Introduction

Bone marrow mesenchymal stem cells (BM-MSCs) participate in niche design and support of hematopoietic cells [1]. They are multipotent, very plastic and extremely tuned to microenvironmental signals. This versatility is hijacked by malignant cells and exploited to the benefit of cancer progression [2]. The appropriation of BM-MSCs is highly evident in multiple myeloma (MM), where the malignant plasma cells accumulate in the BM and maintain intense interactions [3]. Others and we have demonstrated that MM niche BM-MSCs (MM-MSCs) promote MM clone expansion whereas BM-MSCs from normal donors (ND-MSCs) do not [2,3,4]. Despite these cardinal differences it is commonly accepted that MM-MSCs evolve from ND-MSCs exposed to malignant MM cells [3].

The co-existence of BM-MSCs and MM cells encompasses a variety of communication methods such as cell contact, soluble factors, secreted extracellular matrix and vesicles [2]. The robust crosstalk modulates multiple cellular functions including proliferation, migration and survival that all rely on cellular metabolic adaptation and bioenergetic availability [2]. Indeed, alterations in myeloma metabolism of glucose, amino acids and fatty acids are documented and integral to disease pathogenesis and progression [2, 5, 6]. The metabolic shift also allows MM cells adjustments to hypoxic surroundings, evasion of immune surveillance and drug resistance [7, 8]. Evidence that metabolic reprogramming is integral to the MM niche redesign and not the malignant cells alone is readily provided. For instance, increased MM-MSCs’ senescence compared to ND-MSCs that promotes inflammation [9, 10], augmented MM-MSCs osteoclastogenesis that stimulates disease progression and symptoms [11] and the increased risk of MM in people with metabolic syndrome [6, 12]. Metabolic adjustment also supports drug resistance, a major obstacle to myeloma cure [13]. These phenomena advocate for the significance of metabolic modifications in the MM tumor microenvironment (TME) [5, 6]. Despite this understanding and the vital role of BM-MSCs in MM progression, there is a paucity of data regarding the metabolic changes that occur in the MM-MSCs compared to ND-MSCs [12].

Cell metabolism refers to the biochemical processes used to generate energy (ATP) and maintain cells’ survival and growth. The mitochondria are the only source of respiration-mediated ATP production and key to amino acids synthesis and reducing equivalents [13, 14]. Mitochondria also support cell fitness and survival by participating in apoptosis and stress response [13]. The dynamic mitochondrial responses define the capacity of the cell to adapt to various internal and microenvironmental pressures. A delicate and ever changing balance of mitochondrial fusion and fission (dynamics) controls mitochondrial function and facilitates the acquisition of cellular homeostasis [15]. Generally, fusion allows the transfer of gene products for optimal function and fission is needed for cell division and quality control [15].

Indeed, malignant cells modify mitochondrial dynamics and function to better support them [14, 15]. Mitochondrial reprogramming also entails excess ROS production and a proclivity to stimulate quality control programs like mitochondrial unfolded protein response (UPRmt) or mitophagy [14]. MM cells also display decreased OXPHOS, increased dynamics, ROS scavengers and quality control [16]. Yet, in order to comprehend fully cancer metabolism and bioenergetics it is imperative to consider their niche context and take into account the mutualism in cancer cells metabolic dialogue with their surroundings, nicely embodied in the “Reverse Warburg effect” where the mitochondrial function is paramount [17].

In line with this understanding, Marlein et al. showed that MM cells exposed to their BM microenvironment (in vivo) and cultured with MM-MSCs (in vitro) produced more OXPHOS-generated ATP as indicated by oxygen consumption rate (OCR) [18]. They also exhibited higher spare respiratory capacity (SRC), which is the difference between the basal and maximal respiration rates and represents the cells’ reserve ATP production capability [18]. Further studies showed that BM-MSCs traffic mitochondria to MM cells and that mitochondrial trafficking from MM-MSCs promoted recipients’ ATP levels [18, 19]. Giallongo et al. advocated that MM-MSCs known to depend on glycolysis are more prone to transfer their mitochondria to MM cells thereby promoting the recipient cancer cells oxidative phosphorylation [19]. These observations support a function for oxidative respiration in MM cells and the relevance of the state and origin of the trafficked BM-MSCs’ mitochondria (ND-MSCs versus MM-MSCs). Despite this phenomenal progress and the recognition BM-MSCs mitochondrial function is most relevant to MM, to the best of our knowledge, there is no comprehensive comparison of MM-MSCs and ND-MSCs bioenergetics. Our study aimed to address this gap and provide an assessment of mitochondrial function in myeloma and healthy BM-MSCs with emphasis on regulation and potential clinical implications.

