Multiple myeloma cells’ capacity to decompose H₂O₂ determines lenalidomide sensitivity

Multiple myeloma (MM) is a plasma cell neoplasm that produces and secretes monoclonal immunoglobulins.¹  Immunoglobulin synthesis is associated with the formation of disulfide bonds, a reaction catalyzed by the Ero1 flavoprotein in the endoplasmic reticulum (ER), that consumes oxygen and generates H₂O₂.^2,3  Excessive H₂O₂ is deleterious to cells because it activates the oxidative stress response, leading to apoptosis.⁴  Although a cellular antioxidant defense system exists to combat increasing concentrations of reactive oxygen species, such as H₂O₂, this system has a finite capacity, and antioxidant defenses can be overwhelmed by increased production of intracellular H₂O₂.^5,6 

Lenalidomide, a thalidomide analog, is an immunomodulatory drug (IMiD) widely used to treat MM, often in combination with other agents.^7-9  Although cereblon (CRBN) is required for IMiD anti-MM cytotoxicity,¹⁰  the exact mechanism of action of these drugs remains elusive. Recent work has shown that lenalidomide binds to CRBN and promotes degradation of B-cell transcription factors IKZF1 and IKZF3 via ubiquitination and proteasomal degradation, thereby inducing cell death.^11,12  Another recent study identified a ubiquitin-independent physiological chaperone-like function of CRBN that promotes maturation of basigin (BSG; also known as CD147) and solute carrier family 16 member 1 (SLC16A1; also known as MCT1) proteins. IMiDs outcompete CRBN for binding to CD147 and MCT1, leading to destabilization of the CD147-MCT1 complex and mediating cytotoxicity in MM.¹³  However, these studies still fail to explain why MM can be resistant to lenalidomide despite expression of CRBN.^11-13  Even though previous reports indicate that thalidomide teratogenicity is due to oxidative stress^14,15  and that thalidomide analogs induce oxidative killing of leukemic cells,¹⁶  the complete molecular and biochemical mechanisms responsible for lenalidomide-mediated anti-MM activity remain unknown. Therefore, we aimed to analyze whether the IMiD, particularly lenalidomide, given its extensive clinical application, interferes with the antioxidative capacity of MM cells to exert its anti-MM activity.

All human MM cell lines (HMCLs) were maintained in RPMI-1640, supplemented with 10% fetal bovine serum, 100 U/mL penicillin, 100 μg/mL streptomycin, and 2 mM glutamine. All HMCLs were grown at 37°C in a 5% CO₂ incubator. Cell viability was assessed by 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) colorimetric assay.

Whole-cell lysates were prepared from cell pellets using cell lysis buffer (Cell Signaling Technology). Equal amounts of protein extracts were resolved by sodium dodecyl sulfate–polyacrylamide gel electrophoresis. Most gels were run under reducing conditions (by adding dithiothreitol [DTT]), but analysis of immunoglobulin light chain (IgL) dimers required nonreducing conditions (without DTT). Other procedures and antibodies used for this study are listed in supplemental Methods, available on the Blood Web site.

HMCL cells (10⁶ cells/mL in phosphate-buffered saline [PBS]) were or were not treated with H₂O₂ (100 µM) and immediately analyzed for autofluorescence of flavin adenine dinucleotide (FAD; fluorescein isothiocyanate [FITC]−A channel) and reduced nicotinamide adenine dinucleotide and nicotinamide adenine dinucleotide phosphate [NAD(P)H; UV Blue–A channel] with multicolor flow cytometry (LSRFortessa; BD Biosciences). Histogram normalization (FlowJo flow cytometry analysis software v.7.6.5; Tree Star, Inc.) was used to overlay untreated vs treated samples. CD138⁺ patient samples were treated with or without H₂O₂ (100 µM or control) for 30 minutes. Fluorescence readings were obtained for FAD at 450/535- and NAD(P)H at 350/460-nm excitation and emission wavelengths, respectively. All individuals provided written informed consent for the collection and use of samples for research purposes under Institutional Review Board approval and according to the Declaration of Helsinki.

CRBN knockdown cells were the same as previously published.¹⁰  For Bim and IgL-λ knockdown, lentiviral constructs expressing nontargeting (shCtrl) and Bim or IgL-λ short hairpin RNA (shRNAs; Sigma-Aldrich) were used.

