Evidence of a Role for Activation of Wnt/β-Catenin Signaling in the Resistance of Plasma Cells to Lenalidomide

Lenalidomide plays an important role in our chemotherapeutic armamentarium against multiple myeloma, in part by exerting direct anti-proliferative and pro-apoptotic effects. Unfortunately, long-term exposure leads to the development of drug resistance through unknown mechanisms, and we therefore sought to identify pathways that could be responsible for this phenotype. Chronic drug exposure produced myeloma cell lines that were tolerant of the direct effects of lenalidomide, with a degree of resistance of up to 2,500-fold. Gene expression profiling and pathway analysis identified dysregulation of the Wnt/β-catenin pathway as a consistent change across four independent cell isolates, and a pair of primary plasma cell samples. Acute drug treatment also increased β-catenin transcription by 3-fold or more, and both acute and chronic exposure resulted in enhanced accumulation of β-catenin protein by up to 20-fold or more. This produced Wnt/β-catenin pathway activation, as judged by increased activity of a lymphoid enhancer factor/T-cell factor promoter reporter, and enhanced accumulation of the downstream targets cyclin D1 and c-Myc. Components of the β-catenin destruction complex were also impacted by lenalidomide, which suppressed casein kinase 1α expression while augmenting glycogen synthase kinase 3α/β phosphorylation. Stimulation of Wnt/β-catenin signaling with recombinant Wnt-3a, or by overexpression of β-catenin, reduced the anti-proliferative activity of lenalidomide. Conversely, suppression of β-catenin with small hairpin RNAs restored plasma cell sensitivity to lenalidomide. Together, these findings support the hypothesis that lenalidomide mediates activation of Wnt/β-catenin signaling in plasma cells as a mechanism of inducible chemoresistance through effects at the transcriptional and post-translational levels.

Immunomodulatory drugs such as thalidomide and lenalidomide are being increasingly used in the treatment of multiple myeloma (1,2). The anti-angiogenic activity of thalidomide prompted the initial study in patients with advanced myeloma (3), and its encouraging activity there, and in further investigations, led to thalidomide-based regimens becoming a standard of care (1,2). However, its unfavorable toxicity profile motivated efforts to develop derivatives with decreased toxicity (1,2), resulting in the development of lenalidomide (CC-5013) (4,5). Lenalidomide first showed clinical efficacy as a single agent (6,7) and with dexamethasone (Len/Dex) 2 it provided superior outcomes in the relapsed/refractory setting (8,9). Later, Len/Dex showed activity against newly diagnosed myeloma (10,11), and early data suggest a benefit from maintenance therapy with lenalidomide after chemotherapy (12) and stem cell transplantation (13). Finally, Len/Dex may reduce the risk of progression from asymptomatic to symptomatic myeloma (14). Together, these developments argue that lenalidomide will play a crucial role in our chemotherapeutic armamentarium against myeloma for years to come.
The effects of immunomodulatory agents have been divided into three categories: modulation of the host immune system, modification of the tumor microenvironment, and direct inhibition of tumor proliferation and stimulation of apoptosis (2,15). Initial studies to characterize the latter effects noted they inhibited DNA synthesis, promoted G 1 arrest, activated apoptosis, and enhanced p21 expression (16). Induction of p21 correlated with cyclin-dependent kinase-2, -4, and -6 inhibition, and retinoblastoma protein hypophosphorylation (17,18). Lenalidomide may mediate this by epigenetically modifying the p21 promoter chromatin structure through lysine-specific demethylase-1 (19). Other plasma cell effects included inhibition of protein kinase B/Akt (20), suppression of nuclear factor-B (NF-B), down-regulation of inhibitor of apoptosis protein 2 and FLICE, and priming of cells for CD95 receptor-mediated death (15,21).
