Inhibition of Ephrin B2 Reverse Signaling Abolishes Multiple Myeloma Pathogenesis

Ephrin B2 reverse signaling mediated by endothelial cells directly regulates multiple myeloma progression and treatment resistance, which can be overcome through targeted inhibition of ephrin B2 to abolish myeloma.

Abstract

Bone marrow vascular endothelial cells (BM EC) regulate multiple myeloma pathogenesis. Identification of the mechanisms underlying this interaction could lead to the development of improved strategies for treating multiple myeloma. Here, we performed a transcriptomic analysis of human ECs with high capacity to promote multiple myeloma growth, revealing overexpression of the receptor tyrosine kinases, EPHB1 and EPHB4, in multiple myeloma–supportive ECs. Expression of ephrin B2 (EFNB2), the binding partner for EPHB1 and EPHB4, was significantly increased in multiple myeloma cells. Silencing EPHB1 or EPHB4 in ECs suppressed multiple myeloma growth in coculture. Similarly, loss of EFNB2 in multiple myeloma cells blocked multiple myeloma proliferation and survival in vitro, abrogated multiple myeloma engraftment in immune-deficient mice, and increased multiple myeloma sensitivity to chemotherapy. Administration of an EFNB2-targeted single-chain variable fragment also suppressed multiple myeloma growth in vivo. In contrast, overexpression of EFNB2 in multiple myeloma cells increased STAT5 activation, increased multiple myeloma cell survival and proliferation, and decreased multiple myeloma sensitivity to chemotherapy. Conversely, expression of mutant EFNB2 lacking reverse signaling capacity in multiple myeloma cells increased multiple myeloma cell death and sensitivity to chemotherapy and abolished multiple myeloma growth in vivo. Complementary analysis of multiple myeloma patient data revealed that increased EFNB2 expression is associated with adverse-risk disease and decreased survival. This study suggests that EFNB2 reverse signaling controls multiple myeloma pathogenesis and can be therapeutically targeted to improve multiple myeloma outcomes.

Significance:

Ephrin B2 reverse signaling mediated by endothelial cells directly regulates multiple myeloma progression and treatment resistance, which can be overcome through targeted inhibition of ephrin B2 to abolish myeloma.

Graphical Abstract

Introduction

Multiple myeloma is one of the most common hematologic malignancies (1), and although patient survival has increased over the past two decades as a result of modern therapies, multiple myeloma remains an incurable disease (2). Multiple myeloma presents in early stages in the bone marrow (BM) and cells within the BM niche support the pathogenesis of the disease (3–7). For example, BM osteoblasts endow quiescent multiple myeloma cells with colony-forming potential (8, 9) and BM fibroblasts promote multiple myeloma cell growth and survival in the BM (9). Multiple myeloma cells support their own growth and disease progression via production of proangiogenic signals (10–13) and endothelial cells (EC) can support multiple myeloma cell niches within the BM (14). Angiogenic myeloid cells can also drive the pathologic evolution of multiple myeloma via expansion of a multiple myeloma vascular niche (15). Furthermore, immunomodulatory drugs, such as lenalidomide and thalidomide, suppress multiple myeloma growth in patients, at least in part, via antiangiogenic effects (16). However, the precise mechanisms through which BM ECs promote multiple myeloma growth within the niche remain poorly understood.

EPH proteins are transmembrane tyrosine kinase receptors that bind to membrane-anchored ligands, called ephrins, and play an important role in short-distance cell–cell communication during morphogenesis, axon guidance during nervous system development, and vascular development (17–20). Uniquely, EPH–Ephrin complexes can produce bidirectional signaling, thereby affecting the EPH receptor–expressing cell (forward signaling) and the ephrin-expressing cell (reverse signaling), and ephrins can also regulate signals by EPH receptors in the same cell (17). Reverse signaling through ephrin B2 has been shown to regulate diverse organ functions, such as axon synapse formation and chemoattraction (21, 22), pancreatic islet formation (23), and angiogenesis (24). In homeostasis, EPHB4 is expressed by BM ECs and EPHB4 signaling through Ephrin B2 regulates hematopoietic progenitor cell mobilization (25). EPH receptors are also expressed in several human carcinomas (26), and EPH receptor expression has been associated with progression of cancer disease (27). For example, EPHA2 expression has been associated with increased brain metastasis in non–small cell lung cancer (26, 27) and therapy resistance (28). EPHA2 expression also transforms normal breast epithelial cells (29) and promotes HER2 inhibitor resistance (30). Interestingly, expression of the EPH ligand, ephrin B2, has been linked to poor response to chemotherapy, decreased overall survival and disease-free survival in head and neck squamous cell carcinoma, pancreatic and bladder carcinoma (31). Taken together, these studies suggest that EPH–Ephrin signaling may have an important role in regulating the pathogenesis of several human cancers.

Via differential transcriptomic analysis, we discovered that genes encoding EPH receptor tyrosine kinases, EPHB1 and EPHB4, were upregulated in human ECs that support multiple myeloma growth, and that EFNB2, the transmembrane ligand for EPHB1 and EPHB4, is highly expressed by multiple myeloma cells. Targeted inhibition of EFNB2 reverse signaling suppressed multiple myeloma survival and proliferation in vitro, increased sensitivity to chemotherapy and extinguished multiple myeloma growth in vivo. EFNB2 reverse signaling represents an important new mechanistic target in multiple myeloma.

Materials and Methods

Cell lines and culture

Multiple myeloma cells (U266, NCI-H929, RPMI-8226) were purchased from ATCC and cultured in RPMI1640 medium containing 25 mmol/L HEPES, l-Glutamine (Thermo Fisher Scientific, 11875093) and 15% FBS. Cells were incubated in 37°C and 5% CO₂. Genetically modified cell lines were selected using antibiotic resistance and used for experimentation below passage 20 after engineering. Genetically engineered cells (shRNA, dCas9-VPR, and gRNA) were grown in media containing 4 μg/mL of puromycin (Sigma-Aldrich, P9620) and/or 8 μg/mL of blasticidin (Thermo Fisher Scientific, A1113903), depending on the resistance of the plasmids. BMEC-60 cells were provided courtesy of Dr. Ruben Carassco (Dana Farber Cancer Institute, Department of Pathology, Boston, MA; ref. 6). Human umbilical vein ECs (HUVEC) were provided courtesy of Dr. Shahin Rafii (Division of Regenerative Medicine, Weill Cornell Medicine, New York, NY; ref. 32). All other ECs were purchased from ATCC. Flasks or 24-well plates were precoated with 10% gelatin (Sigma Aldrich, G1393–100ML) for 30 minutes, then washed with PBS. BM ECs were cultured in EBM-2 Complete Culture Media (Lonza, CC-3156). Once they were 80% confluent, BM ECs were detached using a mixture of 0.05% Trypsin (Thermo Fisher Scientific, 25200056) and Versene (Thermo Fisher Scientific, 15040066) at a 1:8 ratio. This was inactivated with five times the volume of EBM-2 Complete Culture Media (Lonza). The cells were then centrifuged and passed at a 1:3 ratio. When in coculture conditions, cells were counted by flow cytometry using 123count eBeads Counting Beads (Thermo Fisher Scientific, 01–1234–42) according to the manufacturer's instructions. Mycoplasma testing was performed every 6 months using a PCR-based kit (ATCC, 30–1012K; latest test date August 23, 2023).

Colony-forming cell assay

Multiple myeloma cells (2 × 10³) were seeded in 35-mm gridded culture dishes in MyeloCult M5300 media (StemCell Technologies, 05300) and colonies were scored at day 14.

In vitro luciferase assay

Luciferin solution (50 μL) was added to cells in 24-well plates. RPMI-8226 cells that constitutively express firefly luciferase were utilized for these studies (courtesy of Dr. Alan Lichtenstein, Hematology/Oncology, VA West LA Hospital, Los Angeles, CA). Immediately after the addition of the substrate solution, luminescence was measured using a Luminoskan Ascent Luminometer with Ascent Software (Thermo Fisher Scientific).

