Feldman, Andrew L.

Abstract Background: Peripheral T-Cell lymphomas (PTCLs) constitute approximately 10% of lymphoid malignancies and consist of several distinct entities based on pathologic and clinical characteristics. With the exception of a few subtypes (e.g., ALK-positive anaplastic large cell lymphoma (ALCL) and some primary cutaneous or leukemic forms of PTCL), a majority of PTCLs are aggressive as characterized by poor treatment response, rapid disease progression and poor overall survival. We have shown that landmark timepoints of event-free survival after diagnosis can stratify subsequent overall survival (OS) in diffuse large B-cell and follicular lymphoma. Here we evaluate this approach in newly diagnosed aggressive PTCLs treated with anthracyline-based or related chemotherapy. Methods. Newly diagnosed PTCL patients were prospectively enrolled in the University of Iowa/Mayo Clinic Lymphoma SPORE Molecular Epidemiology Resource (MER) from 2002-2012. Clinical data were abstracted from medical records using a standard protocol. For this analysis, we included patients receiving anthracycline-based or other multiagent chemotherapy for the following PTCL subtypes: ALK-negative ALCL (N=24); angioimmunoblastic T-cell lymphoma (AITL, N=34); PTCL, not otherwise specified (NOS; N=60); enteropathy-associated T-cell lymphoma (EATL, N=8); extranodal NK/T-cell lymphoma, nasal type (ENKTL, N=11); and hepatosplenic T-cell lymphoma (HSTCL, N=1). Patients were prospectively followed, and event-free survival (EFS) was defined as time from diagnosis to progression, re-treatment, or death due to any cause. Landmark EFS timepoints were assessed at 12 (EFS12) and 24 (EFS24) months after the date of diagnosis. Subsequent OS was defined as time from a specific endpoint (diagnosis, event or EFS landmark). Replication was performed in a population-based cohort of T-cell lymphomas diagnosed from 2000-2009 from the Swedish Lymphoma Registry. Results. 138 eligible patients were enrolled in the MER from 2002-2012, the median age at diagnosis was 58 years (range, 19-88), 66% were male, 73% had Stage III-IV disease, and 33% had IPI 0-1. At a median follow-up of 47 months (range 11-120), 87 patients (63%) had an event and 70 patients (51%) had died. From diagnosis, only 60 patients were event-free at 12 months (EFS12 45%). Patients who failed to achieve EFS12 had a poor subsequent OS from event (median OS = 6.8 months, 95% CI: 5.3-14.0, figure 1). In contrast, patients who achieved EFS12 had a favorable subsequent OS (median unreached, figure 2). Of the 427 eligible patients in the Swedish registry, the median age at diagnosis was 66 years (range, 18-88), 63% were male, 68% had Stage III-IV disease, and 25% had IPI 0-1. PTCL subtypes were: ALK-negative ALCL (N=89); AITL (N=80); PTCL, NOS (N=183); EATL (N=44); ENKTL (N=24); and HSTCL (N=7). At a median follow-up of 86 months (range 40-158), 333 patients (79%) had an event and 316 patients (74%) had died. From diagnosis, 183 patients were event-free at 12 months (EFS12 44%). Similar to the MER cohort, Swedish patients failing EFS12 had poor subsequent survival (median OS = 3.7 months, 95% CI: 2.9-5.3, figure 1). Swedish patients achieving EFS12 had a favorable subsequent OS (median OS = 89 months, figure 2). Similar results were obtained when conducting landmark analysis at 24 months after diagnosis (EFS24). Conclusion. Relapse and re-treatment events within the first 12 months of diagnosis are associated with very poor OS in PTCL treated with anthracyclines or related chemotherapy, while patients achieving EFS12 have encouraging subsequent OS. Stratifying patients into prognostically distinct subsets using EFS12 may help focus biologic and biomarker studies. EFS12 has potential as an early endpoint for studies of newly diagnosed PTCL. Further investigation of determinants related to post-EFS12 survival is needed. Disclosures Maurer: Kite Pharma: Research Funding. Cerhan:Kite Pharma: Research Funding. Ansell:Bristol-Myers Squibb: Research Funding; Celldex: Research Funding. Link:Genentech: Consultancy, Research Funding; Kite Pharma: Research Funding. Thompson:Kite Pharma: Research Funding. Relander:Respiratorius: Patents & Royalties: valproate for DLBCL.

