ALBERT

Description: A 77 year-old man initially presented in July 2013 with anemia, splenomegaly, and constitutional symptoms. A bone marrow biopsy revealed a hypercellular marrow with megakaryocytic hyperplasia and atypia and mild reticulin fibrosis, consistent with a diagnosis of primary myelofibrosis (PMF). Cytogenetics revealed a normal karyotype. JAK2 V617F testing was negative. Initiation of treatment with the JAK inhibitor ruxolitinib led to marked symptomatic improvement. The patient was then enrolled in a Phase 2 study with the anti-lysyl oxidase-like-2 (LOXL2) monoclonal antibody simtuzumab, administered via IV infusion every two weeks (while continuing ruxolitinib). He tolerated the simtuzumab infusions well initially, but with the 11th and 12th infusions experienced rigors, hypotension, and hypoxia. This occurred ~8 months after his initial PMF diagnosis. A repeat bone marrow biopsy revealed large aggregates of mast cells comprising 30-40% of the marrow cellularity (with 16% mast cells enumerated on the aspirate). A subset of the mast cells exhibited spindled morphology, and CD25 co-expression was demonstrated by flow cytometry in a subset of CD117-positive cells. Testing for the KITD816V mutation was positive. Tryptase levels were significantly elevated (375 ng/mL). These findings were consistent with a diagnosis of aggressive systemic mastocytosis with an associated hematologic non-mast cell lineage disorder (ASM-AHNMD). Compassionate-use approval for the KIT inhibitor midostaurin was obtained, and treatment with midostaurin (in addition to continuation of ruxolitinib) was initiated. The patient initially reported symptomatic improvement with midostaurin treatment, but after several months his symptoms began to worsen, with a corresponding increase in tryptase (875 ng/mL). A repeat bone marrow biopsy revealed overt evolution to mast cell leukemia (MCL) with 〉 90% mast cell involvement. Based on the presence of an IDH2mutation identified on a clinical next-generation sequencing assay, the patient was evaluated for a clinical trial with the mutant IDH2 inhibitor AG-221. Unfortunately, the patient decompensated and expired before he could enroll in the study. To identify contributing driver mutations and to delineate clonal hierarchy associated with disease initiation and progression in this unique case of PMF with concomitant ASM, exome sequencing was performed on serial samples obtained at the following disease stages:PMF diagnosis (pre-ruxolitinib) (Day 0)PMF on ruxolitinib (before ASM diagnosis) (Day 181)ASM diagnosis (post-anti-LOXL2 antibody, pre-midostaurin) (Day 394)Progression to MCL (on midostaurin) (Day 519)Matched skin (normal) sample Likely driver mutations in IDH2 and SRSF2 were identified at ~40-50% variant allele frequency (VAF) in all samples and were therefore likely present in the founding clone. The KIT D816V mutation was found at 23% VAF at Day 0, then ~40% VAF in all other samples, suggesting it was present in a daughter subclone of the IDH2/SRSF2-containing clone that became dominant over time with disease progression. These findings also suggest that targeting KIT with midostaurin would be unlikely to eradicate the founding clone. Rather, selective targeting of IDH2 and/or SRSF2 could potentially ameliorate both diseases (PMF and ASM/MCL). To provide pre-clinical evidence of the potential utility of targeting IDH2, peripheral blood mononuclear cells obtained at the time of ASM diagnosis were plated in liquid culture in the presence or absence of the mutant IDH2 inhibitor AGI-6780. After 14 days in culture, the differentiation status of the cultured cells was examined by mass cytometry (CyTOF). Treatment with AGI-6780 resulted in a marked enhancement of myeloid differentiation (denoted by CD15 and CD66b expression) along with a corresponding decrease in CD34+ progenitor cells. These effects were not seen in cells cultured in the absence of AGI-6780. These results are consistent with prior studies in acute myeloid leukemia indicating that the beneficial effects of mutant IDH2 inhibition are likely related to inducing differentiation of primitive cells. In summary, this study highlights the capacity of serial genomic analysis to define the clonal architecture that drives disease initiation and evolution, and to distinguish founding vs subclonal mutations to identify the most promising targets for therapeutic intervention. Disclosures Oh: Janssen: Research Funding; CTI: Research Funding; Gilead: Membership on an entity's Board of Directors or advisory committees, Research Funding; Incyte Corporation: Membership on an entity's Board of Directors or advisory committees, Research Funding.

