Mitochondrial-Targeted Decyl-Triphenylphosphonium Enhances 2-Deoxy-D-Glucose Mediated Oxidative Stress and Clonogenic Killing of Multiple Myeloma Cells

Therapeutic advances have markedly prolonged overall survival in multiple myeloma (MM) but the disease currently remains incurable. In a panel of MM cell lines (MM.1S, OPM-2, H929, and U266), using CD138 immunophenotyping, side population staining, and stem cell-related gene expression, we demonstrate the presence of stem-like tumor cells. Hypoxic culture conditions further increased CD138low stem-like cells with upregulated expression of OCT4 and NANOG. Compared to MM cells, these stem-like cells maintained lower steady-state pro-oxidant levels with increased uptake of the fluorescent deoxyglucose analog. In primary human MM samples, increased glycolytic gene expression correlated with poorer overall and event-free survival outcomes. Notably, stem-like cells showed increased mitochondrial mass, rhodamine 123 accumulation, and orthodox mitochondrial configuration while more condensed mitochondria were noted in the CD138high cells. Glycolytic inhibitor 2-deoxyglucose (2-DG) induced ER stress as detected by qPCR (BiP, ATF4) and immunoblotting (BiP, CHOP) and increased dihydroethidium probe oxidation both CD138low and CD138high cells. Treatment with a mitochondrial-targeting agent decyl-triphenylphosphonium (10-TPP) increased intracellular steady-state pro-oxidant levels in stem-like and mature MM cells. Furthermore, 10-TPP mediated increases in mitochondrial oxidant production were suppressed by ectopic expression of manganese superoxide dismutase. Relative to 2-DG or 10-TPP alone, 2-DG plus 10-TPP combination showed increased caspase 3 activation in MM cells with minimal toxicity to the normal hematopoietic progenitor cells. Notably, treatment with polyethylene glycol conjugated catalase significantly reduced 2-DG and/or 10-TPP-induced apoptosis of MM cells. Also, the combination of 2-DG with 10-TPP decreased clonogenic survival of MM cells. Taken together, this study provides a novel strategy of metabolic oxidative stress-induced cytotoxicity of MM cells via 2-DG and 10-TPP combination therapy.


Introduction
Multiple myeloma (MM) is a debilitating plasma cell malignancy. The American Cancer Society estimates that in the United States in 2016 approximately 30,000 new cancer diagnosis and 12,500 deaths will be linked with MM [1]. New pharmacological drugs (i.e. lenalidomide, pomalidomide, bortezomib, carfilzomib, ixazomib, panobinostat) show anti-MM activity and in combination with conventional therapies offer deep clinical remissions [2]. However, most patients eventually relapse with aggressive drug-resistant MM disease [3]. Resistance to chemotherapy and MM relapse has been attributed to persistence of residual disease that contains drug-resistant phenotypes i.e. de-differentiated pre-plasma cells, genetically distinct clones, and stem-like tumor cells [4][5][6]. Although the cancer stem cell (CSC) hypothesis remains controversial in MM, several groups have demonstrated a distinct stem-like sub-population that displays increased clonogenicity, sustained self-renewal, differentiation towards CD138 + mature MM cells, and chemo-resistance [5,7,8]. Therefore, successful therapy requires an approach to induce cytotoxicity in the drug-resistant tumor cells with the overall objective to reduce MM recurrence.
Cancer cells almost universally show metabolic reprogramming with an increased reliance on aerobic glycolysis (Warburg effect) [9]. In cancer cells, an increased glycolytic flux is believed to provide biosynthetic intermediates for cell proliferation and also compensate for upregulated reactive oxygen species (ROS i.e. O 2 •− and H 2 O 2 ) formed as by-product of aberrant electron transport chain (ETC) activity [10][11][12][13]. Glucose metabolism generates pyruvate that can directly detoxify peroxides and NADPH that act as reducing equivalents in thioredoxin and glutathione systems [13][14][15]. Recent studies have shown that targeting the glycolytic pathway results in cytotoxicity in MM cells [16][17][18][19][20]. Similar to cancer cells, adult stem cells and CSCs have been reported to show metabolic reprogramming with altered redox homeostasis [21,22]. Notably, CSCs can exhibit dependence on glycolytic metabolism or mitochondrial oxidative phosphorylation (OXPHOS) that is integral to their survival and self-renewal properties [23][24][25][26]. To the best of our knowledge, the pro-oxidant metabolic state of MM stem-like cells has not been deciphered or used as a target for inhibition of MM disease recurrence. Studies have shown that targeting oxidative metabolism can provide a therapeutic benefit in cancer [12,[27][28][29]. Our published results have established a biochemical rationale (based on intrinsic differences in cancer vs. normal cell metabolism) for manipulating ROS to augment MM cytotoxicity [30,31]. We have also established isogenic MM cell lines with acquired bortezomib resistance and published that bortezomib resistance correlates with upregulated expression of antioxidant enzymes [32]. In the present study, we compromised cellular antioxidant function and increased pro-oxidant production with the aim to selectively induce oxidative stress-mediated cytotoxicity in the MM stem-like fractions. For inhibition of glycolytic and oxidative metabolism, we utilized 2-deoxyglucose (2-DG), which is being actively pursued in clinical trials as an adjuvant to cancer chemotherapy [33,34]. Biochemically, hexokinase converts 2-DG to 2-deoxy-glucose-6-phosphate (2-DG-6-P), an intermediate that cannot be metabolized by phosphoglucose isomerase resulting in the downstream inhibition of glycolysis and induction of oxidative stress [12,35]. Also, besides inhibiting glycolysis, 2-DG can interfere with Nlinked protein glycosylation, induce endoplasmic reticulum (ER) stress, and apoptosis [36].
In this study, we have characterized the oxidative metabolic aberrations of MM stem-like cells and show that compared to bulk tumor cells, these stem-like cells maintain lower steadystate pro-oxidant levels with increased glucose uptake. We show that by combining 2-DG with 10-TPP, we can effectively reduce clonogenic fractions in MM cell lines without increasing cytotoxicity towards normal hematopoietic progenitor cells. The proposed first-in-class combination treatment of 2-DG with 10-TPP provides a novel way to induce cytotoxicity in tumor cells and can potentially inhibit chemo-resistance that is universally and rapidly encountered in MM therapy.