Abstract Abstract 4125 Background: With the exception of angioimmunoblastic T-cell lymphomas, which are thought to derive from follicular helper T cells, little is known about the cell of origin for the most common peripheral T-cell lymphoma (PTCL), PTCL-unspecified (PTCL-U). Following appropriate antigenic stimulation, naïve CD4+ T helper (TH) cells differentiate into effector cells that secrete TH cytokines under the transcriptional regulation of subset-specific transcription factors. Methods: PTCL-U patients (n=53) from a single institution were retrospectively identified and immunohistochemical analysis of Foxp3, T-bet and GATA-3 expression performed on diagnostic biopsy specimens. Immunohistochemical staining was reviewed in a blinded fashion by a hematopathologist and all specimens with greater than 10% staining were scored as positive. Transcription factor expression was correlated with patient characteristics and clinical outcomes. Results: GATA-3 expression was observed in 22 (42%) patients, T-bet expression in 17 (32%) patients and Foxp3 expression in 4 (7%) patients. In order to determine the prognostic significance of GATA-3 or T-bet expression, overall survival was compared between patients with GATA-3 (or T-bet) positive or negative tumors. The median overall survival (OS) was 2 years (95% confidence interval 0.7–12.8 years) for patients with GATA-3 negative tumors, compared with a median OS of 0.9 years (95% confidence interval 0.5–1.2 years, p=0.02) for GATA-3 positive cases. The median overall survival was 1.6 years (95% confidence interval 1.0–6.8 years) for patients with T-bet negative tumors, compared with a median overall survival of 0.7 years (95% confidence interval 0.3–0.9 years, p=0.005) for T-bet positive cases. As a subset of tumors coexpressed GATA-3 and T-bet, both of which are adverse prognostic factors, we examined survival outcomes between those patients with tumors expressing either GATA-3 or T-bet (n=29) and those with tumors which do not express either transcription factor (n=24). The median OS and PFS observed for patients with positive tumors was 0.7 years (95% confidence interval 0.6–1.1 years) and 0.7 years (95% confidence interval 0.5–0.9 years), respectively. In contrast, patients with GATA-3/T-bet negative tumors experienced markedly superior survival, with a median OS and PFS of 3.8 years (95% confidence interval 1.6–12.9 years, p<0.0001) and 2.0 years (95% confidence interval 1.5–12.8 years, p<0.0001), respectively. Four-year estimates of overall and progression-free survival were less than 10% for patients with GATA-3/T-bet positive tumors, whereas 4-year OS and PFS were 49% and 32%, respectively, for those with negative tumors. Both GATA-3 and T-bet expression were associated with advanced age and tumor stage. Therefore, we analyzed GATA-3/T-bet expression as a prognostic factor for survival on both univariate and multivariate analyses, adjusting for pertinent risk factors, including patient age and tumor stage. On univariate analysis, GATA-3/T-bet expression was associated with inferior overall survival (hazard ratio 4.4, 95% confidence interval 2.1–10.4, p<0.0001). Patient age (>60 years), poor performance status (ECOG performance status >1), and stage III/IV disease were also associated with inferior overall survival on univariate analysis. However, when adjusting for these latter adverse prognostic factors on multivariate analysis, GATA-3/T-bet expression remained an independent predictor of poor overall survival in PTCL-U (adjusted hazard ratio 3.2, 95% confidence interval 1.4–8.5, p=0.008). Similarly, when adjusting only for high-risk features (>2 adverse prognostic factors), as defined by the Prognostic Index in PTCL-U (PIT), GATA-3/T-bet expression remained an independent predictor of inferior overall survival on multivariate analysis (adjusted hazard ratio 4.9, 95% confidence interval 2.2–11.5, p<0.0001). Conclusions: GATA-3 and T-bet expression identify subsets of high-risk PTCL-U patients who may benefit from alternative treatment strategies. Disclosures: No relevant conflicts of interest to declare.