Results

MM-MSCs display increased expression of proteins associated with mitochondrial cell metabolism compared to ND-MSCs

In order to understand the differences between normal ND-MSCs and MM-MSCs we initially characterized their protein milieu (n = 3 each; 3332 proteins) using analytical mass spectrometry (proteomics). We determined the differential expression of almost 15% of the proteins in MM-MSCs versus ND-MSCs (unique and significantly increased/decreased in either source). Bioinformatics analysis of all differentially expressed proteins in MM-MSCs versus ND-MSCs (p < 0.05; n = 523) highlighted multiple biological pathways activated in the MM-MSCs (WebGestalt Pathways: KEGG, Panther, Reactome, Wikipathway and Wikipathway cancer). PCA demonstrated a clear segregation between MM-MSCs and ND-MSCs protein expression pattern (81.6%; Fig. 1A). Analysis of the top ten statistically significant pathways annotated in each category demonstrated a substantial enrichment in metabolism-related GO pathways in the MM-MSCs (35% versus 4%)(Fig. 1B, Table 1). A more stringent analysis of the elevated proteins in MM-MSCs (FC ≥ 1.25, p < 0.05; STRING, WebGestalt) demonstrated enrichment in metabolism and mitochondria-associated proteins (Fig. 1C, Table 1). Upregulated proteins in MM-MSCs (FC ≥ 1.25, p < 0.05) were enriched with cellular stress response proteins (STRING)(Table 1). Concordantly, we determined elevated expression of proteins implicated in metabolic stress coping mechanisms such as mitochondrial ribosomal proteins (MRPs)(FC ≥ 2, p < 0.05) and Peroxiredoxin proteins (PRDXs) (FC ≥ 1.5, p < 0.05)(Supplementary table 1) [20,21,22]. Analysis of the 63 proteins categorized under “Mitochondrion” cellular component in MM-MSCs versus ND-MSCs (Webgestalt, STRING) demonstrated an overall increase in the expression of most proteins and a 2.6 FC increase in their median expression (Fig. 1D). Under the “Mitochondrion” category, we determined augmented electron transport chain (ETC) and oxidative phosphorylation in MM-MSCs (Table 1) including increase in complex I components (n = 10)(Supplementary table 2). These results highlight mitochondria metabolic processes and adaptation to metabolic stress as overt features of MM-MSCs phenotype.

Fig. 1: Mass spectrometry analysis of MM-MSCs demonstrated increase in mitochondrial-related metabolic pathways and proteins.
figure 1

A Principal component analysis (PCA) graph presents clear segregation of MM-MSCs and ND-MSCs mass spectrometry data according to source. B The BM-MSCs (ND and MM-MSCs) mass spectrometry data (differentially expressed proteins, p < 0.05 (FDR), N = 3 each) was assessed for cellular component and pathways using WebGestalt. Pie graphs present the percentage of metabolism-related GO pathways (red) out of all cell pathways (Cyan) in BM-MSCs. C STRING functional enrichment analysis of BM-MSCs upregulated proteins (FC ≥ 1.25, p < 0.05) included in the “Mitochondrion” cellular component STRING category (p = 2.16E−17, red nodes). D Median values of 63 protein levels (mass spectrometry) included in the BM-MSCs “Mitochondrion” cellular component category (WebGestalt, GO:0005739, p = 0) presented individually by shape and color.

Full size image
Table 1 Comparative proteomics analysis of MM-MSCs versus ND-MSCs displays significant differences.
Full size table

MM-MSCs demonstrate modified bioenergetics compared to ND-MSCs

Intrigued by the highlighted mitochondria in MM-MSCs proteomics we chose to characterize the energy production processes in ND and MM-MSCs. Using the Seahorse platform we determined MM-MSCs produce higher levels of total ATP than ND-MSCs (↑80%; p < 0.05)(Real-Time ATP Rate Assay)(Fig. 2A). Yet, the excess ATP was primarily attributed to glycolysis (↑110%, p < 0.05)(Fig. 2A). These findings correspond with a previous publication [19]. Since glycolysis takes place in the cytosol outside of the mitochondria, we still deliberated on the distinguishing mitochondrial functions of MM and ND-MSCs as indicated by proteomics. Using the Cell Mito-Stress Test Kit we found that MM-MSCs display increased maximal respiration rate (↑74%; p < 0.0001) and spare respiratory capacity (SRC)(↑170%; p < 0.05) with no change in mitochondrial ATP production rate (NS) compared to ND-MSCs (Fig. 2B, C). Notably, we demonstrated that increased STAT3 in MM-MSCs implicated in SRC control (Supplementary Fig. S1). Increased SRC allows increased cellular versatility in response to bioenergetic demands and stress [23] and we wondered whether MM-MSCs indeed display adjustments to duress and erratic higher ATP demands.

Fig. 2: MM-MSCs display increased bioenergetics, mitochondrial stress adaptation and elevated fusion.
figure 2

ND (cyan) and MM (red) MSCs (n≧5) were subjected to (A) Assessment of cellular glycoATP, mitoATP and TotalATP production rates (Seahorse XF Real-Time ATP Rate Assay Kit). B Assessment of oxygen consumption rates (OCR) and determination of Basal respiration, Maximal respiration and SRC (Seahorse XF Cell Mito Stress Test Kit). C Representative graph of Seahorse XF Cell Mito Stress Test Kit. D Assessment of Proton leak (Seahorse XF Cell Mito Stress Test Kit). E Analysis of mitochondrial membrane potential (TMRE). F Quantification of UPRmt factors’ immunoblotting (ClpX and ClpP). G Representative Immunoblots of UPRmt factors (ClpX and ClpP) and Fusion/Fission regulators (OPA1, MFF and pDRP1S616). H Representative confocal images of BM-MSCs mitochondria (ND and MM, Mitotracker Red CMXros). I Analysis of mitochondria Branches and Total Branch Length (ND and MM confocal images, Imagej/FUJI Mitochondrial Analyzer tool). J Quantification of mitochondrial fusion regulator OPA1 levels (immunoblotting). K Quantification of mitochondrial fission regulators MFF and pDRP1S616 levels (immunoblotting). Immunoblots are cropped for visual ease. All immunoblotting results were normalized to tubulin and are expressed as percent (Mean ± SE, n ≥ 3) of protein levels in MM-MSCs compared to ND-MSCs. Asterisks depicted statistical significance *p < 0.05; **p < 0.01.

