail containing 23 nucleotide-barcoded primary antibodies after undergoing nuclear staining (DAPI). They were then loaded on the stage of an automated inverted fluorescence microscope connected to the robotic fluidic system (PhenoCycler-Fusion, Akoya Biosciences, Marlborough, MA). In addition to the nuclear stain fluorophore-tagged complementary nucleotide sequences (reporters) were used to iteratively reveal three antibodies per cycle as previously described.91,92,93 Auto-fluorescence background was minimized by two blank cycles at the beginning and at the end of the antibody reveal cycles. Automated image acquisition, processing, segmentation, and fluidics exchange were conducted using PhenoCycler-Fusion software (Akoya Biosciences). Imaging was done using a Keyence BZ-X810 inverted fluorescent microscope and staining patterns were confirmed by two pathologists (S.F.Y. and Y.N.). Data analysis Image analysis was performed using QuPath.83 Lymphoma tissue biopsies were annotated by the pathologists. Multiplex images were matched with the corresponding hematoxylin and eosin (H&E) images. Cell segmentation on the selected regions of interest was performed with QuPath using StarDist and scripts supporting open-source tools for Phenocycler analysis according to the guidelines (Akoya Biosciences). Single cell data were extracted to CSV files. Immunological markers were assessed according to their known cell compartment expression of membrane, nucleus, or cytoplasm. To statistically evaluate the difference in marker expression between EBV(+) and EBV(−) groups, we used the Wilcoxon rank-sum test. Given the multiple comparisons inherent in our analysis, we adjusted p-values using the Benjamini-Hochberg correction to control the false discovery rate. This correction ensures the robustness of our findings against the risk of type I errors. Preparation of peripheral blood mononuclear cells (PBMC) PBMC were isolated by Ficoll-Hypaque (Cytiva, # 17-1440-02) density centrifugation from leukoreduction system chambers obtained via the Stanford Blood Center or heparinized peripheral blood obtained via venipuncture with informed consent from healthy volunteers. Isolated PBMC were immediately used or cryopreserved in pure FBS with 10% DMSO at −80°C overnight and in liquid nitrogen for long-term storage. Generation of autologous B lymphoblastoid cell lines (LCL) LCL were generated by infection of freshly isolated B cells from three healthy blood donors using the B95.8 strain of EBV as previously described.59,94 B cell stimulation Cryopreserved PBMC were thawed in the presence of 25U/ml Benzonase nuclease (Millipore Sigma, #E1014-25KU), resuspended in cell culture medium, and rested for 30 min at 37°C with 5% CO2 before use. Rested PBMC were stimulated with 1× eBioscience Cell Stimulation Cocktail containing PMA and ionomycin (Invitrogen, # 00-4970-03) for 5–6 h at 37°C with 5% CO2. Mass cytometry by time-of-flight (CyTOF) Heavy metal-tagged antibodies were purchased pre-conjugated from Fluidigm where available, or conjugated and validated in-house using the Maxpar X8 Antibody Labeling Kits (Fluidigm) or the MIBItag Conjugation Kits (IonPath) according to the manufacturer’s instructions. Each conjugated antibody was validated and titrated to its optimal staining concentration using a combination of primary human PBMC populations and B cell lines. Antibody panels used in each CyTOF experiment are provided in Table S4. Cells were resuspended in 2.5μM Cell-ID Cisplatin (Fluidigm, # 201064) in Maxpar PBS (Fluidigm, # 201058) for 5 min at room temperature (RT) to stain non-viable cells, before being washed in Maxpar Cell Staining Buffer (CSB; Fluidigm, # 201068). In preparation for sample barcoding, cells were fixed in 1× Maxpar Fix I buffer (Fluidigm, # 201065) for 10 min at RT, then permeabilized 1× Maxpar barcode perm buffer (Fluidigm, # 201057) for 5 min at 4°C. Next, sample barcoding was carried out using the Cell-ID 20-Plex Palladium Barcoding Kit (Fluidigm, # 201060) according to the manufacturer’s instructions. Barcoded cells were then pooled and resuspended in Human TruStain FcX Fc Receptor Blocking Solution (BioLegend, # 422302) for 10 min at RT, before undergoing