Viral transduction

A TC10 Cell counter was used to check the viability of the culture populations. All cell lines displayed viability > 95% before undergoing viral transduction. Forty thousand cells were placed in 96-well U-bottom tissue culture plates containing 200 mL of RPMI media supplemented with 1% FBS and 8 mg/mL of polybrene (Sigma Aldrich, TR-1003-G). Cells were plated with varying multiplicity of infections to determine optimal virus dose. At day 2, the virus was washed out and cells were replated with media containing 15% FBS. On day 4, the samples were transferred into selection media containing the appropriate antibiotic(s). The hEFNB2deltaC-EGFP construct was previously described (33), but it was in the pEGFP-N2 vector. The portion encoding hEFNB2deltaC-EGFP was subcloned into pLVX-IRES-Neo (using XhoI and NotI restriction sites) for use in these studies.

To silence EPHB1 expression in human aortic ECs (HAEC), HAECs were cultured with complete endothelial cell growth medium (EBM-2 medium, Lonza) and cells were transfected with lentiviral particles that express human EPHB1 shRNA or nontarget control shRNA and GFP marker (Horizon Discovery). Cells expressing shRNA and GFP were selected with 1 μg/mL puromycin (Sigma Aldrich, P7255–100MG).

EC growth and angiogenesis assay

HAECs (2.5 × 10⁴ cells per well) transfected with EPHB1 shRNA or nontarget shRNA were seeded in 6-well plates in triplicates in complete EBM-2 medium with 1 μg/mL puromycin. Viable cells were counted by Trypan blue exclusion on the indicated incubation time points and cell growth cure was plotted.

To measure EC angiogenesis, HAECs (2.5×10⁴ cells per well) transduced with EPHB1 shRNA or nontarget shRNA were seeded in four replicates onto the top of the EC matrix in 96-well plate according to the instructions of the In Vitro Angiogenesis Assay Kit (Millipore Sigma, ECM625). EC matrix is a solid gel of basement proteins containing laminin, collagen type IV, heparan sulfate proteoglycans, entactin, and nidogen. It also contains various growth factors (TGFβ, FGF) and proteolytic enzymes (plasminogen, tPA, MMPs). HAECs were cultured at 37°C overnight. EC tube formation was inspected and photographed at 50x magnification using a Zeiss inverted light microscope (Carl Zeiss Microscopy), and branch point counting was performed for quantification analysis.

Expression analysis of human BM cells

Human BM cells from cadaveric donors not diagnosed with multiple myeloma were analyzed for EFNB2 expression (Ossium Health) using Ficoll-Paque Plus (Thermo Fisher Scientific, 45–001–749), and washed with RPMI medium with 10% FBS. We sorted CD11b⁺ myeloid cells, CD269 (BCMA)⁺ plasma cells, CD3⁺ T cells, and CD19⁺ B cells using anti-human CD3-BV421 (BioLegend, 317344), CD19-PE (BioLegend, 302208), CD269-AF647 (BioLegend, 357518), and CD11b-FITC (BioLegend, 982614) antibodies; cells were sorted using a BD Influx cell sorter (Becton Dickinson).

qRT-PCR

As previously described (34), RNA isolation was performed using Qiagen Micro RNeasy Kits. RNA concentrations were measured with the Nanodrop 2000 Spectrophotometer (Thermo Fisher Scientific, ND-2000). Five-hundred nanograms of total RNA was reverse transcribed using the High Capacity cDNA Synthesis Kit (Thermo Fisher Scientific, 4368814). Real-time PCR reactions were conducted using TaqMan Universal PCR Master Mix (Thermo Fisher Scientific, 4304437) on an ABI Quant Studio 6-machine instrument with TaqMan predesigned gene expression assays (Thermo Fisher Scientific, 4331182; ref. 34). Primers and probes were produced by ABI and real-time PCR was performed on an ABI Quant Studio 6-machine instrument. All reactions were normalized using GAPDH.

Flow cytometry and Annexin V analysis

BM cells were analyzed by flow cytometry to identify multiple myeloma cells and mouse leukocytes using the following antibodies: APC anti-human CD138 (BioLegend, 352308), APC anti-mouse CD138 (BioLegend, 142506), V450 Rat Anti-mouse CD45 (BD Biosciences, 56050), and PE anti-human CD34 (BD Biosciences, 550761). Discrimination of live and apoptotic cell fractions was performed by flow cytometry using the BD Annexin V-FITC Apoptosis Detection Kit (BD Biosciences, 556570), following the manufacturer's instructions. In addition, 7-amino-actinomycin (7AAD, BD Biosciences, 559925) staining was used to discriminate apoptosis (Annexin V⁺ 7ADD⁻) from necrosis (Annexin V⁺ 7ADD⁺). Data were analyzed on FACS Diva version 8 and FlowJo version 10 software.

Phosflow

Multiple myeloma cells were incubated in serum-free media for 2 hours. Cells were stimulated with either human IgG1 mAb (R&D Systems, MAB9894-SP) or preclustered recombinant mouse EPHB4 Fc Chimera Protein (R&D Systems, 446-B4). Using a 1:1 ratio, human recombinant EPHB4 Fc (R&D Systems, 11307-B4) was preclustered with human IgG1 mAb for 30 minutes at room temperature. Cells were stimulated in a volume of 200 μL at 37°C for 15 minutes using a concentration of 10 μg/mL of EPHB4-Fc and IgG1. Following stimulation, 1 mL of prewarmed 1× Phosflow Lyse/Fix Buffer (BD Biosciences, 558049) was added and incubated for 10 minutes at 37°C. Cells were then washed and resuspended. Five-hundred microliters of Perm Buffer III (BD Biosciences, 558050) was added to each sample and placed on ice for 30 minutes. Samples were washed with 1 mL of Pharmingen Stain Buffer (BD Biosciences, 554656). Cells were separately stained with BD Phosflow Alexa Fluor 647 mouse anti-Stat3 (BD Biosciences, 557815) or Alexa Fluor 647 mouse anti-Stat5 (BD Biosciences, 612599). Cells were stained in the dark overnight at 4°C and analyzed using a FACSCanto (BD Biosciences). Isotype controls were used to delineate nonspecific binding.

Bromodeoxyuridine analysis

Cells were incubated with bromodeoxyuridine (BrdU) for 6 hours and processed according to the manufacturer's protocol for the in vitro labeling of cells (BrdU BD Pharmingen BrdU Flow Kit, 559619). For inhibition of STAT5, cells were treated with 2-[(4-oxo-4H-1-benzopyran-3-yl)methylene]hydrazide 3-pyridinecarboxylic acid (MedChemExpress).

Vk*MYC mouse model

Vk*MYC cells were obtained courtesy of Dr. Leif Bergsagel (Mayo Clinic, Scottsdale, AZ). All Vk*MYC studies were done in mice in a C57BL/6 background. Vk*MYC plasma cell clones were passed from mouse to mouse to preserve the line. Recipient C57BL/6 mice were injected via tail vein with 1–2 × 10⁵ cells. Mice were sacrificed once a monoclonal protein band (M-protein) was detected on serum protein electrophoresis or upon signs of significant morbidity. All mice developed M-proteins by 4 to 6 weeks after tumor cell injection.

Human multiple myeloma–NSG xenograft model

NSG mice (JAX strain no. 005557, The Jackson Laboratory) for human multiple myeloma xenograft studies were purchased from the Jackson Laboratory. Comparative survival studies of xenografted mice were done at least twice independently. A subset of mice were treated with anti-human EFNB2 single chain variable fragment (scFv)(B11)[TAB-390MZ-S(P), Creative BioLabs] or control antibody. Flow cytometric analyses were performed to detect human CD138⁺ cells engrafted in the BM, peripheral blood, and vertebral bones; some samples were lost due to poor viability during processing and analysis. Mice were monitored daily for signs of morbidity. Mice were humanely euthanized when symptoms of morbidity such as weight loss ≥ 15%, immobility, or hunched posture were observed.