Human CRBN complementary DNA was purchased from Thermo Scientific and subcloned into a lentiviral expression vector, pCDH-CMV-MCS-EF1-copGFP (System Bioscience). Lentivirus harboring control vector and CRBN complementary DNA constructs were prepared and used to infect the MM cell line OCIMY5. Infection efficiency was measured by FACScan analysis of GFP expression 3 days after infection. The cells were sorted for GFP expression 14 days after infection. CRBN overexpression was confirmed by immunoblotting.

Lentiviral constructs expressing CRISPR-associated protein 9 and guide RNAs were originally generated from Feng Zhang’s laboratory^17,18  and were obtained from Addgene (Cambridge, MA; detailed in supplemental Methods).

MM cells with similar levels of CRBN expression, but variable IKZF1 and IKZF3 status, can have differential sensitivity to lenalidomide,¹⁹  suggesting other mechanisms for IMiD-mediated cytotoxicity (Figure 1A-B). We hypothesized that a differential capacity to dispose of H₂O₂ might contribute to cellular sensitivity to IMiDs, and we therefore analyzed cellular antioxidative capacity as a predictor of lenalidomide sensitivity.

We measured the capacity of MM cells to decompose H₂O₂ to oxygen and water by using a simple biochemical test that qualitatively measures the amount of oxygen released in vitro after H₂O₂ exposure. We tested several HMCLs that express wild-type CRBN and with different sensitivity to lenalidomide, including the resistant cells RPMI-8226 and JJN3, lenalidomide-hypersensitive MM.1S, and moderately sensitive KMS11. An equal number of cells were treated with an equal amount of H₂O₂, and the release of oxygen was observed and subjectively quantified. This analysis showed that lenalidomide-resistant MM cells had the highest capacity to decompose H₂O₂ compared with lenalidomide-sensitive cells (Figure 1C; supplemental Figure 1A). This differential H₂O₂ decomposition capacity appears to correspond well with external H₂O₂-mediated cytotoxicity; MM.1S was hypersensitive, and RPMI-8226 was resistant to H₂O₂ (supplemental Figure 1B-C).

To develop a quantitative assay of cellular antioxidative capacity, we applied a new strategy that measured total cellular oxidation of reduced flavin adenine dinucleotide (FADH₂) and NAD(P)H after H₂O₂ treatment. Cells with high antioxidative capacity generate more oxidized FAD and NAD(P) after H₂O₂ treatment, and cells with lower antioxidative capacity (ie, those already in a highly oxidized state) show less oxidation of FADH₂ and NAD(P)H with H₂O₂ treatment (Figure 1D; supplemental Figure 1D-E). The autofluorescent properties of oxidized FAD and reduced NAD(P)H allowed us to determine whether H₂O₂ treatment increased oxidized FAD and NAD(P) through increased and decreased autofluorescence, respectively. Cells with greater antioxidative capacity, and therefore resistant to lenalidomide (RPMI-8226 and JJN3), exhibited greater increases in FAD autofluorescence and greater decreases in NAD(P)H autofluorescence than lenalidomide-sensitive cells (MM.1S and KMS11) after treatment with 100 μM H₂O₂ (Figure 1E-F). These results are consistent with the hypothesis that antioxidative capacity determines lenalidomide sensitivity in HMCLs, despite similar CRBN protein expression levels (Figure 1A-F; supplemental Figure 1D-E).

Next, we were interested in determining whether our findings could be clinically useful in predicting which patients are sensitive to lenalidomide. Using a small number of primary patient samples, we found that the antioxidative capacity correlates with lenalidomide sensitivity (Figure 1G; supplemental Figure 1F).