Full size image

MM-MSCs are adapted to increased bioenergetic stress

Supporting this conjecture, we registered an increased proton leak, which is an established indicator of mitochondrial damage, in MM-MSCs compared to ND-MSCs (↑110%; p < 0.05)(Mito-Stress Test)(Fig. 2D). We also determined a decrease in MM-MSCs’ ΔΨm compared to ND-MSCs (↓70%, p = 8.53E−04)(Fig. 2E)(TMRE). The concurrence of low ΔΨm and greater SRC is counterintuitive since the first is correlated with a compromised state and the latter is known to provide cells with added survival capacity, but has been described elsewhere [24]. Further evidence that the MM-MSCs are coping with stress was presented in increased levels of UPRmt major regulators in proteomics of MM-MSCs versus ND-MSCs (↑ClpX 3.7 FC, ↑ ClpP 2.5 FC; p < 0.05). These observations were further verified by immunoblotting of additional samples (ClpX ↑70% and ClpP ↑64%; p < 0.05)(Fig. 2F, G). Encouragingly, a recent report argues that activated UPRmt is correlated with increased expression of fibroblasts’ respiratory complex I [25] as we have witnessed (described above). In our proteomics analysis, we also observed increased expression of NADK2 in MM-MSCs compared to ND-MSCs (2.1FC, p = 0.06). NADK2 is the only mitochondrial enzyme responsible for countering mtROS by NADPH generation marking its unique significance [26]. Thus, we tested the mtROS levels in MM-MSCs versus ND-MSCs but determined no significant difference (MitoSOX). This may suggest that the activated stress response mechanisms effectively control the stress in MM-MSCs. Altogether, our observations so far indicate that the MM-MSCs are distinguished by activated stress coping mechanisms (UPRmt, NADK2, and SRC) that may allow the cells to persist despite their compromised conditions (proton leak, ΔΨm).

MM-MSCs display increased mitochondrial fusion compared to ND-MSCs

Acknowledging mitochondrial dynamics’ role in mitochondrial function we assayed ND and MM-MSCs fission/fusion equilibrium. Using confocal microscopy, we observed more fused mitochondria in MM-MSCs versus ND-MSC (Fig. 2H). Specific image analysis demonstrated 2-fold elevation in MM-MSCs Branches count and Total Branch Length (Mitochondria Analyzer tool)(Fig. 2I). Supporting our visuals we also determined in MM-MSCs increased expression of the critical fusion executer OPA1 (↑68%; p < 0.05)(Fig. 2J) and decreased fission regulators pDRP1s616 and MFF (↓37% and ↓45%; p < 0.05)(Fig. 2K). In summary, we determined increased fusion and decreased fission in MM-MSCs compared to ND-MSCs.

Mitochondrial fusion promotes MM-MSCs glycolysis-dependent ATP production

In order to understand the importance of mitochondrial dynamics to the BM-MSCs state, i.e. ND-MSCs or MM-MSCs we decided to interfere with fission-fusion equilibrium and study the effect on ATP production rates. We assayed ND and MM-MSCs treated with DRP1 inhibitor Mdivi-1 or OPA1 inhibitor MYLS22 (24 h, 50μM determined as effective and non-toxic for both, data not shown) with Mito-Stress Test. Importantly, the fission/fusion balance is such that by inhibiting fusion we actually increase the proportion of fissioned mitochondria and vice-versa [27, 28]. Our results (Table 2) demonstrated that inhibition of DRP1 increased total ATP production rates and SRC in MM-MSCs (↑35% and ↑50%, respectively; p < 0.05)(Fig. 3A, B). The increased total ATP production in Mdivi-1 treated MM-MSCs was attributed to increased glycolysis (↑48%, p < 0.01)(Fig. 3A). In contrast, OPA1 inhibition elevated total ATP production and SRC in ND-MSCs (↑25% and ↑40%, respectively; p < 0.05)(Fig. 3C, D). Again, the increased ATP production was attributed to increased glycolysis (↑50%, p < 0.05)(Fig. 3C). These results suggest that MM-MSCs are energetically more efficient when their mitochondrial equilibrium shifts towards fusion (consistent with their basic starting position) and that ND-MSCs are energetically more efficient with their dynamics tilted towards fission. The reciprocal treatment of MM-MSCs with MYLS22 and ND-MSC with Mdivi-1 did not display substantial bioenergetic effects except for decreased mitochondrial ATP without changes in total ATP levels (↓25%, p < 0.05 in both cell types) since most cellular energy is produced by glycolysis to begin with (>75%, p < 1e−06). Collectively, these observations delineate a role for mitochondrial dynamics in BM-MSCs metabolic programming with distinct differences between ND and MM-MSCs mitochondrial-dynamics/metabolism equilibrium. Interestingly and in contrast to the recognized association between fusion and OXPHOS, our findings portray a specific fusion-dependent regulation of glycolysis in MM-MSCs.

