CK1δ and CK1ε Signaling Sustains Mitochondrial Metabolism and Cell Survival in Multiple Myeloma

Multiple myeloma remains an all but incurable malignancy that is characterized by the accumulation of malignant plasma cells in bone marrow (1). Advances in understanding multiple myeloma biology has led to the development and approval of an expanding list of clinical agents with different mechanisms of action (2–7). Despite these advances, while most multiple myeloma patients respond to initial lines of therapy (1–3 lines), nearly all become refractory to all treatment options. Thus, there remains a critical need to identify new vulnerabilities, especially for patients with late relapsed and refractory multiple myeloma (RRMM; i.e., those with ≥4 lines of therapy) where each new line of therapy affords only months of disease control (8, 9).

Drug resistance to current therapies involves both multiple myeloma cell-intrinsic alterations (10, 11) as well as environmental-mediated drug resistance (EMDR) in the bone marrow (BM) niche (12–15). Several signaling pathways appear to contribute to EMDR, including activation of Notch-1 (16), stroma-derived production of IL6 (17), and HSP70- and CD44-dependent signaling (18, 19).

To identify targetable vulnerabilities that disrupt the EMDR, we performed pharmacoproteomic screens of 31 clinically approved and investigational anti–multiple myeloma agents using validated human multiple myeloma cell lines cultured on a platform that mimics the tumor microenvironment, including patient-derived BM stroma, extracellular matrix proteins, and patient-derived serum from BM aspirates (20). Importantly, screens using this platform with CD138⁺ primary multiple myeloma patient specimens, coupled with known drug pharmacokinetic parameters, accurately predict the response of patients to in-clinic agents, including patient tumors that are resistant to multiple lines of therapy (21).

Casein kinase-1δ (CK1δ) and CK1ε are members of the rogue serine/threonine casein kinase family that have high sequence homology, including in their noncatalytic N- and C-terminal domains that dictate substrate specificity (22). Initially characterized as regulators of circadian rhythm (23), CK1δ and/or CK1ε activity is induced by cell stress and they phosphorylate regulators of the DNA damage response (p53, MDM2), protein folding (Hsp70/90), and apoptosis (Bid; ref. 22). Given these roles, CK1δ and/or CK1ε have emerged as attractive therapeutic targets for refractory and metastatic malignancies, including melanoma, triple-negative breast cancer (TNBC), glioblastoma, and bladder and colorectal cancer (24, 25). For example, inhibition of these kinases using a highly potent and selective dual CK1δ/CK1ε inhibitor SR-3029 compromises tumor cell growth and survival ex vivo, blocks metastasis, and provokes tumor regression in vivo (24, 26–29). Furthermore, in TNBC, CSNK1D, but not CSNK1E, is amplified and highly expressed, and elevated CSNK1D expression is associated with poor outcomes (28). Finally, in TNBC, select silencing of CSNK1D mimics the effects of SR-3029 treatment, establishing that in at least TNBC, CK1δ is the relevant drug target (28).

Here we report the therapeutic potential of targeting CK1δ/CK1ε in both treatment-naïve and therapy-resistant multiple myeloma, and in disabling EMDR. Specifically, unbiased activity-based proteomic profiling (ABPP) and kinase inhibitor screens revealed that: (i) stroma coculture augments CK1δ/CK1ε activity in multiple myeloma cells; (ii) selective inhibition of these kinases compromises the growth and survival of paired naïve and drug-resistant multiple myeloma cell lines (30–34) and of primary multiple myeloma specimens across the disease spectrum; and (iii) targeting CK1δ/CK1ε impairs myeloma potential in vivo. Mechanistically, RNA-seq, untargeted metabolomic profiling, and metabolic analyses revealed that disabling CK1δ/CK1ε signaling leads to metabolic collapse that involves an acute suppression of the citric acid (TCA) cycle and complexes I and IV of the electron transport chain (ETC). Finally, these analyses also revealed that evolution to multidrug resistance in multiple myeloma is associated with increased expression of mitochondrial metabolism genes that are dependent on CK1δ/CK1ε signaling. These findings support a model whereby altered metabolic programming manifested during multiple myeloma progression and therapy resistance can be exploited by targeting CK1δ and CK1ε.

All cell lines were grown in RPMI-1640 medium (Gibco) supplemented with 10% FBS (Gemini Bio-Products, 900–108) and penicillin–streptomycin (Gibco) as described previously (14). All cell lines were routinely tested (every 8 weeks) for Mycoplasma contamination using a MycoAlert Mycoplasma Detection Kit (Lonza). Multiple myeloma cell lines were authenticated for their origin using short tandem repeat (STR) DNA typing according to ATCC guidelines using the GenePrint 10 System (Promega; ref. 35).

Lenalidomide-resistant multiple myeloma cell lines KAS6/R10R, MM.1S/R10R, and U266/R10R were a generous gift from Dr. Robert Orlowski (MD Anderson Cancer Center, Houston, TX; ref. 30). 8226 multiple myeloma cells resistant to bortezomib (8226/B25), doxorubicin (8226/Dox40), and melphalan (8226/LR5), and U266 multiple myeloma cells resistant to melphalan (U266/LR6) have been previously described (36–38). All resistant lines were developed by incremental increased exposure to drugs over time to achieve tolerance levels of 10 μmol/L lenalidomide, 25 nmol/L bortezomib, 400 nmol/L doxorubicin, or 5 μmol/L melphalan. Parental versions of these multiple myeloma cell lines were obtained from ATCC; OPM2 and H929 multiple myeloma cells were obtained from DSMZ-German Collection of Microorganisms and Cell Cultures GmbH. ANBL6 multiple myeloma cells were obtained from Dr. Diane Jelinek (Mayo Clinic, Phoenix, AZ).

Primary multiple myeloma cells were isolated from patients and cultured as described previously (20, 21). Briefly, the involvement of human subjects in this research was limited to the donation of BM aspirates and peripheral blood, as well as to deidentified select clinical data from patients across the spectrum of multiple myeloma disease states. Patients of any age (>18 years old), gender, race, or ethnicity seen at the Moffitt Cancer Center (Moffitt) outpatient clinics or inpatient services were approached via Moffitt standard operating procedures to participate in this research project.

All aspirates and biopsies were collected at the time of a medically indicated procedure and all procedures and samples requested fall under the Institutional Review Board (IRB)-approved Total Cancer Care (TCC; MCC14690) tissues and clinical annotation protocol at the Moffitt Cancer Center. Patients were not asked to provide samples outside the context of their clinical needs. Approximately 10–20 mL of additional aspirate and core biopsies was used for the research. All patients were consented to the IRB-approved MCC14690 Total Cancer Care and the MCC18608 multiple myeloma research protocols at Moffitt Cancer Center.

We and our colleagues obtained written informed consent from the patients with multiple myeloma agreeing to the use of biopsy tissue for research studies. These studies were performed according to the ethical guidelines outlined in the Declaration of Helsinki, the International Ethical Guidelines for Biomedical Research Involving Human Subjects (CIOMS), the Belmont Report, and the U.S. Common Rule.

Patients of any age (>18 years old), gender, race, or ethnicity seen at the Moffitt Cancer Center outpatient clinics or inpatient services were approached via Moffitt standard operating procedures to participate in this research project. Aspirates and biopsies were collected at the time of a medically indicated procedure and patients were not asked to provide samples outside the context of their clinical needs. Approximately 20 mL of BM aspirate biopsies was used for the research for all multiple myeloma specimen–related data. Patient specimens were harvested from consented patients at Moffitt Cancer Center. Myeloma cells were purified from mononuclear fractions from fresh BM aspirate cells using CD138 antibody–conjugated magnetic beads (Miltenyi Biotec, #130–051–301) in our Cancer Pharmacokinetics/Pharmacodynamics Core. CD138-selected cells were in turn used for ex vivo drug testing and molecular analysis.

Ex vivo drug sensitivity screens were performed in BM aspirates harvested from consented multiple myeloma patients at Moffitt Cancer Center and myeloma cells were purified from mononuclear fractions by CD138 affinity chromatography (Miltenyi Biotec). Myeloma cells were then seeded into 384-well plates with primary patient BM stroma in a collagen matrix (bovine collagen I, Advanced BioMatrix) and autologous patient serum derived from BM. Sensitivity to anti–multiple myeloma agents was measured in duplicate using a viability assay based on live cell imaging (EVOS FL Auto, Thermo Fisher Scientific) and a digital image analysis algorithm (21, 39).