As MM cells’ capacity to decompose H₂O₂ was correlated with lenalidomide sensitivity, we tested whether lenalidomide could inhibit H₂O₂ decomposition in HMCLs. Cells were pretreated with the highly specific peroxidase substrate Amplex Red, followed by exposure to increasing concentrations of lenalidomide or vehicle control. Lenalidomide-exposed MM.1S cells showed decreased resorufin (oxidized fluorescent product), indicating that lenalidomide inhibited intracellular H₂O₂ decomposition (Figure 2A). Because lenalidomide was able to inhibit H₂O₂ decomposition, we were interested in evaluating whether other IMiDs (ie, thalidomide and pomalidomide) also inhibit intracellular H₂O₂ decomposition in different HMCLs (Figure 2B). To confirm IMiD-mediated inhibition of H₂O₂ decomposition, we treated different HMCLs with external H₂O₂ and observed (1) increased basal resorufin fluorescence and (2) IMiD-mediated inhibition of external H₂O₂ decomposition by intracellular peroxidases (supplemental Figure 2A). We also showed that all IMiDs analyzed had the same mechanism of action because in vitro decomposition of H₂O₂ by horseradish peroxidase (HRP) (in vitro HRP assay) was similarly inhibited by thalidomide, lenalidomide, and pomalidomide (Figure 2C). Lenalidomide-mediated inhibition of HRP was not concentration dependent (supplemental Figure 2B) and inhibited H₂O₂ decomposition in clinically achievable concentrations. Pomalidomide was the most potent inhibitor, consistent with its perceived clinical potency²⁰  (Figure 2B-C; supplemental Figure 2A). Moreover, we also confirmed that lenalidomide inhibited intracellular H₂O₂ decomposition in samples from MM patients (CD138⁺) at a clinically achievable concentration (Figure 2D).

After showing that lenalidomide inhibited intracellular H₂O₂ decomposition, we sought a biological marker for lenalidomide-mediated, elevated intracellular H₂O₂ in MM. Increased intermolecular disulfide bond formation (protein disulfide dimers) is a consequence and hallmark of elevated intracellular H₂O₂ due to cysteine oxidation.^21,22  Using MM.1S, we observed increased formation of IgL-λ dimers after 3 hours of treatment with lenalidomide or H₂O₂ (supplemental Figure 2C). Because the degree of dimerization was somewhat low after 3 hours of lenalidomide, we subsequently prolonged lenalidomide exposure to enhance dimerization. At the same time, we analyzed the requirement of CRBN for lenalidomide-mediated elevation of intracellular H₂O₂ to induce IgL-λ dimers. We repeated this analysis with lenalidomide-resistant MM.1Sres (this cell line was generated by culturing MM.1S in gradually increasing concentrations of lenalidomide).²³  Our laboratory previously reported that CRBN expression in MM.1Sres is diminished as lenalidomide resistance increases.¹⁰  We treated MM.1S and MM.1Sres cells with lenalidomide for 3 days and observed IgL-λ dimers only in MM.1S (Figure 2E), the lenalidomide-sensitive cell line (Figure 2F). By using other isogenic cell lines, positive and negative for CRBN expression, we showed that lenalidomide treatment caused accumulation of IgL-λ dimers only in CRBN-positive cells (Figure 2G-H; supplemental Figure 2D), further confirming that CRBN-bound lenalidomide is required to inhibit intracellular H₂O₂ decomposition and that the H₂O₂ had to be high to induce IgL dimerization (Figure 2E,G,H; supplemental Figure 2C-D). Dimerization was associated with sensitivity to lenalidomide (Figure 2F,I-J; supplemental Figure 2E). This dimerization was completely eliminated after electrophoresis with DTT (supplemental Figure 2 F), suggesting that lenalidomide acts as a static drug as it has almost the same biological effect (IgL dimerization) at very low and high concentrations (supplemental Figure 2G-H).

Lenalidomide-bound CRBN has been shown to induce proteasomal degradation of IKZF1 and IKZF3.^11,12,24  We hypothesized that lenalidomide-induced IKZF1 and IKZF3 degradation was mediated via H₂O₂-induced protein oxidation cascade. We found that both lenalidomide and H₂O₂ treatment induced IKZF1 protein dimerization (supplemental Figure 3A-D). To confirm that elevated intracellular H₂O₂ caused IKZF1 and IKZF3 degradation, we treated MM.1S cells with H₂O₂ or other drugs also reported to induce oxidative stress (etoposide, doxorubicin, and dexamethasone). Only H₂O₂ and lenalidomide degraded IKZF1 and IKZF3 (degradation evident within 3 hours of treatment; Figure 3A). Moreover, H₂O₂ and lenalidomide-induced IKZF1 and IKZF3 degradation was rescued by proteasomal inhibition with bortezomib (supplemental Figure 3E-G). This finding further supports our hypothesis that lenalidomide and exogenous H₂O₂ degrade IKZF1 and IKZF3 by a common mechanism.

