Abstract Lymphoplasmacytic lymphoma (LPL) is an incurable low-grade lymphoma with no standard therapy. Nine asymptomatic patients treated with a first-in-human, neoantigen DNA vaccine experienced no dose limiting toxicities (primary endpoint, [61]NCT01209871). All patients achieve stable disease or better, with one minor response, and median time to progression of 72+ months. Post-vaccine single-cell transcriptomics reveal dichotomous antitumor responses, with reduced tumor B-cells (tracked by unique B cell receptor) and their survival pathways, but no change in clonal plasma cells. Downregulation of human leukocyte antigen (HLA) class II molecules and paradoxical upregulation of insulin-like growth factor (IGF) by the latter suggest resistance mechanisms. Vaccine therapy activates and expands bone marrow T-cell clonotypes, and functional neoantigen-specific responses (secondary endpoint), but not co-inhibitory pathways or Treg, and reduces protumoral signaling by myeloid cells, suggesting favorable perturbation of the tumor immune microenvironment. Future strategies may require combinations of vaccines with agents targeting plasma cell subpopulations, or blockade of IGF-1 signaling or myeloid cell checkpoints. Subject terms: B-cell lymphoma, Cancer microenvironment, Immunization __________________________________________________________________ Lymphoplasmacytic lymphoma is a B-cell low-grade lymphoma with no approved standard therapy. Here the authors report a non-randomized phase 1 clinical trial performing early intervention with personalized neoantigen vaccines in asymptomatic patients and associating clinical efficacy with successful perturbation of the tumor immune microenvironment. Introduction LPL is an incurable low-grade B-cell lymphoma characterized by the presence of clonal lymphoplasmacytic cells infiltrating the bone marrow as the primary organ and a serum monoclonal protein. IgM-secreting LPL, known as Waldenström macroglobulinemia (WM), is the most common subtype^[62]1. In the absence of end organ damage, patients are considered to have smoldering phase disease. There is no approved standard therapy for smoldering LPL, and patients are generally managed by active surveillance^[63]1. Accordingly, the availability of a well-tolerated therapeutic agent that would enable early intervention to delay progression to symptomatic phase disease and other complications without inducing cross-resistance to subsequent chemotherapies, is desirable. Tumor neoantigens recognized by T cells are emerging as targets for the design of cancer vaccines, and initial clinical trials have demonstrated both safety and efficacy^[64]2–[65]4. One of the first neoantigens tested was the unique B-cell receptor generated by clonal rearrangement of Ig variable region gene sequences by lymphoma cells, rather than somatic mutation, referred to as idiotype^[66]5–[67]8. Idiotype peptides were the dominant neoantigens eluted from HLA molecules in human lymphomas^[68]9,[69]10. Therapeutic idiotype vaccines have been shown to elicit robust CD8 + T-cell immunity in humans, and one randomized, controlled clinical trial demonstrated improved disease-free survival in a minimal residual disease setting following induction chemotherapy in follicular lymphoma^[70]11–[71]15. Here we report on a first-in-human clinical trial of a DNA vaccine, encoding the autologous LPL-derived Ig single chain variable fragment (scFv) fused to human chemokine CCL20 (macrophage inflammatory protein-3, MIP-3α), which was designed to trigger T-cell immunity by targeting the delivery of the expressed fusion protein to antigen-presenting cells^[72]16,[73]17 and associate clinical efficacy in smoldering LPL patients with successful perturbation of the tumor immune microenvironment. Results Patient characteristics Nine patients were enrolled and treated in the trial, three in the 500 µg cohort and six in the 2500 µg cohort. The baseline characteristics of patients are described in Table [74]1. The median age of all patients was 67, and the majority were male (78%). Of eight patients with baseline gene mutation data available, six had MYD88 mutations; of these, one had a CXCR4 WHIM mutation. The median time from diagnosis of asymptomatic LPL to first vaccination was 2.2 years. Table 1. Baseline Patient Characteristics Characteristic (range) Cohort 1 500 µg (n = 3) Cohort 2 2500 µg (n = 6) Median Age 65 (56–71) 69 (61–78) Male Sex 100% 67% ECOG performance status 0-1 100% 100% Time from diagnosis of SWM to 1^st vaccination (yrs.) 8.1 (1.4–8.8) 2.0 (0.7–10.8) Genotype* (no. of patients) MYD88WT/CXCR4WT 1 1 MYD88 L265P/CXCR4WT 2 3 MYD88 L265P/CXCR4WHIM 0 1 Bone Marrow infiltration (%) 30 (25–40) 30 (10–50) Serum IgM (mg/dL) 2900 (814–3150) 3255 (473–7210) Monoclonal Protein (g/dL) 2.3 (0.9–2.8) 2.5 (0.4–6.3) Hemoglobin (g/dL) 13.6 (12.2–14.9) 12.3 (137–353) Platelet count (K/µL) 204 (189–372) 278 (137–353) Beta 2 microglobulin (mg/L) 2.5 (2.0–3.7) 2.5 (1.7–3.7) Albumin (g/dL) 77.7 (76.9–80) 4.1 (3.7–4.4) LDH (U/L) (normal range: 313–618)** 378 (376–405) 337 (201–613) [75]Open in a new tab ^*Genotype not available for 1 patient in Cohort 2. ^** LDH not available for 1 patient in Cohort 2. Safety, tolerability, and response assessment The primary objective was to evaluate the safety profile of the vaccine and to determine its maximum tolerated dose (MTD). All patients successfully completed planned therapy. No patients in either cohort experienced dose-limiting toxicities (DLTs) or Grade 4 adverse events (AEs). Ten months after the last vaccination, LPL-005 developed a grade 3 non-malignant pleural effusion, grade 1 pericardial effusion, and leukocytopenia, accompanied by an increase in rheumatoid factor (23.1 IU/mL [normal range 0.0–15.9]) and an ANA titer of 1:80; all findings resolved within 2 months. Grade 1-2 AEs occurring in 3 or more patients were leukopenia, nausea, myalgias, fatigue, diarrhea, anemia, hyperglycemia, and increased creatinine. Details are provided in Table [76]2. Table 2. Adverse Events Cohort 1: 500 μg dose (n = 3) Cohort 2: 2500μg dose (n = 6) Most common and notable adverse events* Grade 1-2 Grade ≥ 3 Grade 1-2 Grade ≥ 3 Hematologic Neutropenia 0 0 2 0 Anemia 2 0 5 0 Gastrointestinal Nausea 0 0 2 0 Diarrhea 0 0 3 0 General Fatigue 0 0 4 0 Myalgia 0 0 3 0 Back Pain 0 0 2 0 Respiratory Pleural Effusion 0 0 0 1** Nasal Congestion 0 0 2 0 Cough 0 0 3 0 Nervous System Dizziness 0 0 2 0 Peripheral Sensory Neuropathy 0 0 2 0 Cardiovascular Pericardial Effusion 0 0 1 0 Edema 0 0 2 0 Dermatologic Injection Site Reaction 0 0 2 0 Lab Abnormalities Creatinine increase 0 0 3 0 Hyperglycemia 3 0 3 0 [77]Open in a new tab ^* Number of patients experiencing each event. ** The event occurred outside of the DLT window and was not considered a DLT. Using response criteria from the 6th International WM Workshop Consensus Panel, LPL-003 achieved a minor response (MR). The best response for the remaining eight patients was stable disease (SD) (Fig. [78]1B). After a median follow-up period of 90 months for all patients, four patients experienced progression to symptomatic WM, requiring initiation of systemic therapy (LPL-005, -006, -007, and -009) at 29, 8, 32, and 25 months, respectively. LPL-006 experienced early disease progression and was lost to follow-up 8.8 months after the last vaccination before a post-vaccine bone marrow sample could be obtained. All remaining patients are known to be alive. Fig. 1. Single cell RNA-seq analysis reveals vaccine-associated reduction of LPL B-cell but not plasma cell-like subpopulations in the bone marrow. [79]Fig. 1 [80]Open in a new tab A Schematic of personalized chemokine-idiotype DNA vaccines and treatment schedule. B Swimmer plot illustrating clinical responses (patients are designated as LPL-001 through −009) (C) UMAP of