Ethics statement
All procedures involving animals in this study were carried out in accordance with the recommendations set forth in the Guide for the Care and Use of Laboratory Animals of the National Institutes of Health. The protocol was reviewed and approved by the Institutional Animal Care and Use Committee (IACUC), University of Iowa, Iowa (Protocol Number 1312229).
The normal hematopoietic progenitor cells (HPCs) were enriched from C57BL/6 mice (9-11 weeks old, The Jackson Laboratory, Bar Harbor, ME) as previously described [30]. All animal work was performed in accordance with the guidelines of the Institutional Animal Care and Use Committee (IACUC) of the University of Iowa (protocol Number 1312229). Mice were sacrificed by euthanasia with CO 2 gas inhalation followed by cervical dislocation. Bone marrow was harvested and depleted for lineage negative cells using the EasySep™ Mouse HPC Enrichment Kit (#19756, Stem Cell Technologies, Vancouver, BC, Canada). Cell population was confirmed using APC-conjugated c-Kit and PE-conjugated Sca-1 antibodies (#17-1171, #12-5981, eBioscience, San Diego, CA).

Flow cytometric analysis of stem-like MM fractions
For CD138 immunophenotyping, cells were plated at 1.0 x 10 6 /mL for 24 h in RPMI complete medium. Labeling was done with APC-or FITC-conjugated CD138 antibody (#17-1389 or #11-1389, eBioscience) for 30 min on ice. Hoechst 33258 staining (5 μg/mL Sigma-Aldrich, St. Louis, MO) was used to exclude dead cells. Sequential gating for viability, forward and side scatter was done. The CD138 low cells were gated by comparing the treatment samples to the unstained as well as the isotype control. The gate was drawn at the natural breakpoint between the populations, around 10 3 on the Y-axis. The same gate was applied to all samples within a cell line. CD138 low cells (a minimum of 3000 events) were captured on a BD LSR-Violet or LSR-UV (Flow Cytometry Core, UI) using λ ex = 633 nm and λ em = 660/20 nm bandpass filter (APC), λ ex = 488 nm or λ em = 530/30 nm bandpass filter (FITC). Data acquisition and analysis was performed with FlowJo software (Ashland, OR) and the median fluorescent intensity (MFI) of the CD138 low and CD138 high population was calculated and corrected using unlabeled cells. The MFI for the CD138 high population was averaged and used to calculate the fold change between the populations.
SP staining was performed as previous described [46]. Briefly, cells were stained at 37˚C with Hoechst 33342 (5 μg/mL, Sigma-Aldrich) in pre-warmed complete media for 90 minutes. The ABC transport pump inhibitor verapamil (100 μM, Sigma-Aldrich), which inhibits the dye efflux, was added with Hoechst 33342 and used as a control. Cells were pelleted at 4˚C, resuspended in cold Hank's Balanced Saline Solution (HBSS, Thermo Fisher Scientific) containing 2% FBS, and kept on ice prior to analysis. A minimum of 200,000 events were captured on a BD-FACS Aria II flow cytometer (Becton Dickinson) using λ ex = 355 nm, λ em = 450/50 nm bandpass filter (Hoechst blue) and λ em = 675/50 nm bandpass filter (Hoechst red). Flow cytometric data acquisition and analysis was performed with DIVA software (Becton Dickinson). SP cells were identified by comparing to the verapamil treated samples.
Mitochondrial oxidant production was measured by MitoSOX Red staining (5 mM for 15 min at RT, Invitrogen, Thermo Fisher Scientific) after 10-TPP treatment (1 or 2.5 μM, for MM.1S and OPM-2, respectively) [31]. The specificity of MitoSOX staining was confirmed using cells transduced with adenovirus expressing MnSOD (Ad-MnSOD) or control (Ad-CMV) as described by us [31]. Antimycin A (100 μM for 2 h) was used as positive control. These were then washed, placed on ice, and stained with APC-CD138 antibody for 30 min. Samples were analyzed by flow cytometry using λ ex = 488 nm and λ em = 585/42 nm bandpass filter (MitoSOX).

