We have compiled a large panel of HMCLs representing innate sensitivity and resistance to PIs and IMiDs (similar to refractory disease), as well as >10 pairs of parental and clonally-derived PI/IMiD-resistant HMCLs (P vs.R pairs; generated using dose escalation over a period of time) representing acquired or emerging drug resistance, and encompassing the broad spectrum of biological and genetic heterogeneity of myeloma (20). The human myeloma cell lines (HMCLs) were obtained from Dr. Michael Kuehl (NIH), Dr. Leif Bergsagel (Mayo Clinic, Scottsdale, AZ), and Dr. Jonathan Keats (The Translational Genomics Research Institute (TGen) Phoenix, AZ) as described in our earlier publication (20). The clonally related acquired PI-resistant HMCLs were generated from parental cell lines by dose escalation of bortezomib using pulses of once-weekly bortezomib treatment as described previously (21). These HMCLs were used as in vitro model systems to screen and re-purpose drugs for the management of PI-resistant myeloma.

We also found that in vitro drug responses of the four PIs (Bz, Cz, Opz, and Ix) were highly correlated in HMCLs (20). Therefore, we used Ixazomib, a second-generation PI, as a representative PI for our studies.

The lenalidomide (IMiD)-resistant cell line, MM1SLenR, was obtained as a gift from Dr. Keith Stewart, Mayo Clinic. The cell lines were authenticated at source and tested randomly at regular intervals for mycoplasma negativity. HMCLs were maintained in HMCL media supplemented with IL-6 (22).

Bone marrow-derived CD138+ primary myeloma cells were obtained from patients (n=16) following written informed consent at Mayo Clinic with prior IRB approval from the Mayo Clinic review board and used as ex vivo model systems. Written informed consent was received from participants prior to inclusion in the study. Participants have been identified by number, not by name. The primary samples were cultured as total bone marrow and put in culture to retain the patients’ natural BM microenvironment. Cells of interest were CD138+-gated during analysis.

Drugs, reagents, antibodies, and kits are listed in Supplementary Table S1 .

HMCLs were treated with increasing concentrations of CLF, PIs (represented by Ix), and IMiDs (represented by Len) as single agents or in combination for 48h, and cytotoxicity assays were performed using CellTiter-Glo^® Luminescent cell viability assay (Promega Madison, WI). Luminescence was recorded in a Neo2 Microplate Reader (Biotek), and half-maximal inhibitory concentration (IC₅₀) values were determined using by calculating the nonlinear regression using sigmoidal dose-response equation (variable slope). The drug cytotoxicity data and IC₅₀ values were used to compare response and to determine synergy by Calcusyn software (Biosoft; Chou-Talalay’s CI theorem) (23). CI value<0.9 denotes synergism.

Cells were cultured for 48h in the presence of indicated concentrations of compounds, harvested and washed, and incubated with Annexin V-FITC (BD Biosciences) and Propidium Iodide for 15min at room temperature in the dark. Apoptosis was measured by BD LSR II flow cytometry (BD Biosciences).

Caspase-3/7 activity assay was performed on the HMCLs using Caspase-Glo 3/7 luminescent assay kit according to manufacturer’s instructions (Promega Madison, WI). Briefly, 1×10⁶/ml cells were cultured in 6 well plates and treated with either CLF alone and/or combination for indicated times. After harvesting, cells were washed twice with cold PBS and resuspended with Caspase-Glo 3/7 reagents. Reactions were incubated for 1h at 37°C, and luminescence was determined using Synergy 2 Microplate Reader (Biotek). Cell death by apoptosis was also measured by immunoblotting analysis.

HMCLs were seeded in 6-well plates and incubated for 48h following treatment with 10µM CLF. Cells were then harvested, washed, and incubated with Propidium Iodide for 30min at room temperature. Cell cycle progression was measured using a BD LSR II flow cytometer (BD Biosciences). Flow cytometry data analysis was performed using FlowJo software (BD Biosciences).