For activity-based proteomic profiling (ABPP), H929, MM.1S, and OPM2 multiple myeloma cells, and HS5 stroma cells, were plated in either monoculture conditions or in direct contact cocultures (40). For the latter, H929, MM.1S, or OPM2 multiple myeloma cells were grown directly on top of an 80% confluent HS5 stroma cell culture for 24 hours. Nonadherent and monoculture control myeloma cells as well as HS5 monoculture stroma cells were then harvested, rinsed with PBS containing 1 mmol/L sodium orthovanadate (Na₃VO₄), and pellets were snap-frozen in liquid nitrogen and stored −80°C. Similarly, direct contact cocultures of myeloma cells adhered to HS5 cells were rinsed with PBS containing 1 mmol/L Na₃VO₄, harvested by cell scraping, and pellets were snap-frozen in liquid nitrogen and stored −80°C for later analysis.

Sample lysates were prepared per manufacturer's protocol (ActivX ATP probe, Thermo Fisher Scientific). Briefly, cells were lysed using mild IP lysis buffer (Thermo Scientific) and proteins were desalted with Zeba spin columns (Thermo Fisher Scientific) to remove endogenous ATP. An equal amount (1 mg) of total protein was then labeled with desthiobiotinylating-ATP reagent at room temperature for 10 minutes. After reduction/alkylation, buffer exchange to remove excess probe was performed using Zeba spin columns prior to in-solution digestion with trypsin at 37°C overnight. Desthiobiotinylated peptides were purified with high-capacity streptavidin agarose resin at room temperature for 1 hour. After washing, peptides were eluted with aqueous 50% acetonitrile with 0.1% trifluoroacetic acid. Each sample was analyzed in duplicate with LC/MS-MS (RSLCnano and Q Exactive Plus, Thermo Fisher Scientific) to identify peptides and localize sites of desthiobiotinylation. The sample was first loaded onto a precolumn (100 μm ID × 2-cm length packed with C18 PepMap100 reversed-phase resin, 5-μm particle size, 100 Å pore size) and washed for 8 minutes with aqueous 2% acetonitrile and 0.04% trifluoroacetic acid. Trapped peptides were eluted onto the analytic column, (C18 PepMap100, 75 μm ID × 25-cm length, 2-μm particle size, 100 Å pore size, Thermo Fisher Scientific). The 120-minute gradient was programmed as: 95% solvent A (2% acetonitrile + 0.1% formic acid) for 8 minutes, solvent B (90% acetonitrile + 0.1% formic acid) ramped from 5% to 15% over 5 minutes, from 15% to 40% over 85 minutes, then to 90% B over 7 minutes, and held at 90% for 5 minutes, followed by decreasing solvent B from 90% to 5% in 1 minute and reequilibration for 10 minutes. The flow rate on analytic column was 300 nL/minute. Sixteen tandem mass spectra were collected in a data-dependent manner following each survey scan using 60-second exclusion for previously sampled peptide peaks. MaxQuant (version 1.2.2.5) was used for sequence identification and label-free quantification by matching MS/MS data to tryptic peptides from Human entries in the UniProt database considering methionine oxidation and lysine desthiobiotinylation as variable modifications (41).

For 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide (MTT) assays, 1:4 serial dilutions of the indicated compounds were performed prior to 3 × 10⁴ cells being transferred to each well of a 96-well plate. Cells were incubated at 37°C and 5% CO₂ for 72 hours. After incubation, 10 μL MTT Reagent A (Millipore) was added, and plates were incubated an additional 4 hours. Cells were centrifuged for 2 minutes, 1,000 × g, the media discarded, and cells were dissolved in 100 μL of acidic isopropanol. After 30 minutes, plates were read on a Cytation3 Image Reader (BioTek) and the EC₅₀ values for each compound were determined using Prism 7, including logarithmic transformation, normalization, and nonlinear regression (GraphPad Prism 7).

Apoptosis was measured after 24 hours of treatment with vehicle or SR-3029 using Caspase-Glo 3/7 (Promega) according to manufacturer's instructions. Briefly, 1 × 10⁴ cells were transferred to each well of a 96-well white-walled plate and treated in triplicate with vehicle or 250 nmol/L SR-3029 and incubated for 24 hours. For collection, the cell plate and Caspase-Glo 3/7 reagent were allowed to equilibrate to room temperature before addition of the Caspase-Glo 3/7 reagent. The plate was mixed at 300 rpm for 30 seconds, and after 30 minutes, plates were read on a Cytation3 Image Reader and the luminescence of each well was recorded.

To assess effects of dual inhibition of CK1δ/CK1ε on colony-forming potential, 1,000 cells of the indicated multiple myeloma cell lines were plated in 1-mL of methylcellulose (StemCell Technologies) in triplicate in each well of a 6-well plate with 250 nmol/L SR-3029 or vehicle under standard growth conditions. After 10 to 14 days, plates were imaged on STEMVision (StemCell Technologies) and colonies with greater than 50 cells were scored.

To assess long-term effects of SR-3029 treatment on hematopoietic colony-forming potential, C57B6/KaLwRijHsd mice were treated for 8 weeks with 20 mg/kg SR-3029 or vehicle intraperitoneally daily (n = 5/group). Total BM mononuclear cells were isolated from mouse femurs, resuspended in 2% FBS Iscove Modified Dulbecco's Medium (Gibco), and plated in triplicate in MethoCult GF (StemCell Technologies; 2 × 10⁴ cell/well in a 6-well plate) under standard growth conditions. After 10 days, plates were imaged on STEMVision, and colonies were counted.

Myeloid cells (GM-M3534) and pre-B cells (GM-M3630) were plated in Methocult GF (StemCell Technologies) at 2 × 10⁴ cells and 3.5 × 10⁵ cells per well, respectively. The cells were treated with vehicle (DMSO) or the complex I inhibitors rotenone or piericidin at the indicated concentrations. Colonies were counted 7 days after plating.

Lysates were prepared from cells grown in culture dishes as above. Preparation of cell lysates, determination of protein concentrations, and SDS-PAGE gels were described previously (42). Membranes were blocked in Odyssey blocking buffer TBS (LI-COR Biosciences) and incubated overnight at 4°C in blocking buffer plus 0.1% Tween 20 (Fisher BioReagents) and the indicated primary antibody. After washing 3× in TBST, membranes were incubated in appropriate IRDye-conjugated secondary antibody (LI-COR Biosciences) and imaged using the LI-COR Odyssey FC (LI-COR). Membranes were blotted for antibodies specific for β-actin (Sigma, AC-15), CK1ε (BD Transduction; 610445), CK1δ (Abcam, ab48031), PARP (Cell Signaling Technology, 9532S), oxoglutarate dehydrogenase (OGDH; 15212–1-AP), COX1V (11242–1-AP), COX5A (1148–1-AP), and COX5B (11418–2-AP; all from ProteinTech).

Lentiviral vectors expressing nontargeting, CK1δ-, or CK1ε-targeting shRNAs were cloned into the Tet-pLKO-Puro vector using the recommended protocol (43). Lentivirus was generated using HEK293T cells in combination with the Mission Packaging System (Sigma). 8226 and MM.1S multiple myeloma cells were spinfected with optimized viral titers by centrifugation at 2,500 rpm for 90 minutes before overnight incubation. Infected cells were selected in puromycin (1 μg/mL) containing media.

All animal studies were approved by the Moffitt/University of South Florida Animal Care and Use Committee. Six- to 8-week-old NOD-scid/IL2Rγ^−/− (NSG) mice (Jackson Laboratories) were used for human myeloma xenograft studies.

For the subcutaneous (sub-Q) flank multiple myeloma tumor model, 1 × 10⁷ MM.1S cells were injected sub-Q into bilateral flanks of NSG mice in 100-μL of PBS mixed 1:1 with Matrigel (Corning). Mice were randomized into vehicle (10:10:80 DMSO/Tween 80/water) and SR-3029 (20 mg/kg/daily, i.p.) treatment groups once tumors reached 0.1 cm³ (n = 10 mice/group) as determined by caliper measurements (length × width²)/2 (28). Twice weekly caliper measurements were continued until tumors reached 2,000-mm³ or mice became moribund, at which time the mice were humanely euthanized.