Fig. 3: Intervention with fission/fusion balance differentially affects BM-MSCs bioenergetics, mtROS and ΔΨm.
figure 3

ND (cyan) and MM (red) MSCs (n≧3) were treated with Mdivi-1 (mitochondrial fission inhibitor, 50 µM; 24 h), MYLS22 (mitochondrial fusion inhibitor, 50 µM; 24 h), M1 (mitochondrial fusion promoter, 5 µM; 24 h), Ascorbic Acid (AA) (ROS scavenger, 500 µM; 24 h) and analyzed for (A) MM-MSCs glycoATP, mitoATP and TotalATP production rates following Mdivi-1 treatment (Seahorse XF Real-Time ATP Rate Assay Kit) (B) MM-MSCs SRC following Mdivi-1 treatment (Seahorse XF Cell Mito Stress Test Kit). C ND-MSCs glycoATP, mitoATP and TotalATP production rates following MYLS22 treatment (Seahorse XF Real-Time ATP Rate Assay). D ND-MSCs SRC following MYLS22 treatment (Seahorse XF Cell Mito Stress Test Kit). E ND and MM-MSCs mtROS levels following MYLS22 treatment (mitoSOX red). F MM-MSCs mitochondrial membrane potential following MYLS22 treatment (TMRE). G MM-MSCs mitochondrial membrane potential following AA treatment (TMRE). H ND-MSCs mitochondrial membrane potential following M1 treatment (TMRE). I ND-MSCs mtROS levels following M1 treatment (mitoSOX red). J ND-MSCs proton leak following Mdivi-1/MYLS22 treatment (Seahorse XF Cell Mito Stress Test Kit). Results are expressed as percentage (Mean ± SE, n ≥ 3) and normalized to untreated cells represented by control bar (100%) or dashed line. Asterisks depicted statistical significance *p < 0.05; **p < 0.01.

Full size image
Table 2 Effects of fission/fusion modulators on BM-MSCs bioenergetics and fitness.
Full size table

Mitochondrial fusion promotes MM-MSCs robustness

The mitochondria are a major source of intracellular ROS and its production is closely associated with mitochondrial shape and structure regulated by dynamics [29]. We asked whether the higher fusion rate in MM-MSCs that had negligible contribution to mitochondrial energy production, might be more significant to the generation of mtROS (Table 2). Therefore, we inhibited fusion in the highly fused MM-MSCs (MYLS22) and assayed mtROS levels (MitoSOX dye). We determined that MYLS22 treated MM-MSCs displayed decreased mtROS levels (↓35%; p < 0.01) whereas ND-MSCs did not (Fig. 3E). This complies with decreased OXPHOS in MYLS22 treated MM-MSCs (↓24%; p < 0.05). Interestingly, the MYLS22-treated MM-MSCs also displayed decreased ΔΨm (↓47%, p < 0.01)(Fig. 3F), a phenomenon that may indicate that the MM-MSCs’ mtROS are integral to mitochondrial optimization. To test this possibility, we applied the ROS scavenger ascorbic acid (500μM) to MM-MSCs, which reduced mtROS as expected (↓50, p < 0.01). Furthermore, ascorbic acid attenuated ΔΨm in MM-MSCs (↓40%, p < 0.01) (Fig. 3G). We also treated MM-MSCs with fusion promoter M1 but this had no effect on the cells’ ΔΨm/ATP-production/mtROS. Perhaps the basal hyper-fused state of the MM-MSCs mitochondria is limiting their capacity to fuse more, but additional studies are required to explore this possibility. We further tested our conjecture regarding the fusion/mtROS/ΔΨm cascade in the opposite direction by inducing fusion (M1) in the less fused ND-MSCs and testing their response. M1 treatment elevated ND-MSCs ΔΨm (↑24%, p < 0.05) (Fig. 3H) but did not increase mtROS levels (Fig. 3I). These results indicate that the ND-MSCs mitochondrial fusion regulates ΔΨm independently of their mtROS levels in our experimental setup and suggests the association between fusion/mtROS/ΔΨm is exclusive to MM-MSCs. Finally, we assayed the proton leak in BM-MSCs (ND and MM) upon interference with mitochondrial dynamics as an additional all-inclusive indication of mitochondrial function. MYLS22/Mdivi-1 treated ND-MSCs displayed increased proton leak indicative of compromised mitochondrial function (↓50–75%; p < 0.01) (Fig. 3J) whereas M1/MYLS22/Mdivi-1 treated MM-MSCs were unaffected indicating the latter are more resistant to interference in mitochondrial dynamics in terms of OXPHOS efficiency.

Altogether, these results support the critical role of mitochondrial dynamics equilibrium in BM-MSCs, particularly the MM-MSCs. Mitochondrial fusion is essential for mtROS production in MM-MSCs and mtROS promotes mitochondrial integrity (↑ΔΨm). Ergo, the fusion/mtROS/ΔΨm axis has a positive role in MM-MSCs’ form. In fact, by integrating our observations so far we suggest that the bioenergetic wiring of MM-MSCs with higher basal hyper-fused state, increased mtROS and elevated SRC promotes their robustness.