surface antibody staining. An aliquot of cells from each sample was set aside as an unstained control (besides lineage markers necessary for gating) for all markers of interest. All antibody staining steps were carried out in CSB for 30 min at RT. After staining, samples then underwent another fixation step using 1.6% paraformaldehyde in Maxpar PBS for 15 min at RT. DNA was then labeled using 125nM Cell-ID Intercalator-Ir (Fluidigm, # 201192A) in Maxpar Fix and Perm Buffer (Fluidigm, # 201067) and incubated overnight at 4°C. Prior to acquisition, cells were washed in Maxpar Cell Acquisition Solution (CAS; Fluidigm, # 201240), resuspended in a 10% solution of EQ Four Element Calibration Beads (Fluidigm, # 201078) in CAS, and filtered using polystyrene FACS tubes with cell strainer caps (Fisher Scientific, # 08-771-23). Cells were acquired on a Helios mass cytometer (Fluidigm). CRISPR/Cas9 knockout and validation Single-gene knockout cell lines were generated according to the procedure described by Jiang et al.95 Briefly, Gene Knockout Kits v2 (Synthego) comprising three single-guide RNA (sgRNA) sequences were used for each target gene. Cas9-sgRNA ribonucleoprotein complexes (RNPs) were obtained by combining gene-specific sgRNA with Cas9-2NLS nuclease (Synthego) in a 3:1 ratio. RNPs were introduced into cells through nucleofection, using a 4D-Nucleofector System with X Unit (Lonza) and the SF Cell Line 4D-Nucleofector X Kit L (Lonza, #V4XC-2024). Corresponding mock controls for each cell line were performed via nucleofection with Cas9-2NLS in the absence of sgRNA. Nucleofected cells were allowed to recover in fresh culture medium for 72 h at 37°C with 5% CO2. To validate knockout efficiency, genomic DNA was extracted from 0.5-1.0 × 106 cells using QuickExtract (Lucigen, # QE0905T) and subject to amplification via PCR with primers flanking the Cas9 cut site. PCR primers were synthesized at the Stanford PAN Facility and used at a final concentration of 0.5μM. Where applicable, primer sequences are included in Table S7. PCR products were sent for Sanger sequencing and the resulting chromatograms analyzed for gene knockout efficiency against mock controls and unedited cells using the ICE algorithm96 (https://www.synthego.com/products/bioinformatics/crispr-analysis). Formalin-fixed paraffin embedded (FFPE) cell pellets were generated using CD300A KO and mock control cells from AB5 and JB7 cell lines, as previously described.97 4μm sections from each pellet were stained using immunohistochemistry (IHC) for CD300a evaluation. IHC was performed using anti-CD300a antibody (clone MEM-260, mouse monoclonal, ThermoFisher Scientific # MA1-10190). In vitro growth assay for gene knockout cell lines 100μL of cells were seeded in quadruplicate at 0.2 × 106 cells/mL in a 96-well microtiter plate with fresh culture medium at days 10 and 14 post-CRISPR gene editing. After 48h of culture, cell growth was quantified using the CyQUANT Direct Cell Proliferation Kit (Invitrogen/Thermo Fisher, Waltham, MA, USA, #C35011) according to the manufacturer’s instructions. Briefly, 2x detection reagent was prepared in cell culture medium and 100μL was added directly to each well. Cells were incubated for 1 h at 37°C with 5% CO2. Fluorescence was measured at 505nm excitation and 530nm emission on a Spectramax i3x microplate reader (Molecular Devices, San Jose, CA, USA). Background fluorescence signals were calculated using replicate wells containing culture medium only, then subtracted from all other wells. Relative cell growth was derived by expressing background-corrected fluorescence values as a percentage of the corresponding mock-nucleofected controls. Apoptosis assay Cells were seeded in triplicate at 1.0 × 106 cells per well in a 24-well microtiter plate with fresh culture medium. After 72h of culture, cells were washed twice with 200μL of 1× annexin-binding buffer (Thermo Fisher, Waltham, MA, USA, #V13245), then resuspended in 100μL of 1× annexin-binding buffer. Subsequently, 5μL of annexin V- Alexa Fluor 488 (Thermo Fisher, Waltham, MA, USA, #V13245) was added to the cell suspension and left to incubate for 15 min at room temperature in the dark. Cells were then washed twice