Assessment of mitochondrial mass
Cells were plated at 1 x 10 6 cells/mL in RPMI complete medium for 24 h followed by labeling with MitoTracker Green (Molecular Probes, Thermo Fisher Scientific, 0.1 μM) for 30 min at 37˚C. Cells were washed, placed on ice, and stained with APC-CD138 antibody for 30 min. Samples were analyzed by flow cytometry using λ ex = 488 nm and λ em = 530/30 nm bandpass filter (MitoTracker Green).

Quantitative real-time PCR (qPCR) and RT-PCR
Total RNA was isolated using the Direct-zol RNA kit (R2050, Zymo, Irvine, CA) and quantified on a NanoDrop 1000. cDNA was synthesized from 1 μg of total RNA, using the iScript cDNA synthesis kit (#170-8891, Bio-Rad Laboratories, Hercules, CA) as described by us previously [31]. The cDNAs were subjected to qPCR analysis with VEGF primers (5 0 !3 0 , CCTTG CTGCTCTACCTCCAC and CCACTTCGTGATGATTCTGC forward and reverse, respectively, amplicon length of 54 bp), BiP primers (5 0 !3 0 , TGTTCAACCAATTATCAGCAA ACTC and TTCTGCTGTATCCTCTTCACCAGT forward and reverse, respectively, amplicon length of 73 bp), ATF4 primers (5 0 !3 0 , GTTCTCCAGCGACAAGGCTA and ATCCTCC TTGCTGTTGTTGG forward and reverse, respectively, amplicon length of 88 bp), or primers listed in Table 1. Primers design was done utilizing the Universal Probe Library Assay Design Center (Roche, Indianapolis, IN) specifying for intron spanning regions for the respective genes. qPCR was performed in a 10 μL reaction using the synthesized cDNA (10 ng), forward and reverse primers (100 μM), and 0.75x FastStart Universal SYBR Green Master (#0491385 0001, Roche). The assay was plated in triplicate and run on an ABI-7500 Fast and analyzed using the ABI 7500 software v2.3. The C T values for the target genes in all of the samples were normalized to the 18S transcript, and the fold difference (relative abundance) was calculated using the formula 2 -ΔΔ C T and mean value was plotted. In the first step, PCA was applied to genes with more than one probe site to create a unique gene-specific variable. In the second step, PCA was applied to all the gene-specific variables to create a universal genetic signature. In each step, the first principal component was retained. Using Cox proportional hazards regression, overall survival (OS) and event-free survival (EFS) were compared between Q1 and Q2+Q3+Q4 of the genetic signature. Plots of the Kaplan-Meier estimated cumulative probabilities of OS and EFS were constructed (Biostatistics Core, UI).

Clonogenic survival assay
To determine clonogenic potential of unsorted HMCLs, in vitro limiting dilution assay was done as published by us [32]. Cells were plated overnight at 2.5 x 10 5 /mL in RPMI complete medium and treated for 24 h with 10-TPP [0.02 or 0.1 μM (for MM.1S) and 0.2 or 1 μM (for OPM-2)] and/or 2-DG (20 mM). Cells were then plated in a U bottom 96-well plate, cultured for 10 days, and scored. The plating efficiency (PE), survival fractions, and normalized survival fraction (NSF) was calculated for each treatment.

Confocal imaging of 10-TPVP
Cells were plated at 1 x 10 6 /mL RPMI complete medium for 24 h. Mitochondrial imaging was using 10-TPVP, kindly provided by from the Pigge lab (Dr. For improving the quality of 10-TPVP image as well as the co-localization image of 10-TPVP with mitoTracker red, 10-TPVP fluorescence images, postacquisition, were pseudo-colored from blue to green using Adobe software.

Electron microscopy and mitochondrial configuration analysis
HMCLs were enriched for CD138 low and CD138 high cells and visible pellets (containing approximately 10 6 cells) were fixed overnight in 20 volumes of 2.5% gluteraldehyde in 0.1 M sodium cacodylate buffer. Post-fixation was done for 30 min at room temperature with a buffered 1% osmium tetroxide solution reduced with 1.5% potassium ferrocyanide. Samples were stained with 2.5% uranyl acetate and then rinsed and fully dehydrated using increasing concentrations of ethanol. Infiltration of Eponate 812 epoxy resin and ethanol was carried out over several hours to 100% resin and cured overnight in a 60˚C oven. Sections of 80 nm thickness were cut using a Leica UC6 Ultramicrotome II. Grids were then counterstained with 5% uranyl acetate for 3 min and Reynold's lead citrate for 2 min. Samples were imaged using a JEOL JEM-1230 transmission electron microscope at 120 kV (NIH grant 1 S10 RR018998-01, Central Microscopy Research Facility, UI). To perform ultrastructure analysis, all mitochondria were selected and a minimum of approximately 20 mitochondria per cell type were manually scored by 7 unbiased volunteers and assigned to one of two broad categories: condensed (mitochondrial density > cytoplasmic density) or orthodox (mitochondrial density = cytoplasmic density) [54, 55]. The ratio of orthodox/condensed mitochondria was calculated for each sample.