Colony formation potential of untreated and drug-treated HMCLs was accessed using Methylcellulose-based assay according to manufacturer’s instructions (Methocult, Stem Cell Technologies). Briefly, myeloma cells were treated with either CLF alone and/or combination with PI for the indicated time period. Following incubation, cells were harvested, washed with PBS, and methylcellulose-based media (Methocult) was added to cells pellets, and colonies were allowed to grow for 4 weeks in 6-well plates. Images were captured after 4 weeks with an inverted microscope (Olympus, 4x/20x lens with color camera), and cell colony numbers were counted by Image J software (NIH).

Aldehyde dehydrogenase (ALDH) activity was assessed using the Aldefluor assay kit according to the manufacturer’s instructions (Stem Cell Technologies). Briefly, 1×10⁶ myeloma cells were treated with CLF-based regimens for 24h, harvested, and resuspended in 1ml Aldefluor assay buffer containing the ALDH substrate BODIPY-aminoacetaldehyde (BAAA). Negative control samples were treated with 5μl of diethylaminobenzaldehyde (DEAB) as an inhibitor of ALDH1 enzymatic activity. Cells were incubated for 30–45 minutes at 37 °C, then washed twice, and suspended in Aldefluor assay buffer. The brightly fluorescent ALDH⁺ cells were detected by BD LSR II flowcytometry.

For assessing apoptosis in quiescent CD138+ cells, myeloma cells were first stained with carboxyfluorescein succinimidyl ester (CFSE; Invitrogen) for 30 min, then washed and resuspended in RPMI1640 medium and were incubated at 37°C overnight. The next day, cells were treated and maintained further for 4 days with supplementation of the drugs every 48h. Cells were then harvested, washed, and were labeled with CD138-APC and annexin/V-FITC antibodies and then gated into non-dividing (CD138+CFSEbright) and dividing (CD138+CFSEdim) cell populations based on CFSE fluorescence intensity using a flow cytometer. Cells cultured in the presence of colchicine (100ng/ml; Sigma) were used to assess the range of CFSE fluorescence exhibited by cells that were undivided at the end of the culture time.

Side population cells were investigated using DyeCycle Violet (cell-permeable DNA binding) assay according to manufacturer’s instructions (Thermofisher). Briefly, 1×10⁶/ml cells were cultured in 6 well plates and treated with either CLF alone and/or combination for indicated times. After 96h, cells were stained with 10μM Vybrant DyeCycle Violet (DCV; Molecular Probes, Eugene, OR) for 90 min at 37°C. Samples incubated with 100μM verapamil (Sigma-Aldrich) for 30 min were used as positive control. Following dye incubation, cells were washed twice with ice-cold PBS, chilled on ice, and immediately analyzed by flow cytometry.

Cells were treated with CLF-based regiments for 24hr. After 24h, cells were incubated with 10μM DCFDA in RPMI (Phenol red-free) medium at 37°C for 30min, washed twice with phosphate buffer saline (PBS), and ROS production was measured using Synergy 2 Microplate Reader (Biotek).

Myeloma cells treated with CLF-based regimens for 24h were incubated with 2.5 μM DHE (in RPMI) for 15 min in the dark at 37°C. Cells were then washed once with PBS, and red fluorescence was detected by flow cytometry.

CLF-treated (24h) cells were incubated with 5μM JC-1 dye for 15 min in the dark at 37°C and washed twice in PBS, and then analyzed for red and green fluorescence by flow cytometry.

Cells were plated at a density of 4X10⁵/ml cells and were treated with either CLF alone, Ixa alone, or combination. 24hrs after incubation, high-quality RNA was isolated using QIAshredder and RNAeasy mini kit (Qiagen). RNA concentration and integrity were determined using the Nanodrop-8000 and Agilent 2100 Bioanalyzer and stored at -80°C. RNA integrity number (RIN) threshold of 8 was used for RNA-seq analysis. RNA-seq libraries were constructed using Illumina TruSeq RNA sample preparation kit v2. Libraries were then size-selected to generate inserts of ~200bp. Next-generation RNA sequencing was performed on Illumina’s NovaSeq platform using 150bp paired-end protocol with a depth of >20million reads-per-sample.