For a bone-invasive human myeloma model, 4 × 10⁶ MM.1S-luc cells were injected into the tail veins of NSG mice in 100-μL of PBS. Tumor volumes were measured on day 6 by luminescence imaging with the IVIS 200 imager after tail vein injection of luciferin (15 mg/mL; Gold Biotechnology). Average radiance (photons s⁻¹ cm⁻² sr⁻¹) was determined from combined scores from the dorsal and ventral imaging of the entire animal using Living Image analysis software (Xenogen). Mice were then randomized into control vehicle (10:10:80 DMSO/Tween 80/water) and SR-3029 (20 mg/kg/daily, i.p.) treatment groups (n = 10 mice/group). Luminescence imaging as well as assessment of human λ immunoglobulin serum levels were measured weekly. Human λ serum levels were measured by ELISA per manufacturer's protocol (Bethyl Laboratories Inc., E88–116). Mice were humanely euthanized at the onset of hind-limb paralysis, or if they lost >15% of their body weight.

For a syngeneic bone-invasive mouse myeloma model, the murine myeloma cell line 5TGM1-luc was inoculated into 6- to 10-week-old syngeneic C57BL/KaLwRijHsd mice by tail vein injection (1 × 10⁷ cells/mouse). SR-3029 was solubilized immediately prior to intraperitoneal injection with 10% Tween-80/10% DMSO in 0.9% saline and was given at 20 mg/kg daily beginning on day 14 after tumor inoculation. The survival endpoint was hind-limb paralysis. Tumor burden was determined by measuring bioluminescence and IgG2b paraprotein weekly. For bioluminescence measurements, shaved mice were injected intraperitoneally with 3 mg luciferin substrate (Gold Biotechnology) in saline 15 minutes prior to imaging using the IVIS-200 imaging system (Xenogen). Serum was collected from submandibular bleeds and levels of IgG2b were determined by ELISA (Bethyl Laboratories, E99–109).

A reporter plasmid containing the 609-bp OGDH core promoter sequence obtained from the Eukaryotic Promoter Database, upstream of the firefly luciferase gene, pLenti-OGDH-Dual-Luc, and a control plasmid, pLenti-Promoterless-Dual-Luc, were obtained from Applied Biological Materials. This construct also contains a control Renilla luciferase gene driven by the constitutive SV40 promoter. Viral particles containing this construct were produced in HEK293T cells using a third-generation packaging system, Mission Lentiviral Packaging Mix (Sigma). FuGENE HD (Promega) was used to transfect the packaging cell line with packaging, envelope, and luciferase vectors and supernatant containing viral particles was harvested 4 days later. Polybrene (Millipore) was used to enhance infection of HEK293T cells and stable cell lines were produced by puromycin selection. Luminescence was measured 24 hours after SR-3029 treatment with a Cytation3 imager (Agilent) using substrate from the dual luciferase reporter system (Promega) according to the manufacturer's instructions.

RNA was prepared from primary CD138⁺ multiple myeloma patient samples, and from the indicated multiple myeloma cell lines treated with vehicle versus 250 nmol/L SR-3029 for 24 hours, using the NucleoSpin RNA Kit (Macherey-Nagel) as per the manufacturer's protocol. Primary multiple myeloma patient samples were cultured in the top well of a transwell Boyden chamber (Corning Transwell 0.4-μm pore membranes, #3450) with supportive multiple myeloma patient-derived stroma cells in the bottom well. RNA sample quality was verified by RNA Integration Numbers using the Agilent 2100 Bioanalyzer and the RNA 6000 Nano LabChip before library preparation was performed using Nugen Human total RNA-seq library prep kit, to generate cDNA for next-generation sequencing as described previously (44).

For RNA-seq analyses of SR-3029–treated multiple myeloma cell lines the following pipeline was used: paired-end RNA sequencing fastq files were aligned to the hs37d5 human reference genome using the STAR aligner v2.5.3a. Expression counts were summarized at the gene level against the gencode 30 gene model using HTseq v0.6.1. Read counts reported were then normalized to library size estimates using the R package DESeq2 v1.6.3. Finally, differential gene expression for treatment effects was evaluated using DESeq2.

Primary CD138⁺ multiple myeloma patient samples were processed using the following pipeline: Paired-end RNA-sequencing fastq files were aligned to the hs37d5 human reference genome using TopHat 2.0.13. Expression counts were then summarized at the gene level against the refseq hg19 30 gene model using HTseq v0.6.1. Read counts reported were then normalized to library size estimates using the R package DESeq2 v1.6.3, and differential gene expression for treatment effects was determined using DESeq2.

SR-3029- or vehicle-treated multiple myeloma cell lines (2 × 10⁵ cells/well) were washed once and then plated in XFe96 microplates in unbuffered RPMI containing 10 mmol/L glucose, 1 mmol/L sodium pyruvate, 2 mmol/L l-glutamine, and 1 mmol/L HEPES for Mitochondrial Stress Test (MST), and Glycolytic Rate Assays (GRA). Concentrations of compounds that were used were 1 μmol/L oligomycin A, 1 μmol/L FCCP, 500 nmol/L rotenone, 500 nmol/L antimycin A, 10 mmol/L glucose, and 1 mmol/L 2-deoxyglucose (2DG). Data were normalized using Calcein AM.

To assess mitochondria electron transport chain activity, cells were plated in 1× MAS solution with a pH between 7.0 and 7.2 containing 4 mmol/L ADP, 10 mmol/L malate, 10 mmol/L glutamic acid, and 2 nmol/L Plasma Membrane Permeabilizer (PMP), and immediately placed in the XFe96 Analyzer for analysis. Basal readings were taken to analyze complex I activity, 2 μmol/L rotenone followed by 10 mmol/L succinate to measure complex II/III activity, and 4 μmol/L antimycin A followed by 100 μmol/L TMPD/10 mmol/L ascorbic acid to measure complex IV activity. To attach cells to the bottom of the wells, the XF microplates were coated with 50 μg/mL poly-d-lysine.

After 8 and 24 hours of treatment with vehicle or 250 nmol/L SR-3029, MM.1S cells were stained with 100 nmol/L of MitoTracker Green (Thermo Fisher Scientific, M7514) or 25 nmol/L of MitoTracker CMXRos (Thermo Fisher Scientific, M7512) and incubated for 30 minutes at 37°C in RPMI-1640 medium without serum. After two washes with 1× PBS, cells were resuspended in 1× PBS containing 2% serum and 1 nmol/L DAPI (Life Technologies, D1306), and analyzed using a LSR II Flow Cytometer (BD Biosciences). Dead cells were gaited out. Data were collected to a limit of 10,000 events of the population of interest. Geometric means for mean fluorescence intensity were obtained using FlowJo.

Gene expression data were obtained from DNAnexus files containing FPKM and TPM values for 59,368 records. Of these, 19,933 were protein-coding genes, which were further analyzed. For each gene/sample, we calculated log₂(FPKM+10⁻³) and removed any genes whose values for quartile 1 and quartile 3 were the same, that is, any gene must be expressed in at least 25% of samples to be considered in this analysis. The remaining 16,738 genes were Z-normalized across all samples, using the MATLAB function normalize. This step ensures that upon dimensionality reduction analysis (both principal component analysis and t-distributed stochastic neighbor embedding) or enrichment analysis [single-sample gene set enrichment analysis (ssGSEA) or gene set enrichment analysis (GSEA)], that genes with low expression values have equal importance as the high expressing genes; that is, the focus is on relative rather than absolute changes in expression.

To identify genes differentially expressed across the spectrum of multiple myeloma, samples were grouped according to disease state. Student t tests were calculated for each gene/transition [e.g., MYC expression levels in late relapsed multiple myeloma (LRMM) vs. newly diagnosed multiple myeloma (NDMM) samples] and multiple test correction was performed using FDR method (two-stage step-up method of Benjamini, Krieger, and Yekutieli, Q = 1%) calculated using Prism 9 for macOS.

To identify differentially expressed biology that is associated with multiple myeloma progression and refractory disease, for each sample with RNA-seq data we calculated ssGSEA of Kyoto Encyclopedia of Genes and Genomes (KEGG) and cancer hallmark pathways, as well as coexpressing gene clusters (see next section) using the script ssgsea-gui.r from the library SSGSEA 2.0 (github.com/broadinstitute/ssGSEA2.0) for R (version 4.0.2) for macOS. ssGSEA normalized enrichment score (NES) of genesets with P value larger than 0.01 were set to zero, and thus were deemed not enriched. Student t tests were performed with multiple test correction, using Prism 9 for macOS and the FDR method (two-stage step-up method of Benjamini, Krieger, and Yekutieli, Q = 1%).