with 200μL of 1× annexin-binding buffer, resuspended in 200μL of 1× annexin-binding buffer, and incubated with 20μL of DAPI (BioLegend, #422801) at 10 mg/ml for 15 min at room temperature in the dark. Cells were analyzed via flow cytometry on a BD FACSymphony (BD Biosciences) with CellEngine software (cellengine.com, Primity Bio, Fremont, CA). Late apoptotic cells were identified as annexin V+ and DAPI+ while early apoptotic cells were identified as annexin V+ and DAPI-. Cell cycle assay Cells were seeded in triplicate at 1.0 × 106 cells per well in a 24-well microtiter plate with fresh culture medium. After 72h of culture, cells were washed twice with 200μL of PBS, then incubated with 100μL of Fixation/Permeabilization solution from the BD Cytofix/Cytoperm Fixation/Permeabilization Kit (BD Biosciences, # 554714) for 15 min at 4°C. Subsequently, cells were washed twice with 200μL of 1× BD Perm/Wash buffer from the BD Cytofix/Cytoperm Fixation/Permeabilization Kit, before being incubated with anti-human Ki67-Alexa Fluor 647 antibody (BioLegend, # 350510) at a dilution of 1:1000 in 1× BD Perm/Wash buffer for 20 min at room temperature in the dark. Cells were then washed twice with 1× BD Perm/Wash buffer, resuspended in 200μL of 1× BD Perm/Wash, and incubated with 20μL of DAPI (BioLegend, #422801) at 10 mg/ml for 15 min at room temperature in the dark. Cells were analyzed via flow cytometry on a BD FACSymphony (BD Biosciences) with CellEngine software (cellengine.com, Primity Bio, Fremont, CA). Processing of single-cell RNA sequencing (scRNA-seq) libraries Reads from 10x scRNA expression libraries were aligned to human genome assembly GRCh38 (hg19) and quantified using cellranger count (10x Genomics). The filtered feature-barcode matrices containing only cellular barcodes were used for further analysis. Single-cell gene expression matrices were imported into R (v4.0.3) and analyzed using Seurat98 (v4.3.0). Cells were filtered according to the following criteria: number of genes captured within 2 standard deviations of the mean per library; number of UMIs captured >3000 and 10% of cells from either condition and had a log2 fold change of >0.04 between conditions were tested. Mitochondrial and ribosomal genes were excluded from differential expression testing. Genes with adjusted p-value <0.05 and fold change >1.15 were considered to be significantly differentially expressed. Pathway enrichment analysis was performed on identified DEGs using the enrichGO() function from the R package clusterProfiler99 (v4.6.0) with a q-value cut-off of 0.05. Gene ontologies from the biological process, molecular function, and cellular component sub-ontologies were tested. As a second approach the data were preprocessed using Python’s Scanpy package to filter genes and cells based on expression thresholds and annotated mitochondrial genes for exclusion. Subsequently, we normalized the data and performed logarithmic transformation to address technical variance. Using the Wilcoxon rank-sum test, we identified differentially expressed genes between CD300A KO and mock-edited cells, categorizing them as up/down-regulated based on log-fold changes. After merging these differential expression lists from all cell lines with a fold change threshold of 1.15 for refinement, we performed pathway enrichment analysis on these gene sets using GSEApy’s “KEGG 2021 Human” library (Figure S5). Immunohistochemistry of solid tumors developed in in vivo xenograft model of PTLD Tumor tissues developed in vivo from EBV(+) PTLD cells were excised at sacrifice. Resected tumor sections were preserved in 10% neutral buffered formalin and paraffin-embedded before being sliced into 5μm sections (Histo-Tec Laboratory Inc., Hayward, CA). Four sections from each tumor were stained with hematoxylin and eosin (H&E) for morphologic assessment, in situ hybridization (ISH) for detection of Epstein-Barr virus encoded RNA (EBER), and immunohistochemistry (IHC) for CD20, Ki67, and Ly6C evaluation. IHC staining for CD20 and Ki67 was performed using an anti-CD20 antibody (clone L26, mouse monoclonal, 1:1000, DAKO #M0755) or an anti-Ki67 antibody (clone MIB-1, mouse monoclonal, 1:200, DAKO #M7240) on a Leica BOND-III