Abstract Abstract 2702 Background: CXCR5 [chemokine (C-X-C motif) receptor 5; also known as Burkitt lymphoma receptor 1 (BCR1)] is expressed on mature B-cells, subsets of CD4+ and CD8+ T-cells, and skin-derived migratory dendritic cells. CXCL13 is the ligand for CXCR5, and together they are involved in guiding B-cells into the B-cell zones of secondary lymphoid organs as well as T-cell migration. CXCR5 is also expressed in malignant cells in NHL patients. Single nucleotide polymorphisms (SNPs) in CXCR5 have not been extensively studied, although two studies reported that SNPs from this chromosome 11 gene were associated with risk of primary biliary cirrhosis in a European population and with risk of NHL in an Asian population. This study evaluated the role of common germline genetic variation in the risk and prognosis of the common NHL subtypes of diffuse large B-cell lymphoma (DLBCL), follicular lymphoma (FL), mantle cell lymphoma (MCL), marginal zone lymphoma (MZL) and peripheral T-cell lymphoma (PTCL). Methods: Ten SNPs in CXCR5 were genotyped using a clinic-based study of 2694 NHL cases (including 586 DLBCL, 588 FL, 137 MCL, 230 MZL, and 158 PTCL) and 1521 controls enrolled from 2002–2009 in the Molecular Epidemiology Resource of the University of Iowa/Mayo Clinic Lymphoma SPORE. Clinical data were abstracted for all NHL patients and all were prospectively followed for disease progression, retreatment, transformation and death. SNPs were selected from HapMap using a standard tagging approach (r2≥0.80 and minor allele frequency (MAF) ≥0.05) and included SNPs 5kb upstream and downstream from the gene. Genotyping was conducted using a custom Illumina iSelect panel. All SNPs were in Hardy-Weinberg equilibrium in the control group. The most prevalent homozygous genotype was used as the reference group, and each SNP was modeled as having a log-additive effect in the regression model. For risk of NHL, odds ratios (ORs) and 95% confidence intervals were calculated using unconditional logistic regression, adjusted for age and gender. For prognosis, hazard ratios (HR) were calculated for both event-free (EFS) and overall (OS) survival using Cox proportional hazards regression, adjusting for age, treatment type, and IPI (DLBCL, PTCL), FLIPI (FL), or MIPI (MCL). To determine risk of FL transformation to DLBCL, a competing risk Cox model was used. A p<0.05 was declared as statistically significant. Results: Of the 10 CXCR5 tag SNPs in our study, 5 were associated with risk of NHL, with rs1790192 (MAF=0.39 among controls) having the strongest association (OR=1.19, 95%CI 1.08–1.30; p=0.0003). This SNP was strongly associated with FL (OR=1.44, 95%CI 1.25–1.66; p=3.1×10−7), and was also associated with DLBCL (OR=1.16, 95%CI 1.01–1.33; p=0.038) and PTCL (OR=1.29, 95%CI 1.02–1.64; p=0.037) but not MCL or MZL. For FL, none of the SNPs were associated with EFS or OS, although rs1790192 was associated with better outcome among FL patients that were observed (HR=0.64; 95%CI 0.47–0.87; p=0.0039). For DLBCL, none of the SNPs were associated with EFS, while three SNPs were associated with OS: rs523604 (HR=0.79; 95%CI 0.64–0.97; p=0.028), rs12363277 (HR=0.57; 95%CI 0.33–0.99; p=0.046), and rs3922 (HR=0.79; 95%CI 0.65–0.97; p=0.027). The latter SNP was also associated with risk of transformation from FL to DLBCL (HR=0.61; 95%CI 0.40–0.91; p=0.016). For MCL, MZL and PTCL, none of the SNPs were associated with either EFS or OS. Conclusions: Several germline CXCR5 polymorphisms were associated with NHL risk, particularly rs1790192 with risk of FL as well as DLBCL and PTCL. CXCR5 SNPs did not appear to be important in prognosis for most NHL subtypes, although rs1790192 was associated with a better outcome in untreated FL and rs3922 was associated with OS in DLBCL patients and with risk of FL transformation. These results provide additional evidence for a role of host genetic variation in CXCR5 in lymphomagenesis, mainly in risk of disease development, although there may also be a role for CXCR5 in FL and DLBCL prognosis. Replications of these findings in additional studies are needed. Disclosures: Ansell: Seattle Genetics, Inc.: Research Funding.

