ALBERT

Description: Chronic lymphocytic leukemia, a disease that is not curable with standard therapy, is an attractive target for allogeneic (allo) hematopoetic stem cell transplant (HSCT) with demonstrated strong graft vs. leukemia effect. However, optimal selection of patients (pts) in order to maximize outcome and minimize toxicity is still under study. We hypothesized that patient co morbidity, as measured by the HSCT adapted Charlson co morbidity index (CCI) would have a strong effect in this group of patients with older median age. At the Leukemia/BMT Program of BC the CCI has been prospectively calculated for HSCT pts since 2006, and we performed chart review to calculate scores retrospectively for the remaining CLL allo HSCT patients in order to evaluate the impact of this factor on outcome following allo HSCT. Transplant specific data was collected prospectively and entered into an electronic database. Forty pts with CLL proceeded to allo HSCT between Jan 91 and Dec 07, with myeloablative (MA) (n=21) or non-myeloablative/reduced-intensity (NMA/RIC) (n=12/7) conditioning regimen. Median (range) number of prior therapies was 4 (1–7). Twenty-four pts were refractory to fludarabine. Donors were related in 25 cases, unrelated in 15. Median age (range) was 49 yrs (32–57) (MA) and 57 yrs (52–64) (NMA/RIC), with 3 and 13 patients greater than age 55 in the 2 groups respectively. Interval from dx to HSCT was 60 months (range 7–135) (MA) and 90 months (range 18–350) (NMA/RIC). Five yr OS is 55% for the whole group; 51% for the MA group at a median follow-up (med FU) of 7.2 yrs (range 2.8–14.6) and 62% for the NMA/RIC group with med FU of 4.2 yrs (range 0.1–6.8). OS did not differ between the 2 groups (p=0.56). Related and unrelated donors had similar 5yr OS at 60.6 vs. 47.1%, p=0.23. Cumulative incidence of non-relapse mortality at 100 days and 2 years is 10 and 32% for the whole group, 14 and 38% for the MA and 6 and 26% for the NMA/RIC groups. CCI was 0, 1, 2, and 3 or greater for 21, 5, 7, and 7 patients. OS by CCI 0–2 vs. 〉=3 was 63 vs.18% for the whole group (p=0.01), 56 vs. 0% for the MA (p=0.03), and 74 vs. 27% for the NMA/RIC groups (p=0.06). Further analyses will explore the relationship between CCI and NRM, acute and chronic graft vs. host disease, and relapse. Patient numbers in this series are insufficient to evaluate the impact of specific co morbidities. In conclusion, for patients with CLL, who are in general of older age, and who may have other therapeutic options, allo SCT is optimally performed in those with a low co morbidity score. Patients with a CCI of 3 or greater may be preferential candidates for alternate less toxic therapies.

Description: Introduction: FLT-3 internal tandem duplications (ITD) and mutations in the nucleophosmin 1 (NPM1)gene appear to have negative and positive prognostic significance, respectively, in newly-diagnosed patients with acute myeloid leukemia (AML) treated with conventional chemotherapy. The prognostic significance of the D835 point mutation in exon 20 of FLT-3 is uncertain. In this study the relative importance of these abnormalities in predicting outcome was compared between patients receiving chemotherapy-based consolidation (chemo) or allogeneic stem cell transplantation (alloSCT) for AML in first complete remission (CR1). Methods: DNA was extracted from diagnostic blood or bone marrow from 267 AML patients aged 〈 60 years who achieved CR1 with induction chemotherapy and then analyzed for the presence of the ITD by PCR and NPM1 exon 12 and FLT3 exon 20 mutations by direct sequencing. Diagnostic cytogenetic abnormalities were assigned prognostic significance using Medical Research Council (MRC) UK criteria. Patients with intermediate or poor prognostic abnormalities and a sibling donor received alloSCT in CR1. If more than 1 cycle of induction therapy was necessary to achieve CR1 or poor risk cytogenetics were detected, patients without a sibling donor received unrelated donor SCT. All other patients received a minimum of one cycle of chemotherapy consolidation. Most patients received high dose cytarabine and daunorubicin induction therapy with similar consolidation. Median (range) follow-up for the entire group was 870 days (90–6003 days). Results: The overall frequency of mutations was 25%, 10% and 24% for the ITD, D835 and NPM1 mutations, respectively. On analysis of the 230 patients remaining after exclusion of acute promyelocytic leukemia (APL), ITD was significantly associated with poor disease free (DFS), event free (EFS) and overall (OS) survival. The D835 mutation was associated with poor DFS only and no effect of the NPM1 mutations could be detected. The presence of the ITD correlated with normal cytogenetics and a high presenting white cell or blast count. 