The effects of thalidomide on in vivo models have been dissected to identify the mechanisms responsible for its teratogenicity (22,23). Studies in chick embryo models proposed that thalidomide induced reactive oxygen species, which activated bone morphogenetic protein through NF-B inhibition. Dickkopf homolog (Dkk)-1 was induced through reactive oxygen □ S The on-line version of this article (available at http://www.jbc.org) contains supplemental Figs. S1 and S2. 1  species and p53, and possibly from bone morphogenetic protein through the c-Jun-N-terminal kinase (JNK) (24), and in aggregate this suppressed the wingless-type MMTV integration site (Wnt) family member signaling. Inhibition of Wnt reduced ␤-catenin activity, in part through increased glycogen synthase kinase (GSK)-3␤ action, and downstream consequences included stimulation of apoptosis (25). This cell death may cause the limb truncation defects that were seen with thalidomide use in pregnant patients (22,23). Despite the pleiotropic anti-tumor effects of immunomodulatory agents, a minority of patients achieve complete remissions, most ultimately progress after an initial response, and few studies have evaluated potential mechanisms of resistance to thalidomide. In chick, human, and mouse embryo fibroblasts, thalidomide-induced superoxide, and glutathione depletion enhanced the cytotoxicity of thalidomide (25). Conversely, free radical trapping agents, Dkk1 blocking antibodies, and inhibitors of bone morphogenetic proteins, phosphatase and tensin homolog, or GSK-3␤, attenuated apoptosis. The mechanisms of thalidomide resistance in myeloma cells have not been well delineated, however, and no studies of resistance to lenalidomide have been reported.
We therefore sought to examine the mechanisms responsible for lenalidomide resistance, and to do so we developed cell lines that were tolerant of its anti-proliferative and pro-apoptotic effects. Gene expression profiling and pathway analysis identified the Wnt/␤-catenin pathway as being consistently dysregulated. Lenalidomide induced ␤-catenin transcription, activated the downstream lymphoid enhancer factor (LEF)/Tcell factor (TCF) promoter, and enhanced accumulation of ␤-catenin, cyclin D1, and c-Myc protein. Moreover, lenalidomide contributed to Wnt/␤-catenin activation by suppressing casein kinase (CK)-1␣ expression and GSK-3␣/␤ activity. Finally, targeted Wnt/␤-catenin activation suppressed the activity of lenalidomide, whereas ␤-catenin inhibition restored its efficacy. Taken together, these data support the conclusion that, unlike thalidomide, which inhibits Wnt/␤-catenin, lenalidomide induces Wnt/␤-catenin in plasma cells, and this contributes to mechanisms of resistance to this agent.

EXPERIMENTAL PROCEDURES
Cell Lines and Primary Sample-Multiple myeloma cell lines were cultured in RPMI 1640 medium supplemented with L-glutamine, fetal bovine serum, penicillin, and streptomycin (all from Invitrogen). These cell lines were validated through The M.D. Anderson Cancer Center Characterized Cell Line Core Facility. Bone marrow aspirates were collected from patients under a protocol approved by the Institutional Review Board of The University of Texas M.D. Anderson Cancer Center, and informed consent was obtained in compliance with the Declaration of Helsinki. Primary plasma cells were purified by positive selection using magnetic-activated cell sorting with CD138 ϩ MicroBeads (Miltenyi Biotec, Auburn, CA). These cells, as well as interleukin (IL)-6 dependent cell lines (ANBL-6 and KAS-6/1), were supplemented with recombinant human IL-6 (R&D Systems, Minneapolis, MN).
Proliferation and Viability Assay-Cell lines were plated in triplicate in 96-well round-bottom plates with the indicated lenalidomide concentrations. These were incubated for 72 h, at which time medium was replaced with fresh lenalidomidesupplemented medium, and allowed to incubate for another 96 h. In experiments utilizing human recombinant Wnt-3a (rWnt-3a)(R&D Systems), this was added at 25 ng/ml for 2 h prior to lenalidomide. The WST-1 tetrazolium salt (Roche Applied Science) was then added and used according to the manufacturer's specifications. Cleavage of WST-1 occurs through the action of mitochondrial succinate-tetrazolium reductase in viable cells, and the quantity of formazan dye generated is directly related to the number of metabolically active cells. Colorimetric measurements were taken at 450 nm on a Victor 3V (PerkinElmer Life Sciences) plate reader, and analyzed using GraphPad Prism 5 software (GraphPad Software, La Jolla, CA). All data points were represented as a mean Ϯ S.D. from the number of determinations indicated in the figure legends.