Gene expression data was pre-processed, filtered (genes with mean counts<10 were removed), and GEP (CPM – counts per million) data was analyzed further using Partek Flow package to perform differential expression testing to identify GEP signatures. LS (least squares) mean values were calculated for each group as the linear combination or sum of the estimated means from the linear model. LS mean is model-dependent and produces more accurate, unbiased estimate of the group means even in unbalanced data. We used an LS Mean threshold of >=1 for downstream analysis. We used Gene Specific Analysis (GSA) to perform differential gene expression analysis between groups that applies limma, an empirical Bayesian method, to detect the differentially expressed (DE) genes. The advantage of limma compared to traditional t-test is that limma provides a moderated t-test statistic by shrinking the variance statistics, therefore, improving the statistical power. Mean fold-change>|1| and p<0.05 was considered as threshold for reporting significant differential gene expression. Heatmaps were generated using unsupervised hierarchical clustering (HC) analysis based on the top DE genes (DEGs).

Ingenuity pathway analysis (IPA) software was used to identify, i) the most significantly affected molecular pathways, ii) upstream regulator molecules like microRNA and transcription factors, iii) downstream effects and biological processes, and iv) causal networks predicted to be activated or inhibited in response to treatment based on the most significant DE genes (24).

Mass cytometry (CyTOF) analysis is a single-cell proteomics methodology that combines flow cytometry and elemental mass spectrometry. Thirty-seven antibody targets directed against cell surface (neoplastic myeloma) and intracellular markers (immune tumor microenvironment) were utilized for immunophenotyping panels. The antibody markers and respective metal conjugates are described in Supplementary Table S2 . 1-3million untreated or treated (for 24h) HMCLs or primary human myeloma cells were used for CyTOF staining. Cells were stained with cisplatin (1:10,000 dilutions in RPMI1640 media without FBS) at 37°C for 5 minutes, washed with media with 10% FBS, spun at 300 x g, and incubated with media with 10% FBS for 15 min. This was followed by another wash, spin, and the pellet was resuspended in 1x Fix buffer for 10 minutes at room temperature. HMCL samples, along with lyophilized healthy human PBMC (Veri-Cells PBMC: control cells; BioLegend), were washed using Maxpar cell staining buffer (MCSB) (Fluidigm). A total of 16 µL of the surface antibody cocktail was added to each tube containing each cell suspension in 100 µL of MCSB and incubated for 30 minutes at room temperature. Cells were washed with MCSB to remove excess antibodies and stored overnight at -80 C after methanol fixation. Cells were washed with MCSB and incubated with 20 µL of intracellular antibody cocktail for 30 minutes at room temperature, followed by wash and fixation with fresh 1.6% formaldehyde solution for 10 minutes at room temperature. Cells were washed with MCSB and resuspended in 1 ml of Cell-ID™ Intercalator solution (1:4000 dilution in permeabilization buffer, Maxper Fix, and Perm Buffer). All samples were barcoded using cell-ID™ 20-Plex Barcoding Kit (Fluidigm), which were then stained, washed with MSCB, and acquired as a single multiplexed sample using Maxpar Cell Acquisition solution contained calibration beads.

Samples were loaded onto a Helios CyTOF system (Fluidigm) using an attached autosampler and were acquired at a rate of 200-400 events per second. Data were collected as.FCS files using the CyTOF Software (Version 6.7.1014, Fluidigm). After acquisition, intrafile signal drift was normalized to the acquired calibration bead signal using the CyTOF Software.

For CyTOF data analysis, Cytobank software version 7.3.0 (Santa Clara, CA) was used for cleanup of cell debris, removal of doublets and dead cells, and analysis of cleaned fcs files. Clustering, dimensionality reduction (to 10,000 events per file), and the visualization of cell population cluster map were performed by t-SNE. Relative marker intensities and cluster abundances per sample were visualized by a heatmap.