B-lineage cell populations extracted from the total dataset (left). Total cell numbers per cluster are shown for normal B-lineage cells and, specifically, tumor B cells, based on idiotype sequences (right). D Cell frequencies by total B-lineage cluster, comparing pre- and post-vaccine bone marrow for individual patients (n = 8; mean values +/- SD). Two-sided Wilcoxon matched-pairs signed rank test was used (E) Cell frequencies of BCR clonotypes in all B-cell clusters pre- and post-vaccine. Blue = dominant tumor idiotype clone isolated for vaccine production; all other colors indicate normal B-cell clonotypes, except for patient LPL-002 (see Results). F Pooled data from all patients. Volcano plots of differentially expressed genes (adjusted p-value < 0.05) pre- vs. post-vaccine for the major B-lineage clusters containing tumor cells (top). Ingenuity Pathway Analysis of differentially expressed genes (z-score > 2, adjusted log p-value < − 1.3) of canonical pathways (bottom left) and biological processes (bottom right) contrasting tumor mature B-cell (0-2) and plasma cell-like clusters (5,10), respectively. Wilcoxon rank-sum test was used. P-values were adjusted with Bonferroni correction. Colors represent z-scaled pathway activity scores (G) Heatmap of HLA class II gene expression by clonal tumor cells pre- and post-vaccine relative to total cells, grouped by B-lineage cluster (left). Dot plots of HLA family gene expression by clonal tumor cells pre- and post-vaccine in cluster 2 only. H Dual staining immunohistochemistry of post-vaccine bone marrow slides for HLA class II and either CD79a or CD138 to identify mature B- cell and plasma cell-like LPL subpopulations. Normal human tonsils served as a positive control. All images are 200X magnification Scale bar = 50 μm (left). Flow cytometry analysis of HLA-DR expression on plasma cell-like (CD38^highCD19^low) and mature B cell (CD38^lowCD19^high) subpopulations of MWCL-1 and BCWM.1 LPL cell lines (right). Representative result of three independent experiments. Source data are provided as a Source Data file. Reduction in clonal tumor subpopulations and their gene expression pathways after vaccination in the mature B-cell, but not in the LPL plasma cell-like compartment To interrogate vaccine-induced changes directly in the tumor microenvironment (exploratory objective), bone marrow samples were obtained with a median of 3 months (range 1–13 months) after vaccine treatment from all nine patients, except patient LPL-006. We performed single-cell RNA-seq analysis on matched pre- and post-vaccine bone marrow samples, paired with matched single-cell BCR and TCR sequencing (Supplementary Fig. [81]S1B–F). To analyze specific changes in LPL cells following vaccination we separated and re-clustered heterogeneous B-lineage populations and obtained a total of 12 clusters based on differential gene expression (Fig. [82]1C and Supplementary Fig. [83]S2A). To specifically identify clonal tumor cells across various clusters, we matched single-cell BCRs with the previously identified unique tumor idiotype (Ig VH and VL CDR3) sequences used for manufacturing individualized therapeutic vaccines for each patient. LPL is known to consist of distinct clonal B-cell- and plasma cell-like subpopulations^[84]18,[85]19. Clusters 0, 1, and 2, representing mature B cells, were comprised almost entirely of the tumor clonotype, with cluster 1 being the most abundant (Fig. [86]1C, right). Plasmablast-like and mature plasma cells (clusters 5 and 10, respectively) were less abundant but also contained relatively high proportions of tumor clonotypes (Fig. [87]1C, right). Analysis of paired total B-lineage cells pre- vs. post-vaccine showed significantly reduced frequencies post-vaccine for B-cell, but not for the plasma cell-like clusters (Fig. [88]1D). This was most apparent when the analysis was restricted to cluster 1, which contained the vast majority of clonal tumor cells (p < 0.05), However, statistical significance was also retained showing reduced post- vs. pre-vaccine samples when cells from clusters 0 and 1 were grouped together (p = 0.039) and in addition, a trend towards significance was retained when grouping all three mature B-cell clusters (Supplementary Fig. [89]S2B). It is worth noting that cluster 2 exhibits some differences from clusters 0 and 1 with respect to differential gene expression patterns (Supplementary Fig. [90]S2C) and may therefore represent a distinct cell subpopulation that shares some characteristics of plasma cell-like cells, as cluster 2 B cells also exhibited downregulation of HLA class II genes post-vaccine (Fig. [91]1G). Overall, the paired analysis showed a reduction in tumor cells in B-cell compartments post-vaccine in six out of the eight patients, a substantial majority (Fig. [92]1E and Supplementary Fig. [93]S2D), and one of the patients who did not show the reduction of clonal B cells later experienced disease progression (LPL-009). Interestingly, we observed the presence of a second, minor BCR clonotype expressed by LPL-002 (Fig. [94]1E). Further analysis revealed that this clone shared Ig VH family gene usage with the vaccine-targeted tumor idiotype and also showed reduction following vaccination (Supplementary Fig. [95]S2E). Concomitant changes in global gene expression patterns in tumor cells were associated with the reduction of the tumor mature B-cell compartment post-vaccine. Differential gene expression analysis of each of the relevant B-lineage cell clusters revealed a pattern of significant gene downregulation following the vaccine in clusters 0, 1, and 2, but not plasma cell-like clusters 5 and 10 (Fig. [96]1F, top). Among the top downregulated genes were FOS, JUN, ATF3, ATF4, NFKBIA and MAP3K8 which are essential for the growth of B lymphocytes^[97]20,[98]21 as well as genes of the EIF (eukaryotic initiation factor) family (EIF4A1^[99]22, EIF4A2, EIF4A3), GADD34^[100]23, ribosomal protein L family (RPL4, RPL9, RPL13, RPL21, RPL23, PRL27, RPL37, RPL38, RPL10A) and ribosomal proteins family (RPS2, RPS6, RPS9, RPS11, RPS16, RPS20, RPS26, RPS27) which are essential for lymphoma cell proliferation and protein synthesis^[101]24 (Fig. [102]1F, top). Furthermore, pathway analysis based on differentially expressed genes identified signaling pathways significantly reduced post-vaccine known to be critical for B-cell survival, including IL-1^[103]25, IL-6^[104]26, IGF-1^[105]27, and APRIL^[106]28 (Fig. [107]1F, bottom left). BCR^[108]29, PI3K/AKT^[109]30,[110]31, and ERK/MAPK, which are involved in survival-promoting signaling by mutant MYD88 in WM cells, were also significantly downregulated^[111]32,[112]33. Conversely, PPAR signaling, which is known to promote tumor cell apoptosis^[113]34 and ferroptosis cell death pathways^[114]35 were both upregulated by these clusters. Finally, the analysis predicted the overall downregulation of biological processes (z-score > 2, adjusted log p-value < -1.3), including cell survival, viability, proliferation, protein synthesis, and RNA transcription, and upregulation of necrosis (Fig. [115]1F, bottom right). Notably, no global changes were inferred for corresponding plasma cell-like clusters (Fig. [116]1F, center and bottom). These observations suggest that tumor subpopulations of LPL within a single patient may be dichotomous in their response to therapeutic vaccine treatment, with mature B-cell subpopulations more susceptible than plasma cell-like cells. A well-described mechanism of tumor cell resistance to T-cell-mediated killing is the downregulation of expression of HLA family genes, particularly HLA class II genes^[117]36–[118]40. Specifically, B-cell tumor idiotypes are predominantly presented on HLA class II, rather than class I^[119]9. To investigate this possibility, we compared the expression of HLA family genes