Gene Expression Analysis-Cultures of drug-naive parental cell lines and their lenalidomide-tolerant counterparts were harvested in triplicate. In addition, paired primary samples were obtained from a patient prior to initiating lenalidomide therapy, and after the patient had developed clinical lenalidomide resistance. Total RNA was extracted and purified using the RNeasy purification kit (Qiagen, Valencia, CA), and then utilized for labeling and microarray hybridization to U133 plus 2.0 human whole genome microarrays (Affymetrix, Santa Clara, CA) by Genome Explorations, Inc. (Memphis, TN). Quantification of expression levels was determined with Affymetrix Microarray Suite (MAS) 5.0. Alterations in mRNA transcript levels were compared as the fold-difference for drug-naive versus lenalidomide-tolerant samples. These foldchanges were then mapped and analyzed using Ingenuity Pathway Analysis Software (Ingenuity Systems, Redwood City, CA).
Immunoblotting-Cells were harvested and lysed in 1ϫ Cell Lysis Buffer (Cell Signaling Technology, Danvers, MA) supplemented with 1ϫ Protease Inhibitor Mixture, 1ϫ Phosphatase Inhibitor Mixture, and PMSF (Sigma). Cell lysates were clarified, protein concentrations were determined using the DC Protein Assay (Bio-Rad), samples were subjected to electrophoresis on 4 -12 (Invitrogen) or 8 -10% gradient (Bio-Rad) acrylamide gels, and then transferred to nitrocellulose membranes (Bio-Rad). After blocking, they were probed with ␤-Catenin Signaling Mediates Lenalidomide Resistance the respective primary antibodies overnight, then washed and incubated with peroxidase-conjugated secondary antibodies, and protein bands were visualized using the ECL Chemiluminescence Kit (all from GE Healthcare Life Sciences, Piscataway, NJ). Primary antibodies included anti-␤-catenin, anti-CK1␣, and anti-phospho-GSK3␣/␤ (Cell Signaling Technology), and anti-␤-actin (Sigma). Protein bands were measured using ImageJ software (National Institutes of Health, rsbweb.nih.gov/ij), normalized to ␤-actin loading controls, and controls in cells that were either vehicle-treated or wild-type, as indicated.
Promoter Assays-Cells were infected with lentiviral particles containing an LEF/TCF promoter combined with a green fluorescent protein (GFP) reporter construct (SABiosciences, Frederick, MD) in Polybrene (Sigma). Control cells were infected with lentiviral particles containing GFP under control of the cytomegalovirus (CMV) promoter (Sigma). After infection, cells were washed, selected using puromycin (InvivoGen, San Diego, CA), treated with lenalidomide or vehicle for the indicated times, and GFP expression was analyzed using a FACSCalibur flow cytometer (BD Biosciences, San Jose, CA) and FlowJo software (Tree Star, Inc., Ashland, OR). Measurements were made in triplicate, and are presented as the mean Ϯ S.D., whereas statistical comparisons were made using the paired Student's t test, with p Ͻ 0.05 considered as significant, and p Ͻ 0.01 as highly significant.
Flow Cytometry-Cells were collected and fixed in 3% formaldehyde, and either subjected to flow analysis immediately, or stored until later use. Prior to analysis, all samples were permeabilized in ice-cold 100% methanol, and then resuspended in buffer with either anti-␤-catenin or anti-cyclin D1 antibody (Cell Signaling Technology), or anti-c-Myc antibody (Santa Cruz Biotechnology, Santa Cruz, CA). After washing, cells were incubated with Alexa Fluor-488-or Alexa Fluor-647-conjugated secondary antibodies (Invitrogen), and analyzed by flow cytometry with collection of 20,000 events.
Adenoviral and Lentiviral Transduction-Adenoviruses directing the expression of GFP as a negative control, or ␤-catenin (Vector Biolabs, Philadelphia, PA), were infected into myeloma cell lines in triplicate experiments prior to exposure to lenalidomide. Lentiviral particles (Sigma) containing a control, scrambled sequence shRNA (scr-shRNA), or ␤-catenin-targeting shRNAs (␤-cat-shRNA1-2), were transduced overnight into lenalidomide-tolerant cells. These transiently transduced wild-type and drug-resistant cells were then exposed to lenalidomide and analyzed as described in the text.

