Precision Medicine for Relapsed Multiple Myeloma on the Basis of an Integrative Multiomics Approach

Purpose Multiple myeloma (MM) is a malignancy of plasma cells, with a median survival of 6 years. Despite recent therapeutic advancements, relapse remains mostly inevitable, and the disease is fatal in the majority of patients. A major challenge in the treatment of patients with relapsed MM is the timely identification of treatment options in a personalized manner. Current approaches in precision oncology aim at matching specific DNA mutations to drugs, but incorporation of genome-wide RNA profiles has not yet been clinically assessed. Methods We have developed a novel computational platform for precision medicine of relapsed and/or refractory MM on the basis of DNA and RNA sequencing. Our approach expands on the traditional DNA-based approaches by integrating somatic mutations and copy number alterations with RNA-based drug repurposing and pathway analysis. We tested our approach in a pilot precision medicine clinical trial with 64 patients with relapsed and/or refractory MM. Results We generated treatment recommendations in 63 of 64 patients. Twenty-six patients had treatment implemented, and 21 were assessable. Of these, 11 received a drug that was based on RNA findings, eight received a drug that was based on DNA, and two received a drug that was based on both RNA and DNA. Sixteen of the 21 evaluable patients had a clinical response (ie, reduction of disease marker ≥ 25%), giving a clinical benefit rate of 76% and an overall response rate of 66%, with five patients having ongoing responses at the end of the trial. The median duration of response was 131 days. Conclusion Our results show that a comprehensive sequencing approach can identify viable options in patients with relapsed and/or refractory myeloma, and they represent proof of principle of how RNA sequencing can contribute beyond DNA mutation analysis to the development of a reliable drug recommendation tool.


INTRODUCTION
Multiple myeloma (MM) is a mostly incurable malignancy of terminally differentiated plasma cells that affects 6.5 per 100,000 people per year in the United States, making it the second most common hematologic malignancy. 1 Typically, the trajectory of MM is characterized by a pattern of recurrent remissions and relapses, with patients becoming increasingly refractory to treatment. Hallmarks of MM include chromosomal translocations and copy number alterations (CNA). 2 However, the causal drivers of MM pathogenesis are still unclear, and treatment is administered empirically on the basis of recurrence risk rather than genetic events. High-throughput DNA sequencing of patients with MM has revealed wide and remarkable heterogeneity of the mutational spectrum across patients and a complex subclonal structure, 3,4 suggesting that the use of a personalized therapeutic approach is likely to improve the outcomes for myeloma. 5 In the past 4 years, we have focused on the design and development of a computational platform for personalized therapy of relapsed and/or refractory MM, on the basis of a comprehensive patient profile generated from DNA and RNA sequencing. Many institutions have now started precision medicine programs aimed at identifying viable therapeutic options for patients with cancer on the basis of specific Purpose Multiple myeloma (MM) is a malignancy of plasma cells, with a median survival of 6 years. Despite recent therapeutic advancements, relapse remains mostly inevitable, and the disease is fatal in the majority of patients. A major challenge in the treatment of patients with relapsed MM is the timely identification of treatment options in a personalized manner. Current approaches in precision oncology aim at matching specific DNA mutations to drugs, but incorporation of genome-wide RNA profiles has not yet been clinically assessed.
Methods We have developed a novel computational platform for precision medicine of relapsed and/or refractory MM on the basis of DNA and RNA sequencing. Our approach expands on the traditional DNA-based approaches by integrating somatic mutations and copy number alterations with RNA-based drug repurposing and pathway analysis. We tested our approach in a pilot precision medicine clinical trial with 64 patients with relapsed and/or refractory MM.

Results
We generated treatment recommendations in 63 of 64 patients. Twenty-six patients had treatment implemented, and 21 were assessable. Of these, 11 received a drug that was based on RNA findings, eight received a drug that was based on DNA, and two received a drug that was based on both RNA and DNA. Sixteen of the 21 evaluable patients had a clinical response (ie, reduction of disease marker ≥ 25%), giving a clinical benefit rate of 76% and an overall response rate of 66%, with five patients having ongoing responses at the end of the trial. The median duration of response was 131 days.