in tumor cells in relevant B- and plasma-cell clusters. Consistent with previous reports, we observed downregulation of HLA class II family (HLA-DMA, HLA-DMB, HLA-DPA1, HLA-DPB1, HLA-DQA2) gene expression in clusters 5 and 10 containing plasmablast-like and plasma cells in both pre- and post-vaccine samples, but not in B-cell clusters 0 and 1 (Fig. [120]1G)^[121]41,[122]42. Interestingly, like the plasma cell-like clusters, there was also a trend toward downregulation of expression of HLA class II genes post-vaccine in B-cell cluster 2 tumor cells (Fig. [123]1G). Concomitantly, there was downregulation of CD74 in plasma cell-like clusters but no change in any other genes involved in antigen processing and presentation, such as PSMB8, PSMB9, PSMB10, CTSH (cathepsin H), CTSS (cathepsin S), and CIITA. We further confirmed the downregulation of HLA class II proteins by immunohistochemistry on patient bone marrow and flow cytometry of LPL cell lines with established B- and plasma-cell like subpopulations^[124]43 (Fig. [125]1H). Specifically, the results for each patient confirmed the lack of HLA class II in CD138+ plasma cell-like LPL cells, while CD79a+ mature B-cell subpopulations uniformly expressed HLA class II. In contrast, no significant changes were observed in tumor expression of T-cell checkpoint ligands, including PDL1 (CD274), and PDL2 (PDCD1LG2) (Supplementary Fig. [126]S2F). We also observed no significant differences in the expression of genes of the death receptor family among clonal tumor B-cell or plasma cell-like clusters post-vaccine (Supplementary Fig. [127]S2F)^[128]44. Taken together, these observations suggest that plasma-cell subpopulations of clonotypic tumor cells in LPL may exhibit immune evasion to our vaccine therapy by downregulating the expression of HLA genes, rather than by activation of T-cell immune checkpoints. Paired single cell transcriptomics reveals dynamic changes in T cells in the tumor microenvironment following vaccine treatment To investigate vaccine-induced changes in normal immune cells in the bone marrow microenvironment we re-clustered T-cell populations separately and obtained a total of 12 clusters that we identified based on differential gene expression, SingleR software analysis^[129]45, and the expression of defined gene markers (Fig. [130]2A and Supplementary Fig. [131]S3B, C). These T-cell subpopulations were consistent across all patient samples (Supplementary Fig. [132]S1D). We analyzed changes in T-cell frequencies within each cluster, comparing paired pre- vs. post-vaccine bone marrow samples and observed trends toward decrease in cell frequencies of naïve CD4 T cells (cluster 0) and increases in effector memory and terminal effector T cells (clusters 1 and 3, respectively) and statistically significant decrease in MAIT T cell frequencies (cluster 10) (Fig. [133]2B). We also performed differential gene expression analysis on each T-cell cluster, followed by pathway enrichment analysis pre- vs. post-vaccine using IPA software (Qiagen). We observed significant upregulation (adjusted log p-value < − 1.3) of pathways involved in T cell activation, including TCR signaling, PI3K/AKT signaling, integrin signaling and leukocyte extravasation and down-regulation of PD-1/PD-L1 pathway following vaccination in effector T cells (Fig. [134]2C). Notably, there were no obvious changes in frequencies or signaling pathways in Treg (Fig. [135]2B cluster 8, and Fig. [136]2C). Fig. 2. Paired single cell TCR-seq reveals T-cell clonal expansion and activation in the tumor microenvironment following vaccine treatment. [137]Fig. 2 [138]Open in a new tab A UMAP of T-cell subpopulations (left) highlighting cells with TCR sequences detected (right). B T-cell frequencies by cluster, comparing values for individual patients in pre- and post-vaccine (n = 8; mean values +/− SD). A two-sided Wilcoxon matched-pairs signed rank test was used. C Heatmap of selected differentially regulated immune signaling pathways post vs pre-vaccine for each T cell subpopulation. Pooled data from all patients. Colors represent z-scaled pathway activity scores (adjusted log10 p-value < − 1.3). Wilcoxon rank sum test was used and p values were adjusted with Bonferroni correction. D TCR clonotype frequencies pre-vaccine (y-axis) and post-vaccine (x-axis) for each patient. Thresholds of detection are indicated with light gray lines along each axis. Pearson R^2 is shown. E Frequencies of the 20 most prevalent clonotypes within the post-vaccination repertoire, compared with their matching clonotype pre-vaccine. Asterisks indicate newly emerging clones in post-vaccine samples not detectable pre vaccination. A two-sided Wilcoxon matched-pairs signed rank test was used. F Shannon entropy of TCR clonotype repertoires in paired samples pre. vs post-vaccine for each patient (n = 8). Two-sided Wilcoxon matched-pairs signed rank test was used (G) UMAPs of T-cell subpopulations in pre- and post-vaccine samples with highlighted cells in the 20 clonotypes as in (E) in CD8A^+ (left) and CD4^+ (right) populations. H Violin plots of gene expression of selected markers by the top 20 T-cell clonotypes, pre- vs. post-vaccine. A two-sided Wilcoxon rank sum test was used. I Functional tumor idiotype-specific T-cell responses post-vaccination. Bone marrow samples from each patient were stimulated in triplicate (n = 3) with autologous immortalized B cells transfected with either patient-specific tumor idiotype or no antigen as a negative control. Each dot in the volcano plot represents one of 30 individual cytokines assayed. Red dots indicate a significance threshold < 0.05 of two-sided Student t-test and fold change > |1.3|. The one-sided Chi-square test was used to analyze differences between idiotype stimulation and no antigen control. Source data are provided as a Source Data file. n.s. not significant. To analyze the clonal composition of T cells in the microenvironment, we used matched single-cell TCR-seq data (Fig. [139]2A, right panel). A mean of 750 unique T-cell clonotypes (range 141–1596) were identified pre- and post-vaccine for each patient. Comparing post- vs. pre-vaccine samples, we observed expansion of existing clonotypes in all patients except for two of the clinical progressors, LPL-005 and -009 (Fig. [140]2D). Furthermore, among the 20 most prevalent clonotypes detected post-vaccination, the majority increased from low-frequency clonotypes that were present before vaccination, except for clinical progressors LPL-005 and -009 (Fig. [141]2E). New clonotypes were also detected post-vaccination (overall 4.4%), consistent with increased T-cell clonal diversity. It is tempting to speculate that newly emerging clones could represent epitope spreading post-vaccination. Recently, epitope spreading was conclusively verified in human melanoma patients who had received neoantigen vaccines^[142]46. Increasing clonal diversity post-vaccination was also suggested in most patients, as analyzed by individual Shannon entropy scores^[143]47 (Fig. [144]2F). Phenotypically, unique or shared T-cell clonotypes expanded in post-vaccine samples localized primarily to clusters enriched for effector memory or effector T cells (Supplementary Fig. [145]S3C). This same pattern was observed for the 20 most abundant post-vaccine clonotypes (Fig. [146]2G), with cells localized to effector memory T cells and terminal effector T cell clusters. Phenotypically, compared with control total T cells (Supplementary Fig. [147]S3D), these top 20 post-vaccine clonotypes were primarily effector T cells expressing CD4 or CD8, and markers of activation, differentiation, or proliferation, including CD27, CXCR4^[148]48, HLA-DR^[149]49, PIK3RI^[150]50, REL^[151]51, and FKBP1A^[152]52 (Fig. [153]2H). In contrast, clonotypes detected only in a single T-cell were