To identify differentially expressed biology that is associated with ex vivo drug sensitivity or resistance, we used the GSEA software (45) version 3.0. The phenotype chosen was AUC of the ex vivo dose–response curve for all five concentrations of drug, between 0 and 96 hours. Command line options chosen were: -collapse false -mode Max_probe -norm meandiv -nperm 1000 -permute gene_set -rnd_type no_balance -scoring_scheme weighted -rpt_label my_analysis -metric Pearson -sort real -order descending -create_gcts false -create_svgs false -include_only_symbols true -make_sets true -median false -num 100 -plot_top_x 50 -rnd_seed timestamp -save_rnd_lists false -set_max 5000 -set_min 15 -zip_report false -gui false. Libraries used were the KEGG pathways and CANCER HALLMARKS that were obtained from MsigDB (www.gsea-msigdb.org/gsea/msigdb).

The RNA-seq gene set enrichment was performed with GSEA (45) using the HALLMARK gene set collection (46). Additional GSEA was performed using Database for Annotation, Visualization and Integrated Discovery (DAVID; refs. 47, 48).

Cell pellets (∼1–3 million cells) were spiked with a mixture of stable isotope–labeled standards (Cambridge Isotope Labs). An aliquot (1-mL) of prechilled solvent (80% methanol) was added to each sample to extract metabolites and precipitate proteins. The samples were vortexed and incubated at −80°C for 30 minutes. The protein was pelleted by centrifugation at 18,800 × g for 10 minutes at 4°C, and its concentration was calculated using Bradford assays (Coomassie Plus Reagent, Thermo Fisher Scientific) and used for determining the resuspension volume to achieve equal sample loading for untargeted metabolomics. The supernatant containing the metabolites was lyophilized and then resuspended in 20–30-μL of aqueous 80% methanol, depending on the protein concentration of each sample. Liquid chromatography/high-resolution mass spectrometry was performed using a UHPLC (Vanquish, Thermo Fisher Scientific) interfaced with a hybrid quadrupole-Orbitrap mass spectrometer (Q Exactive HF, Thermo Fisher Scientific). An aliquot (2-μL) of each sample was loaded onto a SeQuant ZIC-pHILIC guard column (4.6 mm inner diameter/ID × 20 mm length, 5 μm particle size, Millipore Sigma) that is connected to a SeQuant ZIC-pHILIC column (4.6 mm ID × 150 mm length, 5 μm particle size, Millipore Sigma); both columns were maintained at 30°C. The following solvent system was used for LC-MS analysis: solvent A was aqueous 10 mmol/L ammonium carbonate with 0.05% ammonium hydroxide and solvent B was 100% acetonitrile. A linear gradient was programmed from 80 to 20% B over 13 minutes with a flow rate of 0.250 mL/minute, and then maintained at 20% B for 2 minutes, followed by reequilibration over 5 minutes at a flow rate of 0.250 mL/minute, for a total run time of 20 minutes for each experiment. The Q Exactive HF mass spectrometer was operated in positive and negative ion mode in separate experiments using the following parameters for ionization: sheath gas 50, auxiliary gas 10, sweep gas 1, spray voltage 3.5 kV, 60,000 resolution, 100 ms maximum ion accumulation time, and scan ranges from m/z 60 to m/z 900 or m/z 65 to m/z 900 in the positive or negative mode, respectively.

LC-MS data files were converted to mzXML files using ProteoWizard and were analyzed using MZmine 2.38 (49). Data processing consisted of several steps: mass detection, chromatogram building, smoothing, chromatogram deconvolution, grouping of isotopic peaks, peak alignment, gap filling to fill in missing peaks, duplicate peak removal, peak filtering (retention time range 0.45–17.0 minutes, peak duration range 0.06–2.00 minutes), and custom database search for metabolite identification with m/z tolerance of 6 ppm and retention time tolerance of 0.3 minute.

Cells were grown in RPMI-1640 supplemented with U¹³-C glucose (100% at 2 gm/L; Sigma Aldrich, catalog no. 389374) or U¹³-C glutamine (100% at 300 mg/L; Cambridge Isotope Laboratories cat. #CLM-1822-H-PK) in the presence of vehicle (DMSO) or SR-3029 (250 nmol/L) for 24 hours. Ten million cells were collected and frozen until processed for metabolite extraction. Tracing was performed using five replicates per treatment condition. Metabolites were extracted using established protocols and were separated on LC-MS. The isotopomer peaks were identified using El-MAVEN and the samples were normalized by protein concentration. Individual isotopomers were expressed as average fraction of the total metabolite concentration ± SD. Statistical differences were analyzed using ANOVA.

Under the assumptions of normal distribution, equal sample variance, and independent variables, data analysis for in vitro experiments was performed using GraphPad Prism version 8.4.0 and is presented as indicated in the figure legends. In vivo survival analysis was estimated using the Kaplan–Meier method as implemented in MatSurv (50) using MATLAB R2019b. Log-rank P values for trend was used when appropriate. A P value less than 0.05 was considered statistically significant.

Raw RNA-seq data for BM samples that were isolated from patients with multiple myeloma and for the 8226 and MM.1S multiple myeloma cell lines treated with vehicle or SR-3029 have been deposited at NCBI in the GEO repository (accession number GSE241861). The data underlying all findings of these studies are available from the corresponding authors upon request and are provided as separate source data files.

The patient molecular data from RNA-seq used in this study was generated for the article in preprint (https://doi.org/10.21203/rs.3.rs-1723993/v1). These data are made publicly available through Moffitt Cancer Center's U54 PS-ON/CSBC repository on Synapse (a research data sharing platform that is supported by NIH and that is maintained by Sage Bionetworks) at https://doi.org/10.7303/syn52316987. RNA-seq data from 687 patient samples is available as log₂ FPKM and z-normalized scores. Phenotypic data (disease state and overall survival) associated with these samples and input/output files for GSEA and ssGSEA are made publicly available at https://doi.org/10.7303/syn52339255. Briefly, this repository contains: (i) disease state annotation for 687 multiple myeloma [NDMM/ERMM (early relapsed multiple myeloma)/LRMM] patient samples along with overall survival data for 483 samples; (ii) for ssGSEA, files generated by ssGSEA containing normalized enrichment scores, P values and FDRs in the form of .gct files; and (iiii) for GSEA, files generated by GSEA that correlate phenotype (ex vivo drug sensitivity to SR-3029) and gene expression for two gene sets (cancer hallmarks and KEGG pathways) as compressed folders containing the standard file structure generated by GSEA software, including a summary of the analysis, correlation of individual gene expression, and phenotypes, as well as a detailed enrichment score for each gene set. Input files for GSEA and ssGSEA are also available in this folder in their standard format.

Given the critical need for new therapies in relapsed and refractory multiple myeloma, and the established roles of the BM niche in drug resistance in multiple myeloma (40, 51–54), we carried out pharmacoproteomic screening of a large cast of kinases that are activated in the context of coculture with HS-5 stroma cells. ABPP using three multiple myeloma cell line models, MM.1S, NCI-H929, and OPM2, under mono- and coculture conditions with HS-5 stroma cells, identified kinases whose activity is altered by coculture with BM stroma (Fig. 1A; Supplementary Fig. S1A; see PRIDE/ProteomeXchange PXD025808). This screen identified 176, 136, and 85 kinases that were significantly altered (log₂-fold change; Benjamini P < 0.05) in coculture conditions in MM.1S, NCI-H929, and OPM2 cells, respectively. Of these, 49 kinases were common to the three multiple myeloma models and 87 were shared between two of the three multiple myeloma cell lines (Fig. 1B). In accord with prior phosphoproteomic analyses (40), focal adhesion–associated kinases were upregulated under coculture conditions, including ILK, FAK, and Pyk2/FAK2 (Fig. 1C). Additional kinases whose activity was differentially regulated by stromal interactions included the cyclin-dependent kinases (CDK) CDK1, CDK4, and CDK7, the tyrosine kinases Src and Lyn, the serine/threonine kinases ERK1 and MAPK14 (p38), and the nutrient sensing kinase mTOR (Fig. 1C). Finally, the activity of serine/threonine kinases known to regulate WNT signaling, CK2α, GSK3β, and PKA, and the rogue kinases CK1δ/CK1ε, were also differentially regulated by stroma coculture (Fig. 1C).