Development of Lenalidomide-resistant Cell Lines-Lena-
lidomide-induced anti-proliferative effects were evaluated initially in IL-6-dependent (ANBL-6) and IL-6-independent (U266) multiple myeloma cell lines (Fig. 1, A and B). Cellular viability measurements, determined using the WST-1 reagent as a function of mitochondrial activity and thus cellular proliferation, indicated a strong time dependence for the activity of lenalidomide when plasma cells were exposed continuously to this agent at 1 M. Modest anti-proliferative activity was seen after 24, 48, and 72 h of exposure (Fig. 1, A and B), and substantial effects were not evident until plasma cells had been treated for 96 to 120 h. When a lenalidomide concentration of 10 M was used instead, the time dependence was comparable, although the magnitude of the anti-proliferative activity, and IL-6-independent (U266) (B) cell lines were evaluated for time-dependent viability following continuous exposure to 1 M lenalidomide for up to 7 days. Lenalidomide-resistant cell lines were generated through a serial increase in lenalidomide concentrations added to the culture medium over several months, until the cell lines were capable of maintaining proliferation and cell membrane integrity at 10 M lenalidomide. These new cell lines were named (C) ANBL-6/R10R, (D) KAS-6/R10R, (E) U266/R10R, and (F) MM1/ R10R. Comparison of the sensitivity of lenalidomide-resistant cell lines versus their wild-type counterparts to lenalidomide is shown relative to the IC 50 curves that are displayed as a percentage of cellular viability via mitochondrial reduction of the tetrazolium salt, WST-1. All data points were normalized to the vehicle control (0.01% DMSO), which was arbitrarily set at 100% viability. Samples were run in triplicate, and are represented as the mean Ϯ S.D.
as expected, was greater (supplemental Fig. S1). These effects are likely to be due to lenalidomide itself as opposed to any of its metabolites, because incubation of the drug in the presence of human liver microsomes, recombinant cytochrome P450 isoenzymes, or human hepatocytes did not result in significant drug metabolism, confirming data from human studies (26). Also, the activity of lenalidomide was associated with an increased staining of myeloma cells with an antibody to Annexin V (supplemental Fig. S2), suggesting that its effect was not just anti-proliferative, but also pro-apoptotic. Notably, patient-derived plasma cells typically undergo spontaneous programmed cell death within 48 to 72 h of being removed from the permissive bone marrow microenvironment even in the absence of drug treatment. These observations suggested that primary plasma cells could not be effectively used to model resistance to lenalidomide given its slow onset of action. We therefore chose to utilize thoroughly characterized and established multiple myeloma cell lines to further investigate mechanisms of resistance to lenalidomide.
IL-6-dependent ANBL-6 and KAS-6/1 cells, as well as IL-6independent MM1.S and U266 cells, were exposed to lenalidomide concentrations at the IC 10 for this agent. Concentrations were then serially increased until each cell line maintained membrane integrity during continuous exposure to 10 M, as judged by trypan blue exclusion (data not shown). All cell lines retained their viability (Fig. 1, C-F), and relative fold-resistance was calculated based upon the IC 50 values. For drug-naive ANBL-6 (ANBL-6.wt) cells, for example, the IC 50 was 0.17 M, whereas their drug-resistant counterparts (ANBL-6/R10R) showed an IC 50 Ͼ100 M, indicating an at least 500-fold increased tolerance. Calculations for KAS-6/1, U266, and MM1.S cells indicated an at least 350-, 83-, and 2,500-fold increased resistance, respectively. Interestingly, in ANBL-6/R10R and U266/R10R cells, lenalidomide even seemed to stimulate cell growth at some of the drug concentrations tested, as indicated by an increase in cell viability above 100%. To evaluate the stability of the drug-resistant phenotype, cells were grown in lenalidomide-free medium, first for 7 days, and later for 90 days. When lenalidomide was added back, comparable levels of tolerance were seen compared with cells propagated continuously with lenalidomide using wild-type, drug-naive cells as controls (data not shown). These findings suggested that lenalidomide resistance was a stable phenotype, although the possibility remained that a rapidly inducible, and then reversible event was responsible. Interestingly, lenalidomide-resistant cells displayed a subjective morphological change, and had a greater tendency to form multicellular aggregates than their drug-naive counterparts (data not shown). Also of note, the drug-resistant phenotype appeared to be specific to lenalidomide, because these cells did not show significant cross-resistance to other commonly used anti-myeloma drugs, such as dexamethasone and bortezomib (data not shown).