distributed broadly across all T-cell subpopulations, including naïve CD4 and CD8 T cells, regulatory T cells, Th1/Th2 cells, Th17, and to a lesser extent central memory T cells and effector memory phenotypes (Supplementary Fig. [154]S3C). Notably, the top 20 post-vaccine clonotypes showed a mixed pattern of up- and down-regulation of co-inhibitory molecules genes DUSP2 and TIGIT, respectively, with most, including PDCD1 (PD-1), LAG3 and HAVCR2 (TIM3) showing no significant change post-vaccination (Fig. [155]2H). Taken together, these results suggest that vaccine therapy induced significant expansion and activation of terminal effector and effector memory T cells within the top 20 TCR clonotypes post-vaccine, with little activation of immune checkpoints. Tumor idiotype-specific T-cell immune responses To detect idiotype-specific T cell responses elicited by the vaccine treatment (secondary objective), we analyzed T cells isolated directly from the bone marrow tumor microenvironment. T cells were enriched from each patient’s post-vaccine sample by negative selection and then stimulated with autologous immortalized normal B cells (as antigen-presenting cells, APCs) transfected with either Ig VH and VL sequences (expressed as scFv’s) derived from the respective patient-specific tumor idiotype (used previously for therapeutic vaccine production), or no antigen as a negative control, described previously^[156]53. Multiplex cytokine analysis was performed on culture supernatants. Representative post-vaccination samples are shown on volcano plots for a panel of 30 cytokines for all patients (Fig. [157]2I). The cytokine profiles of 7 of the 8 patients analyzed (87.5% positive immune response) showed a clear preponderance of cytokines secreted in response to idiotype, compared with negative control. Thus, all patients, except LPL-007, reached or trended toward significance (Fig. [158]2I). Although each patient exhibited a unique profile of cytokines, multiple cytokines were shared across patients and included a mixture of both CD4 and CD8 responses. For example, Rantes, IP10, and MIP1α and β are secreted mainly by CD8 + T cells, while IL-2 and IL-13 are secreted mainly by CD4 + T cells, and GM-CSF, IFNγ, and IL-8 can be produced by both^[159]54–[160]65. Taken together these functional data are consistent with T-cell clonal expansion post-vaccination detected by transcriptomic analysis above. Vaccine-induced reduction in cross-talk between immune cell types and tumor cells in the microenvironment To infer and analyze global changes in cell-cell communications in the tumor microenvironment after vaccination, we employed comparative CellChat^[161]66 software to analyze signaling interactions among all major cell types in pre- and post-vaccination bone marrow samples. A cell-cell interaction map was constructed using aggregate sc-RNAseq data from all evaluable patients with five major interaction populations: clonal LPL mature B cells, clonal LPL plasma cell-like cells, T/NK cells, myeloid cells, and normal B progenitor cells as controls. From this cell-cell interaction map, the total number of ligand-receptor pairs contributing to communication between any two interacting cell types was analyzed. We observed that the total number of inferred interactions between the five major cell types in the tumor microenvironment significantly decreased post- compared with pre-vaccine (Fig. [162]3A), with this same pattern consistently observed between individual pairs of cell types (Supplementary Fig. [163]S4A–C). Fig. 3. DNA vaccine significantly alters cell-cell communication networks in the tumor microenvironment. [164]Fig. 3 [165]Open in a new tab A Total numbers of ligand-receptor pair interactions in bone marrow samples pooled from all patients. A two-sided Wilcoxon matched-pairs signed rank test was used to compare pathways pre- and post-vaccine (n = 25 pathways). B Pooled outgoing and incoming interaction strengths between the following cell types in 2D space pre- and post-treatment for all patients: LPL (mature B-lymphoid), LPL (plasma-like), myeloid, T- and NK, and normal B progenitor. Dot size indicates the number of expressed ligand-receptor pairs. Interaction strengths were calculated with CellChat software. C Pooled relative information flows between pairwise pre- and post-vaccine datasets for all signaling pathways, sorted by increasing information flow post-treatment. D Heatmap of relative strengths of all signaling pathways pre- and post-vaccine by cell type. Outgoing and incoming signaling patterns from data pooled from all patients are shown. A two-sided Wilcoxon matched-pairs signed rank test was used to compute p-values comparing signaling patterns within each cell type. E Circle plots of selected signaling pathways and relative contributions of ligand-receptor pairs. Cell types are color-coded, dot size is proportional to the number of expressed ligand-receptor pairs, edge color indicates the source of outgoing signal, and edge weight is proportional to interaction strength. Source data are provided as a Source Data file. To investigate which cell populations contributed to the reduction in inferred interactions, we used network centrality analysis to compare incoming and outgoing interaction strengths (Fig. [166]3B). Interestingly, predicted interaction strengths for myeloid and LPL mature B-cell, but not LPL plasma cell-like populations, were most dramatically reduced post-vaccine. We then analyzed the overall information flow for multiple specific signaling pathways across the pre and post-vaccine datasets^[167]67. Multiple signaling pathways were implicated as active predominantly in pre- but not post-vaccine samples, including pathways such as APRIL^[168]68, which is known to promote B- or plasma cell survival, and others with known roles supporting tumor cell proliferation in solid cancers, such as RESISTIN^[169]69,[170]70, VEGF^[171]71, and IL-10^[172]72, TGFβ and BMP^[173]73 (Fig. [174]3C). Moreover, the IL-6 signaling pathway, which promotes IgM secretion and LPL and plasma cell growth via the JAK/STAT pathway^[175]26 was substantially reduced in post-vaccine samples. The analysis of individual cell types revealed that myeloid cells mainly contributed to the downregulation of the information flow of these signaling pathways (Fig. [176]3D and Supplementary Fig. [177]S4D–O). For example, we observed dramatic reductions in predicted outgoing signals provided by myeloid cells for both the RESISTIN and APRIL pathways, as well as IL-6, associated with their respective ligand-receptor pairs (Fig. [178]3E). Paradoxically, dichotomous upregulation of the insulin-like growth factor (IGF) signaling axis post-vaccine was inferred by plasma cell, but not mature B cell LPL subpopulations, including both autocrine and paracrine pathways, consistent with a potential mechanism of escape by the former (Fig. [179]3D, E). Our scRNAseq data confirmed increased expression of IGF1 among clonal tumor cells post-vaccine in both plasma cell-like clusters (clusters 5 and 10) but not in any B-cell clusters (0, 1, and 2). We also observed an increased proportion of clonal tumor cells expressing IGF1 in cluster 10 (Supplementary Fig. [180]S4P). Vaccine-induced changes in myeloid cell subpopulations in the tumor microenvironment Given that vaccination was associated with significantly reduced cell-cell communication patterns in the tumor microenvironment, most pronounced in outgoing signals provided by myeloid cells to clonal LPL cells, we further analyzed subpopulations of myeloid cells by re-clustering them based on differential gene expression analysis from the combined datasets of pre- and post-vaccine bone marrow cells from all patients. We obtained a total of 9 clusters, based on the differential expression of established marker genes (Fig. [181]4A, B and Supplementary Fig. [182]S5A). Myeloid cell populations were consistent across all individual patient samples (Supplementary Fig. [183]S1D). Fig. 4. Reduced signaling by myeloid cell subpopulations post-vaccination. [184]Fig. 4 [185]Open in a new tab A UMAP of myeloid cells extracted from the total dataset. 