Conclusion
Our results show that a comprehensive sequencing approach can identify viable options in patients with relapsed and/or refractory myeloma, and they represent proof of principle of how RNA sequencing can contribute beyond DNA mutation analysis to the development of a reliable drug recommendation tool. targetable mutations. This is leading to a landmark paradigm shift in cancer therapy, in which treatment may be administered on the basis of the specific genomic alterations observed in a patient's tumor, rather than on the tumor histology or tissue type.
Our approach to precision medicine of relapsed MM critically incorporates RNA sequencingbased drug repurposing. In this article, we present our platform and the results obtained from a precision medicine clinical trial with 64 patients with relapsed MM seen at Mount Sinai. We show that our comprehensive approach can benefit patients beyond an approach that is based on mutations alone. Importantly, although our pipeline is designed and tailored for the specific needs of MM therapy, it can provide a general framework for incorporating RNA sequencingbased drug repurposing in oncology.

Protocol Approvals and Patient Enrollment
The patients were physician referred. Enrollment criteria, which included relapsed myeloma, lack of Food and Drug Administration (FDA)approved therapeutic options, and a prognosis of 6 months of survival, were approved by the Mount Sinai institutional review board (IRB). Informed written consent was obtained from each patient. As part of the IRB-approved genomics protocol, genomics findings were returned to the patient and treating physician in a standardized report to provide interpretative assistance. Patient enrollment started in February 2014 and ended in February 2016. The study was concluded in September 2017.

Sample Processing and Sequencing
Bone marrow (BM) aspirates and peripheral blood (PB) were obtained from patients with MM in the study. Tumor genomic DNA and RNA were obtained from CD138 + cells isolated from BM (Data Supplement). Normal genomic DNA (control) was obtained from granulocytes isolated from PB. Whole-exome sequencing (WES) and RNA sequencing libraries were submitted to Illumina HiSeq2500 for pairedend sequencing (100 base pairs). Targeted sequencing was performed using the Lymphoma Extended targeted next-generation sequencing panel from Cancer Genetic, Rutherford, NJ; Data Supplement). Raw sequencing data are available at National Center for Biotechnology Information Sequence Read Archive (accession number: PRJNA474747).

Data Processing and Bioinformatics Analysis
We designed and implemented a software framework for the definition and execution of data analysis workflows. The DNA workflow processes raw data from paired tumor and normal samples to detect, annotate, and prioritize somatic mutations and CNA and to identify actionable alterations. The RNA workflow processes RNA sequencing (RNAseq) data from tumor samples to identify outlier genes, determine pathway activation, and perform drug repurposing. Extended methods are provided in the Data Supplement.

Overview of the Study and Patient Characteristics
We developed a computational platform for personalized therapy of relapsed and/or refractory MM, on the basis of a comprehensive patient profile generated from DNA sequencing and RNAseq data (Fig 1). We evaluated the feasibility and effectiveness of our approach in an IRBapproved precision medicine clinical trial with 64 participants treated at Mount Sinai (See Methods and Table 1). The study included 39 men (61%) and 25 women (39%), and the median age was 59 years (range, 40 to 85 years). Forty-three patients (67%) had high-risk cytogenetics features such as t(4;14), t(14;16), del(13q), del(17p), and gain of 1q (Data Supplement). Overall, the patients had received a median of seven lines of therapy, with 13 patients (20%) having received > 10 lines of therapy.

Genomic Landscape of Patients With Relapsed MM and Actionable Findings
We obtained DNA data, either WES or targeted sequencing or both, for 55 patients (86%; Data Supplement). The pipeline identified a total of 21,166 somatic mutations in 10,403 genes in 54 of the 55 patients with DNA data available (Data Supplement). The mutational burden, quantified for the 41 patients with WES data and defined as the total number of mutations, ranged from 113 to 1,423, with an average of 527.5 mutations per patient (Fig 2A). Mutations were distributed among 14 different categories according to their nature and position in the genome, the majority of them being located in introns (57%; Fig 2B). Among the mutations with a potential pathogenic impact, missense mutations were the most numerous (17%), followed by nonsense, splice site, and start codon mutations (< 3%), for a total of 4,013 potentially pathogenic mutations in 3,163 genes (Data Supplement). We used the Clinical Interpretations of Variants in Cancer database (https://civicdb. org) to identify actionable mutations (ie, those associated with sensitivity to one or more drugs in one or more cancer types 6