To assess whether any of the identified kinases represented vulnerabilities for multiple myeloma, we performed an ex vivo drug screen in the context of a reconstructed tumor microenvironment (TME) that includes myeloma patient-derived BM stromal cells, a collagen matrix, and multiple myeloma patient-derived BM serum (13, 20, 21). Using this platform, we tested the same 3 multiple myeloma cell lines for their sensitivity to a panel of protein kinase inhibitors (PKI; n = 43) that have overlapping target specificity to kinases altered by stroma interactions in the ABPP data (13, 20, 30). These screens identified six PKIs with potent nanomolar (nmol/L) activity in at least two of the three multiple myeloma models (Fig. 1D; Supplementary Fig. S1B). Five of these PKIs have been tested in preclinical and/or clinical studies in multiple myeloma or other hematologic malignancies, specifically those targeting CDKs (BMS-265246, dinaciclib, and THZ1; refs. 55, 56), mTOR (INK128/MLN0128; ref. 57), and PLK1 (volasertib; ref. 58). In contrast, the anti–multiple myeloma activity of SR-3029, a highly specific dual inhibitor of CK1δ/CK1ε (59), has not been described.

To explore the therapeutic potential of targeting CK1δ/CK1ε, we determined the EC₅₀ of SR-3029 in a panel of naïve and paired isogenic drug-resistant multiple myeloma cell lines using CellTiter-Blue viability assays. Notably, SR-3029 had potent nanomolar activity in 13 of 14 multiple myeloma cell lines examined (Fig. 2A), demonstrating activity in multiple myeloma across different genetic backgrounds. SR-3029 also had high potency against bortezomib proteasome inhibitor–resistant multiple myeloma cell lines (8226, 256 nmol/L vs. 8226/bortezomib, 112 nmol/L) and IMiD-resistant multiple myeloma cells (Kas6, 1205 nmol/L vs. Kas6/R10R, 167 nmol/L; U266, 377 nmol/L vs. U266/R10R, 280 nmol/L; Fig. 2A).

Molecular and cell biological analyses established SR-3029 treatment compromises human (8226, 8226/B25, and MM.1S) and mouse (5TGM1; ref. 60) multiple myeloma cell growth and survival via the induction of apoptosis. First, SR-3029 treatment induced rapid cleavage of caspase-3/7 as measured by Caspase-Glo activity (Fig. 2B). Second, Western blot analysis demonstrated a dose-dependent increase in cleaved PARP following SR-3029 treatment in both human and mouse multiple myeloma cell lines (Fig. 2C). Finally, SR-3029 treatment nearly abolished colony-forming potential of multiple myeloma cells in methylcellulose (Fig. 2D; Supplementary Fig. S1C).

To confirm the deleterious effects of SR-3029 on multiple myeloma cell survival were due to targeting CK1δ and/or CK1ε, 8226 and MM.1S cell lines were transduced with lentivirus expressing inducible shRNAs that selectively reduced their expression (Fig. 2E). Inducible knockdown of either CK1δ or CK1ε impaired the growth of 8226 and MM.1S cells in culture but only had modest effects on their viability, as measured by trypan blue exclusion (Fig. 2F; Supplementary Fig. S1D and S1E). Furthermore, knockdown of either CK1δ or CK1ε impaired growth of both multiple myeloma cell lines in 3D in methylcellulose (Fig. 2G; Supplementary Fig. S1F). Finally, consistent with the cytotoxic effects of their dual inhibition by SR-3029, numerous strategies to dually knockdown or knockout CK1δ and CK1ε via inducible shRNA or CRISPR editing in multiple myeloma cells failed to generate stable cell lines.

Given the potent ex vivo anti–multiple myeloma activity of SR-3029, we next tested its efficacy and safety in well-validated myeloma transplant models. Notably, SR-3029 (20 mg/kg/daily, i.p.) treatment impaired growth of subcutaneous MM.1S tumor xenografts in NOD-scid;IL2Rγ^−/− (NSG) immunodeficient mice (Fig. 3A) and improved their overall survival (Fig. 3B). Similarly, SR-3029 treatment showed potent anti–multiple myeloma activity in a MM1.S-luciferase xenograft model that, akin to multiple myeloma in patients, homes to the BM with monoclonal gammopathy, focal osteolytic bone lesions, and hind-limb paralysis (61). Six days after intravenous inoculation into recipient NSG mice, tumor burden was assessed by bioluminescence and mice were randomized into vehicle or SR-3029 treatment groups. There were significant reductions in tumor burden in SR-3029- versus vehicle-treated animals (Fig. 3C; Supplementary Fig. S2A), as well as significant reductions in human λ serum paraprotein levels (Fig. 3D). Finally, the SR-3029–treated cohort had a significant survival advantage over vehicle-treated mice (Fig. 3E). Thus, SR-3029 has potent anti–multiple myeloma activity in the BM niche, which drives EMDR (12–15).

To assess effects of targeting CK1δ/CK1ε in the context of an intact immune system, we used the syngeneic C57B6/KaLwRijHsd-derived 5TGM1-luciferase multiple myeloma mouse model (60). Fourteen days after inoculation, mice were treated with either vehicle or SR-3029 (20 mg/kg/daily, i.p.). Notably, SR-3029 treatment significantly reduced tumor burden as measured by levels of bioluminescence and serum IgG2b (Fig. 3F and G) and improved overall survival versus the vehicle-treated cohort (P < 0.002; Fig. 3H).

Potential effects of SR-3029 on normal hematopoiesis were also examined in non–tumor bearing C57BL/KaLwRijHsd mice. Consistent with previous safety studies of SR-3029 in healthy C57BL/6 mice (28), an 8-week treatment of these mice with SR-3029 (again daily, i.p., 20 mg/kg) showed no deleterious effects on hematopoietic colony-forming potential (Supplementary Fig. S2B–S2F). Therefore, targeted inhibition of CK1δ/CK1ε is well tolerated in vivo and shows efficacy across three independent transplant models of multiple myeloma.

Given the potent ex vivo and in vivo efficacy of SR-3029, we compared the anti–multiple myeloma activity of SR-3029 to standard-of-care drugs, and to our panel of PKIs, on CD138⁺ primary multiple myeloma patient samples cultured in the TME-reconstructed platform. Seventy-six primary multiple myeloma patient samples, spanning the course of disease from NDMM (treatment naïve), early relapse refractory multiple myeloma (early RRMM; 1–3 lines of therapy), or late relapse/refractory multiple myeloma (late RRMM; ≥4 lines of therapy; Supplementary Table S1), were tested to determine their sensitivity to these agents. Strikingly, 75/76 multiple myeloma patient samples were sensitive to SR-3029 in the nanomolar range, and SR-3029 was more potent in this assay than several current anti-multiple myeloma therapies (Fig. 4A). Furthermore, SR-3029 was one of the most potent PKIs tested (Fig. 4B and C).

Patients with multiple myeloma become progressively refractory to therapeutic options as the disease progresses (9, 62, 63). We therefore assessed if sensitivity to SR-3029 correlated with clinical progression of multiple myeloma. SR-3029 had a similar nanomolar potency regardless of disease progression status (Fig. 4D). Thus, SR-3029 has potent activity against nearly all multiple myeloma patient specimens and, consistent with drug-resistant multiple myeloma cell line data (Fig. 2A), its anti–multiple myeloma activity is not influenced by therapy history.

To gain insights into pathway(s) affected by targeting CK1δ/CK1ε, we treated CD138⁺-selected cells from 5 patients with multiple myeloma with either vehicle or 250 nmol/L SR-3029 for 24 hours in a transwell assay with patient-derived BM stromal cells in the bottom well, and performed RNA-seq. These analyses revealed global transcriptomic changes in response to SR-3029 treatment, including significant (q < 0.01 and |log₂ FC| > 0.585) downregulation of 725 genes and upregulation of 437 genes (Fig. 4E). Pathway enrichment analysis revealed SR-3029 treatment provoked significant downregulation of genes involved in several CANCER HALLMARK signatures related to DNA damage, development, immunity, and key metabolic pathways (Fig. 4F), in particular glycolysis and oxidative phosphorylation (OxPhos). This was also confirmed using the DAVID bioinformatics database, which showed SR-3029 treatment led to significant reductions in the expression of genes that control metabolism, and specifically signatures of the mitochondrion (Fig. 4G).

To support mechanistic and functional investigations, we performed RNA-seq analyses to assess whether the SR-3029–regulated gene signature present in multiple myeloma patient samples was also observed in 8226 and MM.1S multiple myeloma cells treated with SR-3029. These analyses revealed a similar up- and downregulation of genes following SR-3029 treatment of these multiple myeloma cell lines (Fig. 5A).