Serum protein electrophoresis

Blood was collected from mice via submandibular venipuncture into Microtainer tubes (BD Biosciences, 365992), allowed to coagulate at room temperature, and spun for 10 minutes at 2,300 × g. Serum was diluted 1:2 in normal saline buffer and analyzed on a QuickGel Chamber apparatus using precasted QuickGels (Helena Laboratories, 1284), according to the manufacturer's instructions.

Human patient samples

Primary multiple myeloma samples were obtained from BM aspirates of patients with multiple myeloma from the UCLA Ronald Reagan Medical Center under a UCLA Institutional Review Board (IRB)–approved protocol (principal investigator: Joshua Sasine, protocol no. 15–000062) after written informed consent was obtained from each patient. The study was performed in accordance with Declaration of Helsinki guidelines. Deidentified, healthy control BM samples were obtained from AllCells. BM microenvironment cells and CD269⁺ (TNFRSF17/BCMA) multiple myeloma cells were isolated from primary human BM aspirates (34) as follows. Aspirates were diluted 1:1 in sterile PBS, layered over Ficoll-Paque density gradient medium (Thermo Fisher Scientific, 45–001–749) and centrifuged at 1,400 rpm for 30 minutes. Mononuclear cells were collected and column-enriched for CD269⁺ cells using the Miltenyi Biotec human CD269⁺ isolation kit (Miltenyi Biotec, 130–131–092). The negative cell fraction was plated in BMEC media to generate BM ECs from patients with multiple myeloma.

ELISA

One-hundred microliters of mouse peripheral blood was collected in 1.5 mL Eppendorf tubes. After centrifugation, supernatants were collected and analyzed for human lambda light-chain protein levels using an ELISA kit (Thermo Fisher Scientific, E88–116), as per the manufacturer's instructions.

RNA sequencing

RNA was isolated using Qiagen RNeasy Micro kit (Qiagen, 74104). Libraries for RNA-seq were prepared with Nugen Universal plus mRNA-Seq Kit to generate strand-specific RNA-seq libraries. The workflow consisted of poly(A) RNA selection, RNA fragmentation and double-stranded cDNA generation using a mixture of random and oligo(dT) priming, followed by end repair to generate blunt ends, adaptor ligation, strand selection, and PCR amplification to produce the final library. Different index adaptors were used for multiplexing samples in one sequencing lane. Sequencing was performed on NovaSeq 6000 for PE 2×150 bp run. Data quality check was done on Illumina SAV. Demultiplexing was performed with Illumina Bcl2fastq2 v 2.19.1.403 program. The Partek flow (Partek Inc.) was used for bioinformatics methods. Reads were mapped to the latest UCSC transcript set using STAR – 2.7.2a and GRCm38.97 (35). After obtaining gene counts, they were normalized by counts per million. The principal component analysis was applied to the transcript counts. The differential gene expressions were examined. Ingenuity Pathway Analysis (IPA) was used for data analysis (36). For all results of differential gene expression analysis, the P values and fold change (FC) filters were applied. The filter values were P < 0.05 and |FC| > 2 for all differential gene expression results.

Microarray analysis

RNA was isolated using the QIAGEN RNeasy micro kit (QIAGEN). RNA was then sent to the UCLA Clinical Microarray Core (CMC) for processing. Total RNA was assessed for quality with Agilent 2100 Bioanalyzer G2939A (Agilent Technologies) and Nanodrop 8000 spectrophotometer (Thermo Scientific/Nanodrop). The samples were amplified using the NuGEN Ovation RNA Amplification System V2 (catalog no. 3100-A01; NuGEN Technologies Inc.) and were then hybridized to Affymetrix Human Genome U133 Plus 2.0 Array. Affymetrix GeneChip CEL files for the samples assayed were imported into the Affymetrix Expression Console software to perform gene level normalization and signal summarizations. Affymetrix Transcriptome Analysis Console (TAC) software was used to analyze the differential gene expression between the cell types. Robust multiarray average (RMA) values generated in the Affymetrix Expression Console software from this gene list were imported into Partek microarray data analysis software (Partek Inc.) and unsupervised hierarchical cluster analysis was performed on the various gene lists. For all results of differential gene expression analysis, statistical filters were applied: P < 0.05, FDR < 0.05, and |FC|>2. Within the differentially expressed gene set, QIAGEN's IPA (QIAGEN, www.qiagen.com/ingenuity) was applied to identify enriched canonical pathways, diseases, and biological functions.

Immunofluorescence imaging

Seven-week-old NSG mice were injected with 1 × 10⁷ RPMI 8226 multiple myeloma cells at 24 hours following 180 cGy TBI. At week 3 and 5 posttransplant, mice were injected via tail vein with 20 μg AF647 VE-Cadherin antibody (BioLegend, 138006). One hour after antibody injection, mice were euthanized and femurs were collected. Femurs were fixed in 10 mL of 4% paraformaldehyde for 4 hours at 4°C. Femurs were then washed three times with PBS under agitation and then incubated in 10 mL of 0.5 mol/L EDTA solution (pH 7.4) for 24 hours under constant agitation at 4°C. After incubation, the bones were washed three times with PBS and then incubated in a cryoprotectant solution (10% sucrose, 1% PVP in PBS) for 24 hours at 4°C. Femurs were then embedded in OCT medium, frozen on dry ice, and then stored at −80°C. Sections (10 μm) were taken using a Leica CM3050 S. All slides were rehydrated with PBS, permeabilized with 0.30% Triton X-100 in PBS, and treated with a blocking solution (5% BSA, 5% goat serum in PBST). Slides were stained with primary antibodies to BCMA (1:250; R&D Biosystems, MAB107621, and BioLegend, 357518), EPHB4 (1:250, R&D Biosystems, MAB446), EFNB2 (1:100, Invitrogen, MA5–35098), and secondary antibodies against rat (1:1,000, Invitrogen, A11006) and rabbit (1:1,000, Invitrogen, A21428) and DAPI (Thermo Fisher Scientific, D1306). Slides were mounted using Dako Fluorescence Mounting Medium (Agilent, S302380–2). Imaging of femur sections was done on a Zeiss confocal microscope using a 63× objective.

Cytospins were performed on RPMI 8226 multiple myeloma cells and processed for confocal imaging. Slides dried overnight at room temperature and then fixed with 100% methanol for 10 minutes, followed by drying at room temperature for 2 hours. Cytospin slides were then processed identically to the femur slides, starting at rehydration with PBS.

Clinical data analysis

Data were taken from two sources: The Multiple Myeloma Research Foundation (MMRF) CoMMpass trial and the Total Therapy Studies (37). Data from the Total Therapy trials were gathered from Oncomine. The MMRF CoMMpass trial included patients with newly diagnosed multiple myeloma with sufficient tumor material for the comprehensive analysis of each tumor genome and transcriptome. Patients were followed for their clinical outcomes longitudinally for a minimum of 8 years. The clinical study (NCT01454297) opened in July 2011 and includes dozens of sites in the United States and Canada that have enrolled hundreds of patients. The comprehensive analysis of each tumor and matched normal genome involved numerous analyses, among them was RNA sequencing (RNA-seq) to define transcript expression levels and fusion transcripts. An open-access, public clinical and molecular database, the CoMMpass Researcher Gateway (RG; https://research.themmrf.org), is available. Detailed methods used in these trials have been previously published (37–39).