Targeting SRC in MM-MSCs sensitizes co-cultured MM cells to anti-myeloma treatment

Our findings demonstrate that MM-MSCs display a different bioenergetic profile compared to ND-MSCs. We wanted to exploit this unique state to improve anti-MM treatment and specifically target the supportive MM-MSCs. The elevated SRC characteristic of the MM-MSCs afforded such an opportunity.

Initially we tested whether the decrease in MM-MSCs’ SRC compromised the MM-MSCs support of adjacent MM cells. For this we used Venetoclax recently reported to downregulate SRC [23]. Venetoclax was introduced into therapy as a BCL2 inhibitor, a characteristic of myeloma cells with t(11;14) [30, 31]. Since we wanted to exploit only its anti-SRC activity, we used MM cell lines that do not harbor t(11;14), showed low sensitivity to Venetoclax (i.e. RPMI-8226 and MM.1S) [32] and did not respond to 0.5μM Venetoclax administration used in our experiments (Fig. 4A), which is substantially lower than the published IC50 for its anti-BCL2 activity [32]. Moreover, the chosen cell lines RPMI-8226 and MM.1S (termed: MM cells) do not overexpresses BCL2 compared to other MM cell lines [33, 34]. We compared Venetoclax treatment of MM cells co-cultured with MM-MSCs (0.5μM, 72 h) to untreated MM cells cultured alone, treated with Venetoclax, co-cultured with ND-MSCs and co-cultured with MM-MSCs. Firstly, we determined that the relatively low dosage of 0.5μM Venetoclax was indeed sufficient to reduce MM-MSCs SRC (↓65%, p < 0.01, 24 h) with no effect on their viability. Next and as established, we observed that ND-MSCs did not affect co-cultured MM cells’ viability whereas the MM-MSCs increased it (Fig. 4A) (RPMI-8226: ↑25%; MM.1S: ↑62%, p < 0.05). Venetoclax had no effect on viability of MM cells cultured alone (Fig. 4A). Yet, when applied to MM cells co-cultured with MM-MSCs, Venetoclax mitigated their pro-MM effect as evidenced by a relative decrease in MM cells’ viability. MM cells co-cultured with MM-MSCs without the drug served as control group (RPMI-8226ː ↓18%; MM.1S: ↓22%, p < 0.05 respectively)(Fig. 4A). Notably, this decrease in MM-MSCs support of co-cultured MM cells’ due to Venetoclax effect was manifested differentially in RPMI-8226 and MM.1S cells’ count. RPMI-8226 demonstrated elevated cells death (↑77%, p < 0.05, data not shown) (antagonistic \({q}=0.78\)), whereas MM.1S presented reduced total cell counts (↓20%, p < 0.05) (antagonistic \({q}=0.10\)) compared to MM cells co-cultured on MM-MSCs without the drug.

Fig. 4: Venetoclax mitigates MM-MSCs support of adjacent MM cells’ survival and sensitizes them to anti-MM drug Bortezomib.
figure 4

ND and MM MSCs (n ≥ 3) were pre-treated with Venetoclax (Ven) (SRC inhibitor, 0.5 µM; 24 h) then co-cultured with RPMI-8226 or MM.1S MM cell lines (MM cells) in the presence of Ven for 48 h or added with Bortezomib (BTZ)(anti-MM drug, 5 nM; last) for the last 24 h of 72 h of co-culture. Untreated MM cells alone or in co-culture served as controls. A Ven’s effect on co-cultured (with ND/MM-MSCs, 72 h) MM cells’ Viability (Presto-Blue). B, C Ven ± BTZ effect on co-cultured (with MM-MSCs, 72 h) MM cells’ Viability (Presto-Blue) and total cell count (Trypan-Blue). Cell counts and viability were normalized to untreated MM cells cultured alone. Results are expressed as percentage (Mean ± SE, n ≥ 3). Asterisks depicted statistical significance *p < 0.05; **p < 0.005.

Full size image

Recent publications indicate that mitochondrial trafficking from MSCs to MM cells may affect recipient cells’ survival and involve physical interaction between the cells. Therefore, we assessed Venetoclax’s effects on mitochondrial trafficking by evaluating MitoTracker green signal in MM cells co-cultured (24 h) with MitoTracker pre-tagged MM-MSCs. Results demonstrated that Venetoclax did not alter the mitochondrial trafficking rate in our co-culture model. We also assessed cells’ adhesion by counting attached versus un-attached cells with and without Venetoclax and found no difference in adherence rates of MM-MSCs and MM cells. To conclude, these mechanisms seem irrelevant to Venetoclaxs’ mode of action in our experimental setup.