To functionally validate these findings, mitochondrial stress tests (MST) and glycolytic rate assays (GRA) were performed using the Seahorse XFe Analyzer in viable MM.1S, 8226, and 5TGM1 cells treated with vehicle or SR-3029 (Fig. 5B–E, H, and I; Supplementary Fig. S3A and S3B). SR-3029 treatment provoked time-dependent inhibition of basal levels of OxPhos in MM.1S, 8226, and 5TGM1 multiple myeloma cells (Fig. 5B–E; Supplementary Fig. S3A and S3B). Furthermore, inducible knockdown of CK1δ or CK1ε also triggered reductions in the basal respiration rates in both the 8226 and MM.1S multiple myeloma cells (Fig. 5F and G). Interestingly, treatment with the mitochondrial uncoupler FCCP revealed that oxygen consumption rates (OCR) in vehicle-treated multiple myeloma cells were already at maximum levels (Fig. 5B and D; Supplementary Fig. S3A), suggesting a reliance of multiple myeloma cell lines on OxPhos. In addition, SR-3029 treatment led to reductions in glycolytic rates, but these generally followed a collapse in OCR (Fig. 5H and I; Supplementary Fig. S3C and S3D), and this was not observed following CK1δ or CK1ε knockdown (Supplementary Fig. S3E and S3F). There was a transient increase in basal glycolysis in the MM.1S cell line following SR-3029 treatment, but this was not sustained, and rates of both OxPhos and glycolysis were severely impaired by 24 hours (Fig. 5H; Supplementary Fig. S3C).

As there are known effects of stromal cells on multiple myeloma biology, we also tested if the effects of SR-3029 on multiple myeloma cell OxPhos and/or glycolysis were altered by the presence of patient-derived BM stromal cells. Again, SR-3029 treatment abolished OxPhos and impaired glycolysis in both multiple myeloma cell lines irrespective of stroma coculture (Supplementary Fig. S3G–S3J).

MST studies also revealed that SR-3029 treatment led to time-dependent decreases in mitochondrial ATP production in 8226 and MM.1S multiple myeloma cells (Fig. 5J; Supplementary Fig. S3K). Given the observed decrease in OxPhos and mitochondrial ATP levels, we posited that SR-3029 may impact total mitochondrial mass and/or mitochondrial membrane potential. MitoTracker Green staining revealed that SR-3029 treatment led to increases in mitochondrial mass (Supplementary Fig. S3L), yet these mitochondria were impaired in their function, as shown by reductions in mitochondrial membrane potential (Fig. 5K; Supplementary Fig. S3L).

To further explore how CK1δ/CK1ε inhibition compromises multiple myeloma cell metabolism, we performed untargeted metabolomics of both MM.1S and 8226 multiple myeloma cells treated with vehicle versus SR-3029 for 8 or 24 hours. Analysis of the metabolome revealed that SR-3029 treatment provoked significant, global changes in metabolites, and that these changes were generally amplified with time (Supplementary Fig. S4A and S4B). Consistent with an impact on OxPhos, CK1δ/CK1ε inhibition provoked significant reductions in Krebs Cycle (TCA) intermediates (succinate, fumarate, and malate) in both MM.1S and 8226 cell lines treated with SR-3029 (Fig. 6A and B), as well significant reductions in nucleotides, nucleosides, and select amino acids, that is, glutamate, aspartate, and glycine (Supplementary Fig. S4A and S4B).

To investigate the mechanism(s) by which targeting CK1δ/CK1ε suppresses OxPhos and levels of TCA intermediates, we examined the DAVID mitochondrial RNA-seq signature derived from patients in multiple myeloma cell lines. Notably, genes in the DAVID mitochondrial gene signature that are the most profoundly downregulated pathway in SR-3029–treated patient samples (GO:0005739-mitochondrion, Fig. 4G) were also downregulated in SR-3029–treated multiple myeloma cell lines (Fig. 6C). Of note, SR-3029 treatment provoked suppression of OGDH expression in both primary patient samples and multiple myeloma cell lines (Fig. 6D; Supplementary Fig. S5A). OGDH converts α-ketoglutarate to succinyl-CoA, which is necessary for the production of succinate, fumarate, and malate, which are reduced in SR-3029–treated multiple myeloma Cells (Fig. 6A and B; Supplementary Figs. S4 and S5B).

The predominant downregulated mitochondrial signatures of SR-3029–treated multiple myeloma patient samples (Fig. 4F and G), coupled with loss of TCA intermediates upstream of the electron transport chain (ETC) and decreased ATP production (Fig. 5J; Supplementary Figs. S3K, S4, and S5B), suggested there would be effects of targeting CK1δ/CK1ε on ETC activity. To test this hypothesis, MM.1S and 8226 multiple myeloma cells were treated with a plasma membrane permeabilizer (PMP) ± SR-3029 and the activity of each complex was determined using the Seahorse XFe96 Analyzer. SR-3029 treatment selectively impaired, in a dose-dependent fashion, the activity of complex I and complex IV activity in both multiple myeloma cell lines (Fig. 6E–H; Supplementary Fig. S5C and S5D). In contrast, SR-3029 treatment did not inhibit the activity of complex II/III in MM1.S cells but did impair complex II/III activity in 8226 cells (Supplementary Fig. S5E and S5F).

Interestingly, RNA-seq analyses of SR-3029-treated primary multiple myeloma samples and multiple myeloma cell lines revealed that suppression of complex I activity was associated with significant downregulation of several, but not all components of this complex, including NDUFA3, NDUFA7, NDUFB1, and NDUFB5 (Fig. 6C and I; Supplementary Fig. S5G). Thus, targeting CK1δ/CK1ε compromises the ETC at least in part by suppressing the activity and expression of components of complex I.

Furthermore, the observed repression of complex IV activity following SR-3029 treatment of multiple myeloma cell lines (Fig. 6F and H) was concordant with effects of CK1δ/CK1ε inhibition of the expression of these component in multiple myeloma cell lines (Fig. 6C), but the observed effects on expression of components from complex II, III, and IV was not observed in primary multiple myeloma patient samples (Supplementary Fig. S5H-S5J). Finally, qRT-PCR analysis of 8226 and MM.1S multiple myeloma cell lines confirmed that CK1δ/CK1ε inhibition suppressed the expression of most components of complexes I and IV in multiple myeloma cell lines (Supplementary Fig. S5K).

To address whether the observed reductions in OGDH mRNA provoked by SR-3029 was due to effects on transcription, we generated an OGDH promoter-firefly luciferase reporter that also harbors a Renilla luciferase transgene driven by the SV40 promoter (see Materials and Methods). HEK293T cells were stably transduced with this luciferase reporter plasmid or with this reporter plasmid lacking the OGDH promoter (promoterless). Treatment with SR-3029 led to specific reductions in OGDH promoter activity (Supplementary Fig. S5L), suggesting that CK1δ/CK1ε signaling controls OGDH expression at least in part at the level of transcription.

To assess whether SR-3029–mediated repression of complex IV activity (Fig. 6F and H) and the expression of its components that was observed in MM.1S and 8226 multiple myeloma cells (Supplementary Fig. S5I) was associated with reductions in the levels of the corresponding proteins we performed immunoblot analyses. Although inhibition of CK1δ/CK1ε was associated with reductions of OGDH protein levels this was not observed for the COX1V, COX5A, and COX5B components of complex IV in either multiple myeloma.1S or 8226 myeloma cells (Supplementary Fig. S5M). Thus, effects of CK1δ/CK1ε inhibition on complex IV occur at both the mRNA and activity levels.

OxPhos can be supported by glucose oxidation through pyruvate or glutaminolysis and entry of α-ketoglutarate into the TCA cycle. The marked suppression in OxPhos (Fig. 5B–E) and reductions in TCA intermediates (Fig. 6A and B; Supplementary Fig. S4) observed following CK1δ/CK1ε inhibition suggested that this might be associated with differences in glucose and glutamine flux. To test this, 8226 multiple myeloma cells were treated with SR-3029 or vehicle and were labeled with either U-¹³C-glutamine or U-¹³C-glucose to determine the effects of SR-3029 on glutaminolysis or aerobic glycolysis, respectively.

