A Niche-Dependent Myeloid Transcriptome Signature Deﬁnes Dormant Myeloma Cells

SHEILA STEWART 9/1/2020 JOURNAL CLUB

Regular Article

LYMPHOID NEOPLASIA A niche-dependent myeloid transcriptome signature deﬁnes dormant myeloma cells

Weng Hua Khoo,1,2 Guy Ledergor,3,4 Assaf Weiner,3 Daniel L. Roden,5 Rachael L. Terry,1 Michelle M. McDonald,1,6 Ryan C. Chai,1 Downloaded from https://ashpublications.org/blood/article-pdf/134/1/30/1553831/blood880930.pdf by WASHINGTON UNIVERSITY SCHOOL user on 28 August 2020 Kim De Veirman,7 Katie L. Owen,8 Khatora S. Opperman,9,10 Kate Vandyke,9,10 Justine R. Clark,9,10 Anja Seckinger,11 Natasa Kovacic,12 Akira Nguyen,6,13 Sindhu T. Mohanty,1 Jessica A. Pettitt,1 Ya Xiao,1 Alexander P. Corr,1,6 Christine Seeliger,1 Mark Novotny,14 Roger S. Lasken,14 Tuan V. Nguyen,1,6,15 Babatunde O. Oyajobi,16,17 Dana Aftab,18 Alexander Swarbrick,5,6 Belinda Parker,8 Duncan R. Hewett,9,10 Dirk Hose,11 Karin Vanderkerken,7 Andrew C. W. Zannettino,9,10 Ido Amit,3 Tri Giang Phan,6,13,* and Peter I. Croucher1,2,6,*

1Division of Bone Biology, Garvan Institute of Medical Research, Sydney, NSW, Australia; 2School of Biotechnology and Biomolecular Sciences, University of New South Wales (UNSW) Sydney, Sydney, NSW, Australia; 3Department of Immunology, Weizmann Institute of Science, Rehovot, Israel; 4Department of Internal Medicine “T”, Tel Aviv Sourasky Medical Center, Tel Aviv, Israel; 5Cancer Division, Garvan Institute of Medical Research, Sydney, NSW, Australia; 6St Vincent’s Clinical School, Faculty of Medicine, UNSW Sydney, Sydney, NSW, Australia; 7Department of Hematology and Immunology, Vrije Universiteit Brussel, Brussels, Belgium; 8Department of Biochemistry and Genetics, La Trobe University, Melbourne, VIC, Australia; 9Faculty of Health and Medical Sciences, The University of Adelaide, Adelaide, SA, Australia; 10Cancer Theme, South Australian Health and Medical Research Institute, Adelaide, SA, Australia; 11Labor fur¨ Myelomforschung, Medizinische Klinik V, Universitatsklinikum¨ Heidelberg, Ruprecht-Karls-Universitat¨ Heidelberg, Heidelberg, Germany; 12Department of Anatomy and Clinical Anatomy, University of Zagreb School of Medicine, Zagreb, Croatia; 13Immunology Division, Garvan Institute of Medical Research, Sydney, NSW, Australia; 14J. Craig Venter Institute, La Jolla, CA; 15School of Biomedical Engineering, University of Technology, Sydney, NSW, Australia; 16Department of Cell Systems and Anatomy, Long School of Medicine, and 17Mays Cancer Center, University of Texas Health Science Center, San Antonio, TX; and 18Exelixis Inc, Alameda, CA

KEY POINTS The era of targeted therapies has seen significant improvements in depth of response, progression-free survival, and overall survival for patients with multiple myeloma. Despite l Dormant multiple myeloma cells these improvements in clinical outcome, patients inevitably relapse and require further express a unique treatment. Drug-resistant dormant myeloma cells that reside in specific niches within the transcriptome signature. skeleton are considered a basis of disease relapse but remain elusive and difficult to study. Here, we developed a method to sequence the transcriptome of individual dormant l This unique transcriptome signature myeloma cells from the bones of tumor-bearing mice. Our analyses show that dormant controls dormancy, myeloma cells express a distinct transcriptome signature enriched for immune genes and, predicts survival, and unexpectedly, genes associated with myeloid cell differentiation. These genes were identifies new treatment targets. switched on by coculture with osteoblastic cells. Targeting AXL, a gene highly expressed by dormant cells, using small-molecule inhibitors released cells from dormancy and pro- moted their proliferation. Analysis of the expression of AXL and coregulated genes in human cohorts showed that healthy human controls and patients with monoclonal gammopathy of uncertain signif- icance expressed higher levels of the dormancy signature genes than patients with multiple myeloma. Furthermore, in patients with multiple myeloma, the expression of this myeloid transcriptome signature translated into a twofold increase in overall survival, indicating that this dormancy signature may be a marker of disease progression. Thus, engagement of myeloma cells with the osteoblastic niche induces expression of a suite of myeloid genes that predicts disease progression and that comprises potential drug targets to eradicate dormant myeloma cells. (Blood. 2019; 134(1):30-43)

Introduction cytotoxic chemotherapy,6,7 posing an obstacle to cure. Despite the importance of this problem, the nature of dormant MM Multiple myeloma (MM) is a B-cell neoplasm where malignant cells and the factors that govern their behavior have eluded plasma cells disseminate to the skeleton and cause disease.1,2 the ﬁeld. The development of intravital imaging4,8,9 and single- Therapeutic advances have improved complete response rates cell technologies,10 including single-cell RNA sequencing and support the notion that MM may be curable.3 However, (scRNAseq),11 now makes the identiﬁcation and analysis of rare despite the success of targeted therapies, relapse is common. dormant cancer cells possible.12 One mechanism that contributes to relapse is the reactivation of MM cells that reside in a dormant state, within specialized niches in the skeleton.4,5 These long-term, niche-resident Dye-retention strategies, where membrane dyes are shared dormant MM cells evade the immune system and resist with daughter cells during cell division, can mark long-term,

30 blood® 4 JULY 2019 | VOLUME 134, NUMBER 1 © 2019 by The American Society of Hematology nondividing, dormant MM cells.4 When combined with longi- Differential gene expression analysis tudinal intravital imaging, these methods can be used to map the Differential gene expression analysis was performed using fate of dormant cells at single-cell resolution.4 Using these BASiCS19,20 or Seurat.21 approaches, we have shown that MM cells seed the endosteal bone surface4 and interact with bone-lining cells, which Functional annotation and clustering analysis 4,13 induce MM cells to enter a dormant state. Alterations to the Functional annotation was performed using the Database for endosteal bone niche, caused by osteoclasts remodeling the Annotation, Visualization and Integrated Discovery.22,23 Differ- bone surface, displace MM cells from the niche. This releases entially expressed genes identiﬁed by SMART-seq were clus- them from cellular dormancy and causes their selective reac- tered by nonnegative matrix factorization in R.24 t-distributed tivation.4 Thus, MM cell extrinsic factors that alter the stochastic neighbor embedding dimensionality reduction algo- dormant MM cell niche can release cells from dormancy. rithms were performed on the MARS-seq data using Seurat.21 However, our understanding of how the endosteal bone niche Downloaded from https://ashpublications.org/blood/article-pdf/134/1/30/1553831/blood880930.pdf by WASHINGTON UNIVERSITY SCHOOL user on 28 August 2020 reprograms colonizing MM cells to become dormant is limited. Transcription factor prediction analysis We hypothesized that unbiased mapping of the transcriptional Transcription factor prediction analysis was performed on 1492 control programs expressed by dormant cells would identify the differentially expressed genes identiﬁed by SMART-seq using functional pathways that control MM cellular dormancy. iRegulon25 in Cytoscape.26 Networks were visualized using Gephi.27

To address this, we developed a strategy to define the molecular Validation of myeloid markers signature of dormant MM cells in models of dormancy and 5TGM1-eGFP1 myeloma cells were isolated from disease-bearing validate signature genes in patients with MM. We used 2 mice and stained for myeloid markers for index sorting and vali- complementary scRNAseq technologies: SMART-seq for full- dation of myeloid gene expression. Data were acquired and an- length transcriptome profiling and massively parallel scRNA- alyzed using BD Symphony (BD Biosciences). In separate seq (MARS-seq) for high-throughput analysis. We showed that experiments, individual eGFP1DiDhi or eGFP1DiDneg myeloma dormant MM cells express a distinct transcriptome signature cells were index sorted into 96-well plates containing 100 characterized by expression of immune genes typically associ- 5TGM1-luc cells using Aria III (BD Biosciences). Cells were ated with myeloid cell lineage differentiation. Targeting 1 of cultured for 14 days at 37°C with 5% carbon dioxide. GFP1 the genes most highly expressed by dormant cells, the tyrosine colonies were imaged using a Leica DFC9000GT microscope kinase AXL, released MM cells from dormancy and increased and quantified using ImageJ.28 MM burden in vivo. Notably, the dormant transcriptome signature discriminated patients with monoclonal gammopathy of AXL inhibition studies fi uncertain signi cance (MGUS) from patients with overt MM and BKAL mice were injected with DiD-labeled 5TGM1-eGFP1 cells; was associated with a twofold increase in overall survival in 14 days postinoculation, they were treated with cabozantinib clinical trial data sets. (60 mg/kg), BMS-777607 (50 mg/kg), or vehicle daily for 10 days by gavage. Mice were euthanized 25 days postinoculation for flow cytometric analysis of dormant and reactivated cells from Methods bone marrow and spleen. Cell lines 5TGM1-eGFP and 5TGM1-luc MM cells, RM1 prostate cells, and Micro computed tomography analysis of bone MC3T3 osteoblastic cells were maintained as described.4,14,15 structure Bone structure in the femora was assessed using a micro computed tomography scanner (Model 1172; Skyscan) at 50 kV, Mouse models 200 mA, with a 0.5-mm aluminum filter using a pixel size of Atotalof23 106 5TGM1-eGFP murine myeloma cells 4.3 mm. Images were reconstructed using NRecon and analyzed were labeled with the lipid dye 1,1’-dioctadecyl-3,3,39,39- by CTAn software. Three-dimensional models were created tetramethylindodicarbocyanine (DiD) and injected IV into 6- to with Drishti-2.29 Bones were decalcified, sectioned, stained for 8-week-old C57BL/KaLwRijHsd mice (BKAL; Harlan) as tartrate-resistant acid phosphatase, and analyzed as described.4 previously described.4 Animal experiments were approved by the Garvan Institute of Medical Research Animal Ethics Intravital 2-photon microscopy Committee (ARA14/43). Intravital microscopy of dormant and reactivated MM cells in the skeleton was performed as described.4 scRNA-seq Dormant (DiD-retaining eGFP1DiDhi) and reactivated (DiD- Coculture experiments negative eGFP1DiDneg) cells were sorted by fluorescence- A total of 1 3 105 MC3T3 osteoblast-like cells were seeded into activated cell sorting into 384-well plates for SMART-seq or 6-well plates in a MEM media with 10% fetal calf serum/1% MARS-seq and sequenced on Illumina HiSeq2500 or Next- penicillin-streptomycin; 24 hours after seeding, 1 3 105 5TGM1- seq500 platforms (supplemental Methods, available on the eGFP cells were added, and after 72 hours, cells were harvested Blood Web site). External RNA Controls Consortium coverage and GFP1 myeloma cells were analyzed. In separate experi- was calculated.16 Clonotypic sequence of 5TGM1 cells was ments, 1 3 105 MC3T3 cells were seeded in 6-well plates and determined using MiXCR,17 verified using IgBlast,18 and visu- cultured in the presence or absence of media conditioned by alized using MegaAlignPro in DNASTAR. 5TGM1-eGFP cells. Alternatively, 5TGM1-eGFP cells were