Statistical analysis
GraphPad Prism 6.0 software (San Diego, CA) was used for data handling, analysis, and presentation. For all experiments, data depicts average of repeat experiments (n = 2 or 3) run in technical replicates (n = 3) as indicated in figure legends. Statistical significance was determined using either a 2-way ANOVA or two-tailed unpaired t test with a confidence interval of 95%. Microarray data analysis was done using R (http://www.r-project.org). For Kaplan-Meier survival curves, all statistical tests were two-sided and assessed for significance with the SAS 9.3 software package (Cary, NC). P < 0.05 was considered to be statistically significant.

HMCLs contain a distinct stem-like tumor cell population
Low expression of CD138 (syndecan-1, plasma cell differentiation marker) has been used as a phenotypic marker for MM stem-like cells [56][57][58]. In MM.1S, OPM-2, H929, and U266 cells, a distinct sub-population of CD138 low cells (approximately 1-4.2%) was detected by flow cytometry (Fig 1A and 1B). As the robustness of CD138 to correctly identify stem-like MM cells remains debatable [59], side population (SP) staining was also done which an established functional marker of adult stem cells and many types of CSCs including MM [6,[60][61][62]. The SP fractions in MM.1S and OPM-2 cells were~0.4% ± 0.1% and 2.3% ± 0.5%, respectively and verapamil treatment reduced SP fraction by approximately 80% confirming correct SP gating (Fig 1C and 1D). Gene expression profiling has shown a higher expression of induced pluripotent stem cell genes in stem-like CD138than bulk CD138 + cells isolated from HMCLs [63]. We enriched the CD138 low and CD138 high cells from our HMCL panel and examined mRNA expression of stem cells transcription factor genes (OCT4, NANOG), ATP-binding cassette (ABC) transporter family member ABCG2/ breast cancer resistance protein (BCRP), and syndecan-1 (SDC1) by qPCR. For all HMCLs, CD138 low cells showed higher OCT4, NANOG and BCRP expression relative to the CD138 high cells (Table 1). A low SDC1 expression confirmed that CD138 low cells were genetically distinct and not comprised of MM cells that had shed surface CD138 (Table 1) [64]. To assess if SP staining may better represent stem-like MM cells, SP and main population (MP) cells were enriched from MM.1S and OPM-2 cells. For both cell lines, SP cells showed higher OCT4 expression than MP cells (Table 1). Notably, SP cells have been reported to be CD138 -/low in clinical MM samples [65]. Furthermore, CD138 low cells were detected in MM patients [56, 66] and a retrospective study correlated poor prognosis with increased CD138 low cells at diagnosis and relapse [67]. Based on these published results and our findings (Fig 1 and Table 1), we postulate that CD138 immunostaining is a useful way to

Hypoxia increases CD138 low cell population in HMCLs
Studies show that culturing HMCLs under short-term hypoxia (2-3 days) reduces CD138 mRNA and protein expression [68,69]. Also, long-term culturing under hypoxia (30 days) was reported to increase protein and mRNA expression of OCT4, NANOG and SOX2 [68]. Wen et al. [65] reported a 5-fold increase in SP cells (8226 and U266 cells) when cultured under hypoxia for 5 days that was suggested to arise from MP cells without any additional stimuli. To assess hypoxia-mediated alterations in the number CD138 low cells, we cultured HMCLs in 1% O 2 for 3 days. CD138 immunophenotyping showed that hypoxia significantly increased the %CD138 low cells compared to the normoxic cultures (Fig 2A) with a 2.2, 8.1, 5.0, and 1.6-fold increase for MM.1S, OPM-2, H929, and U266 cells, respectively (Fig 2B).
To verify hypoxia-mediated increases in stem-like cell population, we analyzed SDC1, OCT4, and NANOG expression using unsorted normoxic and hypoxic cells; VEGF-A is a wellestablished HIF-1 regulated gene and was included as an indicator of hypoxia. For OPM-2 cells, hypoxia significantly increased OCT4, NANOG while SDC1 expression decreased presumably due to increase in the stem-like cells (Fig 2C). For other HMCLs, a trend towards increased expression of OCT4 and/or NANOG and decreased expression of SDC1 was noted with hypoxia ( Fig 2C). Overall, Fig 2 shows that hypoxia increases CD138 low population and alters stem cell gene expression levels in MM cells. We have used low pO 2 levels to further characterize oxidative metabolism of stem-like MM cells.