9 clusters were generated, and cluster identities were assigned based on differentially expressed gene markers. B Heatmap of expression of selected gene markers for each cluster. C Cell frequencies by myeloid cluster, comparing aggregate values for individual patients (n = 8) in pre- and post-vaccine samples. Data are presented as mean values +/− SD. A two-sided Wilcoxon matched-pairs signed rank test was used. D Dot plots of differentially expressed genes pre- and post-vaccine in selected pathways identified in Fig. [186]3 by cluster. Pooled data from all patients. Significantly downregulated or upregulated genes in post-vaccine samples are highlighted by red and violet rectangles, respectively. A two-sided Wilcoxon rank sum was used. Color intensity and dot size correspond to expression level and relative proportion of positive cells, respectively. E Violin plots of SIRPA and CD47 gene expression pre- and post-vaccination by cell cluster in myeloid and B-cell tumor populations, respectively. Pooled data from all patients. A two-sided Wilcoxon rank sum was used. Source data are provided as a Source Data file. We analyzed changes in cell frequencies per cluster in paired pre- vs. post-vaccine patient samples and observed significant increases in frequencies of CD14^-CD16^+ non-classical monocytes (cluster 3) and concomitant trend toward decreases in the frequencies of CD14^+CD16^+ intermediate monocytes (cluster 4, Fig. [187]4C). Given that monocyte differentiation is believed to proceed from classical CD14^+CD16^- to non-classical CD14^-CD16^+ monocytes via intermediate CD14^+CD16^+ monocytes^[188]74, these results may suggest that monocytes in the tumor microenvironment post vaccination undergo increased differentiation from intermediate to non-classical subpopulations, thereby causing a general skewing away from classical monocytes. This hypothesis was also supported by the trend towards decreasing CD14^+CD16^- classical monocyte frequencies (cluster 0) observed post- vs. pre-vaccination, although these differences did not reach statistical significance. Next, we sought to identify the specific myeloid cell clusters that contributed to the dramatic reductions in outgoing predicted signals provided by myeloid cells to LPL cells by analyzing each of the individual signaling pathways in Fig. [189]3D. Comparing post- vs. pre-vaccine samples, most of the signaling pathways affected were associated with reductions in monocytic subpopulations, particularly cluster 3 non-classical monocytes, and to a lesser extent cluster 0 classical monocytes (Fig. [190]4D). Changes were also observed across many of the pathways for cluster 1 of mature neutrophils, but these were generally of lesser magnitudes. Taken together, these data suggest that vaccination was associated with clear reductions in pro-tumoral outgoing signals provided by non-classical monocytes to LPL cells, but with a paradoxical expansion of this myeloid subpopulation. Finally, because of the availability of potential therapeutic intervention, we also performed cell-cell communication analysis of the CD47-SIRPα pathway which predicted overall decreased signaling after vaccination (Fig. [191]3E), despite increased CD47 expression on at least one lymphoid (B-cell cluster 2) and one plasmacytoid (B-cell cluster 10) LPL tumor subpopulation. SIRPA was observed on cluster 0 classical monocytes pre-vaccination, with no significant change in expression post-vaccination (Fig. [192]4E). Discussion In the absence of any standard treatment for patients with smoldering phase LPL/WM, patients are typically managed by active surveillance alone. The median time to progression to the symptomatic phase is 3.9 years^[193]75. Early intervention with a well-tolerated therapeutic agent, such as a vaccine, that could delay progression to symptomatic phase disease, would therefore be highly desirable. The individualized therapeutic DNA vaccines used in this trial appear to be safe, with no patients experiencing dose-limiting toxicities (DLTs) or Grade 4 adverse events (AEs). Most of the patients experienced potential clinical benefit, with documentation of stable disease or better for a median of 6 years, including one patient who achieved a minor response (MR). The reasons for the lack of more robust objective clinical responses were revealed by direct interrogation of the tumor microenvironment by single-cell transcriptome analysis. Comparing paired pre- and post- vaccine bone marrow samples available from eight of the nine patients, we observed a striking dichotomous pattern of significantly reduced numbers of clonal tumor idiotype-expressing B-cells post-vaccine in the majority of patients, but no change in clonally related plasma cell-like clusters of any patient (Fig. [194]1D). Heterogeneity within LPL, consisting of separate mature B-cell and plasma cell-like subpopulations, has been described^[195]76,[196]77. In our dataset, plasma cell-like cells were detected at lower frequencies, compared with mature B cell-like subpopulations, but partial loss of this fragile cell subpopulation during frozen sample preparation cannot be ruled out^[197]40. Furthermore, there was an associated pattern of downregulation of expression of genes known to be essential for lymphoma cell proliferation and protein synthesis in tumor clusters of mature B cells, but not of plasma cell-like LPL subpopulations (Fig. [198]1F). Pathway analysis predicted global downregulation of signaling pathways known to be critical for B-cell survival and conversely, upregulation of pathways known to promote tumor cell apoptosis and other forms of cell death, in mature B-cell, but not plasma cell-like LPL clusters, consistent with a vaccine-induced antitumor response against the mature B-cell LPL compartment. One recently recognized mechanism of tumor cell death, ferroptosis, is known to be directly mediated by activated CD8 + T cells (Fig. [199]1F, heat map). The observation that most of the top 20 T-cell clonotypes identified after vaccination belong to the effector CD8 T-cell cluster is consistent with this possible mechanism. In addition to direct antitumor effects, immune-mediated tumor suppression may occur indirectly; for example, cytokines secreted by T cells can permanently arrest tumor cell proliferation and induce cell cycle arrest^[200]78. To this point, the most dramatic post-vaccine changes we observed in clusters 0, 1, and 2 were reduced expression of molecules involved in protein synthesis (Fig. [201]1F volcano plots). Downregulation of expression of HLA molecules, particularly class II, has been recognized as one immune-evasion mechanism in cancer^[202]39,[203]40. Indeed, we observed that plasma and plasmablast-like LPL cells expressed low levels of HLA class II genes and protein, similar to their normal counterparts. This represents a potentially intrinsic mechanism of resistance by plasma cell-like subpopulations, as low levels were observed in both pre- and post-vaccine samples (Fig. [204]1G). In contrast, the trend towards downregulation of HLA class II gene expression observed in LPL cluster 2 cells post-vaccine, compared with pre-vaccine, suggests a potential mechanism of immune evasion among mature B-cell subpopulations (Fig. [205]1G, H). In contrast, dichotomous upregulation of the IGF signaling axis post-vaccine was inferred by plasma cell, but not mature B cell LPL subpopulations (Fig. [206]3D, E), suggesting another possible mechanism of acquired resistance to vaccine