A GENE SIGNATURE DEFINING DORMANT MYELOMA CELLS blood® 4 JULY 2019 | VOLUME 134, NUMBER 1 31 cocultured in direct contact or separated by a 0.4-mm transwell coverage per depth (Figure 1D). RNA biotypes showed similar for 24 hours. diversity between dormant and reactivated cells (Figure 1E).

Real-time quantitative polymerase chain reaction All single cells expressed Egfp,confirming that MM cells analysis of myeloid gene expression had been isolated. All eGFP1DiDhi and eGFP1DiDneg cells Nucleic acid isolation and complementary DNA synthesis are expressed Ighg2b immunoglobulin heavy and Igkc immu- described in supplemental Methods and supplemental Table 1. noglobulin light chain genes, consistent with the IgG2b Real-time quantitative polymerase chain reaction was performed k paraprotein secreted by 5TGM1 cells.37,38 Each cell also using Taqman or Syber green primers on a QuantStudio 7 Flex expressed the Ighv1-54 heavy chain and Igkv10-96 light Real-Time PCR System (ThemoFisher Scientific) or 7900HT Real- chain variable region genes, establishing that they expressed Time PCR System (Applied Biosystems), and gene expression the clonotypic B-cell receptor genes (Figure 1F). We also was calculated using the 2-DDCT method. reconstructed the entire variable region gene sequence from Downloaded from https://ashpublications.org/blood/article-pdf/134/1/30/1553831/blood880930.pdf by WASHINGTON UNIVERSITY SCHOOL user on 28 August 2020 single dormant and reactivated cells, which were identical AXL expression in patient samples to the parental cells (Figure 1G). Single cells were shown to Bone marrow aspirates were obtained from patients with MGUS express PR/SET domain 1 (Prdm1) and syndecan-1 (Sdc1), or active MM30 (Royal Adelaide Hospital, Adelaide, Australia). markers of plasma cells, but not na¨ıve B-cell gene markers Bone marrow mononuclear cells were isolated by density gra- Fc receptor, IgE, low affinity II, a polypeptide (Fcer2a), or paired dient centrifugation.31 AXL expression was assessed by flow box gene 5 (Pax5) or the macrophage gene marker myeloid cytometry on CD3811/CD1381 MM plasma cells using a oncogene v-Myb (Myb; Figure 1F). These data demonstrate that LSRFortessa (BD Biosciences). The Royal Adelaide Hospital only bona fide MM cells had been isolated and sequenced. Ethics Committee approved the study (#030206, #131132, and #110304). Gene expression of marrow cell populations was Dormant MM cells express a distinct performed as previously described.32 gene signature Comparison of the transcriptomes of dormant cells and reac- Nearest-neighbor analysis tivated cells identified 1492 differentially expressed genes Nearest-neighbor analysis was performed on the AXL gene (Figure 2A; supplemental Table 2). Gene ontology (GO) analysis (Probeset: 202686_s_at) using the Mayo Clinic Patient Dataset.33 revealed enrichment for biological processes involving response Analysis was performed on the Multiple Myeloma Genomic to interferon g (GO:0034341; P , .0001), immune response (GO: Portal. x2 test of independence was performed using R. 0006955; P , .0001), and inflammatory response (GO:0006954; P , .0001) pathways in dormant cells (Figure 2B). Processes Survival analysis involving DNA transcription (GO:0006351; P , .01), phosphor- Survival analysis was performed on GSE9782-GPL96,34 E-MTAB- ylation (GO:0016310; P , .005), and regulation of transcription 4715, E-MTAB-4717, E-MTAB-5212, E-TABM-937, and (GO:0006355; P , .005) were enriched in reactivated cells, con- E-TABM-1088 data sets32,35,36 (supplemental methods). Univar- sistent with their proliferation status (Figure 2B). Nonnegative iate and multivariate Cox regression models were used to es- matrix factorization analysis39 revealed that the cells clustered into timate hazard ratios (HRs) and 95% confidence intervals (CIs) for 3 distinct groups: 2 dormant cell clusters (C1 and C2) and a single each covariate using the survival package in R. reactivated cell cluster (C3; Figure 2C-E; supplemental Figure 1C). The C1 metagene was enriched with genes associated with in- Quantification and statistical analysis terferon signaling (116 of 168 genes in the interferome v2.01 Statistical analysis was performed by Mann-Whitney test in R. database40) and the myeloid lineage (Figure 2F; supplemental Box and whisker plots show the medians and interquartile ranges Figure 3; supplemental Table 3). Approximately 15% of genes are (IQRs) between 25% (Q1) and 75% quartiles (Q3). The lower under the transcriptional control of Irf7 and Spic (Figure 2G). Thus, whisker is the lowest data point or Q1 2 1.5 3 IQR, whichever dormant MM cells expressed a distinct transcriptome signature is smaller; the upper whisker is the highest data point or Q3 1 characterized by immune-response genes and genes associated 1.5 3 IQR, whichever is greater. with myeloid cell differentiation.

Dormant MM cells express a myeloid Results transcriptome signature Single-cell transcriptomic analysis of dormant and MARS-seq on a larger data set validated the dormant cell reactivated MM cells transcriptome signature. We profiled 1067 eGFP1 cells, in- 5TGM1-eGFP murine myeloma cells14 were labeled with DiD cluding 319 eGFP1DiDhi dormant cells. t-distributed stochastic and injected into BKAL mice (Figure 1A). Dormant MM cells neighbor embedding partitioned the MARS-seq data into were identified by expression of eGFP and retention of DiD 3 clusters. C0 and C1 contained eGFP1DiDneg reactivated cells (eGFP1DiDhi). Reactivated cells that had proliferated and diluted and a limited number of eGFP1DiDhi cells. This reflects the in- theDiDlabelwereeGFP1DiDneg (Figure 1B). Individual clusion of eGFP1DiDhi cells that were recently reactivated but eGFP1DiDhi dormant cells and eGFP1DiDneg reactivated cells had not yet diluted all their label. C2 contained only eGFP1DiDhi were sorted into 384-well plates, and single-cell transcriptomes cells, demonstrating this was the primary dormant cell cluster were generated using SMART-seq and validated against pools (Figure 3A-B). A total of 602 genes were differentially expressed of 20 cells (Figure 1A-B). This enabled us to evaluate ;4 3 103 between cells in C2 and C0 and C1. Functional annotation uniquely mapped genes per cell (Figure 1C). External RNA Controls revealed cell cycle and DNA replication processes were enriched Consortium spike-in controls showed equivalent mean read in C0 and C1 (P , .001), suggesting they were reactivated cells.

