Aberrant chromosomal translocations are seen in ~40% of presenting patients and predominantly involve the IGH locus at 14q32. The five main translocations involving the IGH locus are t(4;14), t(6;14), t(11;14), t(14;16) and t(14;20), which result in over-expression of MMSET/FGFR3, CCND3, CCND1, MAF and MAFB, respectively. In previous clinical trials we have shown that the t(4;14), t(14;16) and t(14;20) are associated with a poor prognosis. In initial sequencing studies of myeloma it has been noted that the spectrum of mutations fall into two groups, one of which is characterised by an APOBEC signature. This signature comprises of C>T, C>G and C>A mutations in a TpC context and comprises only a subset of samples, with the rest having a rather generic mutation signature representing an intrinsic mutational process occurring as a result of the spontaneous deamination of methylated cytosine to thymine. Whole exome sequencing was performed on 463 presentation patients enrolled into the UK Myeloma XI trial. DNA was extracted from germline DNA and CD138+ plasma cells and whole exome sequencing was performed using SureSelect (Agilent). In addition to capturing the exome, extra baits were added covering the IGH, IGK, IGL and MYCloci in order to determine the breakpoints associated with translocations in these genes. Tumor and germline DNA were sequenced to a median of 60x and data processed to generate copy number, acquired variants and translocation breakpoints in the tumor. Progression-free and overall survival was measured from initial randomization and median follow up for this analysis was 25 months. These combined data allow us to examine the effect of translocations on the mutational spectra in myeloma and determine any associations with progression-free or overall survival. Translocations were detected in 232 (50.1%) patients of which 59 patients (12.7%) had a t(4;14), 86 patients (18.6%) a t(11;14), 17 patients (3.7%) a t(14;16), 5 patients (1%) a t(6;14) and 4 patients (0.9%) a t(14;20). MYC translocations were found in 85 patients (18.4%). Using the tiled regions we were able to detect a mutational signature, kataegis, where regional clustering of mutations can be indicative of somatic genomic rearrangements. We found the hallmarks of kataegis in 15 samples (3.2%), where there was enrichment for TpCpH mutations with an inter-mutational distance <1 kb. Where we detected kataegis surrounding MYC there was also an inter-chromosomal translocation involving either IGK or IGL. Interestingly, the partner chromosomes also showed signs of kataegis e.g. in the t(2;8) kataegis was found at IGK and MYC and in the t(8;22) kataegis was found at MYC and IGL. APOBECs are thought to be involved in the generation of kataegis and as such this co-localisation is indicative of APOBEC involvement in the generation of MYCbreakpoints. We found mutation of translocation partner oncogenes, in particular CCND1 was mutated in 10 patients with the t(11;14). There was an association of mutated CCND1 with a poor prognosis when compared with non-mutated t(11;14) patients (OS median of 20.2 months vs. not reached, p=0.005). Mutations were also seen in FGFR3, MAF and MAFB but only in the samples with the respective translocations. The mutations are likely due to somatic hypermutation mediated by AID, an APOBEC family member. We found that t(14;16) and t(14;20) samples have a significantly higher number of mutations compare to the other translocation groups (p=1.65x10-5). These mutations were enriched for those with an APOBEC signature (T(C>T)A, p=9.1x10-5; T(C>T)T, p=0.0014; T(C>G)A, p=0.001; T(C>G)T, p=0.0064), indicating that the ‘maf’ translocation groups are characterized by APOBEC signature mutations, specifically APOBEC3B. When samples are assigned to either an APOBEC or non-APOBEC group the ‘maf’ translocations account for 66.6% of samples in the APOBEC group but only 1.3% of the non-APOBEC group. Here we show three different mutational signatures mediated by the APOBEC family: translocation partner mutation by AID, APOBEC signature mediated by APOBEC3B, and kataegis mediated by an unknown APOBEC family member. We also show for the first time a clinical impact of APOBEC mutations and their association with a poor prognosis. The poor prognosis of this mutational signature is inextricably linked to a high mutation load and the adverse t(14;16) and t(14;20) translocation subgroups. Disclosures Walker: Onyx Pharmaceuticals: Consultancy, Honoraria.