To confirm that external H₂O₂-induced IKZF1 degradation was a rapid process at the protein level, we used an adenoviral vector expressing IKZF1 fused to luciferase (AdIKLuc) or control vector (AdLuc)²⁵  to infect HMCLs. Cells were treated with increasing concentrations of H₂O₂ or lenalidomide for 1 hour, and luciferase activity was measured. We observed a marked decrease only in IKZF1-conjugated luciferase (AdIKLuc) activity within 1 hour of H₂O₂ treatment without affecting control luciferase activity (AdLuc) in MM.1S, MM.1R (resistant to dexamethasone), OPM2, and H929 (Figure 3B), as well as in other HMCLs (supplemental Figure 3H).

To confirm the central role of CRBN in IKZF1 and IKZF3 degradation by the H₂O₂-induced oxidative cascade, we next examined OPM2 isogeneic cells with CRBN wild-type and CRBN-knockdown as well as CRBN-overexpressing OCIMY5 (transfected with wild-type CRBN). We treated OPM2-NT (nontarget shRNA control) and OPM2-shCRBN with lenalidomide and 2 concentrations of H₂O₂ (25 and 50 μM) for 3 hours, and we showed that H₂O₂ mediates IKZF1 and IKZF3 degradation in a CRBN-dependent fashion (Figure 3C). Next, we tested the viability of OPM2-NT and OPM2-shCRBN with increasing concentrations of lenalidomide and H₂O₂ for 3 days. Lenalidomide-induced cytotoxicity was CRBN-dependent, but H₂O₂ was not (Figure 3D-E). OCIMY5 cells overexpressing CRBN also showed enhanced IKZF1 and IKZF3 degradation within 3 hours of lenalidomide or H₂O₂ (50 µM) treatment (Figure 3F). Again, lenalidomide, but not H₂O₂ cytotoxicity, was CRBN dependent (Figure 3G-H), confirming that CRBN isogenic cells exhibit similar H₂O₂ decomposition capacity and sensitivity to external H₂O₂ treatment (Figure 3D,G). However, CRBN is required for lenalidomide to elevate intracellular H₂O₂ (Figure 3E,H). Moreover, IKZF1 and IKZF3 degradation was a consequence of elevated intracellular H₂O₂ through a CRBN-dependent pathway (Figure 3A-H).

To demonstrate that antioxidative capacity determined the degree of IKZF1 degradation in CRBN-positive cells, we used RPMI-8226 and MM.1S, which have high and low antioxidative capacity, respectively (supplemental Figure 1A-C). After 3 hours of treatment with lenalidomide or H₂O₂, IKZF1 degradation was prominent in MM.1S but not in RPMI-8226, even though both cell lines expressed CRBN protein, further indicating that MM antioxidative capacity determined lenalidomide- and H₂O₂-induced IKZF1 degradation (supplemental Figure 3I). It was evident that high H₂O₂ concentration was required to degrade IKZF1 in RPMI-8226 compared with MM.1S due to its higher antioxidative capacity (supplemental Figure 3H).

Next, we sought to identify which H₂O₂-decomposing enzymes could be inhibited by lenalidomide. We performed in vitro assays with glutathione peroxidase and catalase but did not observe any inhibition with lenalidomide (supplemental Figure 3J). Surprisingly, lenalidomide inhibited thioredoxin reductase (TrxR) activity in vitro (Figure 3I). Concomitantly, we asked whether inhibition of TrxR with aurothiomalate (specific inhibitor of TrxR)²⁶  (Figure 3J) or thioredoxin (Trx) inhibition with PX12²⁷  (a major substrate of TrxR for maintaining intracellular H₂O₂ homeostasis; Figure 3K) could induce specific degradation of IKZF1 and IKZF3 in MM cells. We treated MM.1S and OPM2 cells with aurothiomalate and PX12 and showed that, like lenalidomide, aurothiomalate and PX12 also induced IKZF1 and IKZF3 degradation within 3 hours of treatment (Figure 3J-K). These findings clearly indicate that lenalidomide-mediated inhibition of TrxR leads to elevated intracellular H₂O₂ and also impairs the cells' capacity to reduce protein disulfide adducts²⁸  (protein disulfide dimers) in MM.

Lenalidomide-induced IgG-λ dimerization led to decreased secretion and consequent intracellular accumulation of IgG-λ, as evidenced by unchanged IgG-λ messenger RNA (mRNA) expression (supplemental Figure 4A), increased total intracellular IgL protein (Figure 4A), and decreased secretion of IgL-λ (Figure 4B; supplemental Figure 4B).