Cells treated with CLF, PI, or IMiD as single-agent or as a combination for 24h were harvested, washed, and lysed using radioimmunoprecipitation assay (RIPA) lysis buffer containing 50mM Tris-HCl, pH 7.5, 150mM NaCl, 1% NP40, 5mM EDTA, 1mM DTT, phosphatase and protease inhibitors cocktail (Sigma), and incubated on ice for 15min. Samples were then centrifuged at 14000 rpm at 4°C for 30mins. The supernatant was then aspirated, total protein was isolated and quantified using Pierce BCA Protein Assay Kit (Thermo Scientific). Samples were solubilized in SDS-PAGE sample buffer, and equal amounts of protein were loaded per lane of 10% SDS polyacrylamide gels and transferred onto PVDF membranes (Millipore; Billerica, MA). Membranes were blocked in TBS with SuperBlock blocking buffer (Thermo Fisher). Membranes were then incubated with primary antibodies and secondary antibodies in TBS with 0.2% Tween 20 and 2.5% BSA. Immunoreactivity was detected by Chemiluminescent HRP Substrate (Bio-Rad), and the exposed image was captured using a ChemiDoc MP Imaging System (Bio-Rad). Densitometry analysis was performed (in triplicates) using Image J software.

Total RNA isolation from untreated HMCLs and quantification were performed using QIAshredder and RNAeasy mini kit (Qiagen). RNA concentration was determined using the Nanodrop-8000 spectrophotometer, and cDNA was prepared using QuantiTect Reverse Transcription kit (Qiagen). Following reverse transcription, TaqMan gene expression assay was performed using AHR mRNA specific TaqMan primers (TaqMan Real-Time PCR Assays) and TaqMan Fast Advanced Master Mix in CFX96 Touch Real-Time PCR Detection System (Bio-Rad, Hercules, CA). Relative AHR expression in the myeloma lines was calculated using the ΔCt method following normalization with Beta-Actin expression (housekeeping gene).

The drugs were screened in a 7-point, 10-fold dilution, with the highest concentration at 10 µM, and incubated for 24h. Anti-MM activity was assessed through cellular viability, evaluated with CellTiter Glo. The mid-point EC₅₀, percentage of maximum inhibition was calculated using the TIBCO Spotfire^® v.7.0.0 software, and the area under the curve (AUC) was calculated with the GraphPad Prism v.8 software (25). This data was compared to in vitro response profiles and linked through integrated analyses to CyTOF profiles and multi-omics outcomes.

All statistical analyses were performed using R and GraphPad Prism. All tests were two-sided, and p<0.05 was considered statistically significant.

First, we evaluated the single-agent in vitro cytotoxicity of CLF for anti-myeloma activity in our HMCL panel, representing wide variation in PI and IMiD responses (IC₅₀) ( Figure 1A ). We found that CLF alone showed very potent inhibition of cell viability in HMCLs representing sensitive as well as innate and clonally-derived resistant HMCLs. The single-agent IC₅₀ (48h) values of CLF were between 0.2µM- 20.5µM. Next, we compared the link between CLF IC₅₀ of the myeloma cell lines and MM molecular/cytogenetic abnormalities (22). Our results showed no significant association of CLF response with cytogenetic abnormalities (data not shown).

Further, bone marrow-derived CD138-positive PMCs were obtained from patients (n=12) at Mayo Clinic and used as ex vivo model systems. Using our established direct-to-drug screening assay, we screened each patient’s tumor in a Phase 0 assay for drug sensitivity to single-agent CLF (25). The ex vivo CLF EC₅₀ values (1-15 µM; minimum EC₅₀ 1052.1nM, maximum EC₅₀ 15210nM, median EC₅₀ 2408.7nM) were within the in vitro IC₅₀ range. Figure 1B shows representative CLF single-agent ex vivo cytotoxicity plots in myeloma patients.