Introduction The iliac crest is the usual sampling site for minimal residual disease (MRD) monitoring in Multiple Myeloma (MM). However, the disease distribution in the bone marrow (BM) is often heterogeneous. Functional imaging can be used to complement MRD detection at a single site, thereby accounting for asymmetrically distributed disease. Diffusion weighted MRI with background suppression (DWIBS) is a novel functional imaging method that can detect disease in a higher proportion of newly diagnosed MM (NDMM) patients than 18F-fluorodeoxyglucose positron emission tomography (PET), as it is independent of the tumor metabolism. Yet, its performance for monitoring of residual disease has not been described. The aims of this study were 1) to compare DWIBS to PET for the detection of residual disease in patients achieving complete remission (CR), and 2) to test whether DWIBS and PET could complement MRD flow cytometry with a sensitivity of 1x10-5. To address these aims, we investigated 168 NDMM and 33 relapsed patients for whom DWIBS, PET, and MRD were available at the onset of CR during first-line and salvage therapy, respectively. Methods All patients signed written consent in accordance with the Declaration of Helsinki. Residual focal lesions (FLs) were defined as well delineated focal intensities above the surrounding BM background. For DWIBS FLs were considered if restriction could be confirmed on ADC maps. 8-color MRD flow cytometry with a limit of detection of 1x10-5 was available for 83 NDMM and all 33 salvage therapy patients. The Kaplan-Meier method was used for survival analyses. PFS time was measured from onset of CR to relapse or death from any cause or censored at the date of last contact. Paired-end whole exome sequencing of CD138-enriched MM cells was performed on an Illumina HiSeq 2500. Mutations were called from BWA aligned sequencing reads using MuTect. Subclonal reconstruction was done using SciClone. Results Compared to PET, DWIBS detected more CR patients with residual FLs (21% vs. 6%), and the concordance between PET and DWIBS was low. Only 6 of the DWIBS-positive patients also presented with FLs in PET. Yet, 5 patients had PET+/DWIBS- FLs, suggesting that the two techniques are complementary. Both, DWIBS+ and PET+ FLs negatively impacted PFS (p<0.05). For 83 patients MRD data were available. Combining MRD and imaging, residual disease was detectable in 53 patients (64%). The best outcome was seen for 30 double negative (MRD-/Imaging-) patients (3 events with a median follow-up of 3.6 years), the worst outcome was seen for 10 double positive (MRD+/Imaging+) patients (median PFS: 2.1 years). Only 4 of 86 patients were MRD-/Imaging+, indicating that residual FLs are rare in MRD-negative NDMM patients at a sensitivity of 1x10-5. A heterogeneous disease distribution is a common feature of late-stage patients. To test if this increased heterogeneity confounded MRD, we investigated a set of 33 heavily pretreated patients who achieved CR during salvage therapy. Combining MRD and imaging data, we detected residual disease in 25 patients (76%). Of note, the proportion of patients, who were MRD-negative but had residual FLs on functional imaging was significantly higher compared to NDMM (8/16 vs 4/34 patients, p=0.01). At the same time, 10 patients (30%) were MRD+ but Imaging-, supporting the idea that a combined MRD/Imaging approach can improve detection of residual disease and should be used in late-stage patients. To obtain insights in the underlying biology, we performed longitudinal multi-region sequencing of a subset of these CR patients. Our findings support the concept of persistence and progression of multiple spatially separated clones in the BM irrespective of being in an MRD-negative CR. Thereby, focal residual disease could be shown to contribute to relapse. Conclusion DWIBS is a promising tool for detection of residual disease and complements PET. The combination of MRD diagnostics and functional imaging improves prediction of outcome, with double-negativity and double positivity defining groups with excellent and dismal PFS, respectively. Prospective trials using this information to tailor therapy are warranted. From a biological perspective, this study highlights the confounding effects of spatial heterogeneity and limited dissemination of clones within the BM on MRD diagnostics. This may especially be true for patients achieving deep responses during salvage therapies. Disclosures Roy Choudhury: University of Arkansas for Medical Sciences: Employment, Research Funding. Epstein:University of Arkansas for Medical Sciences: Employment. Barlogie:International Workshop on Waldenström's Macroglobulinemia: Other: travel stipend; Millenium: Consultancy, Research Funding; Multiple Myeloma Research Foundation: Other: travel stipend; ComtecMed- World Congress on Controversies in Hematology: Other: travel stipend; Myeloma Health, LLC: Patents & Royalties: : Co-inventor of patents and patent applications related to use of GEP in cancer medicine licensed to Myeloma Health, LLC; European School of Haematology- International Conference on Multiple Myeloma: Other: travel stipend; Celgene: Consultancy, Research Funding; Dana Farber Cancer Institute: Other: travel stipend. Davies:Takeda: Consultancy, Honoraria; Abbvie: Consultancy, Honoraria; Amgen: Consultancy, Honoraria; Celgene: Consultancy, Honoraria; Janssen: Consultancy, Honoraria. Morgan:Celgene: Consultancy, Honoraria, Research Funding; Bristol-Myers Squibb: Consultancy, Honoraria; Takeda: Consultancy, Honoraria; Janssen: Research Funding.