Abstract Background The lymphoma proteome is the phenotypic representation of the underlying genetic and epigenetic makeup of each individual patient's tumor. The proteome is rich in druggable targets and offers a unique opportunity for the hematologist to personalize therapy. We developed a deep proteomic profiling method using ion exchange fractionation and tandem mass spectrometry. In this pilot study, we applied this method to detect differentially expressed proteins in anaplastic large cell lymphoma (ALCL) cell lines with previously known (ALK-positive) and unknown (ALK-negative) therapeutic targets. We then asked whether integrative informatic analysis of these data could be used to predict drug sensitivity in each of the cell lines. Methods To examine reproducibility of our method, proteins were extracted independently from 4 pellets each of FE-PD (ALK-negative) and Karpas 299 (ALK-positive) ALCL cells, reduced with dithiothreitol, alkylated with iodoacetamide, and digested with trypsin. Resulting peptides were separated into 6 fractions using strong anion exchange (SAX) chromatography. Peptides in each fraction were analyzed via shotgun proteomics on a QExactive mass spectrometer. Peptide mass spectra (MS/MS) were matched against a RefSeq human protein sequence database using MyriMatch software. Reversed sequences were added to the database to measure identification false discovery rates (FDRs). IDPicker filtered the peptide identifications at 2% FDR. Proteins with at least two unique peptide identifications and five MS/MS matches were considered to be present in the sample. Filtered protein identifications and corresponding spectral counts were used as input to QuasiTel software, which was configured to use proteins with at least one spectrum per biological replicate. Proteins that were significantly differentially expressed (quasi p-value < 0.05) with an absolute log2 fold-change of at least 0.5 fold were loaded into the Ingenuity Pathway Analysis (IPA) software, and a master list of drugs and corresponding gene targets was assembled using PharmGKB database and the Drug-Gene Interaction Database (DGID). The resulting drug-gene target list was merged with the differentially expressed protein identifications. Candidate targets were validated by Western blot and candidate drugs were assessed in viability assays. Results The SAX-LC-MS/MS method identified 10,111 proteins from all replicate analyses of FE-PD and Karpas 299 samples, and 93% of the identified proteome was detected in all 4 replicate analyses. The detected proteome was well represented by key transcription factors, phophatases, kinases, translation regulators and transmembrane regulators. There were 1369 proteins differentially expressed between the 2 cell lines, 709 up regulated in Karpas 299 cell line and 673 up regulated in FE-PD. Differentially expressed proteins also showed consistent expression across the biological replicates (Figure 1). Our integrated approach to identify candidate targets and drugs “rediscovered” ALK in Karpas299 and unexpectedly identified relative overexpression of the IL2-IL2RA-STAT5A-STAT5B network in ALK-negative FE-PD cells. Western blot confirmed these findings. As expected, 50-100 nM crizotinib (ALK inhibitor) decreased Karpas 299 viability (p =0.016) but had no effect on FE-PD (Figure 2). In contrast, 50-100 nM of the experimental STAT5 inhibitor 573108 (EMD Millipore) decreased FE-PD viability (p=0.002) but had no effect on Karpas 299 (Figure 2). Conclusion The lymphoma proteome is complex with 10,000 proteins and contains druggable targets that can be reproducibly identified using SAX-LC-MS/MS. These targets vary across samples and integrated informatic analysis can predict target-drug combinations that have efficacy in an experimental model. These data suggest that SAX-LC-MS/MS could be used to personalize treatment regimens in lymphoma patients. Figure 1: Spectral counts of 1369 differentially expressed proteins between Karpas299 and FE-PD cell lines were normalized and plotted in a heat map. Red color indicates down regulation and yellow color indicates up regulation. Disclosures: No relevant conflicts of interest to declare.

Abstract Background Next Generation sequencing (NGS) is a powerful tool to identify somatic mutations associated with tumor onset and drug response. While it is well suited for high quality fresh/frozen samples, NGS is not proven for FFPE tissue which is the most common type of clinical specimen. Since the nucleic acids can be readily extracted from FFPE samples for a variety of genomic analyses, a comparative mutational analysis of paired frozen and FFPE tissues is urgently needed. Our long term goal is to establish a lab protocol to detect mutations in FFPE tumors using a targeted capture and sequencing approach for genes of interest. This pilot study focuses on the comparison of FFPE and frozen samples to test the validity of using FFPE tissues in such application. Methods Gene Selection: 128 genes associated with known pathogenic mutations in lymphoma Sample Selection: 9 diffuse large B-cell lymphoma (DLBCL) cases with FFPE, frozen and germline samples, as well as 10 frozen normal lymphatic tissues as references for CNV detections Capture Probe Design: We targeted coding exons and UTR, as well as the evolutionarily conserved intronic regions. The capture probes were designed using the Agilent eArray tool. The titling density of the probes was set to 3 probes overlapping with every base in the target region to improve the capture efficiency in FFPE samples. The least stringent masking of the repeat regions was allowed to include regions with small repeats that are shorter than the length of the sequencing reads (100-bp). In