32 blood® 4 JULY 2019 | VOLUME 134, NUMBER 1 KHOO et al A B Single cell sort Dormant

21 days RNA

Hind Isolation eGFP+/DiD limbs isolated labelled cells Transcriptomics Reactivated DiD eGFP+/ eGFP+/ DiDneg DiDhi GFP Downloaded from https://ashpublications.org/blood/article-pdf/134/1/30/1553831/blood880930.pdf by WASHINGTON UNIVERSITY SCHOOL user on 28 August 2020 C D E

) Dormant Reactivated 2 Dormant Reactivated 100 r = 0.926 r = 0.954 Dormant rs = 0.934 rs = 0.951 Reactivated 75 8 10.0 10 Anti-sense

3 Sense 6 7.5 Protein coding 6 50 Pseudogene

4 5.0 0 % Biotypes, ncRNA Others 25 Reads, 10 Reads, NA 2 2.5 Unique genes, 10 Unique genes,

0 0.0 –10 0

1C 20C 1C 20C log of coverage (counts, read spike–in Mean –10 100 –10 100 Samples Cells Cells Spike–in molecules per reaction (number, log2)

F 9 6 Egfp Prdm1 3 2 2 2 15 6 lghg2b Sdc1 5 2 15 lghv 6 Fcer2a 5 1–54 2

15 4 lgkc Pax5 5 2

Transcripts per million +1, log per million +1, Transcripts log per million +1, Transcripts 4 15 lgkv 2 Myb 5 10–96 Sample Sample Dormant Reactivated Dormant Reactivated

FR1 CDR1 FR2 CDR2 FR3 CDR3 Germ line

C JDV

Figure 1. Single-cell transcriptome of dormant myeloma cells. (A) Experimental workflow for scRNAseq of dormant and reactivated MM cells. eGFP1DiD1 5TGM1 cells were injected IV into BKAL mice. Hind limbs were isolated and individual eGFP1DiDhi and eGFP1DiDneg cells sorted into 96-well plates for scRNAseq. (B) Representative flow plot illustrating the gating of eGFP1DiDhi dormant and eGFP1DiDneg reactivated cell populations for cell isolation and scRNAseq. (C) Distribution of sequencing read coverage (left) and number of uniquely mapped genes (right) from across single (1C) and pooled (n 5 20; 20C) dormant and reactivated cells. (D) Distribution of 92 External RNA Controls Consortium spike-in controls across dormant and reactivated single cells. (E) Distribution of gene biotypes for dormant and reactivated cells. (F) Transcript profile of dormant and reactivated single cells showing the expression level of Egfp and idiotype genes (left) and plasma cell (Prdm1, Sdc1), naive B-cell (Fcer2a, Pax5), and myeloid lineage (Myb) genes (right). (G) Alignment of the VDJ sequence from individual 5TGM1 cells. Each row represents a single cell. C, consensus sequence; D, dormant samples; NA, not annotated; ncRNA, noncoding RNA; R, reactivated samples.

A GENE SIGNATURE DEFINING DORMANT MYELOMA CELLS blood® 4 JULY 2019 | VOLUME 134, NUMBER 1 33 A B D Dormant Reactivated Biological Processes Response to interferon–gamma Immune response Regulation of innate immune response 0.20 metagene Inﬂammatory response C1 0.15 Adaptive immune response Transcription, DNA–templated 0.10 metagene Phosphorylation C3 0.05 Protein phosphorylation Enrichment 0 Cellular response to DNA damage stimulus 0 metagene 0 D C2 0.05 Regulation of transcription, DNA–templated R 0.05 0.10

0.15 Downloaded from https://ashpublications.org/blood/article-pdf/134/1/30/1553831/blood880930.pdf by WASHINGTON UNIVERSITY SCHOOL user on 28 August 2020 0.10 01020 0.20 0 51015 P value, -log10 Transcripts per million + 1, log2 C E Dormant Reactivated C1 C2 C3 C1qa Aif1 100 50 metagene Axl metagene C1 80 ll18bp C1 40 Glul Arl5c 60 metagene metagene 30 Gm43376 C3 40 C3 20 Bace1 1 Creld1 Rmdn2 20 10 Hsph1 Mettl22 0.8 Pih1d2 0 metagene 0 metagene Tmem53 0 C2 0 0 C2 0 0.6 20 5 Lcmt2 10 40 10 20 50 60 15 30 0.4 20 100 80 25 40 100 30 50 0.2 0

F G

Metagene C1

Symbollog2FC Symbol log2FC C1qa 5.03 Fam213b 2.49 Aif1 4.87 Slc43a2 3.55 Axl 4.53 Iﬁ44 4.12 II18bp 5.02 Gng10 3.68 Glul 4.32 Cxcl16 3.28 Mpeg1 3.76 Ptprcap 3.18 H2-Eb1 4.14 Anxa2 4.69 Nr1h3 3.93 Rgs2 3.74 Lilrb4 4.68 Tmed3 4.06 Sdc3 3.86 Igll1 4.61 Fcgr3 2.54 Hpgd 3.51 Oas1a 5.07 Glipr1 4.25 Iﬁt2 4.63 Cd4 2.63 Ly6a 5.37 Cd84 3.57 Ly86 4.42 Gbp2 4.06 Bcl2a1b 4.65 AB124611 3.19 Cd19 5.57 Slc44a2 3.23 Gm4955 4.58 Samd9l 3.05 Hebp1 2.67 Oas1g 3.19 Sirpa 3.72 Fcgr1 3.10 Slfn2 4.66 Pla2g15 3.54 Tnfaip2 2.89 Tifa 3.34 Vpreb3 4.95 Pmp22 4.14 Trancription regulator Regulated by Irf7 only ligp1 4.12 Abcc3 3.39 Oasl2 S100a10 3.86 4.09 Regulated by both Irf7 and Spic Metagene C1

Figure 2. Dormant myeloma cells express a distinct gene signature. scRNAseq analysis of eGFP1DiDhi dormant and eGFP1DiDneg reactivated cell populations using SMART- seq. (A) Heatmap of 1492 differentially expressed genes between dormant and reactivated cells. (B) Biological processes associated with the genes upregulated in dormant (D; gold) and upregulated in reactivated (R; gray) populations. (C) Consensus matrix of nonnegative matrix factorization analysis of 1492 differentially expressed genes between dormant and reactivated cells deﬁnes 2 dormant (gold; C1 and C2) and 1 reactivated (gray; C3) cluster. (D) Distribution of individual cell sample contributions to each of the 3 metagenes. (E) Distribution of all the genes (top) and the top 5 unique genes (bottom) contributing to each metagene. (F) Top 50 genes from the primary (C1) dormant metagene sorted by descending contribution. (G) Genes predicted to be controlled by the Irf7 and Spic transcription factors.

34 blood® 4 JULY 2019 | VOLUME 134, NUMBER 1 KHOO et al In contrast, C2 was enriched for processes, including immune 5A-B).4 This is covered by differentiated cells of the osteoblast system processes (GO:0002376; P , .0001), innate immune lineage. We therefore examined whether osteoblast lineage cells response (GO:0045087; P , .0001), and response to interferon g could induce expression of myeloid transcriptome signature genes. (GO:0034341; P , .005; Figure 3C). Nineteen of the top 20 C2 After 3 days of coculture with MC3T3 cells, the expression of Axl, genes were also identified by SMART-seq. This included genes Mpeg1, Sirpa,andCsf1r was upregulated in 5TGM1eGFP cells associated with myeloid cell differentiation, including the tran- (Figure 5C). Expression of AXL was confirmed by flow cytometric scription factors Irf7 and Spic (Figure 2G), and the adhesion analysis (Figure 5D). This was not observed when cells were cul- molecules and receptors Axl, Fcer1g, Mpeg1, Sirpa, and Vcam1 tured with media conditioned by osteoblasts or separated by (Figure 3D). These are characteristic features of myeloid and a transwell (Figure 5E). RNAseq analysis showed that 1156 genes macrophage lineage cells.41-44 To verify markers, we isolated were upregulated and 1 gene was downregulated by coculture CD1381eGFP1DiDhi dormant cells and CD1381eGFP1DiDneg (Figure 5F). Osteoblasts expressed the binding partners for key reactivated cells for flow cytometric analysis. Surface expression dormancy signature genes, including Gas6 and a4b1; however, Downloaded from https://ashpublications.org/blood/article-pdf/134/1/30/1553831/blood880930.pdf by WASHINGTON UNIVERSITY SCHOOL user on 28 August 2020 of AXL, FCER1G, CSF1R, and SIRPA proteins were all upregulated there were no statistically significant differentially expressed genes in CD1381eGFP1DiDhi dormant cells (Figure 3E). These approaches (Figure 5G). Together, these observations suggest that endosteal demonstrate that dormant MM cells express a suite of genes as- niche cells are able to induce contact-dependent upregulation of sociated with myeloid cell differentiation, which is downregulated as myeloid transcriptome signature genes in MM cells to induce cells are reactivated and transition to active growth. dormancy. Dormant myeloid transcriptome signature originates from donor MM cells and not AXL inhibition releases MM cells from dormancy host-derived myeloid cells One of the most highly upregulated genes in dormant cells was Axl, which was selected for proof-of-concept functional analysis. The expression of myeloid genes in dormant cells raised To inhibit AXL, we used 2 inhibitors: cabozantinib and BMS- the possibility that either host macrophages had engulfed 777607. Cabozantinib inhibits receptor tyrosine kinases, in- dormant MM cells, that MM cells had engulfed host myeloid cluding AXL, c-MET, and VEGF (Figure 6A).48 With the exception cells (cancer cell cannibalism45), or that the MM cells had fused of Axl and Flt3, the targets of cabozantinib were not expressed with myeloid cells to create hybrid host-cancer cells.46 We by dormant or reactivated cells (Figure 6A). Cabozantinib reasoned that an MM cell that had been phagocytosed by treatment decreased eGFP1DiDhi dormant cells and increased a macrophage would not proliferate. To investigate this, we eGFP1DiDneg MM cells, suggesting that inactivation of AXL isolated eGFP1DiDhi dormant cells and eGFP1DiDneg reactivated released cells from dormancy and allowed MM cells to pro- cells, stained them for myeloid cell surface markers, and index liferate (Figure 6B). Importantly, the reduction in numbers of sorted single cells into the individual wells of a 96-well plate to dormant cells was consistent with the number of tumor colonies evaluate their growth potential (Figure 4A). On day 0, all cells required to explain the increase in tumor burden.4 In the spleen, expressed eGFP, and eGFP1DiDhi dormant cells retained the cabozantinib treatment had no effect on dormant cell numbers DiD label (Figure 4B). Evaluation of 1161 individual cells from 3 ex- and only slightly increased MM burden (Figure 6C). To verify this, periments demonstrated the development of eGFP1 colonies we evaluated the effect of BMS-777607, an inhibitor with greater from both eGFP1DiDhi dormant cells and eGFP1DiDneg reac- affinity for AXL and more limited activity against other tyrosine tivated cells (Figure 4B-C). Expression of AXL, SIRPA, and kinases, none of which were expressed by dormant cells49 FCER1G was similar between cells that had formed colonies and (Figure 6D). BMS-777607 treatment also reduced eGFP1DiDhi those that had not (Figure 4D). These data confirm that dormant dormant cells and increased eGFP1DiDneg MM cell burden in the cells could form colonies, arguing they had not been engulfed bone marrow but not the spleen (Figure 6E-F). Given that os- by macrophages. teoclastic resorption releases MM cells from dormancy,4 we determined whether BMS-777607 could increase osteoclasts To determine whether the myeloid signature could arise from and release dormant cells indirectly. BMS-777607 treatment the fusion with host cells, we exploited the knowledge that of naive mice had no effect on osteoclast numbers, the bone the 5TGM1 cells were derived from a female mouse.47 We surface occupied by osteoclasts, or the endosteal bone surface injected DiD-labeled 5TGM1eGFP cells into male and female where dormant cells reside (Figure 6G-I), suggesting this was mice and performed scRNAseq of dormant and reactivated not an off-target effect. Together, this suggests that AXL, as cells (Figure 4E). We then analyzed expression of 1571 a prototypical dormancy gene, is able to control retention Y-chromosome genes. Analysis of positive control cells (RM1 of MM cells in a dormant state and that AXL inhibition releases prostate cells) revealed expression of Y-chromosome genes dormant cells from the endosteal niche, causing reactivation (Figure 4E-H). In contrast, neither dormant nor reactivated and MM growth. MM cells, isolated from either female or male mice, expressed any of the 1571 Y-chromosome transcripts (Figure 4F-H). These data confirm that MM cells, and not host cells, are the source AXL is expressed by human bone marrow plasma of the myeloid transcriptome signature. cells and MM cells We next investigated candidate dormancy genes AXL, CSF1R, Niche-dependent induction of myeloid genes in FCER1G, SIRPA, and VCAM1 in healthy donors and patients dormant MM cells with MM (Figure 7A; supplemental Figure 4). AXL is expressed in We next determined whether the myeloid signature genes were normal CD1381 bone marrow plasma cells. In malignant dis- induced after engagement with the dormant cell niche. Intravital 2- eases, AXL is also expressed, with higher expression in plasma photon microscopy and immunohistochemistry showed that single cells from patients with MGUS than in plasma cells from those dormant MM cells resided adjacent to the endosteal bone (Figure with overt or relapsed MM. AXL is not expressed in memory