Introduction Chromosome instability (CIN) is a driver of copy number aberrations (CNAs) in cancer, and is a major factor leading to tumor heterogeneity and resistance to therapy. By definition, CIN is an increased rate or ongoing acquisition and accumulation of CNAs and not simply the existence of structurally and numerically abnormal aneuploid clones. In multiple myeloma (MM), the most common whole-chromosome CNAs involve either hyperdiploid or non-hyperdiploid clones. Secondary segmental CNAs are associated with high-risk (HR) in MM and involve gains of 1q21 and deletions of 17p (del17p). These types of intra-chromosomal segmental CNAs are also found in the CIN phenotypes of the autosomal recessive (AR) chromosome instability syndromes. These syndromes include Fanconi anemia, Bloom's syndrome, and ICF syndrome (Immunodeficiency, Centromeric instability and Facial anomalies). These chromosome instability syndromes display a spectrum of aberrations characterized by higher rates of chromosomal breaks, chromatid exchanges, quadriradials, and pericentromeric aberrations. In particular, patients with ICF syndrome show a marked increase of 1q12 pericentromeric instability including 1q12 decondensation, triradials, multibranched chromosomes 1q, and 1q micronuclei. ICF patients also show transient 1q aberrations including isochromosome 1q (iso1q) and unbalanced translocations of 1q to 9q and 16q. In MM, we have previously reported increasing pericentromeric instability during tumor progression resulting in increasing CNAs of 1q21 by unbalanced jumping translocations of 1q12 (JT1q12). Strikingly, in a subset of MM patients with 1q21 CNAs of ≥ 5 a distinct cytogenetic phenotype emerges which demonstrates transient 1q12 aberrations including 1q12 decondensation, triradials, and multibranched chromosomes 1q morphologically identical to those seen in ICF patients. In MM this chromosome instability leads to a cascade of increasing clonal 1q21 duplications, iso 1qs, and unbalanced 1q translocations with 16q and 17p, resulting in losses in these receptor chromosomes (RC) and massive intra-clonal CNA heterogeneity. Methods To investigate the cytogenetic impact and progression of high CNAs of 1q21, we performed a comprehensive metaphase analysis of 50 patients showing segmental aneuploidies with 4 or more copies of 1q by G-banding. Locus specific FISH and spectral karyotyping were used to identify the key transient unstable and clonal structural aberrations of 1q12 resulting in segmental aneuploidies in the derivative RCs. Probe for 1q12 (Vysis) was used according to the manufacturer's protocol. Locus specific BAC clones for 1q21 (CKS1B) and 17p (TP53) were prepared and analyzed as previously described (Sawyer et al., Blood 123: 2014). IGH translocations were investigated with IGH break apart probes (Vysis). Results Data for 50 patients including CNAs of 1q21 of ≥ 4, IGH translocations, del(17p), derivative RCs, are presented. The t(4;14) was found in 15 patients, del(17p) in 23, and both aberrations were found in 8 patients. All patients showed unbalanced gains of 1q and deletions of RCs, the most frequent being 7 patients with der(1;16) and 6 with iso1q. In four of the 23 patients with del(17p), the deletion was due to a JT1q12 to 17p. Seven patients with 1q21 CNAs of ≥ 5 showed profound instability involving the 1q12 satellite DNA, demonstrating both transient and clonal aberrations driving the 1q21 CNAs. These aberrations included unstable 1q21 triplications, JT1q12s, iso1q formation with intra-arm 1q12 CNAs, and region specific breakage-fusion-bridge cycle amplifications. Conclusions Among patients with ≥ 5 CNAs of 1q21, a subset develop an acquired HR chromosome instability phenotype with an elevated rate of 1q12 pericentromeric instability characterized by concomitant deletions in 16q, iso1q, del(17p), and intra-arm segmental instability. These patients show pronounced instability in the 1q12 satellite DNA, morphologically identical to ICF syndrome, suggesting hypomethylation of this region as a driver of both 1q21 CNAs and deletions in RCs. We hypothesize that region specific hypomethylation of 1q12 provides the genomic background for the onset of an acquired 1q12 chromosome instability phenotype in MM similar to that found in ICF syndrome. For myeloma patients demonstrating this 1q12 chromosome instability phenotype we propose the term "jumping 1q syndrome." Disclosures Epstein: University of Arkansas for Medical Sciences: Employment. Davies:Amgen: Consultancy, Membership on an entity's Board of Directors or advisory committees; ASH: Honoraria; Abbvie: Consultancy; TRM Oncology: Honoraria; Celgene: Consultancy, Honoraria, Membership on an entity's Board of Directors or advisory committees; Takeda: Consultancy, Membership on an entity's Board of Directors or advisory committees; MMRF: Honoraria; Janssen: Consultancy, Honoraria. Morgan:Takeda: Consultancy, Honoraria; Celgene: Consultancy, Honoraria, Research Funding; Janssen: Research Funding; Bristol-Myers Squibb: Consultancy, Honoraria.