92 and 138 of 230 non-APL patients received alloSCT or chemo, respectively, as consolidation in CR1.Of these 68 in the alloSCT and 95 in the chemo groups had intermediate risk cytogenetics. ITD predicted a poor EFS, DFS and OS for chemo patients regardless of the NPM1 mutation status. For ITD positive patients, relapse risk was higher with chemo in CR1 vs alloSCT (p= 0.007). Among intermediate risk cytogenetic patients 10 of 37 (27%) ITD positive vs 29 of 59 (49%) ITD negative patients are alive after chemo consolidation in CR1 (p0.5). Conclusions: FLT3 ITD predicts a poor OS following chemotherapy as consolidation for AML in CR1. In contrast the results of alloSCT in CR1 are similar for pts with and without the ITD suggesting that alloSCT can overcome the poor prognosis associated with this mutation.

Description: Therapeutic targeting of BCR-ABL with selective ABL tyrosine kinase inhibitors (TKIs) has led to a significant survival benefit for early phase CML. However, TKI monotherapies are rarely curative, with persistence of leukemic stem cells, emergence of resistance and relapses remaining as challenges. To identify differentially expressed and new miRNAs in CD34+ CML stem/progenitor cells that might serve as potential biomarkers and/or therapeutic targets, we have performed Illumina Deep Sequencing to obtain absolute miRNA expression profiles of highly purified CD34+ cells obtained at newly diagnosed stage from six CML patients. Three of the patients were classified retrospectively, after imatinib (IM) therapy, as IM-responders and three as IM-nonresponders. CD34+ cells isolated from five normal bone marrow (NBM) samples were similarly analyzed as controls. Bioconductor DESeq2 analysis revealed 63 differentially expressed miRNAs between CML and NBM samples (adjusted P

Description: Abstract 2341 Background: Growth factors (GFs) that stimulate the proliferation of human hematopoietic stem cells (HSCs) in vitro have been identified, but maintenance of the original stem cell potential has remained a challenge. We have confirmed the observation that UG26–1B6 cells produce factors that enhance the maximum yield in vitro of HSCs achievable with SCF, FLT3-L, IL-3, IL-6, and G-CSF (5 GFs) alone (∼5-fold increase in lympho-myeloid repopulating cells by limiting dilution transplant analysis in NOD/SCID-IL-2Rγc null mice assessed 20 weeks post-transplant). It was therefore of interest to determine whether these different outcomes could be correlated with specific signalling events. Methods: CD34+cell-enriched human cord blood (CB) cells were stimulated with 5 GFs ± UG26 conditioned medium (CM) for defined periods, and then fixed in paraformaldehyde, stained for surface markers, alcohol permeabilized and stained with antibodies specific for multiple intracellular signalling intermediates. Samples were then analyzed using either a CyTOF mass cytometer (34 unique parameters) or a LSR Fortessa cell analyzer (11 parameters). Data analysis was performed using a combination of Cytobank, SPADE, and R using the 'flowCore', 'flowType', 'lattice', 'SamSpectral', and 'gplots' packages. Results: Analysis by CyTOF mass cytometry of 106 CD34+ CB cells after 15 minutes of their stimulation ± 5 GFs ± CM showed that the HSC-enriched subset (Lin−CD34+CD38−CD90+CD49f+cells) exhibited similar responses to the 5 GFs and 5 GFs+CM cocktails (average shifts in median intensity [Δasinh (stimulated-unstimulated)] of 0.32 [-0.14 to 1.96] and 0.37 [-0.02 to 1.78]) with increased phosphorylation of S6, STAT3, STAT5, CREB, IκBa (total protein), AKT, Syk/ZAP70, and ERK1/2. However, the addition of CM did selectively reduce the level of pS6 and pAKT seen with 5 GFs alone, and also caused an increase in pCREB. In addition, activation of pSyk/ZAP70 was detected only in HSCs stimulated with 5 GFs + CM. Phosphoflow analysis of the same subset of cells from a second CB pool confirmed