Description: Background: Follicular lymphoma (FL) is the most common indolent NHL (iNHL), exhibits a variable clinical course, and remains largely incurable. The pathogenesis of FL is complex and involves over expression of Bcl2 via t(14;18) translocation, as well as copy number alterations, recurrent somatic mutations, and changes in the tumor microenvironment. In line with recent publications, we hypothesized that recurrent somatic genomic mutations in FL will be present and may impact FL development, progression, transformation, and clinical outcomes. Methods: To address this, we performed exome sequencing (NimbleGen SeqCap EZ V2.0) of tumor and normal frozen tissue pairs from 24 patients in a discovery cohort with untreated FL (12), relapsed FL (6), or transformed FL/iNHL (6). We developed a custom capture assay (NimbleGen) that targets 7.05 MB corresponding to the coding, 5' and 3' UTR regions of 1717 genes. The custom capture genes included somatic mutations identified in our exome discovery cohort (898 genes) or somatic mutations previously published to be recurrently mutated in B cell NHL (819 genes). Instrument data from the discovery cohort exome and re-sequenced custom capture were combined and analyzed using the McDonnell Genome Institute (MGI) somatic caller pipeline (5 SNV callers, 3 indel callers), filtered (minimum 20X coverage, minimum 2.5% VAF, maximum 10% normal VAF) and manually reviewed. Additionally, the 1717 custom capture strategy was used to sequence an extension cohort consisting of FFPE tumor samples from 80 patients with FL, achieving 〉20x coverage for 〉75% of the targeted region. All discovery and extension samples have clinical annotations that include FLIPI prognostic score, treatment, and clinical outcomes. Results: Combined analysis of exome and custom capture data for the discovery cohort yielded a robust data set with good sequence coverage of 〉78% of the targeted regions with at least 20x depth in all samples and a mean depth of 89x. Based upon somatic mutations identified and manually reviewed using this approach, we conservatively estimate 0.98 mutations per MB in FL. 23 genes were recurrently mutated in 3 or more cases, and an additional 75 genes recurrently mutated in 2 cases in the discovery cohort. Consistent with recent publications (Li H et. al., Blood, 2014; Green MR, PNAS, 2015; Yildiz M et al, Blood, 2015) we confirmed a number of genes that were highly recurrently mutated in FL [TNFRSF14 (50%), Bcl2 (25%), IRF8 (13%), TP53 (13%)] including chromatin modifying genes consisting of histone methyl transferases [KMT2D/MLL2 (58%), EZH2 (13%)], histone acetyltransferases [CREBBP (42%), EP300 (17%)], histone linkers [HIST1H1C (13%), HIST1H1E (8%), HIST1H2BO (8%), HIST1H3G (8%), HIST2H2AC (8%); collectively 42%]. We also confirmed (ATP6V1B2, 13%) and found unreported (ATP6AP2, 8%; ATP6V0A1, 4%; ATP6V1F, 4%) mutations in vacuolar ATPase proton pump genes and P5 or Ca++ ATPase genes (ATP13A2, 4%; ATP13A4, 4%, ATP2B4, 4%;). We confirmed (CD79B, 13%; BCL10, 8%) and found unreported (CD22, 13%) mutations in components of the B cell receptor signaling pathway. The previously unreported recurrent mutations in CD22 were consistent with loss-of function (2 missense, 1 nonsense, 1 frame shift deletion). As a negative regulator of BCR signaling, mutation of CD22 may represent a strategy of to enhance BCR signals in malignant germinal center B cells. We also identified members of the SWI/SNF complex mutated in 33% of this FL cohort: ARID1B (8%), BCL11A (4%), SMARCB1 (4%) in addition to previously reported members BCL7A (12%), SMARCA4 (8%), ARID1A (4%). Somatic mutations were also identified in the Notch pathway: DTX1 (29%), Notch2 (4%), Notch3 (4%), Notch4 (4%). We identified several genes that have not been reported as highly recurrent in FL CXCR4 (42%, mutation calls primarily in RNA), DMD (13%), DNAH9 (13%), FLG (13%), GON4L (13%), PCDH7 (13%), RLTPR (13%), SCN7A (13%), ST6GAL1 (13%). Conclusions: FL genomes harbor a large number of recurrent mutations, consistent with a role in the development and progression of this malignancy. Analysis of the extension cohort and association of recurrently mutated genes and pathways with clinical outcomes is ongoing and will be presented. Disclosures Bartlett: Gilead: Consultancy, Research Funding; Janssen: Research Funding; Pharmacyclics: Research Funding; Genentech: Research Funding; Pfizer: Research Funding; Novartis: Research Funding; Millennium: Research Funding; Colgene: Research Funding; Medimmune: Research Funding; Kite: Research Funding; Insight: Research Funding; Seattle Genetics: Consultancy, Research Funding; MERC: Research Funding; Dynavax: Research Funding; Idera: Research Funding; Portola: Research Funding; Bristol Meyers Squibb: Research Funding; Infinity: Research Funding; LAM Theapeutics: Research Funding.

Description: Key Points Ibrutinib has modest activity in FL with low response rates in rituximab-refractory patients. CARD11 mutations predict for lack of response to ibrutinib.

Description: Background: Composite lymphomas (CLs) describe the rare situation when two morphologically distinct lymphomas occur in the same patient. Prior studies have supported a shared origin of CLs based on shared chromosomal translocations and/or the presence of shared immunoglobulin rearrangements. We used exome sequencing to examine somatic events in two cases of CLs. We hypothesized that each lymphoma will have shared and unique mutations, consistent with the presence of a common lymphoma precursor. Further, we postulated that the shared mutations represent potential initiating events in lymphomagenesis. Finally, because each case includes a Hodgkin (HL) and non-Hodgkin lymphoma (NHL), we hypothesize that the somatic mutations found in HL may offer insight into genomic drivers of HL. Clinical Synopsis:Case 1 was a 57-year-old man who presented with abdominal pain associated with a small bowel obstruction (SBO) and lymphadenopathy (LAD). Pathology showed DLBCL (diffuse large B-cell lymphoma, non-germinal center subtype; stage IV). He was treated with 6 cycles of R-CHOP with complete response. Seven months after the completion of R-CHOP, he presented with left neck LAD. An excisional biopsy revealed classical HL (stage II). Hodgkin and Reed-Sternberg (HRS) cells comprised 10% of the sample. He was treated with 3 cycles of ABVD and involved-field radiation therapy. Case 2 was a 66-year-old man with a history of solitary LAD who presented with increasing LAD, night sweats, and a chest wall mass. An excisional biopsy revealed both follicular lymphoma (FL; grade I-II) and classical HL in distinct areas of the lymph node. HRS cells comprised 30% of the sample. FISH studies were positive for t(14;18). He was treated with ABVD for 6 cycles with partial response. Upon progression, a repeat lymph node biopsy showed persistent HL. He was treated with bendamustine and rituximab for 4 cycles with progressive disease. Methods: We performed whole exome sequencing on formalin fixed paraffin embedded tumor samples and matched normal skin samples. We used our standard somatic variant calling pipeline to call somatic variants. SNV (single nucleotide variant) and indel calls were filtered for basic quality metrics and, using in-house software, were processed through a Bayesian classifier to remove false positive somatic events. Filtered variants were manually reviewed to further verify somatic status. Somatic variants were validated using Ampliseq in Case 1. Results: Sequencing resulted in 〉90% of the target regions with at least 20x in all samples. The mean depth of the NHL and normal samples in Case 1 was 〉65x. To account for the low malignant tumor cellularity of HL, the HL sample of Case 1 was sequenced to a mean depth of 310x. The mean depth for all samples in Case 2 was 〉100x (FL = 188x; normal = 104x; HL = 215x). After filtering, we