addition, boosting parameters were picked to set various levels of probe replication in different regions in order to minimize the local coverage differences (e.g. between regions of different GC contents) Sequencing and Bioinformatics: The target capture and sequencing were performed by the Mayo Clinic Medical Genome Facility. The reads were mapped to Human Reference Genome Build 37 using Novalign, and SNVs were called using GATK. The CNVs were identified using an in-house developed algorithm, patternCNV. Results The designed probes covered 99.65937% of the target regions. We generated 2.2-6.7 Gbp of reads per sample, 57.4-71.5% of which were on target. This equalled an average coverage of 2100-6700 folds which is 10-30 times higher than the minimal coverage recommended by Agilent. Due to this high coverage, we observed duplicate reads that accounted for 7.7-73.5% of the total reads. When we analysed the data with and without the duplicated reads, the concordance of the called SNVs was between 84-93% out of 207-249 mutated positions per trio-sample. There were 7.8-8.9% and 1.1-2.2% unique SNVs per sample by excluding or including duplicate reads, respectively. The dis-concordances were mostly missed calls, where a SNV was observed in only 1 or 2 of the trio samples. The missed calls from frozen samples ranged from 0-10.4% compared to 1.4-10.4% from the FFPE tissues, with 0.88-2.4% more SNVs missed in FFPE. Further analyses showed that all of the missing calls came from the lack of or low coverage of the corresponding positions. There were also differences of the called SNVs between the trio samples. However, this was extremely rare. Only 2 out of the 9 trio samples at a total of 3 positions had disagreements in called SNVs between FFPE and frozen tissues, all due to the allelic imbalance where the percentage of reads supporting the alternative alleles were below 20%. Therefore, this dis-concordance can be removed by back-filling of the read-level information for each position. Unfortunately only 11.9-47.4% of the CNVs called in frozen tissues were identified in FFPE samples, due to the widely various coverage in FFPE samples. The consequent large noises of the log ratio values between the FFPEs and normal references significantly reduced the sensitivity for CNV calling. Conclusions This pilot study compared the performance of SNV and CNV detection in FFPE and paired frozen tissues using a target capture and sequencing approach. With a capture probe design strategized to benefit FFPE samples, we observed SNV detection rates in FFPE that were only slightly lower (0.88-2.4%) than those of frozen tissues due to poor coverage of some positions in FFPE samples. With a proper back-filling step, there was no dis-concordance of the called SNVs between FFPE and frozen samples. However, CNV detections in FFPE were more problematic due to the un-predictable regional coverage in FFPE samples. Disclosures: No relevant conflicts of interest to declare.

Abstract Abstract 2705 Background: Due to similar cytomorphological features and treatment strategies, T lineage lymphomas (T-LBL) and lymphoblastic leukemias (T-ALL) are usually grouped together as a single entity. Traditionally, the difference between T-LBL and T-ALL has been based on percentage of blasts found in the bone marrow (BM): >25% BM blasts and/or presence of peripheral blood (PB) blasts consistent with T-ALL. Recent studies, based on immunophenotypic, cytogenetic and molecular differences, emphasize the clinical uniqueness between T-LBL and T-ALL and argue against grouping them as a single entity. Objectives: To perform a clinicopathologic and cytogenetic comparison between patients with T-LBL and T-ALL in order to identify differences between these disease groups. Methods: We utilized the Mayo Clinic Lymphoma database to identify consecutive adult patients (>18 years) diagnosed with T-cell leukemia and lymphoma from 1980 to 2010. The patients were sub-divided into T-LBL and T-ALL based on BM and PB blasts (absence of BM and PB blasts consistent with T-LBL). Clinicopathological and cytogenetic characteristics were retrospectively reviewed. Result: 50 patients met the criteria for T-cell leukemia/lymphoma (29-T-ALL, 21-T-LBL). 50% of patients in each group were females with median age at diagnosis of 32 years for T-LBL and 29 years for T-ALL. Characteristics associated with a statistically significant survival difference for the cohort included: white blood cell count at diagnosis (>100,000/cmm), platelet count (<50,000/cmm), LDH (>500 U/L), presence of BM blasts and an adverse karyotype (hypodiploidy and MLL rearrangements). ECOG performance status, number of lymph node groups involved, extranodal disease, CNS involvement and constitutional symptoms had no impact on survival. Furthermore, no overall survival difference between patients with T-LBL and T-ALL was identified (figure 1). Among patients with T-LBL, 11 (51%) patients with extranodal disease had shortened survival (p=0.02). CNS involvement was seen in 4 (19%) of T-LBL patients and 5 (17%) of T-ALL patients, and did not affect overall survival (p=0.8). 4 (19%) T-LBL patients and 1 (3%) T-ALL patient had pleural involvement at diagnosis. There were no morphologic, flow cytometric, or immunohistochemical differences noted between the two groups. Conclusion: In a large study from a single institution, we confirm no survival difference between T-LBL and T-ALL, which further provides ongoing support that from a therapeutic standpoint these groups can be viewed as a single entity. However, we hope to further clarify clinical differences by performing microarray and chemokine analysis on available tissue specimens. Disclosures: No relevant conflicts of interest to declare.