the CyTOF results for pAKT, pSTAT5, pCREB and pERK1/2 (variably) following a 15 minute stimulation but not pSyk/ZAP70. After 2 hours of stimulation, the activation patterns were sustained for most targets for cells exposed to 5 GFs + CM but were reduced in 5 GFs alone - particularly pCREB which was decreased to levels in unstimulated cells. After 24 hours, pAKT and pSTAT5 showed continued activation in response to 5 GFs ± CM, whereas pCREB and pERK1/2 signal were reduced to below starting levels. Hartigans' Dip Test for Unimodality revealed significant deviations from unimodality for the mass cytometry HSC data. Principal component analysis (PCA) of the HSC data followed by subsequent clustering also revealed a number of differentially responsive sub-populations including a broadly cytokine non-responsive, non-apoptotic (cleaved PARP−) population, as well as a minority population with high levels of pSrc and pPLCγ, but reduced activation of other pathways. Further analysis using SPADE revealed decreased cytokine responsiveness with higher CD49f intensity. Weak but significant negative correlations between CD49f and pSTAT3, pCREB, and pS6 and positive correlations with pSrc and pPLCγ (holm-adjusted p values

Description: In follicular lymphoma (FL), clinical progression/histologic transformation occurs as a result of emergence of biologically aggressive malignant clones that escape the immune response. Whether such clones are preexisting or arise as a consequence of spontaneous or therapy-induced DNA mutation/genomic instability remains an open question; however, it has become increasingly clear that it will be important to understand the multi-clonal structure of tumors in order to treat them more effectively. As well, tumor-infiltrating immune cells represent a complex and heterogeneous population and are hypothesized to either help or hinder tumor progression. Several studies have searched for prognostic value in various infiltrating immune cell subsets, but results have been inconsistent, likely in part because most studies focus only on a few, incompletely characterized subsets at a time. Furthermore, the potential for co-variance between malignant subclones and the complement of infiltrating immune cells has not yet been explored. We would propose that understanding the entire tumor "ecosystem" in FL is needed to make further strides in predicting biological behavior and ultimately in guiding selection of the most appropriate of available rational therapies. In this study, we sought to use mass cytometry (CyTOF) to characterize both the clonal substructure of malignant populations and the infiltrating immune repertoire with unparalleled breadth, yet at single-cell resolution. We designed and optimized a 2-tube, 40-parameter CyTOF panel in which one tube is dedicated to characterizing population substructure among malignant B-cells and the other to profiling the composition of the infiltrating T-cell repertoire. We accessed viably frozen single cell suspensions from excess lymph node biopsy material remaining after diagnostic flow cytometry have been prospectively banked for the past 2 decades. We stained 12 FL, 23 diffuse large B-cell lymphoma (DLBCL), and 5 reactive lymph node (rLN) samples using the B-cell marker tube, acquired CyTOF data, and performed analyses using tSNE and Complicity mapping algorithms, which led us to make the following observations: First, over half of FL samples contain at least two phenotypically distinct tumor subclones. In contrast, very few of the DLBCL samples that we have examined thus far exhibit definitive phenotypic subclones. Lower intra-tumoral heterogeneity in DLBCL may imply that these tumors represent outgrowth of a highly evolved, dominant clone as compared to a less evolved collection of "untested" clones in FL. Second, tSNE mapping of FL tumors revealed two distinct subtypes, one in which the individual tumors showed highly similar and partially overlapping phenotypes that localized in proximity to normal germinal center B cells ("GC" subtype), and the other which was composed of more phenotypically heterogeneous tumors that were localized more distantly from normal germinal center B cells ("non-GC" subtype). We also stained 6 FL and 5 rLN samples using the T-cell marker tube, acquired CyTOF data, and performed analysis using tSNE