Analysis of the MM Transcriptome for the Identification of Actionable Variations
RNAseq data were available for 60 patients (94%) and were used to determine outlier genes and pathway activation and to perform RNA drug repurposing (Data Supplement). We defined deregulated genes (over-or underexpressed) as actionable if they were reported as predictive of sensitivity to at least one drug in CIViC. We identified a total of 28 actionable genes, 17 over-and 11 underexpressed (Fig 3A; Data Supplement). We then calculated pathway activation by single-sample gene set variation analysis 7 on a set of actionable pathways relevant in MM: XBP1s activation, mammalian target of rapamycin signaling, histone deacetylase (HDAC), DNA repair, interleukin-6 signaling, PI3K/AKT activation, Hedgehog signaling, fibroblast growth factor receptor 3, and mitogenactivated protein kinase (MAPK). Figure 3B presents pathway activation scores calculated for 4 ascopubs.org/journal/po JCO™ Precision Oncology  (20) the 60 patients, whereas Figure 3C illustrates the distribution of activated pathways in the cohort. Each pathway is associated with specific targeted drugs (Data Supplement).
No. of Patients  the L1000CDS2 method, which is based on characteristic direction signatures. 8,9 The rationale behind this approach is that a drug inducing a gene expression profile that is opposite to a patient's profile might be able to revert the disease-associated signature and the phenotype. This approach has been demonstrated successfully in several published cases. [10][11][12] Figure 3D presents the distribution of FDA-approved cancer drugs selected by our analysis in the cohort. The Data Supplement summarizes FDA-approved noncancer drugs options that were selected for at least one patient.

Implemented Recommendations and Response to Therapy
Our pipeline generated recommendations for 63 of 64 patients (98% No. of Patients  Of the 26 treated patients, 21 were evaluable for response (81%). Of these, 11 (52%) received a drug on the basis of RNA profiling, eight (38%) on the basis of DNA profiling, and two (10%) on the basis of both RNA and DNA profiling. Five patients received our recommended drugs alone, whereas for 16 patients, the drugs were either added to previous treatment (n = 10) or administered in combination with other drugs on the basis of physician discretion (n = 6). The clinical benefit rate was 76% (minimal response or greater; Fig 4A). In particular, of the 21 evaluable patients, one (5%) achieved CR; three (14%), very good partial response; and 10 (47%), PR, to give an overall response rate of 66%. Two patients (10%) had minimal response, three (14%) had stable disease, and two (10%) had progressive disease. The median duration of response was 131 days (range, 37 to 372 days). Five patients had still ongoing responses at the end of the study (September 1, 2017), with response durations of 235, 182, 150, 172, and 99 days, respectively ( Fig 4B). Significant (≥ grade 3) nonhematologic toxicities were seen in five patients (24%); these included neutropenic fever, diarrhea, fatigue, and cardiomyopathy ( Table 2).  ISMMS36  ISMMS34  ISMMS56  ISMMS41  ISMMS05  ISMMS42-1  ISMMS23  ISMMS27  ISMMS40  ISMMS08  ISMMS09-1  ISMMS10  ISMMS21  ISMMS20  ISMMS43  ISMMS19  ISMMS01  ISMMS02  ISMMS25  ISMMS26  ISMMS07  - Table 2 for details). Gray bars indicate time receiving a different treatment outside of our recommendation. Red arrow indicates patient had an ongoing response at the end of the study.