Statistical analysis

We utilized GraphPad Prism 6.0 for all statistical analyses and data are shown as means ± SD, unless otherwise indicated. The details of each experiment, including n values and type of statistical test used and P values, are shown in the figure legends. As previously described (34), data were verified for normal distribution and the statistical test utilized was based on the numbers of groups and variables. For one-way and two-way ANOVA analyses, P values were based on multiple test corrections. Sample sizes for in vitro studies were chosen based on observed effect sizes and SEs from prior studies. As we have previously described for animal studies (34), a power test was used to determine the sample size needed to observe a two-fold difference in means between groups with 0.8 power using a two-sided Student t test. Age-matched and sex-matched animals were utilized throughout. Investigators were not blinded and no animals were excluded from the analysis.

Study approval

All animal procedures were performed in accordance with Animal Use Protocol IACUC 009617 approved by the Cedars Sinai Animal Use Committee (principal investigator: J. Chute) and UCLA Protocol 2014–021–13M (principal investigator: J. Chute). All human sample acquisition and analyses were performed in accordance with UCLA IRB 15–000062-CR-00004 (principal investigator: J. Sasine), approved by the UCLA IRB.

Data availability

Raw microarray data have been deposited at NCBI in the Gene Expression Omnibus (GEO) repository at GSE201846. Raw RNA-seq data for this study were generated at the UCLA Technology Center for Genomics and Bioinformatics. Derived data supporting the findings in the study are available in FigShare at hyperlink code https://figshare.com/projects/RNA-seq_of_Multiple_Myeloma_Cells_with_EFNB2_shRNA/191469. The data underlying all findings of these studies are available from the corresponding authors upon request and provided as separate source data files.

The clinical data analyzed in this study were obtained from The Multiple Myeloma Research Foundation (MMRF) CoMMpass trial at https://research.themmrf.org and the Total Therapy Studies (37) at Oncomine.

Results

Human ECs promote multiple myeloma cell growth in vitro

As a model to understand how ECs regulate multiple myeloma growth, we cocultured the human multiple myeloma cell line, U266 (ATCC), with different sources of human ECs. Coculture of U266 cells with HUVECs (32), HAECs, or the BM EC cell line, BMEC-60 (6), for 7 days caused a significant increase in CD138⁺ multiple myeloma cells in culture and multiple myeloma colony-forming cells (40) compared with media alone, whereas coculture with human dermal ECs (HDEC) had no effect (Fig. 1A and B). We observed a similar level of expansion using RPMI-8226 multiple myeloma cells (Fig. 1C). To evaluate whether coculture with ECs could expand human multiple myeloma cells capable of reconstituting disease in vivo, we transplanted the progeny of 1×10⁴ U266 cells at day 7 of coculture with HAECs or HDECs into NOD/SCID-IL2 receptor-gamma null (NSG) mice and compared with the progeny of the same dose of U266 cells cultured with media alone. At 8 weeks posttransplantation, mice transplanted with multiple myeloma cells cultured with HAECs displayed significantly increased engraftment with human CD138⁺ multiple myeloma cells in the BM and approximately 10-fold increased human lambda light chain levels in the PB compared with mice transplanted with multiple myeloma cells cultured with HDECs or media alone (Fig. 1D and E).

Figure 1.

Differential molecular analysis of human ECs that promote multiple myeloma (MM) growth. A, U266 multiple myeloma cells were cocultured with the human ECs shown for 7 days, then stained for CD138 and counted using flow cytometry. Fold change was normalized to media alone group. n = 6/group. B, Following the same conditions as in A, cells were placed into methylcellulose after 7 days and colony-forming cells were quantified. C, RPMI-8226 multiple myeloma cells were cocultured with ECs and counted as in A. n = 6/group. D and E, U266 multiple myeloma cells were cultured alone or with ECs for 7 days and then transplanted into NSG mice. A total of 1 × 10⁴ multiple myeloma cells were cultured in 24-well plates with ECs and the culture progeny were transplanted. Disease burden was quantified on the basis of human CD138⁺ (and mouse CD45⁻) cells in the BM at 8 weeks and by quantification of PB human lambda light-chain levels over time. n = 6/group, AVOVA was used for light-chain analysis. F, Heat map of differentially expressed genes in HAECs versus HDECs. Heat map is relative to the maximum and minimum reads for each gene. G, IPA diseases and functions of HAEC versus HDEC. H, Heat map of selected EC candidate genes with increased expression in HAECs. One-way ANOVA and post hoc t test for all analyses. Error bars, SD. ns, not significant; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

Because HAECs differentially supported the expansion of multiple myeloma cells in vitro, we performed microarray analysis to identify genes with higher or lower expression in HAECs compared with HDECs. Microarray analyses revealed numerous differentially expressed genes in HAECs compared with HDECs (Fig. 1F; Supplementary Fig. S1A). IPA for diseases and functions demonstrated that genes involved in regulation of cancer, organismal injury, cell cycle, and cellular movement were most significantly enriched in the multiple myeloma–supportive HAECs (Fig. 1G). As an additional strategy to identify genes expressed by HAECs that were involved in promoting multiple myeloma growth, we filtered for secreted and cell surface proteins, and proteins for which a receptor is expressed by multiple myeloma cells. Several genes in this category were found to be upregulated in HAECs compared with HDECs, including genes that encode adhesion proteins, chemokines, cytokines, and growth factors (Fig. 1H).

Expression of EPHB1 and EPHB4 in ECs regulates multiple myeloma cell growth

We utilized siRNA knockdown of specific genes that are overexpressed by HAECs coupled with coculture of CD138⁺ multiple myeloma cells to identify HAEC genes important for supporting multiple myeloma growth. To avoid cell line–specific effects, we evaluated both U266 multiple myeloma cells and NCI-H929 multiple myeloma cells. Silencing the expression of several candidate genes in HAECs had no effect on multiple myeloma growth in coculture (Supplementary Fig. S1B) but silencing the expression of EPH receptor B1 (EPHB1) in HAECs suppressed multiple myeloma cell expansion in coculture (Fig. 2A and B). We validated the change in expression seen on the microarray using RT-PCR (Supplementary Fig. S1C and S1D) and siRNA knockdown (Supplementary Fig. S1E). Silencing EPHB1 expression in HAECs using a lentiviral shRNA (sh-EPHB1, Supplementary Fig. S1F) also abrogated the expansion of RPMI-8226 multiple myeloma cells in a bioluminescence assay (41), compared with HAEC cocultures treated with sh-Control (Fig. 2C). More generally, we observed that silencing EPHB1 suppressed HAEC growth in culture and inhibited the angiogenesis capacity of HAECs compared with sh-Control–treated HAECs (Supplementary Fig. S1G–S1I). Further analysis of primary human samples confirmed that EPHB1 expression is relatively low in BM ECs from healthy donors but is significantly increased in BM ECs from patients with multiple myeloma (Fig. 2D).

Figure 2.