MM stem-like cells maintain lower steady-state pro-oxidant levels
We used the H 2 DCF-DA probe to determine endogenous steady-state pro-oxidants levels in CD138 high and CD138 low cells in the HMCLs. In MM.1S and OPM-2 cell lines, CD138 low cells showed a significant lower H 2 DCF-DA probe oxidation (~25-30%) relative to the CD138 high cells (Fig 3A). A similar trend of CD138 low cells maintaining lower steady-state pro-oxidants compared to the bulk CD138 high was noted for H929 and U266 cell lines (Fig 3A). We used the more specific oxidation sensitive PO-1 boronate probe and detected lower H 2 O 2 levels (~40-50%) in CD138 low compared to CD138 high cells for MM.1S and OPM-2 cells (Fig 3B). Notably, in MM.1S and OPM-2 cells, the SP cells also showed reduced PO-1 oxidation compared to the MP cells (Fig 3C) substantiating the novel observation that MM stem-like cells maintain lower steady-state pro-oxidant levels compared to bulk MM cells.
Next, we used mouse bone marrow to obtain HPCs with~2.4 enrichment in the sca1 + c-Kit + cells (data not shown). By H 2 DCF-DA probe labeling, sca1 + c-Kit + cells showed approximately 50% lower probe oxidation relative to the sca1c-Kitcells ( Fig 3D). Overall, Fig 3  results show that MM stem-like cells and sca1 + c-Kit + HPCs maintain lower steady-state prooxidant levels compared to their differentiated counterparts (bulk MM or sca1c-Kitbone marrow cells, respectively). Therefore, a pharmacological approach that selectively induces oxidative stress-mediated killing of MM stem-like cells with minimal toxicity to the normal HPCs is needed to gain a therapeutic benefit in MM.

Poor MM prognosis associates with increased glycolytic gene expression
Clinically, relapse and dismal MM prognosis are associated with elevated glycolytic flux [70][71][72]. To extend this observation, we determined if increased expression of glycolytic genes correlates with clinical outcomes in MM. We queried the microarray dataset and associated clinical information from the TT2 study [73]. Gene signature associated with glycolysis (hexokinase II, aldolase A, triosephosphate isomerase I, glyceraldehyde-3-phosphate dehydrogenase, phosphoglycerate kinase I, pyruvate kinase M2, lactate dehydrogenase A) was used. A significant difference between the lower and upper quartile of glycolytic signature for both OS (Fig 4A) and EFS was seen (Fig 4B). The risk of death for patients with a high glycolytic gene expression was 2.9X that with a low gene expression (95% CI, 1.5-5.5). Additionally, the risk of an event for individuals in the high group was 1.9X that of the low group (95% CI, 1.2-3.0). Therefore, increased glycolytic gene expression appears to serve as an independent predictor of MM patient mortality and targeting glycolytic metabolism could potentially improve survival outcome in MM. To determine if CD138 low stem-like cells display increased glucose uptake, HMCLs were stained with 2-NBDG and CD138-APC and analyzed by flow cytometry. Compared to CD138 high cells, the CD138 low cells showed increased uptake of 2-NBDG by 1.4, 2.5, 1.7, and 1.8-fold in MM.1S, OPM-2, U266, and H929 cells, respectively ( Fig 4C). As increased Oxidative Stress-Mediated Clonogenic Killing of Multiple Myeloma Cells glycolysis is a cellular adaptation to hypoxia [74][75][76] and is also relevant to clinical MM, we next cultured HMCLs in hypoxia for 3 days. A hypoxia-induced further increase in 2-NBDG uptake of 1.1, 1.9 and 1.2-fold was noted in CD138 low cells in OPM-2, U266, and H929 cells, respectively (Fig 4D). For the CD138 high cells, a comparable 2-NBDG fluorescence was noted for hypoxia and normoxia cultures in OPM-2, H929, and U266 cell lines while glucose uptake decreased for MM.1S cells cultured under hypoxia indicating metabolism heterogeneity in different HMCLs (Fig 4D).
We next determined if 2-DG treatment increases steady-state levels of pro-oxidant in MM cells. For this, we measured oxidation of DHE probe as a surrogate for O 2 •− levels [31]. Consistent with the H 2 DCF-DA ( Fig 3A) and PO-1 results (Fig 3B), CD138 low cells showed lower endogenous DHE oxidation compared to CD138 high cells for MM.1S and OPM-2 cells ( Fig  4E), 2-DG treatment increased DHE probe oxidation in both CD138 high and CD138 low subpopulations ( Fig 4E).
To understand the mechanism of 2-DG mediated oxidative stress in MM cells, we performed qPCR analysis of ER stress responsive genes, BiP (binding immunoglobulin protein/ GRP78) and ATF4 (activating transcription factor 4). Since, 2-DG is chemically 2-deoxymannose and is able to inhibit N-linked protein glycosylation [36], mannose plus 2-DG control was included. 2-DG treatment statistically increased BiP and ATF4 expression in MM.1S and OPM-2 cells (Fig 4F). Notably, mannose treatment significantly reduced 2-DG mediated increase in mRNA levels of BiP and ATF4 in both MM cells ( Fig 4F). Next, we examined protein expression of BiP and ER stress markers CHOP (C/EBP homologous protein). Western blotting results were agreement with the qPCR analysis where 2-DG treatment increased BiP and CHOP expression and mannose treatment significantly attenuated ER stress in both MM.1S and OPM-2 cells (Fig 4G). Overall, Fig 4 shows that CD138 low cells may exhibit a preferential reliance on glycolytic metabolism, and 2-DG treatment increases ER stress as well as steady-state O 2 •− levels in MM cells.