Lenalidomide Induces Wnt/␤-Catenin Pathway Dysregulation-To identify mechanisms of lenalidomide resistance, gene expression profiling was performed on the lenalidomide-tolerant myeloma cells and their vehicle-treated, drug-naive counterparts. These data were then analyzed and mapped using Ingenuity Pathway Analysis Software, and this approach identified the Wnt/␤-catenin cascade as the most commonly dysregulated pathway. Indeed, a number of Wnt/ ␤-catenin intermediates were consistently induced at the messenger level in all four independently derived cell lines (Table 1). These included secreted components such as Wnt-3, receptor components such as frizzled Drosophila homolog (Fzd)-4, and downstream intermediates, such as ␤-catenin itself. Consistent with a role for lenalidomide in dysregulating Wnt/␤-catenin, qualitatively similar changes were seen in primary samples from a patient who had lenalidomide-refractory disease (Table 1).
Following the identification of the dysregulation of the Wnt signaling pathway by microarray analysis, we first sought to validate this induction by performing RT-PCR with ␤-catenin-specific probes, and an increased content of ␤-catenin mRNA was seen in the lenalidomide-resistant cells ( Fig. 2A). To determine whether this occurred during acute exposure as well, RT-PCR was performed on previously drug-naive myeloma cells. ANBL-6 cells treated in this fashion for 3 days revealed a lenalidomide dose-dependent increase in ␤-catenin mRNA (Fig. 2B) of up to 3-fold or more. As was the case for both chronically exposed myeloma cells and acutely treated ANBL-6 cells, KAS-6/1 (Fig.  2C) and MM1.S (Fig. 2D) cells treated with lenalidomide for 7 days showed an up to 2-fold ␤-catenin induction. Wnt Pathway Dysregulation Induces Accumulation of ␤-Catenin-Dysregulation of Wnt often leads to ␤-catenin stabilization, and we therefore sought to determine whether lenalidomide increased accumulation of this key protein. For all three cell lines examined, total ␤-catenin levels were increased in the drug-resistant cells propagated continuously in 10 M lenalidomide (R10R) compared with controls (Fig. 3A). Increased ␤-catenin accumulation was also seen in drug-resistant cells grown in the absence of lenalidomide for 7 days (RR), although these levels were lower than in the R10R cells (Fig. 3A). Using the U266 model as an example, compared with wild-type vehicle-treated cells, U266/R10R cells showed a Ͼ6-fold increase in ␤-catenin content, whereas U266.RR cells showed a Ͼ2-fold increase. The acute effects of drug exposure on ␤-catenin content were also studied in ANBL-6 ( Fig. 3B), KAS-6/1 (Fig. 3C), and MM1.S (Fig. 3D) cells. When lenalidomide was added, ␤-catenin content was increased in all three lines in a dose-dependent fashion. In KAS-6/1 cells, for example, ␤-catenin accumulation was seen starting at a lenalidomide concentration of 0.1 M, and increased further through the evaluated range, with a maximal, ϳ20-fold induction over controls.
Accumulation of ␤-Catenin Induces LEF/TCF Activity-Canonical Wnt signaling induces cytoplasmic ␤-catenin, which translocates to the nucleus and acts as a co-activator for the LEF/TCF family of DNA-binding proteins. My-eloma cells were therefore challenged with lenalidomide after infection with a lentiviral LEF/TCF promoter-GFP reporter construct. When control ANBL-6 ( Fig. 4A) or U266 (Fig. 4B) cells infected with a CMV promoter-GFP reporter construct were treated with lenalidomide, no significant promoter activation was seen. In contrast, both lines showed a concentration-and time-dependent LEF/ TCF promoter induction with lenalidomide. Using ANBL-6 cells as an example, lenalidomide concentrations as low as 0.1 M for 3 or 7 days induced LEF/TCF activity by almost 2-fold, whereas concentrations up to 100 M increased activity by Ͼ5-fold (Fig. 3A).