therapy. The IGF axis has been implicated in acquired drug resistance in various hematologic cancers, and selective IGF-1 receptor inhibitors could block tumor cell proliferation and migration and overcome resistance to treatment of multiple myeloma, and lymphomas with bortezomib, EZH2 inhibitors and crizotinib^[207]79–[208]81. Furthermore, our finding that the proportion of tumor cells expressing IGF-1 was also increased in one of two plasma cell-like clusters post-vaccine suggests that clonal selection of IGF signaling-dependent tumor clones cannot be excluded (Supplementary Fig. [209]S4P). The chemokine-antigen fusion vaccine platform was designed to elicit robust T-cell immunity by targeting idiotype antigen delivery to chemokine receptors on antigen-presenting cells^[210]10,[211]11. Previous preclinical mechanistic studies demonstrated the requirement for co-expressing the fusion, rather than chemokine or antigen alone^[212]16,[213]82. Indeed, we observed that vaccine therapy induced dynamic changes in T cells in the tumor microenvironment, consistent with the generation of antigen-specific immune responses. Trends toward increases in effector memory and terminal effector phenotypes post-vaccine (Fig. [214]2B, H) were associated with upregulation of pathways involved in T-cell activation (Fig. [215]2C), expansion of individual T-cell clonotypes (Fig. [216]2D, E), increased T-cell clonal diversity (Fig. [217]2F), and functional LPL idiotype-specific cytokine production (Fig. [218]2I). The effector T-cell populations were mostly enriched for CD8 T cells but also contained a smaller but potentially relevant effector CD4 T cell population (Fig. [219]2G), as idiotype peptides have been reported to be presented on HLA class II molecules^[220]9. Moreover, the expression levels of the CD4 marker significantly increased post-vaccination (Fig. [221]2H), and the cytokine secretion results following idiotype stimulation were consistent with profiles characteristic of both CD4 and CD8 T cells. Finally, we observed the emergence of several new clonotypes post-vaccination, and it is tempting to speculate that these new clonotypes could represent epitope spreading; i.e., CD8 effector T cells recognizing neoantigens. Recent studies have suggested that Treg cells create an immunosuppressive milieu in WM, the most common subtype of LPL^[222]83. However, we observed no obvious changes in frequencies or signaling pathways in Treg that would suggest an effect of vaccination on this subpopulation (Fig. [223]2B cluster 8, and Fig. [224]2C). T cells in the microenvironment also showed a mixed pattern of up- and down-regulation of co-inhibitory pathways (Fig. [225]2C), with most showing no significant change post-vaccination (Fig. [226]2H). Taken together, these results suggest little activation of co-inhibitory molecules by vaccine therapy. Our vaccine also globally altered the levels of the cell-cell communication networks and signaling strength across various other cell populations in the tumor microenvironment. We detected significant downregulation of signaling pathways post-vaccine that likely directly promote the growth of LPL cells, such as APRIL and IL-6 which are known to promote B- or plasma cell survival (Fig. [227]3C–E). IL-6 is produced by both stromal and tumor cells in the LPL microenvironment. Our data suggest a possible mechanism of immune-mediated reduction of autocrine IL-6 production by LPL tumor cells (Fig. [228]3E; and^[229]84), which in turn downregulates the production of APRIL by myeloid cells^[230]85. Other pathways were reduced post-vaccine, such as RESISTIN, which has a known role in supporting the proliferation of solid cancers by binding CAP1 receptors. A role for RESISTIN in supporting LPL has not been previously inferred, but it induced multidrug resistance in human multiple myeloma^[231]69. The bioinformatic analysis also identified a predominant role for myeloid cells in the tumor microenvironment as a source of vaccine-induced, downregulated pro-tumoral signaling to LPL cells. The global signaling pathways affected were primarily associated with monocytic, rather than granulocytic or dendritic cell subpopulations, particularly non-classical CD14^-CD16^+ monocytes, and to a lesser extent, classical CD14^+CD16^- monocytes (Fig. [232]4D). Myeloid cells have been recognized as key component of the immune suppressive microenvironment in solid tumors and this observation has been extended more recently to B-cell tumor microenvironments^[233]86. The trend we observed towards the reduction of classical monocytes may be of particular relevance, as this subpopulation has been recently shown to be pro-tumoral in multiple myeloma^[234]87. In addition, our gene signature analysis revealed that this cluster may contain myeloid-derived suppressor cells (MDSC) which have also been extensively characterized as immune-suppressive (Supplementary Fig. [235]S5B). Finally, reports indicate that an increased pro-inflammatory myeloid signature is an early step in the development of WM and in monoclonal gammopathy of undetermined significance (MGUS)^[236]88. One potential strategy to further overcome resistance to myeloid signaling may be the therapeutic blockade of SIRPα-CD47, an emerging checkpoint utilized by cancer cells to evade immune responses. Despite increased expression of CD47 (“do not eat me” signal) on at least one mature B-cell- (cluster 2) and one plasma cell-like (cluster 10) LPL tumor subpopulation after vaccination, CellChat analysis predicted overall decreased signaling to SIRPα on myeloid cells which expression was confirmed on classical monocytes (Fig. [237]4E). Taken together, these results suggest that vaccine therapy was associated with significant reduction of pro-tumoral signaling by myeloid cells in the LPL microenvironment. Recent improvements in bioinformatics, design, and manufacturing are facilitating the clinical development of individualized neoantigen cancer vaccines. As a prototype, our idiotype neoantigen vaccine demonstrated safety, the ability to significantly reduce clonal mature B-cell, but not plasma cell-like, LPL subpopulations and to favorably perturb the tumor microenvironment. Future functional studies of the pathways affected are needed to confirm the mechanisms of resistance elucidated and to design combination strategies to circumvent them. Such strategies could include adding IFNγ or epigenetic drugs, designed to increase HLA molecule expression on plasma cell-like LPL subpopulations^[238]89 and combining neoantigen vaccines with agents that specifically target plasma cells, such as anti CD38 antibodies^[239]90, or pathways known to promote their growth, such as IGF-1 receptor inhibitors^[240]91. Furthermore, our data suggest that combinations of these vaccines with myeloid cell checkpoint blockade may be worthwhile. Finally, although little activation of co-inhibitory molecules was observed by vaccine therapy, elevated PD-1 ligands on human WM cells and exhausted CD8 T cells in the WM microenvironment have been reported by others^[241]92, suggesting that there may still be a role for therapeutic T-cell checkpoint blockade combined with this vaccine. Methods Experimental design This was an open-label phase I trial conducted at the University of Texas M.D. Anderson Cancer Center ([242]NCT01209871). Patients were enrolled between March 2015 and August 2017. Patients received a series of three DNA vaccinations with autologous MIP-3α fused lymphoma idiotype at 4-week intervals intradermally into both thighs by needle-free injection device (PharmaJet, Golden, CO). Consecutive patients were enrolled in dose cohorts 1 (500 μg) and 2 (2500 μg) according to a standard 3 + 3 statistical design. The trial conformed with the CONSORT principles. All patients were enrolled, and data was collected