CD138 low cells show increased orthodox mitochondria with increased pro-oxidant levels by 10-TPP treatment
To get an idea about the overall bioenergetics of MM stem-like cells, mitochondria ultrastructure was examined using electron microscopy (CMRF, UI). For both MM.1S and OPM-2 cell lines, sorted CD138 low cells showed comparable numbers of orthodox (uncondensed or energized) and condensed mitochondria ( Fig 5A). However, a significantly higher number of condensed (depolarized) mitochondria were noted for CD138 high MM.1S and OPM-2 cells ( Fig  5A). Therefore, the ratio of orthodox:condensed mitochondria was increased for CD138 low cells (0.82 ± 0.09 and 0.92 ± 0.16 for MM.1S and OPM-2 cells, respectively) relative to CD138 high (0.50 ± 0.08 and 0.35 ± 0.14 for MM.1S and OPM-2 cells, respectively). Using flow cytometry, we also noted that the CD138 low cells had higher mitochondrial mass in OPM-2, H929 and U266 cells while the MitoTracker staining was comparable between CD138 low and CD138 high MM.1S cells (Fig 5B). R123 staining showed that relative to CD138 high cells, the CD138 low cells of OPM-2, H929, and U266 exhibit increased Δψ m (Fig 5C) as also reported for other cancer stem cell types [40][41][42]. Overall, we noted that CD138 low cells have more abundant "energized/orthodox" mitochondria with higher mitochondrial mass while CD138 high cells show predominantly condensed "depolarized" mitochondria. Since we measured mitochondrial differences in the bulk and stem-like cells, we next used 10-TPP to target pro-oxidant producing pathways in MM. Treatment with 10-TPP increased pro-oxidant levels in both CD138 high and CD138 low MM.1S cells by 1.4 and 1.9-fold, respectively (Fig 5D). For the OPM-2 cells, 10-TPP treatment resulted in a significant increase in   Fig 5F). Confocal analysis revealed co-localization of the 10-TPVP (blue fluorescence) with the MitoTracker Red staining, suggesting that 10-TPP may effectively localize to mitochondria in MM cells. Treatment with FCCP (mitochondrial uncoupler that dissipates proton gradient) resulted in loss of 10-TPVP signal confirming accumulation of 10-TPVP in mitochondria was driven by Δψ m (data not shown).
We next performed clonogenic assays to examine if 2-DG and/or 10-TPP treatment reduced clonogenic fractions in MM.1S and OPM-2 cells. Treatment with 2-DG induced~20-30% killing of MM.1S and for OPM-2 cell line (Fig 6D). In the MM.1S cells, combination of For ultrastructure analysis, all mitochondria were selected (indicated by *), manually scored, and assigned either condensed or orthodox morphology. Low magnification shows the entire cell with inset used for analysis of mitochondria under higher magnification. *p < 0.01 vs. CD138 low cells. HMCLs were co-stained with APC-CD138 antibody and B. MitoTracker Green or C. Rhodamine 123 and analyzed by flow cytometry. Data is presented as the fold change relative to CD138 high cells. *p < 0.01 vs. CD138 high cells. D. Structure of 10-TPP; 10-TPP-induced H 2 DCF-DA oxidation in CD138 high and CD138 low cells in MM1.S and OPM-2 cells. Data of three independent runs is presented as the fold change relative to CD138 high cells. *p < 0.01 vs. CD138 high cells, # p < 0.01 vs. CD138 low cells. E. HMCLs were transduced with Ad-CMV or Ad-MnSOD, treated with 10-TPP, and MitoSOX oxidation was analyzed by flow cytometry. Data is presented as fold change normalized to control cells expressing Ad-CMV. Antimycin A treatment was used as a positive control. 10-TPP (0.1 μM) with 2-DG showed significantly increased killing compared to 2-DG alone (Fig 6D). The OPM-2 cells showed~10% and 80% killing with 10-TPP (0.2 and 1 μM, respectively) and combination with 2-DG further increased clonogenic killing compared to 10-TPP or 2-DG alone (Fig 6D). Overall, Fig 6 shows that combination of 2-DG plus 10-TPP provides a therapeutic window by showing selective cytotoxicity towards cancer and not normal cells.