U-¹³C-glutamine flux analyses revealed that CK1δ/CK1ε inhibition provokes marked increases in labeling of the TCA cycle intermediates α-ketoglutarate, malate, and citrate that are derived from glutamine (Supplementary Fig. S5N). Furthermore, consistent with an increased reliance on glutaminolysis following CK1δ/CK1ε inhibition, there are also increases in labeling of glutamate, and of aspartate, which can be derived from malate via the aspartate—malate shuttle. This markedly contrasts to the reductions in the steady state levels of these TCA intermediates and OCR following CK1δ/CK1ε inhibition, as shown by metabolomic and Seahorse flux analyses (Figs. 5B–E and 6A and B; Supplementary Fig. S4), suggesting increased anapleurosis, even with lower basal and maximal compensatory rates of OxPhos, is a compensatory response. Consistent with this notion, there is a corresponding reduction in these TCA intermediates derived from glucose following CK1δ/CK1ε inhibition (Supplementary Fig. S5O). These data indicate that following inhibition of CK1δ/CK1ε signaling that multiple myeloma cells upregulate anaplerosis in an attempt to sustain OxPhos. This compensatory response is likely also due to the observed reductions in nucleotides and their intermediates that is manifest following inhibition of CK1δ/CK1ε (Supplementary Fig. S4A and S4B), as aspartate and glutamine are needed for nucleotide biosynthesis.

In contrast to the marked effects on glutaminolysis, U-¹³C-glucose flux analyses revealed that CK1δ/CK1ε inhibition has little effect on the metabolism of glycolytic intermediates. However, it is evident that treatment with SR-3029 negatively affects the conversion of glucose to pyruvate, citrate, α-ketoglutarate, and malate, thereby increasing the dependency on glutamine oxidation (Supplementary Fig. S5O). Collectively, these findings support a model where inhibition of CK1δ/CK1ε is a new strategy to disable multiple myeloma due to their reliance on mitochondrial metabolism.

Recent studies have suggested that increased expression of mitochondrial gene sets is associated with progression from premalignant (MGUS) to newly diagnosed multiple myeloma (64). To assess whether mitochondrial signatures are also associated with disease progression and the evolution of drug resistance, we performed ssGSEA on RNA-seq data from CD138⁺-selected BM samples from 687 Moffitt Cancer Center patient specimens across the multiple myeloma treatment spectrum. This cohort included patients who had NDMM (n = 199), early RRMM (n = 256) and late RRMM (n = 232). ssGSEA revealed a significant increase in the expression of hallmarks of OxPhos, DNA repair, MYC and E2F targets, and the unfolded protein response (UPR) as patients progress in the clinic (Fig. 7A). Notably, ssGSEA revealed that OxPhos genes whose expression is suppressed by SR-3029 treatment ex vivo (Fig. 4F and G) is increased as patients go from NDMM to early RRMM, and again as patients progress from early RRMM to late RRMM (Fig. 7B). Furthermore, when SR-3029 treatment–associated gene signatures were averaged for each pathway (OxPhos, glycolysis, UPR, and DNA repair), there is an increase in the expression of these gene signatures as patients progress in the clinic (Fig. 7C; Supplementary Fig. S6A–S6C). Similarly, there are also increases in expression of genes in the DAVID mitochondrion GO SR-3029 treatment signature with disease progression (Supplementary Fig. S6D). Finally, complex I genes downregulated by SR-3029 treatment also show significant increases in expression as patients progress on therapy (Supplementary Fig. S6E).

SR-3029–associated suppression of gene signatures that are linked to disease progression suggested that CSNK1D and/or CSNK1E expression increases as patients progress on therapy. Indeed, in 483 patients with matched RNA-seq and outcome data (Supplementary Fig. S6F), both CSNK1D and CSNK1E levels increase as patients progressed from NDMM to late RRMM (Fig. 7D). Furthermore, multiple myeloma patients with higher levels of either CSNK1D or CSNK1E had worse outcomes versus those with lower expression levels (P = 0.042 and 0.018, respectively; Fig. 7E and F). After parsing patients into NDMM, early RRMM, and late RRMM, the association between CSNK1E and CSNK1D and poor outcomes is strongest in the late RRMM group (Supplementary Fig. S6G-S6L). Notably, mRNA levels of CSNK1D and CSNK1E are independently associated with patient progression and outcome, as their expression only weakly correlates with one another (Supplementary Fig. S7A). Thus, both of these kinases appear to independently contribute to multiple myeloma disease progression.

To assess the clinical relevance of these findings, we integrated RNA-seq data of primary CD138⁺-selected multiple myeloma patient specimens with their ex vivo sensitivity to SR-3029 (n = 75; Fig. 4D). GSEA identified several hallmark pathways that were enriched with a nominal P < 0.05 and false discovery rate (FDR) q < 0.1 based on multiple myeloma patient specimen resistance to SR-3029 (Fig. 7G). Consistent with SR-3029-treated gene expression signatures observed in 5 multiple myeloma patient samples (Fig. 4F and G), the hallmark pathways OxPhos, DNA repair, UPR, and GO: Mitochondrion negatively correlated with SR-3029 resistance across these 75 patients (Fig. 7G and H; Supplementary Fig. S7B). Of note, an increase in gene expression for the hallmark glycolysis pathway was associated with SR-3029 sensitivity (Supplementary Fig. S7C, nominal enrichment score, NES -1.08, P = 0.310, false discovery rate, FDR = 0.436). Thus, the pathways conferring sensitivity to CK1δ/CK1ε inhibition ex vivo correlate with metabolic processes that are upregulated during multiple myeloma disease progression.

To assess if the SR-3029-suppressed OxPhos and GO: Mitochondrion gene sets (Fig. 4F and G) provide a prognostic index for patients who could potentially benefit from CK1δ/CK1ε-targeting therapy, we analyzed ssGSEA scores from patient RNA-seq data. Using a NES P < 0.05 and FDR q < 0.1, patient outcomes were compared to those that are positively and negatively enriched for each pathway. These analyses revealed that patients positively enriched for either the GO:Mitochondrion or the OxPhos gene sets have a decreased survival probability versus patients negatively enriched in these gene sets (Fig. 7I; Supplementary Fig. S7D).

The suppressive effects of CK1δ/CK1ε on complex I activity and the correlations of mitochondrial metabolism and OxPhos signatures and CSNK1D and CSNK1E expression on disease progression, suggested that other complex I inhibitors would be similarly effective in disabling the growth and survival of multiple myeloma. To test this notion, the multiple myeloma cell lines 8226 and MM.1S were treated with the complex I inhibitors rotenone or piercidin A. Notably, treatment with either of these Complex I inhibitors impaired the growth and survival of multiple myeloma cells (Supplementary Fig. S7E–S7H). However, treatment of primary myeloid and B-cell progenitors with these complex I inhibitors nearly abolished their colony-forming potential (Supplementary Fig. S7I–S7K), indicating that such agents will lack a therapeutic window. Thus, once further developed, CK1δ/CK1ε inhibitors are likely a better therapeutic option for the treatment of relapsed/refractory multiple myeloma.

MYC is a master regulator of cancer cell metabolism (65, 66) and hallmark MYC Targets V1 and V2 are highly enriched in multiple myeloma patient samples that are sensitive to SR-3029 (Fig. 7G). In accord with previous findings showing that high MYC protein levels connote poor multiple myeloma outcomes (67), and in the study cohort presented herein, MYC Target V1 is prognostic for multiple myeloma patient survival (Supplementary Fig. S7L). This appears to reflect increases in MYC protein or activity, as MYC mRNA levels do not correlate with progression or outcome (Supplementary Fig. S7M and S7N). Thus, both a SR-3029 suppressed OxPhos signature and MYC target genes connote poor outcome in multiple myeloma. Interestingly, our RNA-seq analyses of patient samples treated with SR-3029 did not identify MYC target genes as directly affected by CK1δ/CK1ε inhibition, indicating that these are independent pathways that contribute to disease progression.

Despite an expanding list of antimyeloma agents (2–7), nearly all patients with multiple myeloma ultimately succumb to multidrug-resistant disease. Here we establish the serine/threonine kinases CK1δ and CK1ε as therapeutically attractive targets across the disease spectrum of multiple myeloma. Notably, dual targeting of CK1δ/CK1ε provokes metabolic catastrophe in multiple myeloma by rapidly disabling OxPhos, and this is linked to suppressed expression of components of complex I, the collapse of the ETC, and marked reductions in ATP production. Moreover, CK1δ/CK1ε inhibition also leads to reduced expression of OGDH, and to corresponding reductions in downstream intermediates of the TCA cycle. Finally, upregulation of CK1δ/CK1ε-dependent mitochondrial and OxPhos signatures are a hallmark of disease progression and worse patient outcomes. Collectively, these findings suggest the CK1δ/CK1ε-mitochondrial circuit as a priority for therapeutic intervention in multiple myeloma, especially in advanced disease.