A GENE SIGNATURE DEFINING DORMANT MYELOMA CELLS blood® 4 JULY 2019 | VOLUME 134, NUMBER 1 35 A B Cluster C 0 1 2 Biological processes 20 Cited2 Itga4 Creld2 Srp14 Cell cycle 10 Aaas Sub1 Rhob Phlda1 Cell division Adam10 0 Lbr Dync1h1 DNA replication Plin3 tSNE 2 Myc –10 Hist1h2bc Irf4 Cell proliferation Cluster Iglc3 Pold1 G2/M transition of 0 Serp1 –20 Csn3 mitotic cell cycle 1 Pgam2 Top2a 2 Hist1h2ao Cell cycle Tubb5

Prc1 Downloaded from https://ashpublications.org/blood/article-pdf/134/1/30/1553831/blood880930.pdf by WASHINGTON UNIVERSITY SCHOOL user on 28 August 2020 –10 01020 Nusap1 Cell division 2810417H13Rik Hmgb2 tSNE 1 Cenpf Ccna2 DNA replication Kif11 Ube2c Cenpe H2afx Cell proliferation Tpx2 20 Cdk1 G2/M transition of Hmmr Pbk mitotic cell cycle Ccnb1 Rrm2 10 Hist1h1e Immune system process Tmsb4x C1qa Apoe Innate immune response 0 Marcks C1qc Antigen processing and presentation Lyz2 C1qb of exogenous peptide antigen via MHC class II tSNE 2 Fcer1g –10 Mpeg1 Cluster Vcam1 Antigen processing and presentation Group Sirpa 0 Axl 1 D Slc40a1 –20 S100a9 Response to interferon–gamma 2 R H2–Ab1 Lgals3 Irf7 Cd5l H2–Aa 0 20 40 60 –10 01020 H2–Eb1 P value, -log10 tSNE 1

D 25 MARS-seq Dormant Reactivated 20 15

UMI 10 5 0 Axl C1qb Csf1r Fcer1g Irf7 Mpeg1 Sirpa Spic Tyrobp Vcam1

SMART-seq 9

tpm+1, counts tpm+1, 3 2

log 0 Axl C1qb Csf1r Fcer1g Irf7 Mpeg1 Sirpa Spic Tyrobp Vcam1

E 100 100 100 100 100 80 80 80 80 80 60 60 60 60 60 40 40 40 40 40 Frequency 20 20 20 20 20 0 0 0 0 0 AXL FCER1G CSF1R SIRPA CD11b

AXL FCER1G CSF1R SIRPA CD11b 4000 *** 500 *** *** 1500 *** *** 200 3000 3000 400 150 1000 2000 300 2000 MFI 200 100 500 1000 1000 100 50 0 0 0 BM R D BM R D BM R D BM R D BM R D BM: Gr-1+CD11b+ bone marrow D: Dormant R: Reactivated

Figure 3.

36 blood® 4 JULY 2019 | VOLUME 134, NUMBER 1 KHOO et al A B C

Single cell day 0 day 3 day 7 day 14 40 397 384 383 index sort day 0 eGFP+/ 30 DiDhi 21 days Culture 20 37oC day 14 Hind % growth 10 GFP+/DiD limbs

isolated + + + labeled cells eGFP eGFP eGFP / DiDneg DiDhigh Imaging neg 0 DiD DiD +-+ -+ - GFP ++ + +++

Exp 1Exp 2Exp 3

D E Downloaded from https://ashpublications.org/blood/article-pdf/134/1/30/1553831/blood880930.pdf by WASHINGTON UNIVERSITY SCHOOL user on 28 August 2020

GFP DiD CD138 AXL SIRPA FCER1G Male, Male 0.003 MFI : 21083 MFI: 20117 MFI: 17668 MFI: 1098 MFI: 2248 MFI: 945 RM1 prostate cancer mouse Hind MFI : 20931 MFI : 19048 MFI : 17085 MFI : 833 MFI : 1779 MFI : 814 limbs + scRNA- Expression 0.002 eGFP / isolated hi seq of 0.001 DiD Female, Male 5TGM1 myeloma mouse 1571 male 0.000 genes 0.003 MFI : 21276 MFI: 88.3 MFI: 20063 MFI: 145 MFI: 444 MFI: 38 MFI : 220781 MFI : 63.5 MFI : 16077 MFI : 50 MFI : 438 MFI : 30 + Female, Female

Frequency 0.002 eGFP / neg 5TGM1 myeloma mouse 0.001 DiD 0.000 GFP DiD CD138 AXL SIRPA FCER1G Hi, Colonies Hi, No colonies MFI Neg, Colonies Neg, No colonies F G H

RM1 5TGM1 Male 5TGM1 Female 7.5 Eif2s3y RM1 5.0 Erdr1 Eif2s3y Ddx3y 40 Dormant Uty Reactivated 2.5 Kdm5d 0.0 Gm21860 2 Gm21748

Gm20775 2 7.5 Gm4017 Gm20899 Gm28533 5.0 Erdr1 Gm28974 30 Gm29635 2.5 Gm28838 Gm28223 0.0 Gm28705 Gm29568 Gm29148 7.5 Gm28233 Gm28406 20 5.0 Ddx3y Gm28462 Gm29522 2.5 Gm29227 ENSMUSG00000103894 0.0 Zfy1 Gm21857 Gm29045 7.5 Gm28428 ENSMUSG00000103618 10 5.0 Uty Gm29548 log tpm + 1, Aggregrated ENSMUSG00000102835 2.5 Gm29650 Gm29338 Gm21919 0.0 Gm21855 ENSMUSG00000103231 7.5 Gm8525 0 Gm28470 log per million (tpm), Transcripts 5.0 ENSMUSG00000103589 Kdm5d Gm28365 Gm28220 2.5 RM1 Male Female 0.0 5TGM1 RM1 Male Female 02468 5TGM1 Transcripts per million, log2 + 1

Figure 4. Myeloid gene signature originates from donor myeloma cells and not host myeloid cells. (A) Experimental workflow to examine the proliferative potential of dormant myeloma cells. eGFP1DiD1 5TGM1 cells were injected IV into BKAL mice. Hind limbs were isolated, and eGFP1DiDhi and eGFP1DiDneg cells were index sorted into individual wells of 96-well plates. Plates were cultured and imaged over 14 days. (B) Representative images of individual eGFP1 cells isolated from eGFP1DiDhi (top) and GFP1DiDneg (bottom) cell populations at day 0 through day 14. Inset images show eGFP and DiD channels separately. Scale bar 5 50 mm; inset scale bar 5 20 mm. (C) Histogram showing the proportion of wells containing eGFP1 colonies after 14 days from 3 independent experiments. Red bars represent dormant MM cells; green bars represent reactivated MM cells. Numbers are individual cells evaluated from each experiment. (D) Histograms of the aggregated expression of myeloma cells and myeloid markers from index-sorted dormant (gold) and reactivated (gray) cells that formed colonies or those that did not. (E) Schematic workflow for the analysis of expression of 1571 Y-chromosome (chrY) genes from RM1 (positive control) and 5TGM1 scRNAseq data from cells isolated from male and female mice. (F) Heatmap of the top 41 chrY genes detected from raw unfiltered counts from RM1 cells (blue) and 5TGM1 cells isolated from male (green) and female (purple) mice. (G) Aggregated raw unfiltered counts of 41 chrY genes in RM1 single cells isolated from male mice and 5TGM1 single cells isolated from male and female mice. (H) Transcript profile of the top 5 chrY genes from raw unfiltered counts in RM1 single cells and 5TGM1 single cells isolated from male and female mice.