Next, we tested the cytotoxicity of CLF in combination with PIs (represented by Ixazomib; Figures 2A–C ) or IMiDs (represented by Lenalidomide; Figures 2D, E ) in HMCLs representing innate-sensitive (FLAM76, KAS6/1, MM1S), innate-resistance (JIM-3, LP-1), and acquired- PI or IMiD resistance (U266 P/VR, RPMI8226 P/VR, JJN-3 P/VR, and MM1S P/LenR). The CLF+PI and CLF+Len combination index (CI) values calculated using the Calcusyn program were consistently less than 0.9 ( Figure S1 ), indicating synergy (26). Further, CLF improved the therapeutic index of PI and IMiD administration to the cells and decreased the amount of PI/IMiD required to achieve effective responses, as indicated by dose reduction index (DRI) values and predicted decrease in IC₅₀ (nM concentration).

Although the CLF doses used in combination treatments were in the micromolar concentration range, this is within the safe dose range of 0.84-8.4µM, which corresponds to human plasma Cmax of 0.4-4mg/L (10).

We performed Mass Cytometry (CyTOF) analysis to assess CLF-induced changes in phenotypic and functional markers in myeloma cells on a single-cell level and identify unique subgroups that change in relation to disease progression. CyTOF analysis was performed on 77 total samples across 7 Experiments/Batches. This included 4 isogenic sensitive/acquired PI and IMiD resistant pairs (U266, MM1S, RPMI8226, JJN3), 8 innate-sensitive cell lines, and 7 innate-resistant HMCLs. Batch correction was performed for combining samples. Similar clusters across all samples were grouped to compare sub-populations and to calculate the proportion of cells with increases or decreases in markers for each sample. CyTOF analysis revealed distinct cluster of cells defined by elevated cleaved caspase levels in all cell lines and primary samples, which was enriched for cells exposed to high dose CLF ( Figure 3A ).

Myeloma cells are addicted to several proteins like c-Myc, IRF4, and IKZF1. Our pre- vs. post-treatment differential expression analysis using CyTOF ( Figure 3B ) and immunoblotting ( Figure 3C ) revealed shifts in a majority of these markers following clofazimine treatment, including IRF4, IKZF1, IKZF3, CD229, CD27, pS6, pERK, and IkBa.

Furthermore, we had earlier found that CLF also suppresses STAT expression in CML and consequently downregulated stem cell maintenance factors like hypoxia-inducible factor-1α and -2α and Cbp/P300 interacting transactivator with Glu/Asp-rich carboxy-terminal domain 2 (CITED2) (10). Concurrently, we also showed downregulation of STAT5 and HIF-1α in myeloma following CLF treatment ( Figure 3D ).

Next, to confirm whether the loss of cell viability was indeed due to apoptosis (as indicated by CyTOF analysis results; Figure 2 ), we performed annexin V staining using flow-cytometry. CLF induced significant apoptosis either alone or in combination with 15nm Ixazomib in HMCLs ( Figure 4A ). CLF-induced caspase-3 activity was also confirmed by luminescent-based Caspase 3/7 assay (Promega) ( Figure 4B ). We observed that CLF activated caspase-3 and 9 but not 8, indicating that CLF-induced apoptosis was dependent on mitochondria-mediated pathway (Western blotting; Figure 4C ). This is consistent with decrease in the anti-apoptotic/survival markers Bcl2, Mcl-1 and increase in Bax expression ( Figure 4C ). Furthermore, using Caspase 3/7 assay ( Figure S2A ) and western blotting ( Figure S2B ), we demonstrated that CLF treatment results in induction of apoptosis in myeloma lines. When gated for live cells, the percentage of cells in G0/G1, S, G2/M, and sub G0/G1 phases in pre- vs. post- CLF treatment HMCLs are shown in Figure S2C .