at M.D. Anderson Cancer Center. The study protocol is available as a supplementary file. The primary objective was to evaluate the safety profile of the vaccine and to determine its maximum tolerated dose (MTD). Clinical laboratory testing, patient reporting, and physical examination findings were used to evaluate adverse events (AEs), including serious adverse events (SAEs). Toxicities were graded according to the NCI Common Toxicity Criteria v4.0. Dose-limiting toxicity (DLT) was defined as a ≥ grade 2 allergic reaction, ≥ grade 2 autoimmune reaction, and any grade 3 or 4 toxicity except for fever, grade 4 fever, which subsequently required a 50% dose reduction. MTD was defined as the highest dose level in which 6 patients have been treated with less than 2 instances of DLT. Initial disease response was assessed one month after the final vaccination according to International WM consensus panel response criteria from the 6th International Workshop^[243]93. Neither sex nor gender was considered in the study design. Given the small sample size and predominance of male participants, the study was not sufficiently powered to assess outcomes according to sex. A secondary objective was to assess the immunogenicity of the vaccine to generate tumor-specific responses. The exploratory objective was to characterize the tumor microenvironment of LPL bone marrow specimens pre- and post-vaccine and was outlined in the section “Correlative studies” of the study protocol. Patients Eligible patients had a diagnosis of smoldering lymphoplasmacytic lymphoma (LPL) confirmed by tissue diagnosis, with a monoclonal heavy and light chain as determined by flow cytometry. All participants were required to be able to provide informed consent. Patients were excluded if they had a history of autoimmune diseases except for Hashimoto’s thyroiditis, or either a positive antinuclear antibody titer or anti-double strand DNA titer. No form of compensation was provided to participants. Participant sex was determined based on self-report. The conduct of this trial was approved by the University of Texas M.D. Anderson Cancer Center institutional review board (protocol 2009-0465) and was carried out in accordance with the Declaration of Helsinki and the International Conference on Harmonization Guidelines for Good Clinical Practice. Bone marrow aspirates and cryopreservation All patients had up to 10 ml of bone marrow aspirated before treatment for morphological sorting, immunophenotyping, and characterization. A second 15 ml bone marrow aspirate sample was obtained from the contralateral side for additional tumor cell banking for vaccine production and/or correlative analysis. Bone marrow mononuclear cells (BMMNCs) were isolated by density gradient centrifugation using Ficoll. Mononuclear cells were washed three times with 45 mL PBS (800 × g), counted, and viably cryopreserved in 10% DMSO. Generation of individualized DNA vaccines LPL B cells generally comprise > 30% of the total B-lymphocyte population in bone marrow^[244]94. The unique lymphoma idiotype for each patient’s tumor was identified based on the clonal amplification of a predominant Ig heavy and light chain V(D)J sequence^[245]12. MIP-3α fused lymphoma idiotype plasmid DNA vaccines were prepared from each patient’s bone marrow as described previously^[246]95. The plasmid DNA was then amplified and purified from E. coli according to Good Manufacturing Practices (GMP) standards by FUJIFILM Diosynth Biotechnologies U.S.A., Inc., (College Station, TX). Bone marrow sample processing for single cell RNA sequencing Cryopreserved bone marrow samples pre and post-vaccine were processed together for each patient. Cells were thawed at 37 °C and resuspended in culture media. Dead cell removal was performed using the StemCell EasySep Annexin V kit (Cat#17899). Cells were resuspended in PBS with 0.04% BSA, and count and viability was determined using an automated cell counter (BioRad) prior to loading onto 10x Genomics Chip (Chromium Single Cell 5’ Kit). LPL-008 pre and post-samples after thawing, as described above, were used for staining with hashtag antibodies (Biolegend TotalSeq-C0251, Cat #394661, clone LNH94 and TotalSeq-C0252, Cat #393663, clone LNH94) according to manufacturer’s protocol. Pre and post-vaccine samples were mixed in a 1:1 ratio and loaded onto the 10x Genomics chip. Libraries were prepared using a Chromium Single Cell 5’ Kit (10x Genomics) for gene expression, TCR, and BCR, and QC was performed using an Agilent High Sensitivity DNA kit on the Agilent 2100 bioanalyzer. Libraries were sequenced on Illumina HiSeq 2500 (Read1: 26 cycles, i7 index: 8 cycles, Read2: 101 cycles, Sequencing depth: 20,000/ read pairs per cycle for gene expression, Read1: 151 cycles, i7 Index: 8 cycles, Read2: 151 cycles, sequencing depth: 5000/ read pairs per cell for TCR/BCR). Raw sequencing reads were processed using the CellRanger pipeline using default settings (10x Genomics, software version 3.1.0). Single cell RNA-seq data analysis Filtered gene expression matrices generated by the CellRanger pipeline (10x Genomics version 3.1.0) were used for downstream analysis using the Seurat package (version 3)^[247]96,[248]97. Loupe Browser was used to resolve Hashtag identities of cellular barcodes (10x Genomics version 3). Cells were filtered based on total mRNA counts, total genes detected, and mitochondrial content. We obtained an average of 2163 cells per sample (range 830–4680), totaling 36,777 cells in the dataset. Datasets were normalized and 2000 most variable genes were selected. Datasets were merged sequentially using the IntegrateData function in Seurat. To improve clustering resolution, a dataset “A single cell immune cell atlas of the human hematopoietic system” from the Human Cell Atlas portal ([249]https://explore.data.humancellatlas.org/projects/cc95ff89-2e68-4a 08-a234-480eca21ce79) was downloaded on April 24^th, 2020. The Dataset file was a subset to obtain samples of normal healthy bone marrow of the 15 oldest individuals available based on metadata files, subject to identical filtering and processing, followed by sequential data integration as above. Feature scaling, PCA, clustering, and UMAP analysis were performed on merged datasets using the integrated assay. Identification of markers of cell populations was done with the FindAllMarkers function. A total of 25 clusters were identified (Supplementary Fig. [250]S1B–F) which were consistent across all patient samples (Supplementary Fig. [251]S1D). For single-cell BCR and single-cell TCR sequencing, filtered contig annotation CellRanger output files were processed with custom R script and were added to the Seurat object as metadata based on cell barcodes (Supplementary Fig. [252]S1E). For further sub-analysis, B cell (excluding CD3^+ cells), T cell (excluding CD4^-CD8^- cells), and myeloid cell populations were separated out using subset function and re-analyzed with PCA, clustering, and UMAP as described above. Differential gene expression pre vs post vaccine was performed using the FindMarkers function with default parameters. Volcano plots were created using EnhancedVolcano package^[253]98. Pathway analysis and biological processes analysis was performed using IPA software (Qiagen) (adjusted log10 p-value < − 1.3) and visualized using the R pheatmap package. Heatmaps, dot plots and violin plots were generated with Seurat. TCR clonotype analysis was performed using immunarch R package according to instructions^[254]99. Shannon index was calculated on TCR clonotypes using the abdiv package ([255]https://github.com/kylebittinger/abdiv). Identification of cluster identities was guided by analysis using SingleR package^[256]45. Coefficients of determination (R^2) of Pearson correlation between clonotypes pre and post-vaccination were calculated