B cells, in vitro–generated polyclonal plasmablasts, or human showed that a small subpopulation of CD1381CD381 plasma MM cell lines. In the bone marrow microenvironment, AXL is cells expressed AXL (Figure 7B). The proportion of AXL1 cells expressed in mesenchymal cells and osteoblasts only (Figure 7A). was higher in patients with MGUS than those with MM or re- Similar patterns of expression were observed for other myeloid lapsed MM, and there was an inverse correlation between dis- 1 2 dormancy signature genes (supplemental Figure 4). We next ease burden and the frequency of AXL cells (rs 520.689;

Figure 3. Dormant myeloma cells express a myeloid transcriptome signature. scRNAseq analysis of eGFP1DiDhi dormant and eGFP1DiDneg reactivated cell populations using MARS-seq. (A) t-distributed stochastic neighbor embedding (t-SNE) plots of the MARS-seq analysis of 1067 single myeloma cells identifies 3 distinct clusters (0, 1, and 2) (top) and maps eGFP1DiDhi dormant cells to cluster 2 (gold; D) and eGFP1DiDneg reactivated cells to C0 and C1 (gray, R) (bottom). (B) Heatmap of the top 20 genes contributing to each cluster (red, 0; green, 1; and blue, 2). (C) Database for Annotation, Visualization and Integrated Discovery functional annotation of the top 5 biological processes associated with each cluster. (D) Transcript profile of 10 common genes identified by MARS-seq (top) and SMART-seq (bottom) workflows for individual dormant (gold) and reactivated (gray) cells. Horizontal lines are the median; box indicates the IQR. (E) Histograms of expression of myeloid markers (top) and quantitation of mean fluorescent intensity of expression (MFI; bottom) of Gr11CD11b1 bone marrow (BM), eGFP1DiDhi dormant, and eGFP1DiDneg reactivated cells. Mann-Whitney test: ***P , .001. UMI, unique molecular identifier.

A GENE SIGNATURE DEFINING DORMANT MYELOMA CELLS blood® 4 JULY 2019 | VOLUME 134, NUMBER 1 37 AB Downloaded from https://ashpublications.org/blood/article-pdf/134/1/30/1553831/blood880930.pdf by WASHINGTON UNIVERSITY SCHOOL user on 28 August 2020

CDE 3 ** 2 ** ** Axl 1 Axl 1

0 30 **** 1 * 2 Csf1r 1 Csf1r 0 0 ns 25 -  CT * -  CT 2 2 1 1

10 Fcer1g , %

10 Fcer1g + 0 0

1 * of GFP 20 ** + 1 Mpeg1

Mpeg1 Axl 0 0 Fold difference, log difference, Fold Fold difference, log difference, Fold * 8 ** 3 15 5 Sirpa Sirpa 1 2

** ns 8 3 5 Vcam1 10 Vcam1 1 2

5TGM1 5TGM1 5TGM1 5TGM1+MC3T3 5TGM1 5TGM1 5TGM1 5TGM1 + MC3T3 + MC3T3 + MC3T3 + MC3T3 CM Transwell

Experimental replicate Group Exp 1 Exp 2 5TGM1 5TGM1+ MC3T3 Transwell

Exp 3 Exp 4 5TGM1+ MC3T3 Condition Media 5TGM1+ MC3T3 Direct seeding

FG 5TGM1 control 5TGM1 co-culture

15 15 ltgb1 Gas6 + 1 + 1

2 2 Cd47 Vegfa Adam10, 10 10 Pros1 Itga9 5TGM1 Co 5TGM1 MC3T3 Co 5 5

Axl Axl - interacting ligands Transcripts per million, log per million, Transcripts Transcripts per million, log per million, Transcripts Sirpa Sirpa - interacting ligands Itgb2 Vcam1 Vcam1 - interacting ligands 0 0 Itga4

0 5 10 15 0 5 10 15 –2 –1 0 12 Transcripts per million, log2 + 1 Transcripts per million, log2 + 1 5TGM1 only Row Z–score MC3T3 only

Figure 5. Niche-dependent induction of the myeloid gene signature. (A) Intravital imaging of a dormant MM cell (arrow) residing on the endosteal surface and a growing MM colony. Top shows the DiD (red) channel, and bottom shows the eGFP (green) channel. Scale bar 5 50 mm. (B) Immunohistochemical staining for CD138 showing an individual 5TGM1 tumor cell (arrow) adjacent to the endosteal bone surface. Scale bar 5 50 mm; inset scale bar 5 20 mm. (C) Real-time quantitative polymerase chain reaction (RT-qPCR) of myeloid marker genes from 5TGM1 cells cultured with and without MC3T3 cells. (D) Flow cytometric analysis of AXL expression by 5TGM1 cells cultured with and without MC3T3 cells. Symbols denote different experimental replicates. (E) RT-qPCR of myeloid marker genes from 5TGM1 cells cultured alone, with MC3T3 conditioned media (CM), and with MC3T3 cells seeded directly or separated by a transwell. (F) Scatter plot of genes expressed by 5TGM1 cells cultured alone or with MC3T3 cells (left); heatmap of differentially expressed genes (right). (G) Scatter plot of genes expressed by MC3T3 cells cultured alone or with 5TGM1 cells. (C-E) Each point represents an independent experiment. Horizontal lines are the median; box indicates the IQR. Mann-Whitney test: *P , .05, **P , .01, ****P , .0001. ns, not signiﬁcant.

38 blood® 4 JULY 2019 | VOLUME 134, NUMBER 1 KHOO et al A B C Vehicle Cabo Vehicle Cabo Vehicle Cabo Vehicle Cabo Cells * **** ns * 0.0125

Daily 90 0.0100 75 Cabo (60mg/kg) / Veh 0.015 , % gavage , % + + 0.0075 Day 0 Day 14 Day 24 Day 25 80

2 0.010 50 of live, % of live,

8 % of live, of GFP D of GFP 0.0050 + + hi 6 hi R GFP GFP

4 DiD 70 0.005 DiD 0.0025 2 25 tpm+1, log tpm+1, 0 0.0000 Kdr Met Ret Kit Flt4Axl Flt3 Flt1 Tek Downloaded from https://ashpublications.org/blood/article-pdf/134/1/30/1553831/blood880930.pdf by WASHINGTON UNIVERSITY SCHOOL user on 28 August 2020 Vehicle Cabo Vehicle Cabo Vehicle Cabo Vehicle Cabo

IC50 0.035nM7.0nM 14.3nM Treatment Treatment Treatment Treatment

D E Vehicle Vehicle F Vehicle Vehicle BMS BMS BMS BMS Cells ns ns ** **

Daily 70 0.07 0.20 40 BMS (50mg/kg) / Veh , % gavage , % + + 0.06 60 Day 0 Day 14 Day 24 Day 25 30 0.15 2

0.05 % of live, 8 % of live, of GFP of GFP +

D +

hi 20 6 hi 50 R GFP 4 0.04 GFP DiD DiD 0.10 2 10 40 tpm+1, log tpm+1, 0 0.03 Axl Mst1r Met Tyro3 Vehicle BMS Vehicle BMS Vehicle BMS Vehicle BMS

IC50 1.1nM 4.3nM Treatment Treatment Treatment Treatment

G Naive Vehicle Naive Vehicle H I Naive Vehicle Naive Vehicle Naive BMS Naive BMS Naive BMS Naive BMS ns ns ns ns 25 100 15 25

20 2 20 75 10 15 15

50 10 10 BV/TV, % BV/TV, 5

5 Bone surface mm 5 Osteoclast surface, % Osteoclast surface, 25 Osteoclast number, N/mm Osteoclast number, 0 0 0

Vehicle BMS Vehicle BMS Vehicle BMS Vehicle BMS Vehicle BMS

Figure 6. Inhibiting AXL releases myeloma cells from dormancy. (A) Experimental design investigating carbozantinib (cabo) treatment on dormant myeloma cells (top) and transcript expression levels of cabo targets in dormant (D) and reactivated (R) cells (bottom). (B) Impact of cabo treatment on dormant cells (left) and myeloma burden (right) in the bone marrow. (C) Impact of cabo treatment on dormant cells (left) and myeloma burden (right) in the spleen. (D) Experimental design investigating BMS-777607 (BMS) treatment on dormant myeloma cells (top) and transcript expression levels of BMS targets in dormant (D) and reactivated (R) cells (bottom). (E) Impact of BMS treatment on dormant cells (left) and myeloma burden (right) in the bone marrow. (F) Impact of BMS treatment on dormant cells (left) and myeloma burden (right) in the spleen. (G) Impact of BMS and vehicle treatment on osteoclast surface (%) (left) and number (per mm) (right). Individual points represent a single mouse. (H) Three-dimensional micro computed tomography reconstructions of distal femora from representative mice treated with BMS or vehicle. (I) Impact of BMS and vehicle treatment on trabecular bone volume/total volume (BV/TV; %) (left) and the bone surface 2 (mm ) (right). Data represented by median and IQR. Mann-Whitney test: *P , .05, **P , .01, ****P , .001. IC50, half maximal inhibitory concentration; ns, not signiﬁcant.