Loss of EPHB1–Ephrin B2 signaling impairs multiple myeloma pathogenesis. A and B, Mean fold change of CD138⁺ cells in cocultures of U266 (A) and NCI-H929 (B) multiple myeloma (MM) cells with HAECs for 5 days. Multiple myeloma cells were treated with si-EPHB1 or si-Control. Fold change is relative to input (n = 6/group). C, Average radiance of RPMI-8226 multiple myeloma cells cocultured for 1 week with HAECs transduced with sh-EPHB1 or sh-Control. Multiple myeloma cells were tagged with luciferase and luminescence was the readout for multiple myeloma cell expansion. n = 6/group. D, Comparison of EPHB1 expression by qRT-PCR across multiple human tissues: BMECs from a healthy donor (BMEC), the BMEC-60 cell line (BMEC-60), primary multiple myeloma patient ECs (multiple myeloma BM EC1 and 2) and HAECs. EPHB1 expression is shown relative to normal BM ECs and normalized using GAPDH. n = 5/group. E, Expression of EFNB2 in U266 multiple myeloma cells, NCI-H929 multiple myeloma cells, primary multiple myeloma patient cells (MMPT1 and 2), and RPMI-8226 multiple myeloma cells. Data represent fold change in expression by qRT-PCR following reverse transcription, normalized using GAPDH. Median value for U266 was used as a basis for relative expression. F, Cell counts of RPMI-8226 multiple myeloma cells over time, following treatment with shRNAs targeting EFNB2 (sh-EFNB2) or sh-Control. Cells were grown in absence of ECs. Results are mean ± SD using three independent shRNA sequences. n = 5/group, analyzed with ANOVA. G, Average radiance of RPMI-8226 multiple myeloma cells at day 5 of coculture with BMECs following treatment with sh-EFNB2 or sh-Control. n = 6/group. H, Left, representative flow cytometric analysis of Annexin V⁺7AAD⁺ (necrotic) and Annexin V⁺7AAD⁻ (apoptotic) cells at 4 weeks of culture of RPMI-8226 multiple myeloma cells treated with sh-EFNB2 or sh-Control. Right, mean levels of total Annexin V⁺ cells within each treatment group. I, Mean percentages of Annexin V⁺ cells within RPMI-8226 multiple myeloma cells treated with sh-EFNB2 or sh-Control and exposed to bortezomib (Bor), dexamethasone (Dex), or lenalidomide (Len) for 24 hours. n = 6/group. J, Mean %BrdU⁺ RPMI-8226 multiple myeloma cells after 4 hours of BrdU incubation under the conditions shown. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

We next tested whether Ephb1 expression is also increased in BM ECs in transgenic Vk*MYC mice, which develop indolent myeloma as a consequence of activation-induced deaminase-dependent MYC activation in germinal B cells (42, 43). We found that Ephb1 expression is increased significantly in BM ECs from Vk*MYC mice that have active multiple myeloma compared with BM ECs from unaffected control mice (Supplementary Fig. S2A). Of note, BM ECs in Vk*MYC mice also displayed increased expression of Ephb4 compared with BM ECs from control mice (Supplementary Fig. S2A). EPHB4 was also the most highly expressed EPHB receptor in the human BM EC line, BMEC-60 (Supplementary Fig. S2B). shRNA-mediated silencing of EPHB4 in BMEC-60 cells reduced the capacity of BMEC-60 cells to support multiple myeloma growth in coculture (Supplementary Fig. S2C and S2D).

Silencing EFNB2 suppresses multiple myeloma growth

Both EPHB1 and EPHB4 are expressed by ECs and their signaling capacity is activated by binding to transmembrane ephrin B ligands expressed in adjacent cells (44). In the human genome, there are five EPHB receptors that bind to three transmembrane ephrin B ligands (44). Interactions are promiscuous within each A or B class of receptors and ligands. EPHB1, for example, regulates neuronal growth by binding EFNB2, but can also interact with other ephrin B ligands (45).

We detected expression of EFNB2, in three multiple myeloma cell lines and two primary multiple myeloma patient cell populations, whereas EFNB1 was not detectable, and EFNB3 was expressed at a low level in U266 cells (Fig. 2E; Supplementary Fig. S3A). Of note, analysis of cadaveric human BM cells revealed that EFNB2 is highly expressed by human B and T cells, with lower expression in normal plasma cells, CD34⁺ hematopoietic stem/progenitor cells and myeloid cells (Supplementary Fig. S3B). Because RPMI-8226 multiple myeloma cells displayed the highest levels of EFNB2 expression, we treated RPMI-8226 cells with shRNAs targeting EFNB2 (Supplementary Fig. S3C); this caused a nearly 70-fold reduction in growth of RPMI-8226 cells at day 12 of culture in all three conditions (Fig. 2F). Multiple myeloma cells also express EPHB receptors (Supplementary Fig. S3D and S3E), which could interact with EFNB2 and stimulate its reverse signaling capacity. Importantly, when we cultured human RPMI-8226 multiple myeloma cells harboring shRNA targeting EFNB2 (sh-EFNB2) with BMEC-60 cells, this abrogated BMEC-60–mediated expansion of multiple myeloma cells in 5-day culture, showing that BM EC–mediated expansion of multiple myeloma cells in culture was dependent on EFNB2 signaling (Fig. 2G). Silencing EFNB2 was also associated with increased multiple myeloma cell death compared with sh-control–treated multiple myeloma cells (Fig. 2H). Silencing EFNB2 in multiple myeloma cells also augmented multiple myeloma cell death in 24-hour culture with antimyeloma therapies, including the proteasome inhibitor, bortezomib, dexamethasone, or the immunomodulatory drug, lenalidomide (Fig. 2I). Silencing EFNB2 also decreased multiple myeloma proliferation measured by 4-hour BrdU incorporation (Fig. 2J).

Inhibition of EFNB2 suppresses multiple myeloma growth in vivo

Because silencing EPHB1 and EPHB4 expression in ECs suppressed human multiple myeloma growth in coculture and inhibition of EFNB2 in human multiple myeloma cells inhibited multiple myeloma cell growth in vitro, we sought to determine whether EFNB2-expressing multiple myeloma cells would colocalize with EPHB-expressing BM ECs in vivo. We chose EPHB4 because it is the preferred binding partner for EFNB2 and is highly and specifically expressed on sinusoidal BM ECs (25). Immunofluorescence analysis of RPMI-8226 multiple myeloma cell cytospins demonstrated coexpression of B-cell maturation antigen (BCMA) and EFNB2 (Fig. 3A). To evaluate for in vivo colocalization, we transplanted NSG mice with 1 × 10⁷ RPMI-8226 multiple myeloma cells at 24 hours post 180 cGy TBI and then immunostained femurs at 5 weeks. As shown in Fig 3B, VE-cadherin (Cdh5) and EPHB4 expression colocalized in BM ECs lining sinusoidal blood vessels in NSG mice (25). At 5 weeks posttransplant, BCMA⁺EFNB2⁺ multiple myeloma cells were detected abundantly and adjacent to Cdh5⁺ blood vessels (Fig. 3C).

Figure 3.

Inhibition of ephrin B2 suppresses multiple myeloma growth in vivo. A, Representative immunofluorescence images of RPMI-8226 multiple myeloma (MM) cells stained for BCMA (green) and EFNB2; merged image with DAPI (blue) nuclear stain at right (red). Scale bar, 100 μm. B, Representative immunofluorescence images of VE-cadherin (Cdh5) expression (white) and Ephb4 colocalization (green) in BM blood vessels in NSG mice. Scale bar, 100 μm. C, Representative images of femur cross sections from NSG mice at 5 weeks posttransplant of RPMI-8226 multiple myeloma cells; sinusoidal blood vessel expressing VE-cadherin (Cdh5; white), multiple myeloma cells expressing BCMA (green), and EFNB2 (red) and merged image shown at right. Scale bar, 100 μm. Arrows, multiple myeloma cells adjacent to sinusoidal vessel. D, Left, representative flow cytometric analysis of human CD138⁺ multiple myeloma cells in the BM of NSG mice at 5 weeks posttransplant of 5 × 10⁶ RPMI-8226 cells treated with sh-EFNB2 or sh-Control. SSC, side scatter. Right, mean percentages of CD138⁺ multiple myeloma cells in the BM of transplanted mice. n = 8/group. E, Percent survival of NSG mice after transplantation of 5 × 10⁶ RPMI-8226 cells transduced with sh-EFNB2 or sh-Control. Log-rank test, P = 0.0012, n = 12/group. F, Mean percentages of Annexin V⁺ RPMI-8226 cells at 24 hours posttreatment with EFNB2 scFv (n = 6/group). G, Top, schematic representation of experiment. Bottom left, representative flow cytometric analysis of human CD138⁺ multiple myeloma cells in the BM of NSG mice at 4 weeks posttransplantation of 3 × 10⁶ RPMI-8226 cells and intravenous administration of EFNB2 scFv or IgG on days +4, 6, 8, 10, and 12. Bottom right, mean percentages of CD138⁺ multiple myeloma cells in the BM of transplanted mice. One-way ANOVA was utilized for analyses of three or more groups; Student two-sided t test was utilized otherwise. Error bars, SD. **, P < 0.01; ****, P < 0.0001; ^#, P < 0.05; ^###, P < 0.001. (G, Created with BioRender.com.)