identified 60 and 133 variants in 57 and 101 genes in Case 1 and Case 2, respectively. In Case 1, we identified three sites inTP53, TNFRSF14, and RASAL2 that were shared. We also identified variants in HIST1H2AG, KMT2D, and STAT3 unique to the DLBCL sample, and two mutations in PTPRT unique to the HL sample of Case 1 (Figure 1). In Case 2, we identified shared variants in TNFRSF14 and HIST1H2BF. We also identified two distinct BCL10 mutations in the FL and HL samples of Case 2 (Figure 1). Conclusions: From the shared somatic mutations identified, we infer that a shared lymphoma precursor for each case is likely, and the shared mutations may be early initiating events. In both cases, a shared nonsense mutation was found at TNFRSF14. TNFRSF14 is recurrently mutated in NHL, and recurrent deletions involving TNFRSF14 have been described in HL. Further work is needed to understand how TNFRSF14 alterations drive HL vs. NHL. Finally, it is intriguing that the HL samples contain independent mutations in PTPRT and BCL10. PTPRT is a known cancer gene, and BCL10 has been shown to be associated with the pathogenesis of B-cell NHLs. These data provide new hypotheses for loci involved in HL pathogenesis. Figure 1 Variant allele frequencies (VAFs) in CL samples from Case 1 (A) and Case 2 (B). (A) shows the VAFs of confirmed variants in the DLBCL vs. HL. (B) shows VAFs of FL vs. HL. Variants of note are highlighted in black. The lack of separation of sites in Case 2 may be the result of some admixture of the two lymphomas, which were sequenced from cores taken from the same lymph node. Figure 1. Variant allele frequencies (VAFs) in CL samples from Case 1 (A) and Case 2 (B). (A) shows the VAFs of confirmed variants in the DLBCL vs. HL. (B) shows VAFs of FL vs. HL. Variants of note are highlighted in black. The lack of separation of sites in Case 2 may be the result of some admixture of the two lymphomas, which were sequenced from cores taken from the same lymph node. Disclosures No relevant conflicts of interest to declare.

Description: The myelodysplastic syndromes (MDS) are the most common cause of bone marrow failure in adults, with an incidence of 40,000 cases per year. Next-generation sequencing of candidate genes has led to major advances in the description of the genetic landscape of MDS, identifying recurrently mutated genes and cellular pathways involved in disease pathogenesis. However, use of targeted panels indicates more comprehensive, unbiased sequencing techniques may yet identify additional recurrently mutated genes or cellular pathways important in MDS. Diseases caused by defects in telomere maintenance (telomeropathies) are variable in clinical and genetic presentation but often involve bone marrow failure. We hypothesized that acquired mutations in telomerase maintenance genes may be a recurrent event in MDS. This is significant as identification of recurrent somatic mutations in telomerase maintenance genes would provide further insight into MDS pathogenesis and identify a potential druggable pathway for MDS patients, as novel agents targeting the telomerase pathway are currently in clinical development. First, we identified three adults who presented in middle age with MDS and idiopathic pulmonary fibrosis (IPF), two of which also had a family history of IPF. All three patients had shortened telomeres (

Description: The investigators present their analysis of primary cells from patients with human T-cell leukemia virus 1–associated adult T-cell leukemia/lymphoma treated in a phase 2 clinical trial with nivolumab to elucidate mechanisms of hyperprogression that halted the trial after just 3 patients received a single treatment.

Description: Key Points FLs harbor more recurrent mutations in the BCR signaling pathway, SWI/SNF complex, and histone genes than previously known. Novel recurrent mutations affecting BTK, SYK, and HVCN1 may have therapeutic and prognostic implications for FL.