We postulated that the inhibition of IgG-λ protein secretion (due to H₂O₂-induced high dimerization) led to an ER stress response in CRBN-positive cells. After 3 days of treatment with lenalidomide, an ER stress response occurred in OPM2-NT but not in CRBN-knockdown cells (Figure 4C-D). Immunoblotting showed decreased XBP-1 unspliced and increased XBP-1 spliced proteins in CRBN-positive (nontarget) but not in CRBN-knockdown cells (Figure 4C). Another ER stress marker, GRP78/BiP, accumulated after lenalidomide treatment in CRBN-positive cells (Figure 4C). To further confirm that lenalidomide induced ER stress, we analyzed mRNA splicing of XBP-1 after lenalidomide treatment and saw that splicing was more evident in OPM2-NT than CRBN-knockdown cells (Figure 4D) after 3 days of lenalidomide treatment. The inositol-requiring enzyme 1 protein level markedly decreased in CRBN-positive cells after lenalidomide treatment (Figure 4C), indicating elevated inositol-requiring enzyme 1 endoribonuclease activity during ER stress.²⁹ 

We detected minimal apoptosis in OPM2-NT cells after 3 days of drug treatment, shown by low levels of poly(adenosine 5′-diphosphate-ribose) polymerase (PARP) cleavage (Figure 4C). Therefore, we treated OPM2-NT and OPM2-shCRBN cells with lenalidomide for 6 days and analyzed the PARP levels. Immunoblots showed nearly complete degradation of PARP proteins (cleaved PARP could not be detected because it was completely degraded by day 6) in CRBN-positive, lenalidomide-sensitive cells (Figure 4E). Moreover, to better understand lenalidomide-mediated IgL dimerization and accumulation as a mediator of time-dependent progressive ER stress, we treated CRBN-overexpressing OCIMY5 cells with lenalidomide and showed progressive IgL-λ dimer accumulation, ER stress, and PARP degradation (a sign of later-stage apoptosis; Figure 4F), whereas the OCIMY5 parent cell line was not responsive to lenalidomide (Figure 2G,I). We analyzed other isogenic HMCLs, MM.1S (lenalidomide sensitive) and MM.1Sres (lenalidomide resistant), and found lenalidomide-induced, ER stress–mediated accumulation of Bip protein in CRBN-positive cells but not in CRBN-negative cells (Figure 4G). We further showed that wild-type CRBN-expressing RPMI-8226 with high antioxidative capacity (Figure 1A-F) was resistant to lenalidomide because of no elevated intracellular H₂O₂-mediated IgL dimerization and ER stress (Bip accumulation) even though IKZF1 and IKZF3 were degraded at day 3 of treatment (Figure 4H).

Moreover, the limited IgL knockdown in OPM2 and H929 cells reduced lenalidomide sensitivity (supplemental Figure 4C,E,G) and decreased IgL-λ dimerization after 3 days of lenalidomide treatment, compared with control cells (supplemental Figure 4D,F,H). Next, we asked whether NAC treatment could reduce lenalidomide-induced IgL dimerization and cytotoxicity, even though NAC did not rescue lenalidomide-induced IKZF1 degradation (supplemental Figure 5A). MM.1S cells treated with NAC (20 mM) exhibited reduced IgL dimerization and also rescued from lenalidomide-induced growth inhibition (supplemental Figure 5B-C). We further noted that increasing concentrations of NAC shown some degree of cell proliferation inhibition on its own in the MM.1S cell line (supplemental Figure 5D).

Bim activation induced apoptosis after lenalidomide treatment in CRBN-positive MM cells. CRBN-expressing and CRBN-knockdown OPM2 cells were treated with lenalidomide for 3 days, and cell lysates were immunoblotted and probed for various pro- and antiapoptotic proteins. BH3-only protein Bim was activated after lenalidomide-induced ER stress (Figure 5A). Bim has 3 isoforms, BimS, BimL, and BimEL; their differing proapoptotic potencies are partly due to differences in their interactions with the dynein motor complex.³⁰  We observed accumulation of Bim, especially BimEL, after lenalidomide treatment in CRBN-positive, lenalidomide-sensitive cells. Mcl1 and Bcl2 antiapoptotic proteins did not change markedly after lenalidomide treatment (Figure 5A). We analyzed other CRBN-negative and CRBN-positive isogenic cell lines and confirmed Bim activation (Figure 5B). To confirm Bim involvement in lenalidomide-induced apoptosis, we used stable shRNA expression to knockdown Bim in OPM2 cells. We established 2 different OPM2 clones (numbers 73 and 75) with downregulated Bim and treated them with lenalidomide. Because lenalidomide induced late apoptosis in OPM2 cells, we performed a day-4 viability assay. Both Bim knockdown clones were less sensitive to lenalidomide than control cells (Figure 5C). Moreover, Bim knockdown cells underwent less PARP degradation compared with control cells (Figure 5D). Bim knockdown did not inhibit IgL-λ dimerization and accumulation or subsequent ER stress after lenalidomide treatment. Bim knockdown cells also showed elevated Bip protein accumulation (a marker of ER stress), like nontarget (control) cells after lenalidomide treatment, but ER stress–mediated apoptosis was diminished in Bim knockdown clones (Figure 5D). Thus, BimEL appeared to be a downstream effector of lenalidomide-induced, ER stress–mediated apoptosis in MM.