Similar to findings in TNBC (28), we show elevated CSNK1D expression in multiple myeloma connotes poor outcomes and that CK1δ inhibition by SR-3029 treatment provokes rapid apoptosis. However, the sensitivity of TNBC to SR-3029 is selectively due to its ability to suppress CK1δ and the WNT–β-catenin pathway (28), whereas in multiple myeloma both CK1δ and CK1ε contribute to multiple myeloma growth and survival via their control of metabolism. Furthermore, SR-3029 sensitivity and resistance in multiple myeloma patient samples is not associated with WNT/β-catenin signaling, but rather clearly aligns with the control of mitochondrial metabolism. Thus, CK1δ/CK1ε signaling circuits in cancer are context specific.

Upregulation of CSNK1D and CSNK1E is manifest in LRMM versus NDMM patients, yet the mechanisms underlying this control are not yet resolved. Our preliminary analyses of mutations in regulatory regions of these genes in myeloma patient samples, and of their chromatin structure using assay for transposase-accessible chromatin using sequencing to define open chromatin regions, have thus far not revealed an association with increased levels of CSNK1D and CSNK1E mRNAs seen in LRMM. However, using chromatin immunoprecipitation, we have observed increases in H3K27 acetylation during multiple myeloma disease progression that are associated with increased CSNK1D and CSNK1E mRNAs in LRMM patients (data not shown), but increased patient numbers will have to be interrogated to affirm these findings are significant. Transcription factors whose expression is upregulated in LRMM versus NDMM patients, and which have binding motifs in the promoter-regulatory regions of CSNK1D and CSNK1E, include CTCF, YY1, and RAD21 (cohesin) that are components of the topologically associated domain epigenetic machinery that is essential for tissue differentiation and cancer (68). However, their roles in controlling CSNK1D and CSNK1E expression will have to be validated by genetic studies. Regardless, the data are most consistent with upregulation of CSNK1D and CSNK1E during myeloma progression being driven by epigenetic regulation of enhancer activity and associated machinery.

The contribution of altered metabolism to multiple myeloma drug resistance is largely unknown. To date, EMDR-associated drug resistance in multiple myeloma has been linked to HIF1α and LDHA activity (69), and increased mitochondrial metabolism has been reported in proteasome inhibitor–selected multiple myeloma cell lines (70). In addition, the sensitivity of multiple myeloma cells to the BCL2 inhibitor venetoclax is associated with reduced ETC activity (71). Our ssGSEA analysis of RNA-seq from 687 multiple myeloma patient samples (across the disease spectrum) indicates mitochondrial metabolic programs are upregulated during progression on therapy. Furthermore, RNA-seq studies of both ex vivo–treated multiple myeloma patient samples and multiple myeloma cell lines revealed that inhibiting CK1δ/CK1ε disables mitochondrial metabolism and has potent activity against both treatment-naïve and drug-resistant multiple myeloma cells.

Mechanistically blocking CK1δ/CK1ε in multiple myeloma provokes rapid metabolic catastrophe by inhibiting OxPhos and glycolysis. Although it is unclear how SR-3029 treatment inhibits glycolysis, the profound effects on OxPhos are clearly linked to the suppression of complex I components and to reductions in the activity of complexes I and IV of the ETC. These effects are also linked to OGDH suppression and to corresponding reductions in succinate, fumarate, and malate intermediates of the TCA cycle. Future studies are aimed at defining the mechanism(s) by which CK1δ/CK1ε signaling controls the expression of OGDH and complex I components, and the activity of complex IV.

Finally, there is a clear correlation between the expression of CSNK1D, CSNK1E, and OxPhos genes with both disease progression and decreased survival in patients with multiple myeloma. Thus, the development of clinical candidates targeting these kinases, or other agents that specifically disrupt OxPhos, may show therapeutic benefit. Consistent with this notion, we observed limited spare respiratory capacity in multiple myeloma cells, and our preclinical data and those of others (72) suggest that complex I–directed agents having a suitable therapeutic window may show benefit in impairing disease progression.

M.B. Meads reports a patent for “A Model of Clinical Synergy in Cancer” pending and a patent for “A Multiomic Approach to Mathematical Modeling of Gene Regulatory Networks in Multiple Myeloma” pending. P. Sudalagunta reports grants from Multiple Myeloma Research Foundation outside the submitted work; in addition, P. Sudalagunta has a patent for “A Model of Clinical Synergy in Cancer” pending and a patent for “A Multiomic Approach to Mathematical Modeling of Gene Regulatory Networks in Multiple Myeloma” pending. R. Renatino Canevarolo reports a patent for WO2021108551A1 pending and a patent for WO2022217136A1 pending. J.M. Koomen reports grants from NCI during the conduct of the study and grants from Bristol Myers Squibb outside the submitted work. A.S. Silva reports a patent for “A Model of Clinical Synergy in Cancer” pending, a patent for “A Multiomic Approach to Mathematical Modeling of Gene Regulatory Networks In Multiple Myeloma” pending, and a patent for “Methods for Assessing Cell Viability or Predicting Cell Response to a Treatment Using Cell Movement” issued. J.L. Cleveland reports grants from NCI/NIH during the conduct of the study; in addition, J.L. Cleveland has a patent for US 10,603,322 B2 issued. K.H. Shain reports grants from Abbvie and Karyopharm and personal fees from BMS, Janssen, GSK, Amgen, and Adaptive outside the submitted work; in addition, K.H. Shain has a patent for PCT/US2020/024217 issued and a patent for PCT/US2022/024217 issued. No disclosures were reported by the other authors.

K.L. Burger: Conceptualization, formal analysis, validation, investigation, methodology, writing–original draft, writing–review and editing. M.R. Fernandez: Formal analysis, validation, investigation, methodology, writing–review and editing. M.B. Meads: Conceptualization, data curation, formal analysis, validation, investigation, methodology, writing–original draft. P. Sudalagunta: Data curation, formal analysis, investigation. P.S. Oliveira: Data curation. R. Renatino Canevarolo: Data curation. R.R. Alugubelli: Data curation. A. Tungsevik: Data curation. G. De Avila: Data curation. M. Silva: Validation, investigation. A.I. Graeter: Validation, investigation. H.A. Dai: Formal analysis. N.D. Vincelette: Validation, investigation. A. Prabhu: Validation, investigation. D. Magaletti: Validation, investigation. C. Yang: Validation, investigation. W. Li: Validation, investigation. A. Kulkarni: Data curation. O.A. Hampton: Data curation. J.M. Koomen: Formal analysis, validation, investigation. W.R. Roush: Resources. A. Monastyrskyi: Resources. A.E. Berglund: Formal analysis, investigation, visualization, methodology, writing–original draft. A.S. Silva: Data curation, formal analysis, validation, investigation, visualization, writing–original draft. J.L. Cleveland: Conceptualization, formal analysis, supervision, funding acquisition, investigation, writing–original draft, writing–review and editing. K.H. Shain: Conceptualization, formal analysis, supervision, funding acquisition, investigation, writing–original draft, writing–review and editing.

The authors sincerely thank our multiple myeloma patients and their families for donating their samples for research purposes. They also thank the members of the Cleveland, Shain, and Silva laboratories, and the Multiple Myeloma Working Group for scientific discussions. The authors are also grateful for the expert assistance and service of the Proteomics & Metabolomics, Molecular Genomics, Translational Research, Comparative Medicine, Flow Cytometry, and Biostatistics and Bioinformatics Cores of Moffitt Cancer Center, and the Cores of M2Gen, Inc. This work is supported by grant NCI PSOC grant U54 CA193489 (to. A.S. Silva and K.H. Shain), by the Cortner-Couch Endowed Chair for Cancer Research from the University of South Florida School of Medicine (to J.L. Cleveland), by personal donations from Lisa France Kennedy (to. J.L. Cleveland), by NRSA F32 CA203217 (to M.R. Fernandez), by personal donations from the Pentecost Family Myeloma Research Center (to K.H. Shain and A.S. Silva), by NCI Comprehensive Cancer Grant P30-CA076292 to the H. Lee Moffitt Cancer Center & Research Institute, and by monies from the State of Florida to the H. Lee Moffitt Cancer Center & Research Institute.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.