Having established that targeting MM-MSCs’ SRC with Venetoclax reduces their support of MM cells we asked whether Venetoclax could enhance the anti-MM activity of a key drug such as Bortezomib, particularly when co-cultured with MM-MSCs. Thus, we repeated the co-culture of MM cells/MM-MSCs/±Venetoclax but this time with Bortezomib supplementation (5 nM) for the last 24 h of the 72 h co-culture. As expected, results demonstrated that Bortezomib decreased MM cells’ viability and cell count (RPMI-8226: ↓35% and ↓36%; MM.1S: ↓50% and ↓31%, p < 0.05, respectively) and supplementation of Venetoclax had no cumulative effect on the Bortezomib treated MM cells cultured alone (Fig. 4B, C). Co-culture of MM cells on MM-MSCs afforded protection from Bortezomib emphasized by increased cell count and viability (RPMI-8226: ↑47% and ↑24%, p < 0.05; MM.1S: ↑49%, p < 0.05 and ↑60% p < 0.005, respectively) (Fig. 4B, C). All of these observations are well established. Newly, the combined supplementation of Venetoclax and Bortezomib to MM cells co-cultured on MM-MSCs decreased the MM cell line’s total cell count (RPMI-8226: ↓46%; MM.1S: ↓17%, p < 0.01) compared to Bortezomib only treated MM cells/MM-MSCs co-culture (Fig. 4C). This effect was attributed to reduced MM cells’ proliferation as witnessed in reduced live cell count (RPMI-8226: ↓20% and MM.1S: ↓26%, p < 0.05) and decrease in cell viability (RPMI-8226: ↓20% p = 0.1 and MM.1S: ↓73%, p < 0.05) with no increase in dead cells’ count ruling out prominent myeloma toxicity (data not shown). Taken together, our results show that by targeting MM-MSCs SRC we downregulate their support of adjacent MM cells and enhance the efficacy of Bortezomib.

Discussion

This study addressed the bioenergetic reprogramming of BM-MSCs in the myeloma TME and identified unique wiring that enabled specific therapeutic targeting. MM-MSCs display increased mitochondrial fusion responsible for augmented SRC, glycolytic ATP and resilience to stress. By targeting the MM-MSCs’ SRC with Venetoclax we attenuated their support of MM cells rendering the latter more sensitive to anti-MM treatment with Bortezomib.

MSCs bioenergetics are vital for their plasticity [35]. Ergo, MSCs’ bioenergetics/plasticity is integral to their adaptation to the malignant niche and cooperation with the neoplastic process [36]. In support of this acumen, we have determined that the MM-MSCs are bioenergetically distinguished from ND-MSCs and regulated by unique mitochondrial dynamics, suggesting the MM-MSCs bioenergetic setup is central to their acquired pro-MM identity.

We showed that MM-MSCs are under greater stress than ND-MSCs, and have activated various coping mechanisms. The MM-MSCs display increased proton leak and decreased ΔΨm. Mitochondria proton leak reflects the uncoupling of oxygen consumption and OXPHOS due to dissociation between ΔΨm generation and its use for mitochondria-dependent ATP synthesis. Indeed, proton leak is higher in dysfunctional mitochondria as evidenced in multiple cellular models [37]. Convolutely, mtROS (produced by OXPHOS) may promote damage-induced proton leak and proton leak may limit ROS production by attenuating OXPHOS. Furthermore, enhanced oxygen consumption can compensate for the proton leak to maintain ATP production levels [37]. We observed unchanged mtROS levels and increased oxygen consumption in MM-MSCs along with the increased proton leak. Furthermore, we have demonstrated that the mtROS is instrumental to upholding the ΔΨm in the MM-MSCs. As a conservative conclusion we may deduce the increased MM-MSCs’ proton leak indicates that their mitochondria are less efficient in oxygen-dependent ATP production than ND-MSCs but is compensated by increased OCR and additional ATP produced by glycolysis. MM-MSCs mtROS kept at bay by response mechanisms is vital to the cells’ ΔΨm and suggests ROS is needed for cell survival [38]. Indeed, upholding ΔΨm is instrumental to efficient ion and protein transport and escape from elimination as part of quality control [39].

Cells managing mitochondrial stress employ various coping mechanisms. We determined activated UPRmt, which is a transcriptional stress response [40, 41]. Increased SRC is another mitochondrial stress-coping mechanism that allows cells greater flexibility in their energetic response to stimuli and demands [23]. SRC is reserve energy the cell can access upon demand to maintain function under dynamic conditions such as stress or synthetic burdens [23]. We have identified prominently higher SRC in MM-MSCs compared to ND-MSCs. The high SRC in the MM-MSCs may accommodate metabolic and biosynthetic demands and provide a means to achieve greater fitness particularly under the extreme condition of the MM niche such as oxygen and nutrient shortages [9, 42]. The interesting conjunction of low ΔΨm and greater SRC was previously described in long-lived memory CD8 + T cells [24]. Altogether, we propose that the MM-MSCs coping with increased stress in the TME have converted into a more resilient BM-MSCs population [43,44,45].

To the best of our knowledge, the identification of the MM-MSCs bioenergetic adaptation to stress is novel and opens the field for new therapeutic approaches. We applied this insight to a proof-of-concept co-culture experiment with Venetoclax. Venetoclax, a known selective BCL-2 antagonist, was recently indicated it to affect electron transport chain activity and decrease SRC as well [23, 46, 47]. By repurposing Venetoclax and targeting SRC, we compromised MM-MSCs support of co-cultured MM cells and sensitized MM cells without BCL-2 amplification to Bortezomib treatment in the presence of MM-MSCs. This combined application demonstrates the clinical relevance of Venetoclax to myeloma treatment across the board and underscores the dual function of the drug as anti SRC in the BM niche as well as anti BCL-2 agent in MM cells. The crosstalk of the malignant cells with their TME (immediate or systemic) is a well-recognized cause of evolving drug resistance and the target of much research [48, 49]. Our strategy not only reassigns existing drugs but also selectively targets the MM-MSCs’ general vitality thereby stemming their capacity to support MM cells regardless of the support mechanism in use.