The genetic basis underlying the inherited risk of developing multiple myeloma (MM) is largely unknown. To examine the impact of rare protein altering variants on the risk of developing MM we analyzed high-coverage exome sequencing data on 513 MM cases and 1,569 healthy controls, performing both single variant and gene burden tests. We did not identify any recurrent coding low-frequency alleles (1–5%) with moderate effect that were statistically associated with MM. In a gene burden analysis we did however identify a promising relationship between variation in the marrow kinetochore microtubule stromal gene KIF18A, which plays a role in control mitotic chromosome positioning dynamics, and risk of MM (P =3.6×10−6). Further analysis showed KIF18A displays a distinct pattern of expression across molecular subgroups of MM as well as being associated with patient survival. Our results inform future study design and provide a resource for contextualizing the impact of candidate MM susceptibility genes.

932 Introduction Multiple myeloma (MM) is a heterogeneous disease. The discovery of a class of small non-coding RNAs (miRNAs) has revealed a new level of biological complexity underlying the regulation of gene expression. It may be possible to use this interesting new biology to improve our ability to risk stratify patients in the clinic. Methods We performed global miRNA expression profiling analysis of 163 primary tumors included in the UK Myeloma IX clinical trial. miRNA expression profiling was carried out using Affymetrix GeneChip miRNA 2.0; expression values for 847 hsa-miRNAs were extracted using Affymetrix miRNA QC tool and RMA-normalized. There are also 153 matching samples with gene expression profiles (GEP) and 72 matching cases with genotyping data available for integrative analyses. GEP was generated on Affymetrix HG-U133 Plus 2.0 and the expression values were RMA normalized; genotyping was performed on Affymetrix GeneChip Mapping 500K Array and the copy number values were obtained using GTYPE and dChip and were inferred against normal germ-line counterpart for each sample. Results Firstly we have defined 8 miRNAs linked to 3 Translocation Cyclin D (TC) subtypes of MM with distinct prognoses, including miR-99b/let-7e/miR-125a upregulation and miR-150/miR-155/miR-34a upregulation in unfavourable 4p16 and MAF cases respectively as well as miR-1275 upregulation and miR-138 downregulation in favourable 11q13 cases. The expression levels of the miRNA cluster miR-99b/let-7e/miR-125a at 13q13 have been shown to be associated with shorter progression free survival in our dataset. Interestingly unsupervised hierarchical clustering analysis using these 8 miRNAs identified two subclusters among 11q13 cases, which have differential effect on overall survival (OS). We then evaluated the association of miRNA expression with OS and identified 3 significantly associated miRNAs (miR-17, miR-18 and miR-886-5p) after multiple testing corrections, either per se or in concerted fashion. We went on to develop an “outcome classifier” based on the expression of two miRNAs (miR-17 and miR-886-5p), which is able to stratify patients into three risk groups (median OS 19.4 months vs 40.6 months vs 65.3 months, log-rank test p = 0.001). The robustness of the miRNA-based classifier has been validated using 1000 bootstrap replications with an estimated error rate of 1.6%. The miRNA-stratified risk groups are independent from main adverse fluorescence in situ hybridization (FISH) abnormalities (1q gain, 17p deletion and t(4;14)), International Staging System (ISS) and Myeloma IX treatment arm (intensive or non-intensive). Using the miRNA-based classifier in the context of ISS/FISH risk stratification showed that it can significantly improves the predictive power (likelihood-ratio test p = 0.0005) and this classifier is also independent from GEP-derived prognostic signatures including UAM, IFM and Myeloma IX 6-gene signature (p < 0.002). Integrative analyses didn't show enough evidence that the miRNAs comprising the classifier were deregulated via copy number changes; however, our data supported that the mir-17-92 cluster was activated by Myc and E2F3, highlighting the potential importance of Myc/E2F/miR-17-92 negative feedback loop in myeloma pathogenesis. We developed an approach to identify the putative targets of the OS-associated miRNAs and show that they regulate a large number of genes involved in MM biology such as proliferation, apoptosis, angiogenesis and drug resistance. Conclusion In this study we developed a simple miRNA-based classifier to stratify patients into three risk groups, which is independent from current prognostic approaches in MM such as ISS, FISH abnormalities and GEP-derived signatures. The miRNAs comprising the classifier are biologically relevant and have been shown to regulate a large number of genes involved in MM biology. This is the first report to show that miRNAs can be built into molecular diagnostic strategies for risk stratification in MM. Disclosures: No relevant conflicts of interest to declare.