DISCUSSION
Here, we reported our integrated multiomics approach for personalized therapy of MM and its application in a pilot precision medicine clinical trial with 64 patients with relapsed MM seen at Mount Sinai. As patients with MM progress through advanced disease and receive multiple lines of therapy, they are left with fewer and fewer treatment options. The results of our study show how a precision medicine approach incorporating RNA sequencing may identify viable and effective therapeutic options beyond the current FDA-approved armamentarium in MM. Our pipeline generated recommendations from a larger pool of FDA-approved cancer drugs (MM and non-MM) for 63 of 64 patients in the trial (98%; Fig 5).
Treatment was implemented in 40% of patients, and 81% of these were evaluable (Table 2).
Remarkably, 62% of the evaluable patients received a drug on the basis of RNA profiling. Two drugs that we repurposed successfully from other cancers in our study were trametinib and venetoclax, the former selected on the basis of DNA and the latter on the basis of RNA findings. Trametinib is a MEK1/2 inhibitor approved by the FDA in combination with dabrafenib for unresectable or metastatic melanoma and non-small-cell lung cancer carrying mutations in BRAF (V600E/V600K). Recent studies have shown activating mutations in NRAS, KRAS, and BRAF in MM, making the MAPK pathway a significant therapeutic target also in MM. [3][4][5] A retrospective study has demonstrated clinical activity of trametinib in patients with MM with RAS-mutated tumors. 14 Eleven of 21 evaluable patients in our study received trametinib, either alone or in combination with other drugs. Six of them had clinical response, with a median progression-free survival of 110 days, and two patients had ongoing responses at the end of the study. Venetoclax is a BH3 mimetic that acts as a Bcl-2 inhibitor and is approved only for patients with chronic lymphocytic leukemia carrying deletion of 17p. 15 Phase I clinical trials reported the efficacy of venetoclax in relapsed and/or refractory MM, where patients with high BCL2 expression had a higher ORR than did patients with low BCL2 expression. [16][17][18] Our pipeline selected venetoclax for eight patients on the basis of BCL2 expression, and all of them showed clinical response, with a median progression-free survival of 161 days, and five patients had ongoing responses at the end of the study. Overall, these results compare favorably to the efficacy seen in several novel agents for MM, including pomalidomide, daratumumab, and elotumumab, as well as targeted precision medicine approaches such as the NCI-MATCH (National Cancer Institute Molecular Analysis for Therapy Choice) trial, in which patients with specific solid tumors received treatment on the basis of actionable DNA mutations. 19 Our trial also identified the challenges in implementing NGS-based recommendations in a real-world setting. In 60% of the patients with recommendations, treatment was not implemented, either because of insurance denial of the drug or because of rapid progression of disease before sequencing results were available. Insurance denial represents a significant limitation in the implementation of personalized therapy that is based on genomic findings. Our results suggest a need for specialty pharmacies and insurers to evolve with the technology to optimally help patients.
The second limiting factor in our study was the time required for sequencing. The average turnaround time from sample collection to sequencing data was 6 weeks. However, most patients with relapsing MM experience rapid progression and need immediate treatment. To address this, we are now using rapid-run sequencing, which provides results in a few days. This may be more expensive initially but may become cost effective with greater use.
Of the 21 treated patients who were evaluable, five did not respond to the recommended therapy. Multiple factors should be considered to improve the accuracy of treatment prediction.
One key factor is clonal heterogeneity. MM is characterized by a branching pattern of clonal evolution, in which distinct subclonal tumor populations may evolve by selective pressure of therapy. 5,20,21 We have recently started to investigate the impact of clonal heterogeneity on therapy selection, by extending our analysis to include clonality assessment on the basis of WES and/or single-cell RNA sequencing (scRNAseq). Integration of both WES and scRNAseq may provide a more comprehensive profile of a patient's tumor, enabling drug repurposing at the subclonal level.
Another key factor in optimal therapy prediction is the BM microenvironment. MM cells are strongly dependent on the surrounding microenvironment, which promotes their homing, growth, survival, and migration, as well as their resistance to drugs. 22 Therapies targeting both the cellular and the noncellular BM compartments are available and are currently used in MM, although not in a targeted manner. Profiling the microenvironment by scRNAseq and/ or mass cytometry, which allows simultaneous measurements of up to 50 markers at single-cell resolution, may help dissect the nontumor compartment and better inform targeted therapy. We are currently investigating the feasibility of including this feature in the next generation of our platform. In particular, the use of scRNAseq in clinical precision oncology is limited by several challenges, including increased cost compared with bulk sequencing (approximately 10 times higher), high sensitivity to sample purity, and dropouts (ie, undetected transcripts due to low amounts of RNA sequenced within individual cells), as well as data processing and interpretation. In fact, the analysis of scRNAseq data requires the use of specialized tools, which are still in their development stage. Nevertheless, scRNAseq represents a promising and powerful technology that will likely complement current precision medicine strategies in the near future.
A third key factor to consider concerns the accuracy of the predictions. This involves sequencing data, reference data in the knowledge base, and the algorithms used. We are working on extending our knowledge base to incorporate MMspecific data, including transcriptional effects of novel agents, to quickly and effectively translate robust research findings into clinical practice. To further improve drug recommendations, we are implementing validation of the in silico findings using both in vitro (micro-C3) and in vivo (PDX) models. 23,24 This will also contribute to reducing both the costs and the toxicities associated with ineffective treatments. Finally, we are going to equip our platform with a machine learning flow, which will implement interactive learning techniques to refine the predictions on the basis of therapy outcome and physician's opinion.
In conclusion, here we have described our sequencing-based precision medicine platform for relapsed and/or refractory MM and reported the results of a pilot clinical trial to assess the feasibility and usefulness of this approach. The trial has allowed us to test and reveal the accuracy of our platform and to understand the pitfalls and limitations of the approach, laying the foundation for our next-generation precision medicine framework. Overall, the results demonstrate feasibility and early efficacy, providing a basis for expanding NGS-guided personalized therapy integrating RNA-and DNA-based drug repurposing for patients with cancer.