Description: Multiple Myeloma (MM) is a malignancy of plasma cells that affects over 30,000 Americans every year. Despite advances in the treatment of the disease, approximately 12,000 American patients will still die of MM in 2019. One of the mainstays of treatment for MM is the immunomodulatory and antiangiogenic drug lenalidomide; which is used in induction therapy, maintenance therapy and treatment of relapsed disease. Although not fully elucidated, lenalidomide's mechanism of action in MM involves the drug binding to Cerebelon (CBN) and leads to the subsequent degradation of the Ikaros (IKZF1) and Aiolos (IKZF3) transcription factors (TF). These TFs play important regulatory roles in lymphocyte development. Despite lenalidomide's importance in MM treatment, several groups have reported that MM patients treated with lenalidomide rarely go on to develop B-cell acute lymphoblastic leukemia (B-ALL). The genetics and clonal relationship between the MM and subsequent B-ALL have not been previously defined. Importantly, it is not clear if the MM and B-ALL arise from the same founding clone that has been under selective pressure during lenalidomide treatment. As deletions in IKZF1 are common in B-ALL, one could hypothesize that lenalidomide's mechanism of action mimics this alteration and contributes to leukemogenesis. We sequenced the tumors from a cohort of seven patients with MM treated with lenalidomide who later developed B-ALL. These data did not show any mutational overlap between the MM and ALL samples-the tumors arose from different founding clones in each case. However, several genes were recurrently mutated in the B-ALL samples across the seven patients. These genes included TP53, ZFP36L2, KIR3DL2, RNASE-L, and TERT. Strikingly, five of the seven patients had a TP53 mutations in the B-ALL sample that was not present in the matched MM sample. The frequency of TP53 mutations in our cohort was much higher than that reported in adult de novo B-ALL patients which can range between 4.1-6.4% (Hernández-Rivas et al. 2017 and Foa et al. 2013). Utilizing CRISPR-Cas9 gene editing, we disrupted the Zfp36l2 or Actb in murine hematopoietic stem cells (HSCs) of mice with or without loss of Trp53. We performed our first transplantation experiment in which the cohorts of mice have loss of Trp53 alone, loss of Zfp36l2 alone, loss of both Trp53 and Zfp36l2, or a control knockout (KO) of Actb. To characterize the disruption of Zfp36l2 alone and in combination with Trp53 we analyzed the hematopoietic stem and progenitor cell compartments in the bone marrow of the above transplanted mice. In mice with a loss of Zfp36l2 there is a decrease in Lin- Sca-1+ c-Kit+ (LSK), short term-HSC (ST-HSC), and multipotent progenitors (MPP). This decrease was not observed in the mice with a loss of both Trp53 and Zfp36l2, where instead we noted an increase in monocyte progenitors (MP), granulocytes-macrophage progenitors (GMP), and common myeloid progenitors (CMP) cells. In this Trp53 Zfp36l2 double loss model we also noted a decrease in B220+ B-cells that was not seen in the Zfp36l2 alone. In this cohort of Trp53 Zfp36l2 loss, we characterized B-cell development through hardy fraction flow cytometry, and identified a decrease in fractions A and B/C (pre-pro and pro-B-cells, respectively) as compared to Zfp36l2 or Actb alone. As lenalidomide does not bind to Cbn in mice, we used the human B-ALL NALM6 cell line to test if treatment with lenalidomide will lead to a selective growth advantage of cells with the same genes knocked out versus wild-type control cells grown in the same culture. We hypothesize that lenalidomide treatment selectively enriched for pre-existing mutated cell clones that evolved into the B-ALL. Preliminary data in NALM6 cells with a loss of TP53 demonstrate a slight increase in cell number at day 7 compared to a RELA control. These experiments will be repeated with concurrent ZFP36L2 and TP53 mutations as well as ZFP36L2 alone. Treatment-related disease is a key consideration when deciding between different treatment options, and this project aims to understand the relationship between MM treatment and B-ALL occurrence. It may be possible to identify MM patients who are at-risk for B-ALL. For example, MM patients who harbor low-level TP53 mutations prior to lenalidomide treatment could be offered alternative treatment options. Disclosures Barnell: Geneoscopy Inc: Current Employment, Current equity holder in private company, Membership on an entity's Board of Directors or advisory committees. Wartman:Novartis: Consultancy; Incyte: Consultancy.