with GraphPad Prism (version 10) and a cutoff value of 0.6 was used^[257]100. Flow cytometry analysis MWCL-1 cell line was a kind gift of Asnell lab, Mayo Clinic^[258]101. BCWM.1 cell line was a kind gift of Treon Lab, Dana Farber Cancer Institute^[259]32. Both cell lines were cultured in RPMI Medium 1640 (Gibco Cat# 72400-047) supplemented with 10% fetal bovine serum (Omega Scientific, Cat# FB-12). Staining for flow cytometry was performed as described previously^[260]53. Briefly, cells were washed in Stain Buffer (BD Biosciences Cat# 554657), blocked with Fc blocking reagent (Human TruStain FcX, Biolegend Cat# 422302), followed by staining with fluorophore conjugated antibodies anti-HLA-DR AlexaFluor 647 (BD Biosciences, Cat# 563591, clone Tu39, lot 8087988, 5ul per sample), anti-CD38 PE-Cy7 (BD Biosciences, Cat# 560677, clone HIT2, lot 2129025, 5ul per sample), anti-CD19 V450 (BD Biosciences, Cat# 560354, clone HIB19, lot 4015718, 5ul per sample). Cells were analyzed on BD LSRFortessa. Data analysis and visualization were performed in FlowJo version 10.8.1. Immunohistochemistry analysis For dual staining immunohistochemistry for HLA II/CD79a and HLA II/CD138, each antigen-antibody interaction was distinctly visualized through color-specific detection. For both combinations, HLA II antigens were identified using clone CR3/43 (Dako Cat # M0775) at a dilution of 1/800. For HLA II/CD79a staining, CD79a was detected using clone D1X5C (Cell Signaling Technology Cat # 13333) at a dilution of 1/100, with CD79a staining presented in purple and green for HLA II staining. For HLA II/CD138 staining, CD138 was detected using clone B-A38 (Ventana Cat # 760-4248) in a ready-to-use format, visualized in green in contrast to the purple HLA II staining. These stains were run on the Ventana Discovery platform. Bone marrow T cell stimulation and cytokine 30-Plex human panel Cells were thawed at 37 °C and washed with ImmunoCult-XF T cell media (StemCell Cat#10981) supplemented with 1 U/mL of DNAse (Thermo Sci Cat# EN0521). Cells were centrifuged and resuspended in EasySep buffer (StemCell Cat# 20144), followed by exposure to CD20 magnetic microbeads (Miltenyi Biotec Cat#130-0910104). T cells were enriched by negative selection (EasySep Cat#10981) and resuspended in T cell media supplemented with IL-2 at 50 U/mL (clinical grade), IL-7 at 25 ng/mL (Miltenyi Biotec, Cat# 130-095-361) and IL-15 at 25 ng/mL (Miltenyi Biotec, Cat# 130-095-762) to a concentration of 2 × 10^6 cells/mL. 125 µl of cell suspension were plated in V-bottom 96-well plates. Autologous patient-derived immortalized B cells were generated and passaged as described previously^[261]53,[262]102. Cells were suspended in PBS, irradiated with 1500 rads for 5 min, spun down, and resuspended in Neon Buffer R (Invitrogen). mRNA encoding the patient’s unique Idiotype, scFv-MITD were prepared as described previously^[263]53, and cells were electroporated with or without mRNA using Neon MPK5000 system (Invitrogen) with settings: pulse voltage: 1150 V, pulse width: 30 ms, pulse number: 2. Electroporated APC were resuspended in T cell media supplemented with IL-2, IL-7 and IL-15 to a final concentration of 2 × 10^6 cells/mL and 125 µl of cells were added to V-bottom 96-well plate containing T-cell suspensions. Cells were co-cultured in a 37 °C incubator for a total of 3 days after which cells were spun down and supernatants were snap-frozen at -80 °C. Supernatants were used for 30-Plex human panel analysis according to the manufacturer’s protocol, and the data are presented as volcano plots, with red dots indicating significance threshold < 0.05 and fold change > |1.3|. A chi-square test was performed to analyze differences between idiotype stimulation and no antigen control using GraphPad Prism (version 10). Cell–cell communication analysis The CellChat package was used to infer cell–cell communications between the following cell types via interaction-network analysis: LPL (mature B-lymphoid), LPL (plasma cell-like), myeloid, T- and NK, and normal B progenitor cells. A Seurat object was used as input for CellChat following standard protocols ([264]https://github.com/sqjin/CellChat. NetAnalysis_signalingRole_heatmap) and was used to compute the comparison of overall signaling pathways in pre- vs. post-vaccine samples. Circle plots were generated via netVisual_aggregate, vertex.size = groupSize, respectively. In circle plots, the edge color indicates the source of the outgoing signal, and the edge weight is proportional to interaction strength. Incoming and outgoing strength were calculated via the CellChat function with default parameters. The overall information flow for a given signaling pathway was defined as the sum of the communication probability among all ligand-receptor pairs of cell groups in the inferred cell-cell communication network and was generated via CellChat rankNet. Statistical comparisons of relative interaction strengths were performed by the Wilcoxon signed-rank test. Statistical analyses Analysis of cell frequencies pre and post-vaccination was performed using a two-sided Wilcoxon matched-pairs signed rank test, comparing paired values for individual patients pre- and post-vaccine (n = 8 pairs of independent biological samples). Data are presented as mean values +/− SD. Differentially expressed genes were identified using the Wilcoxon rank sum test, and p values were adjusted with Bonferroni correction. Ingenuity Pathway Analysis (Qiagen IPA) was performed based on differentially expressed genes (z-score > 2, adjusted log p-value < − 1.3). Multi-IHC stain was performed once and with parallel controls showing appropriate staining due to limited availability of patients’ bone marrow slides. Flow cytometry analysis is a representative result of n = 3 independent biological experiments. Pearson coefficients of determination (R^2) were calculated using clonotype frequencies pre- and post-vaccine for each patient individually. A two-sided Wilcoxon matched-pairs signed rank test was used to compare frequencies of the 20 most prevalent T-cell clonotypes identified within the post-vaccination T-cell repertoires, compared with their matching clonotype pre-vaccine. Shannon entropy for paired samples pre- vs post-vaccine was compared using a two-sided Wilcoxon matched-pairs signed rank test (n = 8 pairs of independent biological samples). A Two-sided Wilcoxon rank sum test was used to compare the expression of select gene markers from the total single-cell RNA-seq dataset pre- and post-vaccine. Experiments of functional tumor idiotype-specific T-cell responses post-vaccination were performed in triplicate (n = 3 technical replicates) once (LPL-1, 2, 3, 4, 8, 9) due to limited patient sample availability or twice (LPL-5, 7) with similar results. Significance was calculated using a two-sided Student t test. The one-sided Chi-square test was used to analyze differences between idiotype stimulation and no antigen control. A two-sided Wilcoxon matched-pairs signed rank test was used to compare total interacting pathways pre- and post-vaccine (n = 25 pathways). A two-sided Wilcoxon matched-pairs signed rank test was used to compute p-values comparing signaling patterns within each cell type (n = 35). No statistical method was used to predetermine the sample size. No data were excluded from the analyses. The experiments were not randomized. The Investigators were not blinded to allocation during experiments and outcome assessment. Reporting summary Further information on research design is available in the [265]Nature Portfolio Reporting Summary linked to this article. Supplementary information [266]Supplementary information^ (48.7MB, pdf) [267]Peer Review File^ (395KB, pdf) [268]Reporting Summary^ (105.5KB, pdf) Source data [269]Source data^ (214.8KB, xlsx) Acknowledgements