Proteasome inhibitors, including bortezomib, are also reported to induce oxidative stress and inhibition of oxidized protein degradation in MM and related cancers.^31-35  Recently, proteasome inhibitors are reported to directly induce oxidative and ER stresses in MM cells via transcriptional repression of a gene encoding mitochondrial thioredoxin reductase (TXNRD2).³⁶  Therefore, we postulated that dual treatment with lenalidomide and bortezomib is more clinically effective due to mutually enhanced oxidative stress.

We treated OPM2 with lenalidomide, bortezomib, and lenalidomide plus bortezomib for 3 days to identify better described downstream signaling. We found that concurrent administration of lenalidomide and bortezomib augmented the accumulation of IgL dimers and Bim (Figure 6A). Interestingly, bortezomib treatment alone also increased IgL-λ monomers and dimers, further indicating induction of oxidative stress and inhibition of degradation of oxidized immunoglobulins. Because lenalidomide-mediated MM cytotoxicity is a slow process (Figure 4F) compared with bortezomib, we propose that lenalidomide should precede bortezomib in the clinic; we pretreated MM cells with lenalidomide and then treated them with bortezomib, showing increased sensitivity (Figure 6B). Next, we analyzed whether CRBN-positive cells were more prone to apoptosis after lenalidomide pretreatment followed by bortezomib treatment. OPM2-NT and OPM2-shCRBN cells were pretreated with lenalidomide for 2 days, and cells were washed and then treated with bortezomib for an additional 2 days. Lenalidomide pretreatment enhanced bortezomib-induced PARP cleavage in OPM2 CRBN-positive cells, but cleavage was reduced in CRBN knockdown cells (Figure 6C), suggesting that when cells are resistant to lenalidomide, the synergistic effects with bortezomib also cease.

We wanted to extend our observations to pomalidomide, a third in class IMiD. We treated OPM2 cells with lenalidomide and also with 2 low concentrations of pomalidomide and found that pomalidomide was even more potent than lenalidomide in its induction of immunoglobulin light chain dimerization as an indicator of increased intracellular H₂O₂ elevation, induction of ER stress (Bip accumulation), and Bim-mediated apoptosis (Figure 6D).

We hypothesized that if CRBN is required for lenalidomide to inhibit TrxR, using direct TrxR inhibitor or targeting Trx, the downstream effector could potentially obviate CRBN and in some cases overcome resistance. TrxR inhibitors and Trx inhibitors have previously been reported by others to have activity in MM cells.^37-39 

The direct TrxR inhibitor auranofin, an organic gold compound used in clinic as treatment of rheumatoid arthritis,⁴⁰  was very effective in inducing cell death in MM cell lines, in a CRBN-independent fashion. Using MM.1R isogeneic cells with and without CRBN, and the CRBN-overexpressing OCIMY-5 cell line (transfected with wild-type CRBN), we showed that auranofin treatment caused cytotoxicity irrespective of CRBN expression, while lenalidomide was only effective in CRBN-expressing cells (Figure 7A-C). As with lenalidomide, the sensitivity of CRBN-expressing cell lines to auranofin correlated with their basal antioxidative capacity (MM.1S cells are sensitive as they have lower antioxidative capacity, whereas RPMI-8226 is resistant due to its high antioxidative capacity) (Figure 7D). We also found that targeting Trx with the direct inhibitor PX12 at a clinically achievable concentration of 20 µM⁴¹  overcame this cellular antioxidative capacity and was cytotoxic in both RPMI-8226 and MM.1S cell lines (Figure 7E); both express CRBN but have differential sensitivity to lenalidomide (Figure 1A-E). This effect was CRBN independent. Accordingly, isogenic MM cell lines with and without CRBN protein expression showed equal sensitivity to PX12 (Figure 7F-G). Clinically achievable concentrations of auranofin and PX12 induced cell signals similar to those of lenalidomide, including degradation of IKZF1, IKZF3 in MM.1S cell line (Figure 7H) and also in CRBN isogenic MM.1R cells (Figure 7I).