platform (Leica Biosystems, Buffalo Grove, IL). IHC staining for murine Ly6C was performed using an anti-Ly6C antibody (clone EPR27220-23, rabbit monoclonal, Abcam # ab314120). EBER-ISH was performed on a Ventana Benchmark Ultra XT (Roche/Ventana Medical Systems, Tucson, AZ) instrument. Morphologic and immunohistochemical evaluation of tumors was performed by two pathologists (S.F.Y. and Y.N.) blinded to tumor type. Tumors were assessed morphologically for the percentage of necrosis to the total tumor area, degree of pleomorphism (categorized into mild, moderate, and marked), mitotic activity (1+: <5, 2+: 5–10, and 3+: >10 mitotic figures per high power field (600× magnification) evaluated across at least 5 high power fields), and extent of Ly6C+ monocyte infiltration (1+: 1–4 cells per 300× magnification, 2+: 5–29 cells per 300× magnification, 3+: ≥30 cells per 300× magnification). The percentages of tumor cells expressing CD20, Ki67, and EBER were also reported. Quantification and statistical analysis Unless otherwise specified, all data and statistical analysis was carried out using the R programming language100 (v3.4.2-v4.2.2). Figures were generated using the R packages ggplot2101 (v3.4.1), ggbeeswarm102 (v0.7.1), and/or pheatmap103 (v1.0.12). Statistical analyses were carried out using the R package ggpubr104 (v0.6.0). Comparison between experimental groups was performed using the two-tailed non-parametric Wilcoxon rank-sum test (Mann-Whitney U test), unless otherwise specified. Statistical analysis of data from in vivo xenograft mouse model Statistical analysis of data generated using our in vivo xenograft mouse model was performed using GraphPad Prism (version 9.3.1.471 for Windows; GraphPad Software, San Diego, California USA, www.graphpad.com). Comparison of tumor volumes between CD300A KO and mock control groups was performed using 2-way repeated measures ANOVA; comparison between CD300A KO and mock control groups within each time point was corrected for multiple hypothesis testing using the original Benjamini-Hochberg FDR method. Pearson’s chi-square test of independence was conducted for statistical analysis of IHC staining data from tumor xenografts isolated at sacrifice. p values were corrected for multiple hypothesis testing using the original Benjamini-Hochberg FDR method. Statistical analysis of IHC staining data from primary PTLD FFPE tissues For statistical analysis of IHC staining data from primary PTLD FFPE tissues, the data points from the three staining categories were first combined into two categories (low vs. high). Then, Pearson’s chi-square test of independence was conducted between each EBV-related variable and individual markers of interest. Having this binary staining classification allows us to more clearly infer the directionality of association between EBV gene product expression or latency stage and the marker staining intensity if the chi-square test indicates an association between both variables assessed. p values were corrected for multiple hypothesis testing using the original Benjamini-Hochberg FDR method. CyTOF data pre-processing and analysis Mass cytometry.fcs files were concatenated, bead-normalized, and debarcoded with the R package premessa (https://github.com/ParkerICI/premessa/), which implements the MATLAB-based bead normalization method developed by Finck et al.105 Normalized and debarcoded data were then uploaded onto CellEngine (cellengine.com, Primity Bio, Fremont, CA). B cells were gated as CD19(+) non-B lineage(−) cells from live singlets in all samples (Figure S2F) for inclusion in downstream analyses. Data from events within the B cell gates for each sample were downloaded and further analyzed using the R package CATALYST.106,107 Raw expression values were transformed with an inverse hyperbolic sine (arcsinh) transformation using a cofactor of 5. For each analysis, phenotypic markers that did not exhibit positive staining within the B cell population for all included samples were removed. For visualization purposes, the median expression values of individual phenotypic markers for each sample were computed (where applicable) before the expression values of each marker were scaled between 