Description: Introduction Very few studies have assessed the incidence and risk factors associated with CNS progressive disease (PD) in elderly DLBCL. The CNS IPI (international prognostic index) uses the standard IPI with an additional point for renal/adrenal (R/A) involvement to define CNS risk in DLBCL patients (pts). The CNS IPI is validated in younger DLBCL pts (Schmitz et al, 2016) but little is known about CNS risk in the elderly. Few data exist outwith a single study of pts 〉80 years (y) pooled from 2 LYSA mini CHOP trials (Cabannes-Hamy et al 2018). CNS PD risk from 270 pts 〉80y was 1.8% at 2y and 3% overall. CNS PD risk did not differ according to CNS IPI in this study. No real world data outside of these trials exists. Few studies have assessed the risk of CNS prophylaxis (intrathecal (IT) +/- high dose intravenous (IV) methotrexate (HDMTX)) in the elderly. Methods Data on consecutively treated new DLBCL pts ≥70y were retrospectively collected across 6 UK centres (2009-2018). We compiled a detailed database of CNS-IPI, extranodal disease sites, CNS prophylaxis, adverse events related to IV HDMTX, rate and timing of CNS PD, outcome post CNS PD and associations with CNS PD risk. Results 529 pts received Rituximab (R) chemotherapy with curative intent (513 R-CHOP/attenuated R-CHOP (97.0%), 13 R-PMitCEBO, 3 R-GCVP). 80.7% (427/529) received no CNS prophylaxis, 14.7% (78/529) received intrathecal (IT) MTX prophylaxis, 2.1% (11/529) received HDMTX only, and 2.5% (13/529) received both IV HDMTX and IT MTX. The median age of all pts was 76.7y and median follow up 2.9y (range 0.4 - 10.9y). The CNS IPI was 1 in 11.8% (60/510), 2 in 21.2% (108/510), 3 in 24.3% (124/510), 4 in 26.1% (133/510), 5 in 12.9% (66/510) and 6 in 3.7% (19/510). 19 pts were unclassifiable due to a missing LDH (Table 1). Extranodal sites (ENS) were noted in 64.7% (342/529). Median ENS were 1 (range 0-5). Commonest sites were bone 24.2% (128/529), liver 9.6% (51/529), renal 7.6% (40/529), and gastric/oesophageal 7.0% (37/529). R/A involvement was present in 9.3% (49/529). Pts with higher CNS IPI were more likely to receive CNS prophylaxis (IT and/or MTX); CNS IPI 0-1; 13.3%, 2-3; 10.9%, 4+; 29.9% (p

Description: Idelalisib and duvelisib are FDA approved oral inhibitors of PI3K delta and PI3K delta/gamma, respectively. Previously, we reported a decrease in regulatory T cells (Tregs) associated with autoimmune hepatotoxicity in untreated CLL patients receiving idelalisib-ofatumumab. We hypothesized that a Th17 inflammatory phenotype may be involved. To directly assess Th17s, we performed IL-17A and IL-17F intracellular staining in CD4 and CD8 T cells from this cohort (no / low toxicity group (grade 0-2 AEs) n=4; high toxicity group (grade 3+ by CTCAE) n=6), and found that the percentage of CD8 cells positive for IL-17F was significantly higher at cycle 2 in patients with high tox compared to patients with low tox (p = 0.0095). Interestingly, pre-treatment, patients who developed high tox had higher IL-17F expression in CD8s compared to CD4s compared to patients with low or no tox. The ratio of IL-17F staining in CD8 to CD4 T cells was significantly higher in patients who developed high tox compared to patients with low tox at the pre-treatment timepoint (p = 0.0381). Furthermore, IL-17A and IL-17F in CD8 T cells were also higher at baseline among mutated IGHV patients (p = 0.1048 and p = 0.0095 respectively), who were also at much greater risk of autoimmune toxicity. These data implicate the Th17 pathway in the autoimmune toxicity of idelalisib. Additionally, single cell mass cytometry (CyTOF) on relapsed/refractory patients treated with idelalisib (n=10) and duvelisib (n=2) showed that the percentage of Tregs