MSCs predominantly rely on glycolysis for ATP production, making changes in glycolytic rates more significant for total cellular ATP levels than alterations in OXPHOS rates. Concordantly, we observed that the higher ATP levels in MM-MSCs compared to ND-MSCs originated from glycolysis. We also established a direct connection between mitochondrial dynamics in BM-MSCs and ATP generation through glycolysis. Promoting mitochondrial fission increased glycolytic ATP production in ND-MSCs, consistent with the idea that fused mitochondria support OXPHOS better, while increased fission reduces available fused mitochondria, thus promoting glycolysis [15]. In contrast, we have also determined that MM-MSCs produce more glycolysis-ATP when fusion is stimulated, which is not readily explained by existing views. Nevertheless, reports presented that OPA1 associates between fusion and glycolysis in transplanted MSCs and that leptin promotes MSCs glycolysis, drug resistance and survival by OPA1-mediated fusion [50]. Altogether, published data and our observations support the existence of a regulatory signaling pathway in MSCs that connects mitochondrial fusion machinery (OPA1) to the MSCs’ glycolytic pathway in atypical microenvironments such as MM TME.

Another report showed that OPA1-mediated fusion prevents mitochondrial release of pro-apoptotic proteins suggesting that MSCs’ mitochondria hyperfusion is primarily essential for restraining apoptosis and not energy production [51]. We propose that increased MM-MSCs fusion facilitates both cell survival under stress as well as increased glycolysis and ATP production. Interestingly, both MM-MSCs and ND-MSCs produce most of their energy by glycolysis but with different mitochondrial dynamics. It is tempting to suggest that the increased MM-MSCs mitochondrial fusion supports their survival under duress due to their compromised state whereas the ND-MSCs being healthier do not depend on fusion as much. Since the mitochondria in both MM and ND-MSCs are primarily elongated, we may consider the increased fusion state in MM-MSCs as hyper-fusion. Interestingly, hyper-fusion corresponds with differentiation and loss of pluripotency characteristic of MM-MSCs [52].

The phenotypical plasticity of cancer cells is a recognized trait necessary for tumor development, progression, spread and drug resistance [53, 54]. It is also a trait of stem cells including BM-MSCs [55]. The role of bioenergetics in phenotypical plasticity is only now emerging [35] and our observations support its involvement in the myeloma niche. We have shown that MM-MSCs acquire unique bioenergetics features that allow them to survive under stressful conditions and support the MM cells. We further demonstrated that these unique traits allow selective and effective targeting of this support and afford a means to promote the therapeutic efficacy of current anti-MM drugs. Future studies should endorse the development of new therapies that exploit the MM niche bioenergetics and the TME generally.

Materials and methods

Cell culture

BM samples obtained from the femur head of consecutive normal donors, undergoing elective full hip replacement surgery and newly diagnosed MM patients’ BM aspirates taken for medical purposes at Meir Medical Center, Israel. BM-MSCs were isolated and propagated as previously described [4]. Newly purchased MM cell line RPMI-8226 (ATCC, Manassas, VA, USA) and BM-MSCs (1-2 passage) were cultured as previously [4] and routinely screened for mycoplasma contamination. Co-cultures included BM-MSCs pre-seeded in 96 wells plate (104 cells, overnight) overlaid with RPMI-8226 or MM.1S (2 × 104 cells) for 72 h with various treatments. RPMI-8226 and MM1.S cultured alone served as control.

Fluorescent microscopy

BM-MSCs cultured on glass slides, labeled (MitoTracker Red CMXRos, Invitrogen, Waltham, MA, USA) and imaged by SP8 inverted confocal microscopy equipped with a Leica HC PL APO CS2 × 63/1.4 NA objective (Leica Microsystems, Wetzlar, Germany). BM-MSCs mitochondrial Branch length and number were assayed with ImageJ/FIJI Mitochondria analyzer tool [56].

Proteomics

BM-MSCs (N = 5 each) were assayed for protein content by analytic mass spectrometry (LC-MS/MS; Q Exective HF mass spectrometer, Thermo Fisher Scientific, Waltham, MA, USA) at the Smoler Protein Research Center (Technion, ISRAEL).

Bioinformatics

Mass spectrometry Log LFQ intensities (Perseus software) were statistically analyzed for Principal Components Analysis (PCA) and differentially expressed proteins (ND-MSCs vs MM-MSCs; FC ≥ 1.25, p < 0.05) using Partek Genomics suite (v 6.6; http://www.partek.com/pgs). Protein lists were analyzed for enriched gene ontology and pathways using three different bioinformatics tools (Webgestalt, ToppGene and STRING)[20,21,22]. Upregulated proteins (FC ≥ 1.25, p < 0.05) in MM-MSCs were investigated for enriched metabolic pathways and biological processes.

Bioenergetics

Real-time ATP production and oxygen consumption rates (OCR) measured using Seahorse XF Real-Time ATP Rate and Cell Mito Stress Test Kits, respectively according to manufacturer’s instructions (Seahorse XFe96 Analyzer, Agilent Technologies, Santa Clara, CA, USA). Briefly, BM-MSCs seeded in 96-well Seahorse assay plates (104 cells/well, overnight) then incubated for additional 24 h with different treatments and analyzed. Readings normalized by cell count and analyzed by XFe wave software.