Discussion
Clinical course of MM is fraught with emergence of refractory disease and high mortality [2,3]. A growing body of evidence suggests that a minor, therapy-resistant, clonogenic sub-population of tumor cells is most likely the principal roadblock to curing MM [4,8]. However, the identity of these stem-like MM cells remains only partially resolved [5,7]. Matsui et al. [57] identified a CD138 -ve memory B-cell like clonogenic fractions in HMCLs and clinical MM samples. Besides CD138 expression, functional markers such as SP staining and aldehyde dehydrogenase activity (Aldefluor assay), mRNA expression of induced pluripotent stem cells genes (i.e. OCT4, NANOG, SOX2), drug resistance and chromosomal instability genes (i.e. NEK2, BTK, RARα2), BCRP, and/or ALDH1A1 expression has been used to identify stem-like MM cells [7,8]. In the present study we have used two established markers of MM stem-like cells (CD138 immunophenotyping and SP staining) to characterize the oxidative metabolic properties of stem-like cells.
Based on CD138, OCT4, NANOG expression, we report that culturing cells under hypoxic conditions results in "increased MM stemness". We have focused on hypoxia as it is the primary regulator of stemness for normal adult stem cells and CSCs [78], constitutes a specific feature of MM bone marrow [79,80], and is being used as therapeutic target for MM [81]. Moreover, hypoxic-inducing factors (HIFs, HIF-1α and HIF-2α) are stabilized in MM patients with prognostic relevance [82]. Also, HIF-1 contributes to metabolic shift via upregulation of glucose transporters and glycolytic enzymes, and can aid in re-establishing redox homeostasis and cell survival under prolonged hypoxia [83]. However, besides hypoxia, CD138 down-regulation on MM cells within the bone marrow microenvironment can also occur via interaction with bone marrow stromal cells and interleukin 6 [61,84,85].
New approaches for targeting MM stem-like cells are under development like inhibitors of RARα2 or its downstream Wnt and Hedgehog signaling pathways and bruton tyrosine kinase (BTK) inhibitor [63,86]. We have previously utilized redox-based chemotherapeutics to elevate mitochondrial ROS levels and attain selective killing of MM cells [30,31]. We have also published on the role of antioxidant network (endogenous or driven by IL6−NFκB pathway in intrinsic or acquired drug resistance [31,32]. In this study we investigated the unique pharmacological targets, related to the oxidative metabolism, in MM stem-like cells such that oxidative stress-mediated cytotoxicity can be selectively induced in the drug-resistant tumor cells relative to normal stem cells. Recent studies have provided insight on the redox status of CSCs where low levels of intracellular ROS are reported in the CSC fractions of leukemia, breast and prostate cancer [87][88][89]. We detected low steady-state pro-oxidant levels in stem like MM cells (CD138 low vs. CD138 high cells and SP vs. MP cells). CSCs can maintain lower ROS levels via a preferential reliance on glycolytic metabolism [90][91][92][93][94][95]. We found that the CD138 low cells have increased 2-NBDG uptake relative to the bulk MM cells. We will extend these studies by performing real-time measurement of mitochondrial respiration and glycolysis (Seahorse Flux Analyzer) in MM stem-like cells enriched form cell lines and patient samples. Besides increased glycolysis, increased expression of antioxidant systems can also confer lower ROS levels in CSCs [88,96]. Ongoing studies in the lab are utilizing sorted stem-like cells to perform mRNA analysis of thiol regulatory proteins.
Bulk MM cells have been reported to overexpress glucose transporters and to undergo apoptosis upon pharmacological inhibition of glucose metabolism [16][17][18][19][20]. However, targeted inhibition of glycolytic metabolism in MM stem-like cells for therapy has remained unexplored. Microarray results with human MM biopsies suggest that increased expression of glycolytic genes may predict poorer overall survival outcomes. However, a tight correlation between gene transcription and pathway functionality may always not hold true [97]. Further studies in primary MM samples are required to show a direct correlation between glycolysis and overall survival outcomes.
Based on the observation that MM stem-like cells have lower steady-state ROS levels and increased 2-NBDG uptake, we selected 2-DG to inhibit the glycolytic pathway. We used a 2-DG concentration (20 mM) [98] that is in excess to the glucose concentration in the cell culture media (RPMI-1640, 11 mM glucose). In HMCLs, 2-DG treatment increased steady-state levels of O 2 •− in both stem-like and bulk tumor cells. Pharmacological manipulation of glucose metabolism (using dichloroacetate that inhibits pyruvate dehydrogenase kinase) has been shown to incur specific cytotoxicity on glioma stem cells and not normal neural stem cells [92]. Biochemically, 2-DG blocks pyruvate production by glycolysis and partially inhibits the pentose phosphate pathway, reducing NADPH regeneration by 50% and increasing cellular oxidative stress in cancer cells [99,100]. Also, 2-DG treatment induces ER stress via inhibiting N-linked protein glycosylation [36]. Notably, MM is a plasma cell neoplasm with extensive ER network required for antibody secretion [101] and undergoes unfolded protein response activation and cell death by prolonged ER stress [102,103]. Also, high ER function in MM has been shown to result in an increased passive leak of Ca 2+ into cytoplasm [104]. Constitutive Ca 2+ transfer from the ER to mitochondria is essential for maintenance of cellular bioenergetics [105]. Therefore, besides inhibiting oxidative metabolism, 2-DG treatment in MM may also disrupt ER−mitochondrial Ca 2+ homeostasis. This may result in a feed-forward cycle [106,107], enhancing oxidative stress-mediated killing of stem cells in combination with other mitochondrial redox-modulating chemotherapies.