and Scaffold mapping algorithms. We found the relative abundance of T-cells in FL samples was comparable to rLN (36.1% vs. 37.4% of total cells, p=0.90); however, most of the FL samples showed elevated CTL fractions (20.1% vs. 13.0% of total T-cells, p=0.02). Using Scaffold mapping, we were able to resolve a dramatic diversity of T and T/NK cell subsets, each of whose abundance varied substantially from one sample to the next. For example, the majority of CTL cells in rLN samples exhibited a relatively homogeneous, naïve-like phenotype, while CTL cells in FL typically contained a heterogeneous mixture of activated and exhausted subsets. The T-helper compartment exhibited a similar picture with relative population homogeneity within rLN samples, contrasting with heterogeneity within FL samples. Composite analysis correlating B- and T-cell features within the same FL specimen revealed that loss of HLA-DR and CD124 on malignant cells was correlated with CTL exhaustion, suggesting a potential immune escape mechanism. Our findings illustrate that novel information is revealed by 40-dimensional CyTOF analysis and provide insight into important, but understudied aspects of intra-tumoral heterogeneity and variable host immune response in lymphoma. Disclosures Scott: Janssen: Consultancy; Celgene: Consultancy; BC Cancer Agency: Patents & Royalties: Inventor on a patent licensed to NanoString Technologies; Roche: Honoraria.

Description: Induction chemotherapy followed by autologous stem cell transplant (ASCT) has become standard treatment for symptomatic myeloma patients. Our strategy has been to use the Dex component of the VAD protocol alone for 2–4 cycles as induction. Between January 1998 and May 2005, 406 pts were referred to us for treatment of multiple myeloma. We retrospectively reviewed all pts in order to determine the rate and depth of response to Dex, to determine if response predicted overall survival (OS) or progression free survival (PFS) and assess factors predictive of response. Of 406 pts, 124 were excluded due to prior therapy, no treatment or lack of response data. Analysis was performed on 282 pts all of whom received single agent Dex as primary therapy for newly diagnosed myeloma given at 40mg on days 1–4, 9–12, 17–20 every 28 days for at least two cycles. 40% were female and 60% male and median age was 56. Myeloma subtypes were: IgA 21%; IgG 56%; light chain (LC) 17%, nonsecretory 2% and others 1.4%. Cytogenetic information was available in 51% of pts with 53 confirmed as having a deletion on chromosome 13 (del13q). Response to Dex induction therapy was as follows; nCR [no M protein, IF not performed] 4%, PR1 [〉90% reduction in M protein/involved Ig subtype] 10%, PR2 [〉50% reduction] 43%, MR [〉25% reduction] 20%, no response (NR) [within 25% of base level] 14% and progressive disease (PD) [increase by 〉25%] 7%. 241 proceeded to transplantation; 212 underwent auto SCT with a melphalan based conditioning regimen after stem cell mobilization with cyclophosphamide and GCSF, 29 had an allo SCT. We evaluated the effect of del13q, International Staging System (ISS), b2Microglobin (b2M), sex, age, Salmon-Durie stage, hypercalcemia, Ig subtype, creatinine and hemoglobin at diagnosis (Hgbdx) on response to Dex, OS and PFS from treatment initiation. Univariate analysis showed that OS was significantly affected by Hgbdx 〉/≤10g/dl, ISS score and b2M 〉/≤3.5 while PFS was influenced by b2M

Description: Patients (pts) with relapsed (REL) transformed non-Hodgkin lymphoma (NHL) have poor outcome with conventional therapies. Although the role of autologous (auto) stem cell transplantation (SCT) has been previously described, there is scarce literature about allogeneic (allo) SCT in this setting. Forty pts with REL composite low/intermediate (L/I) NHL (25 transformed, 8 composite (same site) and 7 discordant (different sites)) underwent allo-SCT Jan ’89 to June ’05. Fifteen pts (38%) received stem cells from unrelated donors (UD) (12 matched and 3 with one antigen mis-match [1AgMM]). Six of 15 pts (40%) in the UD group are alive including one pt post 1AgMM SCT. 25 pts (62%) had matched sibling allo-SCT and 5 (20%) pts are still alive. Twenty-nine of 40 pts (73%) died, with REL NHL (n=15, 38%) or treatment related mortality (TRM) (n=14, 35%). The 2 and 5-year probability of OS were 