decreases with treatment (p = 0.0405 for idelalisib), consistent with our data from the previously untreated idelalisib cohort. To understand this decrease in Tregs, we tested whether PI3K inhibition skews the differentiation of T cell subsets. Naïve CD4 T cells from healthy individuals were isolated and differentiated toward a Treg or Th17 phenotype in the presence of 10 µM idelalisib, duvelisib, or DMSO. Treg differentiation was determined by measuring intracellular FOXP3 by flow cytometry and Th17 differentiation was evaluated by RORγT intracellular staining. Idelalisib treatment significantly decreased Treg differentiation from a median of 53.8% with DMSO to 43.4% (n=4; p = 0.0002), and increased Th17 differentiation from 24.7% with DMSO to 37.2% (n=5; p = 0.0005). Similarly, with duvelisib, Treg differentiation decreased significantly from a median of 53.8% with DMSO to 47.1% (n=4; p = 0.0005), and Th17 differentiation increased from 24.7% with DMSO to 36.9% (n=5; p = 0.02). These results indicate that both idelalisib and duvelisib can skew T cell differentiation toward a more inflammatory phenotype in vitro. Additionally, the effects of idelalisib and duvelisib on proliferation of CD3 cells from healthy PBMCs were evaluated with 3 days in-vitro treatment using the cell division tracker dye CellTrace Violet. The proliferation index was calculated using the ModFit algorithm to quantify the increase in cell number. The proliferation index of CD3 cells in the presence of 1 µM idelalisib and 1 µM duvelisib decreased by a median of 19.3% (n=4; p = 0.0031) and 28.6% (n=4; p = 0.0028) respectively, compared to DMSO; duvelisib led to greater inhibition of T cell proliferation in vitro than idelalisib (n=4; p = 0.0241) which may impact on risk of irAEs. To further identify predictors of toxicity with PI3Ki therapy and to characterize differences between idelalisib and duvelisib, we are analyzing CyTOF data on a cohort of 13 patients treated on a DFCI upfront trial with 7 days duvelisib lead-in, then FCR. With one week of duvelisib monotherapy we found an increase in ICOS expression on CD8 T cells in the low tox group (n=5; p = 0.0495) but not in the high tox group (n=8; p = 0.2245). Furthermore, cluster analysis of the CyTOF data by RPhenograph and FlowSOM demonstrated significant changes associated with high tox in 3 cell population clusters showing markers indicative of Tregs. In conclusion, frontline idelalisib treatment appears to increase a Th17 phenotype while decreasing Tregs, which we showed in vitro to be likely related to the drug altering naïve CD4 T cell differentiation. We hypothesize that an imbalance of Tregs and Th17 T cells is contributing to irAEs in some idelalisib treated patients. Preliminary data with duvelisib suggest alternative mechanisms which we are still analyzing, along with performing cytokine analysis and immunohistochemistry on patient biopsies to further validate these findings. Disclosures Pan: Gilead Sciences: Employment, Equity Ownership. Brown:Teva: Honoraria; Verastem: Consultancy, Research Funding; Sun Pharmaceuticals: Research Funding; TG Therapeutics: Consultancy; Pfizer: Consultancy; Pharmacyclics: Consultancy; Sunesis: Consultancy; Octapharma: Consultancy; Morphosys: Other: Data safety monitoring board; Juno/Celgene: Consultancy; Kite, a Gilead Company: Consultancy, Research Funding; Loxo: Consultancy, Research Funding; Novartis: Consultancy; Gilead: Consultancy, Research Funding; Dynamo Therapeutics: Consultancy; Genentech/Roche: Consultancy; BeiGene: Consultancy; Catapult Therapeutics: Consultancy; Acerta Pharma: Consultancy; AstraZeneca: Consultancy; AbbVie: Consultancy; Invectys: Other: Data safety monitoring board; Janssen: Honoraria.