Mitochondrial membrane potential (ΔΨm)

Evaluation of ΔΨm performed using TMRE Mitochondrial Membrane Potential Assay Kit (Cayman Chemical, Ann Arbor, MI, USA) according to manufacturer’s instructions. Briefly, cells were pre-seeded in black-wall, 96-well culture plates (104 cells/well, overnight), subjected to treatment (24 h), washed (PBS), suspended with 25 nM TMRE (30 min, 37 °C, dark) and fluorescence was captured (Infinite-200 microplate-reader, iControl software; TECAN, Männedorf, Switzerland). FCCP treatment (10 μM) served as control for ΔΨm reduction.

Mitochondrial ROS (mtROS)

The mtROS levels were quantified by measuring the fluorescence of MitoSOX™ Red according to the manufacturer’s instructions (Invitrogen). Briefly, BM-MSCs pre-seeded on black-wall 96-well culture plates (104 cells/well, overnight, incubated with different treatments (24 h), stained with 750 nM MitoSOX™ Red (30 min, 37 °C, dark), washed (PBS) and assayed for fluorescence as described above. Rotenone treatment (0.25 μM) served as the control for mtROS presence [57].

Immunoblotting

Proteins lysates were immunoblotted as described previously [16,17,18] using rabbit/mouse anti-human: MFF (#86668), DRP1 (#8570), phospho-DRP1S616 (#4494), OPA1 (#80471), ClpP (#14181), STAT3 (#9139), phospho-STAT3Y705 (#9145); β-Actin (#8457) (CST, Danvers, MA, USA); phospho-STAT3S727 (ab32143), ClpX (ab168338) (Abcam, Waltham, MA, USA); α-Tubulin (T5168) (Sigma, Burlington, MA, USA).

Cell adhesion

RPMI-8226 MM cells (3 × 104 cells/well) were co-cultured on MM-MSCs (1.5 × 104 cells/well) with/without Venetoclax (0.5 μM). As control, RPMI-8226 MM cells cultured alone. After 24 h, un-attached cells collected following gentle washing with PBS leaving the attached cells in the wells. The collected un-attached cells counted (Advia) and attached cells visualized and counted (microscope). Adhesion ratios were determined as: 100% - [(un-attached cells/total cells)*100)]. To exclude the proliferative effect on cell populations, all cultured wells were validated for no change in total cell numbers.

Mitochondrial trafficking

MM-MSCs (1.5 × 104 cells/well) pre-tagged with MitoTracker Green FM Dye (MTG) (Invitrogen, Waltham, MA, USA) according to manufacture instructions then left for 3 days of dye leakage. Next, RPMI-8226 MM cells (3 × 104 cells/well) co-cultured on MM-MSCs (1.5 × 104 cells/well) with/without Venetoclax (0.5 μM). As controls, RPMI-8226 MM cells were cultured alone or co-cultured with unstained MM-MSCs. After 24 h, RPMI-8226 MM cells were collected and analyzed for MTG positivity (488 nm laser, 530/30 nm filter) using flow cytometry (Navios EX, Beckman Coulter, Indianapolis, IN, USA). MTG ∆MFI was determined as: Venetoclacx treated co-cultured cells/untreated co-cultured cells.

Cell count and viability

RPMI-8226 and MM.1S MM cells were counted as previously [4] and their viability tested (PrestoBlue Cell Viability Reagent, Thermo Fisher scientific) according to manufacturer’s instructions.

Inhibitors and drugs

Mdivi-1, mitochondrial fission inhibitor (475856, 50 μM, Sigma) [58]; MYLS22, mitochondrial fusion inhibitor (HY-136446, 50 μM, MedChemExpress, Princeton, NJ, USA) [59]; M1, mitochondrial fusion promoter (SML0629, 5 μM, Sigma) [60]; Ascorbic acid, mtROS scavenger (A4544, 500 μM, Sigma) [61, 62]; Venetoclax, SRC inhibitor (HY-15531, 0.5 μM, MedChemExpress) [63] and Bortezomib (BTZ) (C070410-00, 5 nM, Medomie Pharma, Petah Tikva, IL) [64]. Mdivi-1, MYLS22, M1 and Venetoclax were dissolved in DMSO (<0.1% v/v) and did not show any effect on treated cells compared to media.

Statistical analysis

All experiments were conducted at least three separate times. Student’s paired t-tests were applied in analyses of differences between two cohorts and multivariate One-Way-ANOVA test for more. An effect was considered significant when P ≤ 0.05. In drug combination assays, an antagonistic effect was determined by the drugs’ interaction formula: \({\rm{q}}={\rm{P}}({\rm{A}}+{\rm{B}})/({\rm{P}}\left({\rm{A}}\right)+{\rm{P}}\left({\rm{B}}\right)-{\rm{P}}\left({\rm{A}}\right)\times {\rm{P}}\left({\rm{B}}\right))\) \(({\rm{q}}\, < \,0.85\), antagonist; \({\rm{q}}\, > \,1.15\), synergist; \(0.85\, < \,{\rm{q}}\, < \,1.15\), additive) [65].