To determine whether inhibition of EFNB2 could affect multiple myeloma cell function in vivo, we transplanted NSG mice with 5 × 10⁶sh-EFNB2–treated RPMI 8226 multiple myeloma cells or equal dose of sh-Control multiple myeloma cells and measured multiple myeloma cell engraftment at 5 weeks. NSG mice transplanted with sh-EFNB2–treated multiple myeloma cells displayed significantly lower engraftment of human CD138⁺ multiple myeloma cells compared with mice transplanted with sh-Control–treated multiple myeloma cells (Fig. 3D). This was associated with a significant increase in the percent survival of NSG mice transplanted with sh-EFNB2–treated multiple myeloma cells compared with that of mice transplanted with sh-Control–treated multiple myeloma cells (Fig. 3E).

Because silencing EFNB2 expression suppressed multiple myeloma growth in vivo, we sought to determine whether administration of a highly specific single-chain variable fragment targeting EFNB2 (46) could also inhibit multiple myeloma cell growth. Treatment of RPMI-8226 cells in vitro with scFv-EFNB2 increased multiple myeloma cell death at 24 hours compared with control cultures (Fig. 3F). To evaluate the in vivo effects of scFv-EFNB2 in a human multiple myeloma xenograft model, we irradiated NSG mice with 180 cGy and injected 3 × 10⁶ RPMI-8226 multiple myeloma cells via tail vein, followed by intravenous administration of scFv-EFNB2 (5 mg/kg) or vehicle on days +4, +6, +8, +10, and +12. At day +30, mice treated with scFv-EFNB2 displayed 50% lower engraftment of human CD138⁺ multiple myeloma cells in the BM compared with control mice (Fig. 3G).

EFNB2 expression promotes multiple myeloma survival, proliferation, and chemotherapy resistance

To understand the mechanisms through which EFNB2 promotes multiple myeloma growth, we performed RNA-seq on multiple myeloma cells treated with sh-RNA-EFNB2 versus multiple myeloma cells treated with sh-Control (Fig. 4A; Supplementary Fig. S4A). Silencing EFNB2 in RPMI-8226 cells caused the downregulation of several pathways including those involved in cellular movement, cell death and survival, cellular assembly, cell cycle, and DNA repair (Supplementary Fig. S4B). On the basis of the transcriptomic profiles of multiple myeloma cells treated with shRNA-EFNB2 compared with multiple myeloma cells treated with sh-Control, RNA-seq Upstream Regulator IPA predicted various regulators that could be responsible for the transcriptomic differences, including E2F4, IFNG, CCND1, STAT1, E2F1, STAT3, CDKN1a, and STAT5a (Fig. 4B). Further analysis revealed that numerous STAT5 target genes were downregulated in multiple myeloma cells treated with sh-EFNB2 (Fig. 4C). Of note, STAT proteins can be directly activated by ephrin B proteins (47–50), are drivers of multiple myeloma (51, 52), and associate with protection of multiple myeloma via the BM microenvironment (53–55). We found that silencing EFNB2 in multiple myeloma cells decreased levels of phospho-STAT5 compared with sh-Control–treated multiple myeloma cells (Fig. 4D). Conversely, we did not detect a change in phospho-STAT3 or phospho-S6K in response to silencing EFNB2 in multiple myeloma cells (Supplementary Fig. S4C).

Figure 4.

The effects of ephrin B2 on multiple myeloma are mediated by STAT5. A, Heat map of differentially expressed genes by RNA-seq of sh-EFNB2 versus sh-Control cells. B, Upstream regulator analysis from IPA of RNA-seq data. C, Selected STAT5 target genes in the RNA-seq data. D, Left, representative analysis of phospho-STAT5 in RPMI-8226 cells treated with sh-EFNB2 or sh-Control in presence of EPHB4-Fc. Right, mean fluorescence intensity (MFI) of pSTAT5 in the sh-EFNB2- and sh-Control–treated multiple myeloma cells. E, Numbers of U266 cells in culture over time when transduced with gRNA1-EFNB2 and gRNA2-EFNB2. F, Left, representative histograms of BrdU incorporation over 6 hours in U266 cells transduced with the labeled construct. Right, %BrdU⁺ cells in each group. G, Left, representative flow cytometric analysis of Annexin V⁺7AAD⁺ (necrotic) and Annexin V⁺7AAD⁻ (apoptotic) cells. Right, mean levels of total Annexin V⁺ cells within each treatment group. H, Mean percentages of Annexin V⁺ U266 cells engineered with the indicated construct and exposed to bortezomib (Bor) or dexamethasone (Dex) for 24 hours. n = 6/group. I, Representative histograms of p-STAT5 in U266 cells transduced with gRNA-EFNB2 or gRNA-control, with and without stimulation by EPHB4-Fc. IgG (isotype) was used as a negative control for EPHB4-Fc. Right, mean MFI of pSTAT5 in each treatment population. J, Multiple myeloma cells transduced with gRNA-EFNB2 or gRNA-control were cultured with increasing concentrations of the STAT5 inhibitor 2-[(4-oxo-4H-1-benzopyran-3-yl)methylene]hydrazide 3-pyridinecarboxylic acid (STAT5i) for 24 hours, then BrdU incorporation analysis was performed. n = 6/group. One-way ANOVA used for analyses in which there were three or more groups, Student two-sided t test for the other analyses. Error bars, SD. ns, not significant; *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

Because silencing EFNB2 in multiple myeloma cells suppressed multiple myeloma growth in vitro and in vivo, we next examined whether enforcement of EFNB2 expression in multiple myeloma cells could increase multiple myeloma pathogenesis. For EFNB2 overexpression studies, we utilized the U266 multiple myeloma line that demonstrated the lowest level of EFNB2 expression (Fig. 2E; Supplementary Fig. S5A) and overexpressed EFNB2 via the CRISPR dCas9-VPR system (Supplementary Fig. S5B; ref. 56). Transduction of multiple myeloma cells with the dCas9-VPR and gRNA-EFNB2 caused a >30-fold increase in EFNB2 expression (Supplementary Fig. S5C) that resulted in a 10-fold increase in multiple myeloma cells in 14-day culture compared with gRNA-control–treated multiple myeloma cells (Fig. 4E).

Multiple myeloma cell proliferation as measured by BrdU incorporation was also significantly increased in gRNA-EFN2 multiple myeloma cells compared with control multiple myeloma cells (Fig. 4F). In addition, gRNA-EFNB2 multiple myeloma cells displayed significantly decreased Annexin⁺ apoptotic cells compared with gRNA-control multiple myeloma cells (Fig. 4G). Importantly, gRNA-EFNB2 multiple myeloma cells demonstrated markedly decreased sensitivity to chemotherapy agents, bortezomib and dexamethasone (Fig. 4H). At the molecular level, gRNA EFNB2 multiple myeloma cells demonstrated increased STAT5 phosphorylation compared with gRNA control multiple myeloma cells (Fig. 4I), and multiple myeloma cell proliferation in culture in response to EFNB2 overexpression was abrogated by treatment with a STAT5 inhibitor (Fig. 4J). These results suggest that EFNB2-mediated multiple myeloma growth is dependent on STAT5 activation. We did not detect a difference in STAT3 phosphorylation or phospho-SRC in multiple myeloma cells in response to enforced EFNB2 expression (Supplementary Fig. S5D).