Abstract Tumors arise and evolve by iterative steps of mutation, subclonal selection, and clonal expansion due to growth advantage of the fittest subclones and external mutation induction and selection pressure from radiation and drug therapies. Various studies have shown that greater clonal complexities in primary tumors are correlated with poor clinical outcome of tumor progression and/or drug resistance. Recently, we have reported that very few of the driver mutations, all of which were clonal, detected in primary diffuse large B-cell lymphoma (DLBCL) were associated with DLBCL outcome as measured by 24-month event-free survival (EFS24). In the current project, we identified and studied subclonal mutations in primary DLBCL tumors and assessed their associations with EFS24, as well as the role of activation-induced cytidine deaminase (AID) in genomic mutations occurred in DLBCL tumors. The detection of subclonal mutations is still a significant bioinformatics challenge. In addition, identification of clonal mutations in samples with lower tumor purities faces similar challenges due to the low concentration of reads supporting the mutant alleles. The current methods of somatic mutation calling are not sufficiently sensitive to identify the low concentration mutations in tumor DNA sequencing data. We implemented a bioinformatics workflow for low-concentration and subclonal mutation detection which is based on the positional read pile-up data and a back-fill approach (PUB). The pile-up files were first generated using the re-aligned and re-calibrated BAM files after read alignment using Burrows-Wheeler Aligner (BWA). The variant positions with alternative bases were annotated by attributes defined in the Variant Quality Score Recalibration (VQSR). PUB then used a boosting method and a generalized linear model (GLM) to train a model of 'good quality' variants using common variants from HapMap, and prioritized and called clonal and subclonal variants based on the trained model. The VQSR attributes related to alternative allele depth were less-weighed in order to call subclonal mutations. The somatic mutations were then identified by Fisher's Exact Test using sequencing depths of the reference and alternative alleles from paired tumor and germline sequencing data. The exome sequencing data of paired tumor and peripheral blood from 48 newly diagnosed DLBCL patients were analyzed using PUB. Thirty-six of the 48 patients achieved EFS24 with the other 12 patients experienced primary treatment failure. Most of the tumors studied had tumor purities between 40-70%. PUB identified substantially higher number of somatic mutations, both clonal and subclonal, compared to those detected using existing somatic callers. We observed that the prevalence of mutations in previously reported driver genes were higher using thresholds of mutation concentration ≥ 5% and Fisher's Exact Test p ≤ 0.05, including EZH2 (mutated in 26% of DLBCL tumors analyzed by PUB, compared to 12.7% as previously reported), MLL2 (72% vs. 31%), CD79B (22% vs. 14%), TNFRSF14 (36% vs. 20%), MEF2B (22% vs. 16.4%), CARD11 (46% vs. 21.8%), and MYD88 (18% vs. 11%). In addition, other genes involved tumorigenesis that have not been previously linked to DLBCL also harbored both clonal and subclonal mutations with substantial prevalence, including FGFR3 (12%), KIT (24%), and ATM (28%). Furthermore, association of genes displaying clonal and subclonal mutations with EFS24 identified potential biomarkers for DLBCL outcome. Among these, two genes EPGN and ASTE are involved in epithelial growth factor receptor signaling and had association p values of 0.0001 and 0.0016, respectively. The oncogene MAFB was also associated with EFS24 (p= 0.0016). We searched for AID induced mutations among all identified variant positions and concluded that there was no evidence of AID site enrichments compared to a simulated data set. In summary, we developed and applied a sensitive bioinformatics pipeline for the identification of both clonal and low concentration somatic mutations in primary DLBCL exome sequencing data which further revealed the clonal complexity of the primary DLBCL tumors. Several of the genes were identified as potential biomarkers for DLBCL outcome. Disclosures Maurer: Kite Pharma: Research Funding. Ansell:Bristol-Myers Squibb: Research Funding; Celldex: Research Funding. Link:Genentech: Consultancy, Research Funding; Kite Pharma: Research Funding. Cerhan:Kite Pharma: Research Funding.