0 and 1 across all samples, using the 1st and 99th percentiles as the lower and upper boundaries respectively. Principal component analysis was performed on median marker expression using the prcomp() function in R and individual loadings overlaid with using the R package ggbiplot (https://github.com/vqv/ggbiplot) (Figure 5A). To visualize expression patterns of phenotypic markers amongst the various B cell lines, cells were equally subsampled from each B cell line and shown on a UMAP plot generated using expression of all 27 phenotypic markers expressed in at least one B cell line (Figure 5C). Clustering and manual annotation of healthy B cell subsets To further classify healthy B cells into distinct functional subsets, 12 phenotypic markers described to functionally profile B cell subsets were selected.40,41 Using these 12 markers, both stimulated and unstimulated B cell samples were clustered as described in the CATALYST workflow.106,107 Briefly, cells were over-clustered into 100 microclusters via implementing FlowSOM, then meta-clustered into 10 clusters with ConsensusClusterPlus. Meta-clusters were then manually assigned to known B cell subsets, as previously characterized40,41 (Figure S2F). Clusters that could not be unambiguously characterized and assigned to known B cell subsets (clusters 5, 8, and 9) were removed from further analysis. Mapping of B cell lines to healthy B cell subsets To profile EBV(+) and EBV(−) B cell lines in relation to healthy B cell subsets, the median expression value for individual phenotypic markers was computed for each cell line and B cell subset. Median expression values were then scaled within each marker between 0 and 1, bound by the 1st and 99th percentiles. Given the distinct separation of the B cell subsets from the B cell lines, 0–1 scaling was carried out independently for the B cell subsets and the B cell lines in order to overlay their relative expression distributions and map the B cell lines to the B cell subsets. Unsupervised hierarchical clustering was then carried out on all samples and the Pearson correlation coefficient computed between samples in a pairwise manner, using the scaled median expression of all 27 selected phenotypic markers. Published: December 9, 2024 Footnotes Supplemental information can be found online at https://doi.org/10.1016/j.xcrm.2024.101851. Contributor Information Purvesh Khatri, Email: pkhatri@stanford.edu. Olivia M. Martinez, Email: omm@stanford.edu. Supplemental information Document S1. Figures S1–S6 and Tables S3, S4, and S6–S8 Table S1. Gene expression datasets from EBV(+) and EBV(−) PTLD tumor biopsies used in multi-cohort analyses, related to Figure 1 Table S2. Differentially expressed genes in EBV(+) PTLD compared to EBV(−) PTLD, and EBV(+) B cell lines compared to EBV(−) B cell lines, related to Figure 1 Table S5. Pathway overrepresentation analysis of genes downregulated in CD300A KO compared to mock control EBV(+) B cell lines, related to Figure 6 Document S2. Article plus supplemental information References Associated Data Supplementary Materials Document S1. Figures S1–S6 and Tables S3, S4, and S6–S8 Table S1. Gene expression datasets from EBV(+) and EBV(−) PTLD tumor biopsies used in multi-cohort analyses, related to Figure 1 Table S2. Differentially expressed genes in EBV(+) PTLD compared to EBV(−) PTLD, and EBV(+) B cell lines compared to EBV(−) B cell lines, related to Figure 1 Table S5. Pathway overrepresentation analysis of genes downregulated in CD300A KO compared to mock control EBV(+) B cell lines, related to Figure 6 Document S2. Article plus supplemental information Data Availability Statement B cell line microarray data and CD300A KO scRNA-seq data have been deposited at National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) as GEO: GSE279598 and GEO: GSE279989, respectively, within superseries GEO: GSE279991 and are publicly available as of the date of publication. This study does not report original code. All software and algorithms used in this study are publicly available. Any datasets or additional information required to reanalyze the data reported in this paper is available from the lead contact/corresponding authors upon request.