P , .001; Figure 7B). This indicates that AXL and proteins Finally, we analyzed expression of dormant MM cell signature encoded by myeloid dormancy signature genes may identify genes in the transcriptome of plasma cells from a clinical trial in a subpopulation of dormant human MM cells. which patients were treated with dexamethasone or bortezomib.34 Median survival was increased by almost twofold in Dormant myeloid transcriptome signature genes patients with high expression of the dormant cell transcriptome may be associated with survival signature, when compared with patients with low expression of To determine the role of AXL as an exemplar dormant signature signature genes (392 vs 723 days; P , .0001; Figure 7E). This was gene in disease evolution, we used nearest-neighbor analysis to independent of treatment and not observed with the reactivated identify AXL-coregulated genes (Figure 7C). Hierarchical clus- cell transcriptome signature (supplemental Figure 5A). This was tering showed that expression of AXL-coregulated genes could validated in an independent data set (Figure 5B),32,35,36 where discriminate controls and patients with MGUS from patients overall survival increased from 77 months in patients with low with MM (Figure 7C-D). This suggests that AXL and coregulated expression of the dormancy gene signature to 111 months in genes play an important role in MM. patients with high expression of the signature (P 5 .00057;

A GENE SIGNATURE DEFINING DORMANT MYELOMA CELLS blood® 4 JULY 2019 | VOLUME 134, NUMBER 1 39 A AXL expression B BM microenvironment B–cell lineage MM 4 rs = -0.689 Disease present P < 0.001 MGUS (n = 7) 15 absent MM (n = 10) 2 rMM (n = 9) 10

, log % , 0 +

5 AXL –2 Expression, arbitrary units arbitrary Expression, 0 008802001154 128 185 17 1 57005355810 144 580 73 53 –4 Downloaded from https://ashpublications.org/blood/article-pdf/134/1/30/1553831/blood880930.pdf by WASHINGTON UNIVERSITY SCHOOL user on 28 August 2020 –2.5 0.0 2.5 5.0 7.5 OB OC CD3 PPC MM CD14 MSC CD34 MBC AMM rMM BMPC MGUS HMCL CD138+CD38+, log %

C D

VSIG4 SLCO2B1 MS4A4A CD163 MRC1/MRC1L1 CD163 MS4A6A C1QA MSR1 APOE XLKD1 VCAM1 EPB41L3 CTSB SLCO2B1 VSIG4 LIPA SLCO2B1 MS4A6A P2RY13 AXL TGFBI FCGR3A/FCGR3B CD163 P2RY13 CD163 XLKD1 C1QA MS4A4A VCAM1 MRC1/MRC1L1 APOE MSR1 AXL EPB41L3 FCGR3A/FCGR3B TGFBI CTSB SLCO2B1 LIPA Healthy Control (n = 62) MM (n = 101) MGUS (n = 22) MM (n = 101) AXL Expression AXL Expression

High Low Total High Low Total HC 14 1 15 MGUS 12 10 22 MM 27 74 101 MM 20 81 101

P value < 0.0001 P value = 0.002

E F All treatments Dormancy Signature Albumin Co-variate HR 95% CI P value 1.00 Dormancy Signature Dormancy Signature 0.334 0.233-0.480 <0.001 High Low Study Code 0.75 mean Myeloma Score Albumin 0.946 0.926-0.967 <0.001 C Reactive Protein Study Code 0.985 0.960-1.012 0.278 0.50 Beta 2 Microglobulin mean Myeloma Score 1.020 1.014-1.034 <0.001 Age at Randomization C Reactive Protein 1.010 1.001-1.011 0.029 Co-variates 0.25 P < 0.0001 Number of Prior Lines Beta 2 Microglobulin 1.000 0.991-1.017 0.563

Survival probability Treatment, Bortezomib Age at Randomization 1.010 0.998-1.030 0.087 0.00 Sex, Male Number of Prior Lines 1.160 1.087-1.240 <0.001 0 300 600 900 1200 Mayo Clinic Treatment 1.270 0.881-1.843 0.199 Proliferative Index Sex 1.330 0.961-1.853 0.085 Time (days) 0.5 1.0 1.5 2.0 Mayo Clinic 1.370 0.959-1.960 0.084 Number at risk Hazard ratio Proliferative Index High 132 98 63 16 0 Low 132 79 34 70 G Dormancy Signature 0 300 600 900 1200 Albumin Cox Adj Co-variate HR 95% CI P value Time (days) Study Code Dormancy Signature 0.447 0.254-0.789 0.005 mean Myeloma Score Albumin 0.952 0.929-0.982 0.001 C Reactive Protein Study Code 1.001 0.957-1.047 0.278 Beta 2 Microglobulin mean Myeloma Score 1.024 1.008-1.041 0.003 Age at Randomization C Reactive Protein 1.005 0.999-1.011 0.117 Co-variates Number of Prior Lines Beta 2 Microglobulin 0.999 0.984-0.982 0.004 Treatment, Bortezomib Age at Randomization 1.014 0.991-1.039 0.239 Sex, Male Number of Prior Lines 1.071 0.959-1.197 0.225 Mayo Clinic Treatment 0.893 0.537-1.485 0.663 Proliferative Index Sex 1.465 0.940-2.286 0.092 123 Mayo Clinic 2.183 1.426-3.342 <0.001 Hazard ratio Proliferative Index

Figure 7. Myeloid transcriptome signature predicts disease progression. (A) Gene expression proﬁling of isolated bone marrow (BM) cells, CD141 monocytes (CD14), osteoclasts (OCs), mesenchymal stromal cells (MSCs), osteoblasts (OBs), T cells (CD3), hematopoietic stem cells (CD34), memory B cells (MBCs), in vitro–generated polyclonal plasmablastic cells (PPCs), BM plasma cells (BMPCs), and CD1381 cells from patients with MGUS, asymptomatic MM (AMM), MM, and relapsed MM (rMM). Blue represents the

40 blood® 4 JULY 2019 | VOLUME 134, NUMBER 1 KHOO et al supplemental Figure 5A). Event-free survival was also greater signaling also control breast cancer metastasis in bone.53 To- with high signature expression (P 5 .0048; supplemental gether, these data suggest that it is the endosteal niche that Figure 5B). Univariate and multivariate Cox regression analyses regulates a repertoire of myeloid genes in myeloma cells that of clinical covariates demonstrated the dormancy signature control cellular dormancy. score was a prognostic marker (HR, 0.33; 95% CI, 0.23-0.48; P , .001 and HR, 0.44; 95% CI, 0.25-0.79; P 5 .005, respectively; Inhibition of AXL in solid cancers decreases tumor burden, Figure 7F-G). The replication of the prognostic value of the particularly in combination with conventional treatments.54-56 dormancy transcriptome signature in 2 independent random- Although the mechanism is unclear, AXL has been implicated in ized controlled trials makes it unlikely that this was a false- invasion and metastasis.57 However, the present study suggests positive signal. Nevertheless, we also examined the MMRF that AXL inhibition releases cells from dormancy and sensitizes CoMMpass data repository. This is a prospective, longitudinal, them to chemotherapy, providing an alternative mechanism of observational study involving patients receiving different action. In MM, cabozantinib showed no anti-MM activity as Downloaded from https://ashpublications.org/blood/article-pdf/134/1/30/1553831/blood880930.pdf by WASHINGTON UNIVERSITY SCHOOL user on 28 August 2020 treatments. Genes in the dormancy signature were detected in a single agent.58 However, in experimental models, cabozantinib patients in the CoMMpass data set (supplemental Figure 6). combined with bortezomib showed greater efficacy than either However, when patients were stratified based on high or low agent alone.59 Although our data establish AXL as a regulator expression of the dormancy signature, there was no statistically of MM cell dormancy, the dormancy gene signature has iden- significant difference in overall survival (supplemental Figure 6). tified many other candidate molecules. Analysis of these mol- This could be due to the fact that patients and treatments are ecules will likely reveal the architecture of dormancy control and heterogeneous, unlike in the gold-standard randomized con- additional drug targets. Furthermore, AXL acts through a range trolled trials. However, collectively, this suggests that the of downstream signaling pathways, the targets of which may also dormant MM cell transcriptome signature maybe a marker of prove to be potential therapeutic targets. disease progression and overall survival. In patients with MM, predicting who will relapse has proven challenging. This study showed that dormancy signature genes Discussion could identify subpopulations of dormant cells in patients. Using In this study, we defined a dormant MM cell transcriptome AXL as proof of concept, plasma cells in patients contained signature, a catalog of genes that are expressed by dormant MM a rare subpopulation of AXL1 cells. Patients with MGUS had cells but switched off when they are reactivated and transition to a greater proportion of AXL1 cells than patients with MM, actively growing MM. The transcriptome signature was enriched whereas patients with relapsed disease had the least number for genes expressed by myeloid cells and included the tran- of AXL1 cells. This is consistent with the AXL1 subpopulation scription factors Irf7 and Spic, which are predicted to control being reactivated and playing a role in disease relapse. In ;15% of signature genes, including the exemplar genes Axl, support of this, expression of AXL nearest-neighbor genes Vcam1, and Sirpa. Inhibition of AXL reduced dormant cells and distinguished MM patients from patients with MGUS. In- increased MM burden, showing these genes are functionally terestingly, a subgroup of patients with MGUS clustered with important. The absence of effects in the spleen argues that patients with MM, raising the possibility that they are at high risk dormancy control pathways may be under bone niche–specific of progression. High expression of the dormant MM cell tran- regulation. scriptome signature was also associated with improved survival in some cohorts when compared with patients with low dor- In MM, cells of the osteoblast lineage play an important role in mancy gene signature expression. Although this is consistent the dormant cell niche.4,5,13 These cells retain both murine and with evidence that proliferation is an independent prognostic human myeloma cells in a dormant state, and release from factor in MM,60 unlike our single-cell data, we cannot exclude the osteoblast niche control allows them to proliferate to form possibility that this reflects rare contaminating myeloid cells in myeloma colonies.4,5,13 Culturing MM cells with osteoblast lin- these data sets. Notably, the dormancy transcriptome signature eage cells induced expression of dormancy genes in a contact- was a superior predictor of disease progression than many dependent manner. This argues that engagement with the conventional biomarkers. endosteal niche switches on expression of myeloid genes to control dormancy. In support of this, osteoblasts express GAS6, In summary, this study has defined a dormant MM cell tran- the ligand for AXL, and prostate cancer cells, which disseminate scriptome signature, enriched with myeloid differentiation– to the same endosteal niche, also express AXL and, through related genes that control cellular dormancy in MM. This also binding GAS6 on osteoblasts, enter dormancy.50 Myeloid pro- revealed the importance of the endosteal niche in inducing MM genitor cells also reside in the endosteal niche,51 and the AXL- cell dormancy. This dormant cell transcriptome signature pro- GAS6 axis controls the survival and self-renewal capacity of vides a platform on which to build an understanding of the role of chronic myeloid leukemia stem cells.52 Irf7 and interferon dormant cells in disease evolution and relapse and develop