Our data demonstrate that EFNB2 expression strongly promotes multiple myeloma pathogenesis. To assess whether EFNB2 augmented multiple myeloma growth via forward signaling through its EPH receptor, we simultaneously silenced EPHB4 expression in gRNA-EFNB2 multiple myeloma cells. At 72 hours posttreatment with shRNA-EPHB4, we observed a modest increase in BrdU incorporation in gRNA-EFNB2 multiple myeloma cells compared with shRNA-control–treated cells (Supplementary Fig. S6A). However, gRNA-EFNB2 multiple myeloma cells treated with shRNA-EPHB4 displayed no difference in cell growth compared with gRNA-EFNB2 multiple myeloma cells treated with shRNA-control (Supplementary Fig. S6B). We also observed no change in pSTAT5 levels in gRNA-EFNB2 multiple myeloma cells treated with shRNA-EPHB4 compared with control cells (Supplementary Fig. S6C). These results suggest that EFNB2 does not augment multiple myeloma growth via forward signaling through EPHB4.

EFNB2 reverse signaling controls multiple myeloma pathogenesis

We have shown that silencing of EFNB2 expression suppresses multiple myeloma growth in vitro and in vivo, but this approach does not discriminate whether EFNB2 forward or reverse signaling is the essential mechanism of action. To determine whether EFNB2 forward or reverse signaling predominates in multiple myeloma, we used a mutant EFNB2 construct (delta cytoplasmic EFNB2, dEFNB2; Fig. 5A), encoding EFNB2 residues 1–265, which includes the extracellular region, transmembrane helix and the first 15 residues of the intracellular region; dEFNB2 lacks the cytoplasmic portion responsible for reverse signaling, which is replaced by EGFP (33). We transduced U266 multiple myeloma cells with dEFNB2 or control vector and measured multiple myeloma pathogenicity. U266 multiple myeloma cells transduced with dEFNB2 displayed decreased cell growth in culture compared with control multiple myeloma cells (Fig. 5B). Commensurate with this, dEFNB2 expression repressed BrdU incorporation in multiple myeloma cells (Fig. 5C). Expression of dEFNB2 also substantially increased multiple myeloma cell death and suppressed STAT5 activation compared with control multiple myeloma cells (Fig. 5D and E).

Figure 5.

Loss of ephrin B2 reverse signaling abrogates multiple myeloma pathogenesis. A, Plasmid maps of overexpression constructs for dEFNB2 mutant lacking the cytoplasmic reverse signaling domain and GFP control. B, Numbers of U266 cells in culture over time when transduced with dEFNB2 or GFP control. C, Left, representative histograms of BrdU incorporation over 6 hours in U266 cells transduced with the labeled construct. Right, %BrdU⁺ cells in each group. D, Left, representative flow cytometric analysis of Annexin V and 7AAD staining of multiple myeloma cells transduced with dEFNB2 or GFP control; right, mean levels of total Annexin V⁺/7AAD⁻ cells and Annexin V⁺/7AAD⁺ cells within each treatment group. E, Left, representative histograms of phospho-STAT5 in U266 cells transduced with the labeled construct. Right, quantified phospho-STAT5 mean fluorescence intensity (MFI) in each group. F, Mean percentages of Annexin V⁺ U266 cells engineered with the indicated construct and exposed to bortezomib (Bor) or dexamethasone (Dex) for 24 hours. n = 6/group. G, Left, representative flow cytometric analysis of human CD138⁺GFP⁺ multiple myeloma cells in the BM of NSG mice at 7 weeks posttransplant of 1 × 10⁵ U266 cells transduced with dEFNB2 or GFP control. Right, mean percentages of human CD138⁺GFP⁺ multiple myeloma cells in the BM of transplanted mice. n = 6–7/group. H, Mean percentages of CD138⁺GFP⁺ multiple myeloma cells in the spine and PB in NSG at 7 weeks after transplantation of U266 cells transduced with dEFBN2 or GFP control (n = 5–10/group). I, Survival of NSG mice transplanted with 1 × 10⁵ dEFNB2-treated multiple myeloma cells or GFP control multiple myeloma cells (n = 10 control; n = 11 dEFNB2 mice). One-way ANOVA was used for analyses in which there were three or more groups, Student two-sided t test for the other analyses. Error bars, SD. **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

To evaluate the therapeutic relevance of inhibiting EFNB2 reverse signaling in multiple myeloma cells, we measured multiple myeloma cell apoptosis during 24-hour culture with and without bortezomib or dexamethasone, coupled with dEFNB2 transduction. The combination of dEFNB2 expression and bortezomib treatment caused approximately three times increased multiple myeloma cell death compared with bortezomib alone (Fig. 5F; Supplementary Fig. S6D). Similarly, dEFNB2 expression coupled with dexamethasone treatment nearly doubled multiple myeloma cell death compared with dexamethasone alone (Fig. 5F). These results suggest that blockade of EFNB reverse signaling is mechanistically nonredundant with bortezomib and dexamethasone effects and appears to synergize with proteasome inhibitors for the treatment of human multiple myeloma. These data also suggest a dominant-negative effect caused by the competition of overexpressed dEFNB2 with endogenously expressed EFNB2 for binding to EPHB receptors, resulting in inhibition of endogenous EFNB2 reverse signaling.

To determine the effect of silencing EFNB2 reverse signaling on multiple myeloma growth in vivo, we compared the repopulation of U266 multiple myeloma cells following transplantation into NSG mice. NSG mice transplanted with 1 × 10⁵ U266 multiple myeloma cells (transduced with GFP control) engrafted robustly, with mean > 60% CD138⁺GFP⁺ multiple myeloma cell engraftment detected at 8 weeks posttransplant (Fig. 5G). In contrast, mice transplanted with an equal dose of U266 multiple myeloma cells transduced with dEFNB2 displayed < 0.1% CD138⁺GFP⁺ multiple myeloma cell engraftment at 8 weeks, suggesting that EFNB2 reverse signaling is essential for multiple myeloma repopulation in vivo (Fig. 5G). Analysis of PB and thoracic spine of transplanted mice also demonstrated nearly undetectable CD138⁺GFP⁺ multiple myeloma cells in mice transplanted with dEFNB2-expressing multiple myeloma cells compared with mice transplanted with equal doses of control multiple myeloma cells (Fig. 5H). Commensurate with these results, PB lambda light-chain levels were significantly lower in mice transplanted with dEFNB2-expressing multiple myeloma cells compared with mice transplanted with control multiple myeloma cells (Supplementary Fig. S6E). Survival studies demonstrated 100% survival of NSG mice transplanted with dEFNB2-treated multiple myeloma cells, whereas no mice transplanted with GFP control multiple myeloma cells survived to 100 days (Fig. 5I).

EFNB2 expression is associated with poor prognosis in multiple myeloma

Because our studies suggested that EFNB2 signaling regulates multiple myeloma cell growth and survival, we interrogated publicly available databases to determine whether EFNB2 expression is associated with specific outcomes in patients with multiple myeloma. Reviewing the Total Therapy 2 and 3 studies (37) and the CoMMpass data (Multiple Myeloma Research Foundation, https://research.themmrf.org/, accessed on October 20, 2021; n  =  773 patients), we discovered that EFNB2 expression was increased in tumors from higher risk cytogenetic categories, including the t(4;14) translocation and gain of chromosome 1q populations (Fig. 6A and B). Conversely, the favorable risk (“standard risk”) category, hyperdiploid multiple myeloma, displayed lower EFNB2 expression (Fig. 6C). Stage III multiple myeloma also demonstrated higher EFNB2 levels compared with stage I or II disease (Fig. 6D). Increased expression of EFNB2 is also associated with decreased survival in patients with multiple myeloma (Fig. 6E). Among patients in the CoMMpass trial, those with the highest decile of EFNB2 expression demonstrated decreased progression-free survival and decreased overall survival compared with all other patients (Fig. 6F and G). These results suggest that EFNB2 expression is a poor prognostic marker in patients with multiple myeloma. Of note, analysis of the Total Therapy dataset did not demonstrate an association between EPHB1 or EPHB4 expression and prognosis in multiple myeloma.