Abstract Background: CLL patients (pts) treated with novel targeted therapies continue to experience disease progression. Approximately 7-15% of relapsed CLL treated with ibrutinib or venetoclax developed Richter's syndrome (RS), an aggressive high grade lymphoma transformation. Standard chemotherapy has limited efficacy in RS. We hypothesized that checkpoint blockade would re-establish anti-tumor immune response in progressive CLL/RS and here we aim to update the clinical data and correlative analysis of MC1485: a phase 2 study of PD-1 blocking antibody pembrolizumab in relapsed/refractory CLL and RS. Methods: Relapsed/refractory CLL including RS enrolled in the CLL arm of MC1485 trial was reported here. The primary end point of this study is overall response rate (ORR). Pts with prior allogeneic stem cell transplant were excluded. Pembrolizumab 200 mg was administered intravenously every 3 weeks. NCI CTCAE v4.0 and IWCLL 2008 criteria were used for non-hematological adverse events (AE) and for CLL grade hematological AE. Tumor expression of PD-1 and PD-L1 were assessed with standard immunohistochemical (IHC) staining. Percentage expressions of individual antigens were analyzed with whole-slide scanning followed by image analysis with Image-Pro software (Media Cybernetics). Fluorescence in situ hybridization (FISH) was performed on available CLL/RS tissue sections to assess copy number of chromosome 9p region that contains PD-L1 and PD-L2. Results: 25 pts including 16 CLL and 9 RS (biopsy-proven large cell lymphoma) were enrolled in CLL arm. The median age was 69 years (46-81). Twelve (48%) pts had del(17p) or monosomy 17 or TP53 mutation. The median number of prior therapies was 4 (1-10). 96%, 72% or 48% of pts had received alkylator and anti-CD20 antibody, purine analog or anthracycline. Fifteen (60%) pts had prior ibrutinib and 12 (48%) progressed clinically while receiving ibrutinib, including 6 who progressed to RS and 6 who developed progressive CLL. The median number of pembrolizumab doses that pts received was 3 (1 to 21), administered over a median treatment duration of 11 weeks (1 to 56). Drug-related AE occurred in 21 pts (88%) with the most common ones being neutropenia in 9 (37%), cough in 7 (29%) and dyspnea in 6 (25%) pts. Drug-related grade 3 or above AE occurred in 9 (38%) pts with most common ones being thrombocytopenia (5, 21%), dyspnea (2, 8%), and fatigue (2, 8%). The most common immune-related AE was liver enzyme elevation (12%, 8% G3) and was reversible with therapy interruption or steroid therapy. Based on investigator assessment using the Revised Response Criteria for Lymphomas and IWCLL 2008 criteria, one RS pt had complete response (CR, 4%), 3 RS had partial responses (PR, 12%). Among 9 RS pts, 1 CR (11%), 3 PR (33%), 3 SD (33%) and 2 PD (22%) were observed. Thus the ORR in RS was 44%. No CLL pts achieved a CR/PR, 3 had SD and 9 had progressive disease (PD). The ORR of all pts was 16%. Of the 4 RS who had CR/PR, responses occurred early after 2 cycles of therapy and median duration of therapy for CR/PR pts was 8.3 months with data cut-off by June 30th 2016 (Figure 1A). The causes for therapy discontinuation in RS were PD (3, 33%), alternate therapy (2, 22%), and thrombocytopenia due to increased marrow CLL (1, 11%). Increased marrow CLL was observed in 2 RS pts with PR to RS phase of disease assessed by PET, thus the protocol was amended to allow addition of idelalisib/ibrutinib to control the underlying marrow CLL. For RS pts who had a CR/PR or SD, variable nodal responses were observed (Figure 1B). In particular, 6 RS who had prior ibrutinib experienced CR or PR or SD with nodal responses. After a median follow-up of 10.2 (1.9 to 16.1) months, the median overall survival (OS) for all pts was 10.7 months (95% confidence interval [CI], 5.4 to not reached). The 6 month OS rate for RS and CLL pts were 73% and 59%, respectively. Biomarker assessment using IHC analysis on 10 pts with available tumor/nodal tissues (6 RS and 4 CLL) showed an increased expression of PD-L1 (p = 0.02) and a trend for an increased expression of PD-1 (p = 0.1) in the group of pts with CR/ PR versus pts in the group with no clinical response. One out of 10 tested showed polysomic for chromosome 9p. Conclusion: Pembrolizumab had an acceptable safety profile in CLL and RS patients. We confirmed that pembrolizumab has substantial therapeutic activity in RS. Single-agent pembrolizumab does not appear to have clear activity in CLL. Disclosures Ding: Merck: Research Funding. Parikh:Pharmacyclics: Honoraria, Research Funding. Shanafelt:GlaxoSmithKline: Research Funding; Genentech: Research Funding; Janssen: Research Funding; Pharmacyclics: Research Funding; Celgene: Research Funding; Cephalon: Research Funding; Hospira: Research Funding.