39% [95% CI: 26–58%] and 23% [10–49%] respectively; of EFS 36% [24–55%] and 23% [11–46%]. Univariate analysis (UVA) showed presence of residual (res) NHL prior to allo-SCT as a poor prognostic factor for OS and EFS with p value of 0.029 (HR=2.2) and 0.011 (HR=2.5) respectively. No survival difference was seen between patients treated with related or UD stem cells. On multivariate analysis (MVA), development of aGVHD grades 2–4 had a significantly negative impact on both OS and EFS (for OS: p=0.006, HR=1.46, for EFS: p=0.01, HR=1.37) as did res disease prior to SCT (for OS: p=0.04, HR=2.2, for EFS: p=0.02, HR=2.4). Age 〉46 years was a significant risk factor for TRM on UVA (p=0.02, HR=3.3) and MVA (p=0.07, HR=2.9). MVA also showed development of cGVHD as a significant risk factor for TRM (p=0.001, HR=1.8). No difference in TRM was seen between related and UD SCT. Pts with res NHL at SCT had a higher risk of REL NHL (p=0.008, HR=5), as did pts who were chemotherapy resistant (no change in the tumor mass) pre-SCT (p=0.002, HR=3.9). Inclusion of TBI in the conditioning regimen significantly reduced the risk of REL NHL (p=0.01, RR=0.22). Prior rituximab therapy was associated with reduced REL risk, although this was not statistically significant (p=0.06, HR=0.82). CGVHD was significantly protective against REL in MVA (p=0.03, HR= 0.8), and res NHL at SCT was associated with increased REL (p=0.003, HR= 7). Stem cell dose (mononuclear cells) 〉3X108/kg was the only significant risk factor for development of cGVHD in UVA (p=0.04, HR=0.4) and MVA (p=0.05, HR=0.4). In conclusion, this is the largest report on the utility of allo-SCT (inculding UD-SCT) in patients with REL composite L/I-NHL. Survival was similar in pts receiving stem cells from related and unrelated donors. UVA and MVA analysis revealed multiple significant risk factors for survival, TRM, development of cGVHD and relapse which should be taken into consideration for future allo SCT pts. Development of chronic, but not acute GVHD was protective for NHL relapse, supporting the action of a graft vs lymphoma effect in this population.

Description: The effect of prior exposure to rituximab (ritux) on relapse (REL) and survival following stem cell transplantation (SCT) in patients (pts) with REL composite low and intermediate grade non-Hodgkin lymphoma (L/I-NHL) is unknown. Fifty-four pts with REL L/I-NHL underwent high-dose chemotherapy (CT) and SCT Jan ’89 to June ’05. Follow-up is complete to April ’6 with a median of 32 mo. Ritux was added to CT regimens since 2001 and given at 375 mg/m2. Eighteen pts received ritux at initial diagnosis of NHL or after REL with salvage CT prior to SCT. Pts proceeded to high dose CT with allogeneic (allo) (n=12) or autologous (auto) SCT (n=6). This group was compared with a group of pts not receiving ritux pre-SCT (n=36)(Table 1). Eleven pts (61%) are alive in the ritux group compared with 11 pts (31%) in the non-ritux group. The 2 and 4-y OS for pts who received ritux were 52% and 52%, vs 36% and 24% (p=.12, RR=.52) for pts who did not receive ritux. The EFS were significantly different with 2 and 4-y EFS for the ritux group 56% and 56% vs 24% and 18% (P=.038, RR=.42) for the non-ritux group (figure 1). No effect was seen on TRM. The risk of REL post SCT was significantly lower in the ritux group (p=.017)(figure 2). Two of 18 pts (11%) had NHL REL in the ritux vs 18 of 36 (50%) in the non-ritux group. The hazard rate of REL in pts who received ritux was 20% that of pts in the non-ritux group. Significance maintained in multivariate analysis (p=.016). No impact was seen on graft-versus-host disease (GVHD). Fifty percent and 61% of pts developed acute GVHD grades 2–4 and 50% and 64% of pts developed chronic GVHD in the ritux and non-ritux groups respectively. In conclusion, prior treatment with ritux in pts with relapsed composite L/I-NHL undergoing SCT was associated with reduced risk of REL and improved survival without associated increase in toxicity. Further analysis is underway to clarify the nature of this important finding. Table 1: Clinicopathological characteristics of ritux and non-ritux groups Parameter$ Rituximab group, n=18 (%) Non-rituximab group, n=36 (%) $ all p values are 〉0.1. *at diagnosis Median age at SCT 48 y 44 y M:F 2.6:1 1.6:1 Allo-SCT 12(67) 28(78) BM involvement* 10(56) 22(61) B symptoms* 7(39) 10(28) Prior purine analogue therapy 8(44) 11(31) Residual disease prior to SCT 9(50) 14(39) TBI in conditioning 13(72) 30(83) Figure Figure Figure Figure