Previously, it has been reported that CLF imparts its anti-bacterial actions by generating reactive oxygen species (ROS), particularly superoxides and hydrogen peroxide (H₂O₂). DCFDA (Sigma), which shows fluorescence when it is oxidized, was used to measure intracellular ROS. Cellular superoxide anions were measured by using the fluorescent dye DHE (Sigma). Mitochondrial membrane potential was assessed using JC-1 (Sigma). JC-1 is a cationic carbocyanine dye that accumulates in mitochondria. Representative plots demonstrate that we observed induction of cellular superoxide anions ( Figure 4D ) and intracellular ROS production ( Figure 4E ) that causes mitochondrial membrane depolarization following CLF treatment in the cell pair RPMI8226P/VR ( Figure 4F ). Similar results were also obtained for the U266P/VR cell line pair ( Figures S3A, B ).

We observed CLF alone significantly reduced colony number as well as colony size when compared to control or Ixazomib. Combination of CLF with Ixazomib further reduced the colony numbers. Importantly, CLF significantly reduced colony formation in PI-resistant lines (RPMI826 P/VR and U266 P/VR) ( Figure S4 ).

Figures 5A , S5A show baseline ALDH (aldehyde dehydrogenase) activity in PI-resistant HMCLs compared to sensitive HMCLs. While no/very low ALDH activity was observed in the PI-sensitive line FLAM76, we found ALDH activity is remarkably higher (>95%) in clonally derived PI-resistant HMCLs RPMI8226-VR, U266-VR compared to innate resistant LP-1 cell line. CLF caused a significant decrease in ALDH activity, and its effect was comparable with control ( Figure 5A ).

We performed carboxyfluorescein succinimidyl ester (CFSE) assay to evaluate if CLF alone or in combination with PIs could erode quiescent CFSE+ cells in parental (P) and clonally derived PI-resistant (VR) cells, RPMI8226P, RPMI8226VR, U266P, and U266VR. While PIs failed to reduce CFSE-bright (non-dividing) cells, CLF alone or in combination with Ixazomib significantly reduced their number and increased CFSE-dim (dividing cell) population ( Figures 5B , S5B ). Remarkably, evaluation of apoptosis in these cells revealed that CLF alone caused apoptosis in both CFSE^bright and CFSE^dim cells while combining CLF+Ixazomib caused a more robust effect amounting to near-obliteration.

We next gated and selected side population (SP) cells from main populations (MP) using DyeCycle violet, pre- and post- CLF treatment. We found that baseline %SP is higher in resistant cells as compared to parental cells, as shown in ( Figures 5C , S5C ). Notably, CLF alone or in combination (CLF+PI) reduced SP in PI- resistant HMCLs ( Figures 5C , S5C ). A summary of the results is provided as bar plots in Figure 5D .

Genome-wide transcriptome (next-generation mRNA sequencing or RNAseq) analysis showed 864 genes differed significantly at baseline between the CLF-sensitive and the CLF-resistant groups (p<0.05; fold-difference≠1). Among these, 524 genes had least squares (LS) Mean of 1 or higher in both sensitive and resistant groups. IPA analysis revealed Phagosome Maturation (p=2.45E-05), DNA Methylation and Transcriptional Repression Signaling (p=6.80E-04), Autophagy (p=1.51E-03), BEX2 signaling pathway (p=1.34E-03) as the top canonical pathways associated with CLF sensitivity.