Abstract Abstract 1567 Fusions of the tyrosine kinase domain of JAK2 with multiple partners occur in leukemia/lymphoma where they are believed to promote JAK2-oligomerization and autonomous signalling although the underlying mechanisms remain unclear. Affected entities are candidates for therapy with JAK2 signalling inhibitors. Among 200 peripheral T-cell lymphomas surveyed, we identified only two with JAK2 amplification and none with JAK2 translocations, confirming their rarity in T-cell neoplasias. Here we describe the genomic, transcriptional and signalling characteristics of PCM1-JAK2 formed by t(8;9)(p22;p24) in cell lines established at indolent and aggressive phases of a cutaneous T-cell lymphoma. To investigate signalling, PCM1-JAK2 was subjected to lentiviral knockdown which inhibited 7 genes most upregulated in t(8;9) cells, notably SOCS2/3. SOCS3, but not SOCS2, was also upregulated in a chronic eosinophilic leukemia bearing PCM1-JAK2. Conversely, expression of GATA3, a key T-cell developmental gene silenced in aggressive phase cells, was partially restored by PCM1-JAK2 knockdown. Activation of the tumor suppressor SOCS3 by PCM1-JAK2 may follow structural alteration of JAK2 affecting the ternary complex it forms with SOCS3 and receptor proteins. Treatment with JAK2 inhibitor (TG101348) to which MAC-1/2A/2B cells were conspicuously sensitive mimicked knockdown results, highlighting JAK2 as the active moiety. PCM1-JAK2 signalling required pSTAT5 as reported for JAK2V617F or ETV6/TEL-JAK2, thus extending the paradigm of STAT5 activation by JAK2 alterations in hematopoietic malignancies. MAC-1/2A/2B are the first JAK2– translocation cell line models. Our data support further investigation of SOCS2/3 as signalling effectors, prognostic indicators and potential therapeutic targets in cancers with JAK2 rearrangements. Disclosures: No relevant conflicts of interest to declare.

Abstract BACKGROUND: Diffuse large B-cell lymphoma, not otherwise specified (DLBCL), though defined as one disease entity, consists of two main cell-of-origin (COO) subtypes, germinal center B-cell like (GCB) and activated B-cell like (ABC), with the latter predicting a significantly worse prognosis using conventional chemotherapy. Up to 40% of all patients diagnosed with DLBCL will experience relapse, and outcomes are generally poor in this setting. In this study, we examine the proportion of GCB vs. non-GCB subtypes at initial diagnosis and at time of relapse, and assess the incidence of reclassification from one subtype to another. METHODS: Patients were prospectively enrolled in the University of Iowa/Mayo Clinic SPORE Molecular Epidemiology Resource (MER) within 9 months of diagnosis and followed for relapse, retreatment, and death. All relapse events were validated by review of the medical record. This analysis includes patients diagnosed with DLBCL who underwent initial treatment and subsequently experienced relapse as confirmed by biopsy. Immunohistochemical staining data including CD10, Bcl-6, and Mum-1, were collected from available pathology reports and classified to either GCB or non-GCB subtypes as predicted by the Hans algorithm. We compared subtype classification and individual immunohistochemical data both at the time of diagnosis and relapse. The overall subtype agreement between the two diagnostic time points was assessed. RESULTS: 1023 patients with newly diagnosed DLBCL were enrolled in the MER from 2002 - 2012. Of those, 249 had documented disease relapse. Of those analyzed, a total of 43 patients had COO classification based on the Hans algorithm from tissue biopsy at both diagnosis and relapse. Thirty patients were characterized as GCB DLBCL on initial presentation, and upon relapse 28 were noted to be GCB, while 2 were noted to be the non-GCB subtype. Of the 13 that were characterized as non-GCB on initial presentation, 10 remained non-GCB while 3 were reclassified as GCB. The overall agreement of DLBCL COO phenotypes between initial diagnosis and relapse was 88%. Seven percent of those initially diagnosed as GCB were reclassified as non-GCB subtype; in contrast, 23% of those initially diagnosed as non-GCB had changed to GCB upon relapse. Similar analysis was performed on individual immunohistochemical staining factors including CD10, Bcl-6, and Mum-1, and we observed overall agreement of 88%, 87%, and 91%, respectively. CONCLUSIONS: A majority of patients with relapsed DLBCL exhibit similar COO phenotypes at initial presentation and at relapse, and the incidence of reclassification from one subtype to another is uncommon. This suggests that the initial treatment regimen for DLBCL rarely alters the basic cancer phenotype when immunohistochemistry laboratory technical factors and pathologist interpretive discrepancies are excluded as causes for this change. However, reclassification from GCB to non-GCB subtypes is observed in a minority of cases, which may affect the overall outcome and response to treatment. Further studies are needed using gene expression to categorize COO and to examine the factors surrounding the time of relapse for possible harbingers of relapse as well as the behavior and outcome of reclassified DLBCL. Table Table. Disclosures Ansell: BMS, Seattle Genetics, Merck, Celldex and Affimed: Research Funding.