Description: Background: Recent work in both hematologic malignancies and solid tumors has supported the notion that human cancers exhibit marked intra-tumoral heterogeneity (ITH). Results from next generation sequencing (NGS) studies support that subclonal DNA mutations underlie genotypic ITH within a single tumor since the majority of sequence variants are present in only 5-50% of reads for a given tumor sample, and single cell analyses have shown that individual tumor subclones may be ancestrally related in a complex branching hierarchy, suggesting that therapy failures and progressive disease likely arise by Darwinian selection for more aggressive or therapy resistant clones. It has become increasingly clear that it will be important to understand the multi-clonal structure of tumors in order to treat them more effectively. In this study, we sought to use time-of-flight mass cytometry (CyTOF) to explore clonal phenotypic substructure in diffuse large B-cell lymphoma (DLBCL), a diagnostic entity notorious for clinical heterogeneity. Methods: We examined viably frozen single cell suspensions from diagnostic lymph node biopsy samples received for flow cytometric analysis at the BC Cancer Agency. We have thus far acquired CyTOF data from 25 cases of DLBCL using a two-tube, 40-parameter panel encompassing a total of 58 different markers including both surface and intracellular antigens that were selected to reveal heterogeneity within the malignant B-cell population. For each sample acquisition, we included "spiked-in" control cells from pooled reactive (non-malignant) lymph node samples to control for staining variation between antibody/reagent lots and also run-to-run CyTOF instrument drift, facilitated by a CD45 antibody "barcoding" approach. We analyzed the data using a combination of viSNE, Isomap, and PhenoGraph analysis packages. Results: Analysis of individual tumor samples readily distinguished between malignant and residual normal B-cell populations, and also revealed distinct subpopulations among malignant cells of varying degrees of relatedness to one another. These subpopulations were then sorted from one another by conventional FACS from parallel vials of cryopreserved cells using lower dimensional sorting strategies derived from the 40-parameter CyTOF data. Sorted subpopulations will be analyzed by targeted amplicon sequencing for single nucleotide variants identified from whole exome sequencing data obtained from unsorted material to explore the hypothesis that these may represent genotypic subclones. 
Analysis of multiple tumor samples at once yielded several observations. First, B-cells from reactive lymph nodes and non-malignant B-cells within patient lymphoma specimens reproducibly cluster atop one another, indicating highly similar if not identical phenotypic profiles. Second, the majority of patient DLBCL tumors form cohesive individual clusters, separate and distinct from one another, suggesting that the 40-dimensional panel defines cell populations with sufficient resolution such that each patient's tumor can be uniquely identified. Third, individual DLBCL tumors do not aggregate in tight proximity with one another to the extent that we observe among patient follicular lymphoma (FL) samples, suggesting DLBCL represents a broader diversity of phenotypic classes. Fourth, there is local, but loose aggregation of ABC versus GCB subtypes, but there are also clear outliers and areas of intermingling between ABC and GCB tumors, as defined by immunohistochemistry. Finally, a subset of DLBCL tumors exhibit minor subpopulations that map apart from their corresponding "parent" tumor populations, but yet overlap one another, raising the possibility of divergent evolution away from (or alternatively convergent evolution towards) a common tumor archetype. Conclusions: 
Taken together, these observations support that novel information can be derived from CyTOF data with important implications for our understanding of both intra- and inter-tumoral heterogeneity in DLBCL. Disclosures Scott: Celgene: Consultancy, Honoraria; NanoString: Patents & Royalties: Inventor on a patent that NanoString has licensed.