Figure 7 (continued) absence of expression; red represents expression according to the presence-absence calls with negative probesets algorithm. Data represented by box and whisker plots. (B) Relationship between expression of AXL and CD1381CD381 determined by ﬂow cytometry for plasma cells isolated from patients with MGUS, MM, and rMM. (C) AXL nearest-neighbor analysis of CD1381 plasma cells in healthy controls (HC; green) and patients with MM (brown) from the Mayo Clinic microarray data set. Heatmap of 20 genes coexpressed with AXL in HC compared with MM (top) and the x2 test of the association of AXL expression in HC vs MM (bottom). (D) AXL nearest-neighbor analysis of CD1381 plasma cells in MGUS (orange) and MM (brown) from Mayo clinic microarray data set. Heatmap of 20 genes coexpressed with AXL in MGUS compared with MM (top) and the x2 test of the association of AXL expression in MGUS vs MM (bottom). (E) Kaplan-Meier analysis of dormancy signature genes and overall survival in the Millennium clinical trial microarray data set. (F) Univariate Cox regression analysis of the dormancy gene signature (red) and disease biomarkers. Forest plot of the HRs and 95% CIs (left) and summary table (right). (G) Cox-adjusted analysis of the dormancy gene signature (red) and disease biomarkers. Forest plot of the HRs and 95% CIs (left) and summary table (right).

A GENE SIGNATURE DEFINING DORMANT MYELOMA CELLS blood® 4 JULY 2019 | VOLUME 134, NUMBER 1 41 rationally informed therapies to eradicate these cells and pre- A. Swarbrick, B.P., D.H., K. Vanderkerken, A.C.W.Z., P.I.C., and T.G.P. re- vent disease relapse. vised the manuscript; and all authors read and approved the manuscript.

Conflict-of-interest disclosure: D.A. is an employee of Exelixis Inc. The Acknowledgments remaining authors declare no competing financial interests. The authors acknowledge the Biological Testing Facility, the Australian fi BioResources Centre, and the Kinghorn Cancer Centre for Clinical Genomics ORCID pro les: G.L., 0000-0003-0443-6066; R.L.T., 0000-0002-6260- and thank Rob Salomon, the Garvan-Weizmann Centre for Cellular 7372; K.V., 0000-0002-1033-849X; A.S., 0000-0002-2771-712X; A.P.C., Genomics, and Nancy Anders and the Small Animal Imaging Facility. 0000-0002-8450-012X; T.V.N., 0000-0002-3246-6281; A.S., 0000-0002- 3051-5676; T.G.P., 0000-0002-4909-2984; P.I.C., 0000-0002-7102-2413. This work was supported by Janice Gibson and the Ernest Heine Family Foundation, the National Health and Medical Research Council Correspondence: Peter I. Croucher, Garvan Institute of Medical Re- (NHMRC; 1139237, 1104031, and 1140996), a Kay Stubbs Cancer Research search, 384 Victoria St, Sydney, NSW 2010, Australia; e-mail: Downloaded from https://ashpublications.org/blood/article-pdf/134/1/30/1553831/blood880930.pdf by WASHINGTON UNIVERSITY SCHOOL user on 28 August 2020 Grant, a Cancer Council New South Wales Grant, Peter and Val Duncan, [email protected]; and Tri Giang Phan, Garvan Institute of the Ann Helene Toakley Charitable Endowment, the Deutsche Medical Research, 384 Victoria St, Sydney, NSW 2010, Australia; Forschungsgemeinschaft (SFB/TRR79, TP B1, B12, M9), and the e-mail: [email protected]. German Federal Ministry of Education (CLIOMMICS; 01ZX1609A) and (CAMPSIMM; 01ES1103). K.D.V. is a postdoctoral fellow of Fonds Wetenschappelijk Onderzoek–Vlaanderen; T.G.P. is an NHMRC Senior Footnotes Research Fellow (1155678). Biospecimens were provided by the South Submitted 18 October 2018; accepted 8 April 2019. Prepublished Australian Cancer Research Biobank, which is supported by the Cancer online as Blood First Edition paper, 25 April 2019; DOI 10.1182/ Council SA Beat Cancer Project, Medvet Laboratories Pty Ltd, and the blood.2018880930. Government of South Australia.

*T.G.P. and P.I.C. are joint senior authors.

Authorship RNA-seq data are available at SRA: PRJNA453652. Contribution: P.I.C. and T.G.P. conceived the study; W.H.K., G.L., A.W., R.L.T., M.M.M., R.C.C., K.D.V., K.L.O., K.S.O., K. Vandyke, A. Seckinger, The online version of this article contains a data supplement. N.K., D.R.H., C.S., M.N., R.S.L., T.V.N., B.P., D.H., K. Vanderkerken, A.C.W.Z., I.A., P.I.C., and T.G.P. designed and performed experiments; J.R.C., S.T.M., J.A.P., Y.X., and A.P.C. performed experiments; B.O.O. There is a Blood Commentary on this article in this issue. and D.A. provided reagents; W.H.K., G.L., A.W., D.L.R., R.L.T., M.M.M., R.C.C., K.D.V., K.L.O., K.S.O., K. Vandyke, A. Seckinger, N.K., A.N., Thepublicationcostsofthisarticleweredefrayedinpartbypage T.V.N., B.O.O., D.A., B.P., D.H., K. Vanderkerken, A.C.W.Z., I.A., P.I.C., charge payment. Therefore, and solely to indicate this fact, this article is and T.G.P. analyzed and interpreted data; P.I.C., T.G.P., and W.H.K. hereby marked “advertisement” in accordance with 18 USC section wrote the manuscript; W.H.K., G.L., A.W., A. Seckinger, B.O.O., D.A., 1734.

REFERENCES 9. Lo Celso C, Lin CP, Scadden DT. In vivo im- 17. Bolotin DA, Poslavsky S, Mitrophanov I, et al. aging of transplanted hematopoietic stem and MiXCR: software for comprehensive adaptive 1. Kumar SK, Rajkumar V, Kyle RA, et al. Multiple progenitor cells in mouse calvarium bone immunity proﬁling. Nat Methods. 2015;12(5): myeloma. Nat Rev Dis Primers. 2017;3:17046. marrow. Nat Protoc. 2011;6(1):1-14. 380-381. 2. Ghobrial IM. Myeloma as a model for the 10. Chattopadhyay PK, Gierahn TM, Roederer M, 18. Ye J, Ma N, Madden TL, Ostell JM. IgBLAST: process of metastasis: implications for ther- Love JC. Single-cell technologies for moni- an immunoglobulin variable domain se- apy. Blood. 2012;120(1):20-30. toring immune systems. Nat Immunol. 2014; quence analysis tool. Nucleic Acids Res. 2013; 3. Nijhof IS, van de Donk NWCJ, Zweegman S, 15(2):128-135. 41(web server issue):W34-W40. Lokhorst HM. Current and new therapeutic 11. Nguyen A, Khoo WH, Moran I, Croucher PI, 19. Vallejos CA, Marioni JC, Richardson S. strategies for relapsed and refractory multiple Phan TG. Single cell RNA sequencing of rare BASiCS: Bayesian Analysis of Single-Cell Se- myeloma: an update. Drugs. 2018;78(1): immune cell populations. Front Immunol. quencing Data. PLOS Comput Biol. 2015; 19-37. 2018;9:1553. 11(6):e1004333.