Next, we performed analysis of kinetic changes in gene expression patterns between untreated (baseline) vs. treated (24hrs post-treatment) cells. Figure S6 shows a heatmap of all the genes with LS mean>=1 that were significant (ANOVA p<0.05) in either CLF vs. CON and/or CLF+IXA vs. CON. Our pair-wise GSA analysis for CLF vs. CON showed that 172 differentially expressed (DE) genes displayed p-value less than 0.05 (|fold-change|>1) in response to single-agent CLF treatment, 24 hours following drug exposure to 5uM CLF. Among these, 45 genes had LS Mean>=1. Figure S7 shows volcano plots for each pair-wise comparison using gene-specific analysis/GSA. Figure 6A shows a heat map of the top DE (CON vs. CLF) genes. When single-agent CLF-induced kinetic changes for each HMCLs were considered separately - 714, 426, 128, 33, and 10 genes were differentially regulated in RPMI8226, FLAM76, JIM3, U266, and LP1, respectively at |fold-change|>2 (p<0.05). This suggests that the number of highly expressed genes decreases with increase in drug resistance, which is consistent with our earlier observation that GEP data may be highly influenced by the DE genes expressed in sensitive HMCLs (22). The Venn diagram in Figure 6B shows 46 significant (p<0.05) genes were common between all the CLF vs. CON signatures. IPA analysis ( Figure 6C ) based on these shared gene signatures revealed mitochondrial dysfunction and oxidative phosphorylation as top canonical pathways and predicted downregulation of MYC as the top regulatory gene.

On the other hand, 279 genes changed significantly between untreated vs. CLF+Ixa combination-treated samples (p<0.05; fold-difference≠1). 179 genes had |fold-change|>2 (p<0.05). Among these, 114 genes had LS Mean of 1 or higher in both groups. Figure 6D shows a heatmap of the top 50 genes associated with CLF+Ixa combination treatment. Seventy-one genes were common between the two comparisons (CLF vs. Control and CLF+Ixa vs. Control), as shown in the Venn diagram in Figure 6E .

IPA analysis based on top DE genes revealed Autophagy (p=2.59E-04), Unfolded protein response (p=3.11E-04), and ER stress pathway (p=4.31E-04) as the top canonical pathways associated with CLF treatment gene signatures ( Figure 7A ). BAG2 signaling (p=7.07E-36), FAT10 signaling (p=5.75E-34), polyamine regulation (p=6.61E-33), inhibition of ARE-mediated mRNA degradation pathway (p=8.47E-24) and protein ubiquitination (p=7.56E-22) were among the top canonical pathways for CLF+PI combination treatment ( Figure 7B ). Upstream regulator prediction revealed NR1H4, PPP1R15B, and UCP1 as top upstream regulars for CLF treatment and RICTOR ( Figure S8 ) and NFE2L2 for combination treatment. Results from the RNAseq and IPA analysis were confirmed using immunoblotting ( Figure 7C ) as well as CyTOF analysis of ~40 different CLF treatment-induced markers in HMCLs, and PMCs, as shown earlier.

Next, we compared the anti-myeloma efficacy of CLF with that of other Peroxisome Proliferator Activated Receptor Gamma (PPARg) agonists - rosiglitazone and pioglitazone. Single-agent CLF was found to be the most potent among all the PPARg agonists tested ( Figure 8 ).

An earlier study had reported that the anti-myeloma effect of CLF was due to its inhibition of the AHR-polyamine metabolism axis (12). We evaluated AHR expression using qRT-PCR (TaqMan Real-Time PCR Assays) in 10 myeloma cell lines and compared AHR expression with CLF cytotoxicity. Relative AHR expression (ΔCt) in the myeloma lines was calculated following normalization with Beta-Actin (housekeeping gene). Since the U266 VR cell line showed the lowest relative AHR expression, we used the ΔCt value of this line as the baseline to compute relative fold gene expression of all the HMCLs using the 2^-ΔΔCt method (27). The Scatter plot in Figure S9A shows that AHR expression was not significantly correlated with CLF response in myeloma cell lines (Spearman r = -0.14; p-value = 0.75). Furthermore, Figure S9B shows the cytotoxicity data of the AHR antagonist StemRegenin 1 (SR1) in FLAM76, LP1, U266P, and U266VR cell lines. As represented by IC₅₀ values, we found that treatment with SR1 was dependent upon baseline AHR expression (AHR-negative U266 P/VR and FLAM76 lines were highly resistant to SR1 compared to LP1; Figure S9C ), while CLF treatment response was not correlated with AHR expression ( Figure S9A ).