Figure 6.

Ephrin B2 expression associates with clinical outcomes in patients with multiple myeloma. A–D,EFNB2 expression in patients with t(4;14) (A), amplification of chromosome 1q (B), hyperdiploid multiple myeloma (C), and high-stage disease (D) in two independent clinical studies. E, Patients with multiple myeloma expressing EFNB2 above the median had decreased survival in the Total Therapy 2 and 3 trials. Log-rank analysis, P = 0.024. F and G, A patient cohort in the CoMMpass data defined by the highest decile of EFNB2 expression has decreased progression-free survival (F) and overall survival (G). Log-rank analysis, P = 0.013 and 0.034, respectively. Student two-sided t test for all analyses unless otherwise stated. Error bars, SD. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001.

Discussion

These studies provide the first evidence that EFNB2 reverse signaling is essential for multiple myeloma growth and pathogenesis. Ephrins are a class of cell-membrane bound ligands for the EPH receptor tyrosine kinases with demonstrated functions in regulating mammalian development (57, 58), axon regeneration (59, 60), and angiogenesis (33). Ephrin–EPH receptor signaling is unique in that both forward signaling through the EPH receptor and reverse signaling through the ephrin can occur in cells (44, 57, 61). EPH–Ephrin signaling has also been shown to be promiscuous within EPH receptor–Ephrin A or B subclasses (62). Preliminary studies have suggested a potential role for ephrins in regulating the pathogenesis of glioblastoma, melanoma, breast cancer, and other cancers (44, 49, 63, 64), and in regulating the BM microenvironment response to multiple myeloma in immune-deficient mice (65). EPHA3 expression was shown to be upregulated in BM ECs from patients with multiple myeloma and in multiple myeloma cells, and silencing EPHA3 expression inhibited both the angiogenic capacity of myeloma patient–derived ECs and the adhesion and growth of multiple myeloma cells (66, 67). In vitro studies also showed that multiple myeloma cells express EPHA4 and that EPHA4 regulates multiple myeloma cell adhesion and proliferation via activation of Akt signaling (68). Our studies provide evidence that ephrin B2 reverse signaling directly promotes multiple myeloma pathogenesis and that targeted inhibition of ephrin B2 reverse signaling abolishes multiple myeloma growth in vivo.

In our gene expression analysis of human ECs that support multiple myeloma growth in vitro and in vivo, we identified EPHB1 and EPHB4 as overexpressed in multiple myeloma–supportive ECs. In complementary studies, we determined that the gene encoding one of the cognate ligands for EPHB1 and EPHB4, EFNB2, is upregulated in multiple myeloma cell lines and primary multiple myeloma cells from patients. These results suggest that vascular ECs may drive multiple myeloma pathogenesis via activation of EFNB2 reverse signaling through high expression of EPHB receptors. EFNB2 reverse signaling has been shown to promote colorectal cancer chemoresistance and progression through activation of the Src–ERK pathway (69). Similarly, activation of EFNB2 reverse signaling has been implicated in the progression of oral squamous cell carcinoma (70), glioblastoma (64, 71), and perivascular invasion and migration, in part by mediating anchorage-independent cytokinesis through RhoA (72). Here, silencing EFNB2 expression in multiple myeloma cells inhibited multiple myeloma survival and proliferation, increased sensitivity to chemotherapy and abrogated multiple myeloma cell engraftment in vivo. However, silencing of EFNB2 expression could suppress forward and reverse signaling in multiple myeloma cells, so this approach did not resolve which EFNB2 signaling pathway was the dominant mechanism of multiple myeloma pathogenesis.

Forward signaling via Ephrin–EPHB interactions has been shown to regulate EC function and vasculogenesis, which could contribute to multiple myeloma pathogenesis. Specifically, EPHB2 and EPHB4 are required for vascular morphogenesis and EC chemotaxis in response to SDF-1 gradients (73). EFNB2-mediated activation of EPHB2 or EPHB4 forward signaling has also been shown to augment EC chemotaxis in response to SDF-1 (73). Utilizing an implant angiogenesis mouse model, systemic administration of an EPHB4 kinase inhibitor was shown to inhibit VEGF-driven vessel angiogenesis in vivo (74). To resolve the precise contributions of EFNB2 reverse signaling and forward signaling in multiple myeloma, we utilized a mutant dEFNB2, which lacks reverse signaling capacity (33). We discovered that specific inhibition of EFNB2 reverse signaling in multiple myeloma cells increased multiple myeloma cell death, decreased proliferation, increased sensitivity to chemotherapy, and eliminated multiple myeloma growth in vivo. Conversely, shRNA-mediated silencing of EFNB2 in multiple myeloma cells, which interrupts EFNB2-reverse signaling and EFNB-forward signaling through EPHB receptors, slowed multiple myeloma disease progression in vivo, but all mice ultimately succumbed to multiple myeloma. These results suggest that EFNB2-mediated forward signaling through EPHB receptors may negatively regulate multiple myeloma growth. Taken together, our data demonstrate conclusively that EFNB2 reverse signaling drives multiple myeloma pathogenesis.

In summary, we have provided anatomic and functional evidence that suggests the importance of EPHB4–Ephrin B2 reverse signaling in EC-mediated regulation of multiple myeloma pathogenesis. Furthermore, targeted inhibition of ephrin B2 reverse signaling in multiple myeloma cells abolishes multiple myeloma growth in preclinical models. Corresponding clinical data suggest that that ephrin B2 expression correlates with adverse prognosis in patients with multiple myeloma. Given the substantial progress that has been made in the development of decoy proteins and single-chain variable fragments that target EPHB4 and ephrin B2, respectively (46, 75), the ephrin B2–reverse signaling pathway represents an important new mechanistic and therapeutic target in multiple myeloma.

Supplementary Material

Supplementary Data

Supplementary Figures

Authors' Disclosures

J.P. Sasine reports personal fees from Kite Pharma/Gilead, Genmab, and personal fees from Autolus outside the submitted work. E.B. Pasquale reports grants from NIH during the conduct of the study; in addition, E.B. Pasquale has a patent 7999069 issued, a patent 20100183510 issued, and a patent 7582438 issued. J.P. Chute reports grants from National Institute of Allergy and Infectious Diseases and grants from National Heart, Lung, and Blood Institute during the conduct of the study. No disclosures were reported by the other authors.

Authors' Contributions

J.P. Sasine: Conceptualization, formal analysis, investigation, writing–original draft, writing–review and editing. N.Y. Kozlova: Formal analysis, investigation, methodology. L. Valicente: Formal analysis, investigation. J. Dukov: Formal analysis, investigation. D.H. Tran: Formal analysis, investigation. H.A. Himburg: Formal analysis, investigation, writing–original draft. S. Kumar: Formal analysis, visualization. S. Khorsandi: Formal analysis, investigation. A. Chan: Formal analysis, investigation. S. Grohe: Formal analysis, investigation. M. Li: Formal analysis, investigation. J. Kan: Formal analysis, investigation. M.E. Sehl: Resources, investigation. G.J. Schiller: Resources, writing–review and editing. B. Reinhardt: Formal analysis, investigation. B.K. Singh: Resources, investigation. R. Ho: Resources, investigation. P. Yue: Formal analysis, investigation. E.B. Pasquale: Resources, writing–review and editing. J.P. Chute: Conceptualization, resources, supervision, funding acquisition, methodology, writing–original draft, writing–review and editing.