In this study, we have used unsorted cells, stained for CD138 expression to differentiate between mature and stem-like MM cells, and directly compare glucose uptake, mitochondrial density, and mitochondrial membrane potential in these sub-populations. To better characterize differences in glycolytic rate and OXPHOS capacity, further studies are required to assess differences in glycolytic rates (lactate plus pyruvate formation) and OXPHOS capacity (oligomycin sensitive oxygen consumption) in stem-like vs. bulk MM cells. Also, it remains to be seen if CD138 low cells show dysregulated expression of PGC-1α (peroxisome proliferator-activated receptor γ coactivator 1) resulting in increased mitochondrial mass. Notably, altered PGC-1α expression has been linked to metabolic adaptation and survival of cancer stem cells [108,109].
Recent reports showing that normal stem cells and induced pluripotent stem cells retain young mitochondria with high metabolic capacity [54, 110,111]. Similarly, physiological aging may occur in cancer where CSCs show hyperpolarized Δψ m [24,44,45]. We report that the CD138 low stem-like cells have increased mitochondria numbers, Δψ m , and also higher percentages of mitochondria with orthodox configuration. It can be speculated that these mitochondrial properties of stem-like cells could be linked to increased metabolic demands for differentiation to mature MM cells that are more proliferative and highly secretory cells. Further studies using primary clinical samples may be useful to validate the changes in mitochondrial properties seen in stem-like cells of HMCLs.
Several mitochondria-targeted agents have developed by using the synthetic lipophilic TPP cation that can selectivity target cancer cells relative to normal cells [39]. Screening of a commercially available library of 46 different TPP-based compounds (specs.net) provided 10-TPP as one of the lead agents with high and selective cytotoxicity in melanoma cell lines (unpublished results). 10-TPP (C 28 H 36 BrP, 483.5) contains an aliphatic 10-carbon chain that is believed to uncouple mitochondria via embedding in the inner mitochondrial membrane [38]. A TPP derivative Mito-VES (vitamin E succinate) has shown selectivity towards breast cancer cells via mitochondrial localization and increasing ROS levels from ETC complex II [112]. Also, Mito-VES with a 11 carbon aliphatic chain length is most effective in mitochondrial ROS generation (compared to 5, 7 or 9 carbon linker) [43]. Also molecular modeling of Mito-VES indicated that TPP + anchors at the matrix face of the inner mitochondrial membrane while 11-C linker enables a close positioning of VES with the UbQ site of complex II [112]. Based on mitochondrial alterations in MM stem like cells (increase mitochondrial number, Δψ m , and orthodox:condensed mitochondria ratios) we hypothesized that 10-TPP treatment can potentially increase oxidative stress and cytotoxicity of MM stem-like cells. For mitochondrial imaging, we used an aggregation-induced emission active triphenylvinylpyridine (TPVP) agent with 10-carbon chain (10-TPVP, C 35 H 40 BrN, 554.6) [53,113]. Our results suggest that 10-TPP can potentially localize to mitochondria and increases cellular pro-oxidant levels as well as mitochondrial O 2 •− level in MM cells.
It is established that tumor cells often maintain redox homeostasis via multiple pathways and can adapt to chronic metabolic oxidative stress [114]. We therefore used a two-pronged approach that simultaneously targets oxidative (2-DG and 10-TPP) and glycolytic metabolism (2-DG) to reduce clonogenic fraction in MM cell lines presumably via metabolic oxidative stress-mediated cytotoxicity. To determine therapeutic efficacy of the proposed drug combinations, we used normal HPCs and noted that the DG+10-TPP treatment was minimally toxic towards these normal cells. Clinically, 2-DG is well-tolerated in humans and is being actively pursued as an adjuvant to cancer chemotherapy [33,115]. However, chronic ingestion of 2-DG in rats showed increased mortality with cardiac pathology [116]. Therefore, clinical development of 2-DG for MM therapy will require careful assessment of potential long-term normal tissue toxicity. Also, 10-TPP has a similar chemical structure as mitochondrial CoQ10 (MitoQ) that has been considered safe for long-term administration in human trials [117,118]. In preliminary studies, oral dosing of 10-TPP (100 μM for 17 days) caused no evidence of hematological toxicity in mice (data not shown). By in vivo limiting dilution assay, future studies will combine 2-DG (400 mg/kg intraperiponeally 5 days per week for 2 weeks)[119] with 10-TPP to determine if combination therapy can deplete MM clonogenic fractions and improve therapy outcome.
In conclusion, we show that altered oxidative metabolism can be used as a novel indicator of stem-like cells in MM. We report that MM stem-like cells may have a preferential reliance on glycolytic metabolism with increased mitochondrial numbers, Δψ m , and orthodox mitochondrial configuration. The proposed biochemical approach of combining 2-DG with 10-TPP is expected to provide a therapeutic window to selectively eliminate MM and drugresistant residual disease while sparing normal cells, leading to more effective MM therapy.  (8)