IMiDs bind to CRBN, and this binding is a prerequisite for teratogenic and anti-MM activity.^10,42  Here, we show that the presence of wild-type CRBN does not necessarily indicate sensitivity to IMiDs; its presence is necessary, but not sufficient. Our data showed that cellular antioxidative capacity determines lenalidomide sensitivity in MM cells expressing CRBN (Figure 7J). Accordingly, CRBN-expressing cells that have the highest capacity to decompose H₂O₂ were resistant to lenalidomide. Moreover, we developed a new method for determining total cellular antioxidative capacity, which can be used to predict which MM patients might benefit from IMiDs-based therapy. Furthermore, we discovered that the Trx/TrxR system-mediated intracellular H₂O₂ decomposition is inhibited by lenalidomide. Future studies are required to understand how it occurs and whether other peroxidases (eg, Trx-dependent peroxidases) are also inhibited by IMiDs. We are confident that CRBN retains IMiDs in intracellular compartments, induces chemical changes in the cellular redox status, and inhibits intracellular H₂O₂ decomposition. We and others have found that CRBN levels increase after lenalidomide treatment, so it likely follows that the intracellular concentration of IMiDs will increase with time. Future studies are required to elucidate how CRBN expression is required for IMiDs to optimally inhibit TrxR. These discoveries also support the use of TrxR and/or Trx inhibitors for future MM therapy,³⁸  which can bypass the CRBN requirement. Moreover, how some MM cells with similar CRBN expression have different antioxidative capacities necessitates further studies to better understand MM redox balance. Finally, lenalidomide-mediated elevation of intracellular H₂O₂ is particularly toxic to MM because of their high production of immunoglobulin proteins. Our findings not only could help identify patients who are most likely to benefit from IMiDs but might also facilitate the design of new drugs that diminish cellular antioxidative capacity and might increase sensitivity to IMiDs.

The authors thank Carole Viso and Tammy Brehm-Gibson for assistance with fluorescence-activated cell sorting analysis and Jessica Smith for help with experimental laboratory procedures. Bonnie Schimek helped edit figures and designed the graphic illustration. They thank the Center for Individualized Medicine, Mayo Clinic, for their funding contribution (S.S.) for developing biomarkers for predicting IMiDs response in MM. S.S. thanks the Mayo Clinic Specialized Program of Research Excellence Multiple Myeloma for Career Enhancement Award.

This work is supported by National Institutes of Health, National Cancer Institute grant R01 CA83724, Eastern Cooperative Oncology Group grant CA 21115T, and grants from Predolin Foundation, Mayo Clinic Cancer Center, and the Mayo Foundation.

Contribution: S.S. designed and performed research, analyzed data, and wrote the manuscript; R.F. was the principal investigator, oversaw the project, and wrote the manuscript with S.S.; Y.X.Z. and C.-X.S. generated initial constructs for CRBN knockdown and overexpression and validated them in vitro; A.K.S., P.L.B., M.C., S.C.P., E.B., S.A.V.W., and G.J.A. provided necessary reagents and experimental advice.

Conflict-of-interest disclosure: S.S. and R.F. have filed a US patent professional application for the method of using cellular antioxidative capacity to predict clinical response to IMiD treatment. R.F. has received a patent for the prognostication of MM based on genetic categorization of the disease. R.F. has received consulting fees from Celgene, Genzyme, BMS, Bayer, Lilly, Onyx, Binding Site, Novartis, Sanofi, Millennium, and AMGEN. He also has sponsored research from Onyx and is a member of the Scientific Advisory Board of Adaptive Biotechnologies. R.F. is a Clinical Investigator of the Damon Runyon Cancer Research Fund.

Correspondence: Rafael Fonseca, Division of Hematology and Medical Oncology, Mayo Clinic, 13400 E Shea Blvd, Scottsdale, AZ 85259; e-mail: fonseca.rafael@mayo.edu.