Description: Follicular lymphoma (FL) is an indolent, but incurable malignancy as most patients eventually experience progressive disease. We hypothesized that clonal heterogeneity and patient-specific immune responses would contribute to variable clinical outcomes and that understanding the complexity of the entire tumor "ecosystem" would allow us to better match patients with specific types of tumor- and immune-targeted therapies. In this study, we performed 38-dimensional single-cell phenotyping by mass cytometry (CyTOF) to simultaneously characterize both the substructure of malignant B cell populations as well as the T cell microenvironment in a cohort of 77 diagnostic patient FL biopsies and 35 benign reactive LN (rLN) biopsies. We first applied the t-distributed Stochastic Neighbour Embedding (t-SNE) algorithm to explore intra- and inter- tumoral heterogeneity among malignant B cell populations. t-SNE mapping of individual samples showed that more than a third of FL samples contain at least two phenotypically distinct tumor subpopulations, supporting the notion of multi-clonal tumor architectures presumably due to ongoing clonal evolution. Batched analysis combining all 77 FL cases together with 35 rLN samples revealed two distinct tumor subtypes comprising about 25% (type "A") and 10% (type "B") of total FL samples, respectively, with individual tumors within each subtype showing highly similar and partially overlapping phenotypes. Mapping the same data using Uniform Manifold Approximation and Projection (UMAP), a dimensional reduction algorithm similar to t-SNE but preserves global structure more accurately, revealed that type A tumors localized in close proximity to normal germinal center (GC) B cells, thus fulfilling conventional expectations as to the histogenesis of FL. In contrast, type B tumors localized more closely to pre-GC B cells, implying the existence of an alternate histogenic path in FL. Importantly, we also performed single-cell RNA-Seq on a subset of FL cases which independently confirmed the type A vs type B distinction in whole transcriptomic space. We next analyzed matching T cell data using a modified Statistical Scaffold algorithm in order to place distinct subsets in context with conventionally defined normal T cell populations. Clustering analysis using multi-layer phenograph performed on T cells from all FL and rLN samples combined yielded hundreds of small, but phenotypically distinct populations that were then annotated according to the nearest conventionally defined T cell subset. These imputed designations were used as features to perform hierarchical clustering of samples which revealed 3 major clusters. Cluster1 was characterized by mostly naive T cell populations and contained the majority of rLN samples. Cluster2 was characterized by more differentiated effector T cell populations and was dominated by FL samples. Samples within Cluster2 could be further divided into Tfh, Treg and Th1-rich subgroups. Cluster3 was characterized by a diverse T cell environment including naive, memory and differentiated effector subsets and contained a mixture of rLN and FL samples. Integrative analysis correlating B- and T- cell features revealed type B FL tumors were associated with a Tfh-rich immune landscape. Taken together, these data reveal pervasive phenotypic heterogeneity in both malignant and immune cell compartments of patient FL samples and suggest that defining tumoral subtypes as well as the status of the local immune response within individual samples will support more refined diagnostic classification and highlight functional interactions most amenable to therapeutic targeting. Disclosures Gascoyne: NanoString: Patents & Royalties: Named Inventor on a patent licensed to NanoString Technologies. Scott:Celgene: Consultancy, Honoraria; Roche: Research Funding; NanoString: Patents & Royalties: Named Inventor on a patent licensed to NanoString Technologies, Research Funding; Janssen: Research Funding. Steidl:Juno Therapeutics: Consultancy; Seattle Genetics: Consultancy; Nanostring: Patents & Royalties: patent holding; Bristol-Myers Squibb: Research Funding; Tioma: Research Funding; Roche: Consultancy.