4. Lawson MA, McDonald MM, Kovacic N, et al. 12. Giustacchini A, Thongjuea S, Barkas N, et al. 20. Vallejos CA, Richardson S, Marioni JC. Beyond Osteoclasts control reactivation of dormant Single-cell transcriptomics uncovers distinct comparisons of means: understanding myeloma cells by remodelling the endosteal molecular signatures of stem cells in chronic changes in gene expression at the single-cell niche. Nat Commun. 2015;6(1):8983. myeloid leukemia. Nat Med. 2017;23(6): level. Genome Biol. 2016;17(1):70. 692-702. 5. Croucher PI, McDonald MM, Martin TJ. Bone 21. Butler A, Hoffman P, Smibert P, Papalexi E, metastasis: the importance of the neigh- 13. Chen Z, Orlowski RZ, Wang M, Kwak L, Satija R. Integrating single-cell transcriptomic bourhood. Nat Rev Cancer. 2016;16(6): McCarty N. Osteoblastic niche supports the data across different conditions, technologies, 373-386. growth of quiescent multiple myeloma cells. and species. Nat Biotechnol. 2018;36(5): Blood. 2014;123(14):2204-2208. 411-420. 6. Sosa MS, Bragado P, Aguirre-Ghiso JA. Mechanisms of disseminated cancer cell 14. Oyajobi BO, Muñoz S, Kakonen R, et al. 22. Huang W, Sherman BT, Lempicki RA. dormancy: an awakening field. Nat Rev Detection of myeloma in skeleton of mice by Bioinformatics enrichment tools: paths toward Cancer. 2014;14(9):611-622. whole-body optical fluorescence imaging. the comprehensive functional analysis of large Mol Cancer Ther. 2007;6(6):1701-1708. gene lists. Nucleic Acids Res. 2009;37(1):1-13. 7. Ghajar CM. Metastasis prevention by targeting the dormant niche. Nat Rev Cancer. 2015; 15. Power CA, Pwint H, Chan J, et al. A novel 23. Huang W, Sherman BT, Lempicki RA. 15(4):238-247. model of bone-metastatic prostate cancer in Systematic and integrative analysis of large immunocompetent mice. Prostate. 2009; gene lists using DAVID bioinformatics 8. Price TT, Burness ML, Sivan A, et al. Dormant 69(15):1613-1623. resources. Nat Protoc. 2009;4(1):44-57. breast cancer micrometastases reside in specific bone marrow niches that regulate their 16. Wu AR, Neff NF, Kalisky T, et al. Quantitative 24. Gaujoux R, Seoighe C. A flexible R package transit to and from bone. Sci Transl Med. 2016; assessment of single-cell RNA-sequencing for nonnegative matrix factorization. BMC 8(340):340ra73. methods. Nat Methods. 2014;11(1):41-46. Bioinformatics. 2010;11:367.

42 blood® 4 JULY 2019 | VOLUME 134, NUMBER 1 KHOO et al 25. Janky R, Verfaillie A, Imrichova´ H, et al. and mechanisms of upfront resistance. Front efficacious inhibitor of the Met kinase super- iRegulon: from a gene list to a gene regulatory Immunol. 2018;9:1676. family. J Med Chem. 2009;52(5):1251-1254. network using large motif and track collec- tions. PLOS Comput Biol. 2014;10(7): 37. Radl J, Croese JW, Zurcher C, Van den Enden- 50. Shiozawa Y, Pedersen EA, Patel LR, et al. e1003731. Vieveen MH, de Leeuw AM. Animal model GAS6/AXL axis regulates prostate cancer of human disease Multiple myeloma. Am invasion, proliferation, and survival in the 26. Shannon P, Markiel A, Ozier O, et al. J Pathol. 1988;132(3):593-597. bone marrow niche. Neoplasia. 2010;12(2): Cytoscape: a software environment for in- 116-127. tegrated models of biomolecular interaction 38. Garrett IR, Dallas S, Radl J, Mundy GR. A networks. Genome Res. 2003;13(11): murine model of human myeloma bone dis- 51. Wilkins BS. Histology of normal haemopoiesis: 2498-2504. ease. Bone. 1997;20(6):515-520. bone marrow histology I. J Clin Pathol. 1992; 45(8):645-649. 27. Bastian M, Heymann S, Jacomy M. Gephi: an 39. Brunet JP, Tamayo P, Golub TR, Mesirov JP. open source software for exploring and ma- Metagenes and molecular pattern discovery 52. Jin Y, Nie D, Li J, et al. Gas6/AXL signaling nipulating networks. https://gephi.org/ using matrix factorization. Proc Natl Acad Sci regulates self-renewal of chronic myeloge- publications/gephi-bastian-feb09.pdf. USA. 2004;101(12):4164-4169. nous leukemia stem cells by stabilizing Downloaded from https://ashpublications.org/blood/article-pdf/134/1/30/1553831/blood880930.pdf by WASHINGTON UNIVERSITY SCHOOL user on 28 August 2020 Accessed 22 July 2018. b-catenin. Clin Cancer Res. 2017;23(11): 40. Rusinova I, Forster S, Yu S, et al. Interferome 2842-2855. v20: an updated database of annotated 28. Schneider CA, Rasband WS, Eliceiri KW. NIH Image to ImageJ: 25 years of image analysis. interferon-regulated genes. Nucleic Acids 53. Bidwell BN, Slaney CY, Withana NP, et al. Res. Nat Methods. 2012;9(7):671-675. 2013;41(database issue):D1040-D1046. Silencing of Irf7 pathways in breast cancer cells promotes bone metastasis through 41. Veillette A, Thibaudeau E, Latour S. High 29. Limaye A. Drishti: a volume exploration and immune escape. Nat Med. 2012;18(8): expression of inhibitory receptor SHPS-1 and presentation tool. Proc. SPIE 8506, Develop- 1224-1231. its association with protein-tyrosine phos- ments in X-Ray Tomography VIII. 2012; phatase SHP-1 in macrophages. J Biol Chem. 85060X. doi: 10.1117/12.935640. 54. Goyette MA, Duhamel S, Aubert L, et al. The 1998;273(35):22719-22728. receptor tyrosine kinase AXL is required at 30. Rajkumar SV, Dimopoulos MA, Palumbo A, multiple steps of the metastatic cascade 42. Spilsbury K, O’Mara MA, Wu WM, Rowe PB, et al. International Myeloma Working Group during HER2-positive breast cancer pro- Symonds G, Takayama Y. Isolation of a novel updated criteria for the diagnosis of multiple gression. Cell Reports. 2018;23(5):1476-1490. macrophage-specific gene by differential myeloma. Lancet Oncol. 2014;15(12): cDNA analysis. Blood. 1995;85(6):1620-1629. e538-e548. 55. Ludwig KF, Du W, Sorrelle NB, et al. Small- molecule inhibition of Axl targets tumor 43. Liongue C, Hall CJ, O’Connell BA, Crosier P, 31. Martin SK, Diamond P, Williams SA, et al. immune suppression and enhances chemo- Ward AC. Zebrafish granulocyte colony- Hypoxia-inducible factor-2 is a novel regulator therapy in pancreatic cancer. Cancer Res. stimulating factor receptor signaling of aberrant CXCL12 expression in multiple 2018;78(1):246-255. promotes myelopoiesis and myeloid cell mi- myeloma plasma cells. Haematologica. 2010; gration. Blood. 2009;113(11):2535-2546. 95(5):776-784. 56. Palisoul ML, Quinn JM, Schepers E, et al. Inhibition of the receptor tyrosine kinase AXL 44. Bournazos S, Wang TT, Ravetch JV. The role 32. Seckinger A, Delgado JA, Moser S, et al. restores paclitaxel chemosensitivity in uterine and function of Fcg receptors on myeloid Target expression, generation, preclinical ac- serous cancer. Mol Cancer Ther. 2017;16(12): cells. Microbiol Spectr. 2016;4(6): doi: tivity, and pharmacokinetics of the BCMA- 2881-2891. 10.1128/microbiolspec.MCHD-0045-2016 T cell bispecific antibody EM801 for multiple myeloma treatment. Cancer Cell. 2017;31(3): 45. He MF, Wang S, Wang Y, Wang XN. Modeling 57. Antony J, Huang RY. AXL-driven EMT state as 396-410. cell-in-cell structure into its biological signifi- a targetable conduit in cancer. Cancer Res. cance. Cell Death Dis. 2013;4(5):e630. 2017;77(14):3725-3732. 33. Chng WJ, Kumar S, Vanwier S, et al. Molecular dissection of hyperdiploid multiple myeloma 46. Duelli D, Lazebnik Y. Cell fusion: a hidden 58. Lendvai N, Yee AJ, Tsakos I, et al. Phase IB by gene expression profiling. Cancer Res. enemy? Cancer Cell. 2003;3(5):445-448. study of cabozantinib in patients with relapsed 2007;67(7):2982-2989. and/or refractory multiple myeloma. Blood. 47. Asosingh K, Radl J, Van Riet I, Van Camp B, 2016;127(19):2355-2356. 34. Mulligan G, Mitsiades C, Bryant B, et al. Gene Vanderkerken K. The 5TMM series: a useful expression profiling and correlation with in vivo mouse model of human multiple my- 59. Suominen MI, Fagerlund KM, Alhoniemi E, outcome in clinical trials of the proteasome eloma. Hematol J. 2000;1(5):351-356. Rissanen JP, Halleen JM, Aftab DT. Effects of inhibitor bortezomib. Blood. 2007;109(8): combination treatment with cabozantinib and 3177-3188. 48. Yakes FM, Chen J, Tan J, et al. Cabozantinib bortezomib in the 5TGM1 murine multiple (XL184), a novel MET and VEGFR2 inhibitor, myeloma model. In: Proceedings from the 35. Seckinger A, Meissner T, Moreaux J, et al. simultaneously suppresses metastasis, angio- 106th Annual Meeting of the American As- Clinical and prognostic role of annexin A2 in genesis, and tumor growth. Mol Cancer Ther. sociation for Cancer Research; 18-22 April multiple myeloma. Blood. 2012;120(5): 2011;10(12):2298-2308. 2015; Philadelphia, PA. Abstract 786. 1087-1094. 49. Schroeder GM, An Y, Cai ZW, et al. Discovery 60. Hose D, Reme` T, Hielscher T, et al. 36. Seckinger A, Hillengass J, Emde M, et al. of N-(4-(2-amino-3-chloropyridin-4-yloxy)-3- Proliferation is a central independent prog- CD38 as immunotherapeutic target in light fluorophenyl)-4-ethoxy-1-(4-fluorophenyl)-2- nostic factor and target for personalized and chain amyloidosis and multiple myeloma: as- oxo-1,2-dihydropyridine-3-carboxamide risk-adapted treatment in multiple myeloma. sociation with molecular entities, risk, survival, (BMS-777607), a selective and orally Haematologica. 2011;96(1):87-95.

A GENE SIGNATURE DEFINING DORMANT MYELOMA CELLS blood® 4 JULY 2019 | VOLUME 134, NUMBER 1 43