Activation of LEF/TCF can lead to induction of downstream targets such as cyclin D1 and c-Myc, and myeloma cells were therefore examined using a FACS-based assay. Consistent with earlier findings (Fig. 4), U266/R10R and MM1/R10R lenalidomide-resistant cells showed a 3.6-and 5.8-fold increased ␤-catenin content, respectively, compared with drug-naive cells (Fig. 5A). When propagated in 10 M lenalidomide, drug-resistant MM1.S cells showed a 2.3-fold increase in cyclin D1 protein (Fig. 5B). A smaller, 1.2-fold increase was seen in lenalidomide-treated U266 cells, possibly because they harbor the t(11;14) translocation, which induces high cyclin D1 levels at baseline, and may disrupt the input of the Wnt/␤-catenin pathway. Finally, MM1/R10R and U266/ R10R cells had an increased content of c-Myc by 3.1-and 3.9-FIGURE 2. Acute exposure to lenalidomide induces ␤-catenin transcription. RT-PCR was performed to detect ␤-catenin or ␤-actin transcripts from cells propagated continuously in lenalidomide (R10R), or R10R cells that were removed from lenalidomide for 1 week (RR), to detect ␤-catenin, and comparisons were made to vehicle-treated, drug-naive cells (WT) (A). RT-PCR products were separated by acrylamide gel electrophoresis, and detected by ethidium bromide staining. The left panel shows representative agarose gels, and densitometry was performed to calculate the increase in ␤-catenin after correcting for ␤-actin, which was normalized to the WT cells, which were set at 1.0. Drug-naive ANBL-6 (B), KAS-6/1 (C), and MM1.S (D) cells were exposed to the indicated lenalidomide concentrations for 72 (ANBL-6) or 168 h (KAS-6/1, MM1.S). Each top panel provides a representative photomicrograph, and densitometry was performed to calculate the ␤-catenin/␤-actin ratio. Values for each cell line are provided in the lower panels after normalization to the vehicle-treated control, which was arbitrarily set at 1.0. APRIL 1, 2011 • VOLUME 286 • NUMBER 13 fold, respectively, compared with controls (Fig. 5C). These data support the hypothesis that lenalidomide induces the Wnt/␤-catenin pathway, and activates LEF/TCF with consequences on downstream targets such as cyclin D1 and c-Myc.

␤-Catenin Signaling Mediates Lenalidomide Resistance
Lenalidomide Suppresses CK1␣ and GSK3␣/␤-Increased ␤-catenin transcription would contribute to its intracellular accumulation, but the induction at the protein level seemed out of proportion to the increase in mRNA, suggesting contributions from other mechanisms. Constitutive cytoplasmic ␤-catenin levels are controlled by a "destruction complex" that includes CK1␣ and GSK3␣/␤. In the absence of Wnt, ␤-catenin is phosphorylated by CK1␣ at Ser 45 , which primes  . The Student's paired t test was used to determine statistical significance, and ** denotes p Ͻ 0.005, whereas *** denotes p Ͻ 0.001. ANBL-6 (A) and U266 (B) cells were also infected with lentiviral particles directing expression of a CMV-GFP control reporter, and the impact of lenalidomide is shown.
Wnt/␤-Catenin Modulates Lenalidomide Sensitivity-To more directly examine the role of the Wnt/␤-catenin pathway in lenalidomide resistance, we sought to use molecular approaches to induce its activity. Extracellular Wnt acts by binding to the Fzd receptor and LRP co-receptor, which then propagate a signal to disrupt the ␤-catenin "destruction complex" (28). We therefore preincubated drug-naive KAS-6/1 and ANBL-6 (Fig. 8A, right and left panels, respectively) for 2 h with rWnt-3a, and then exposed them to lenalidomide. Such a Wnt/␤-catenin pathway stimulation significantly reduced the ability of lenalidomide to exert anti-proliferative effects, as judged by an increase in the IC 50 values of at least The lower panels display semiquantitative analyses of the mean fluorescent intensity for each R10R cell line compared with the WT controls, which were arbitrarily set at 1.0. Each dotted line represents 1.0, and values above that level represent increased expression of the proteins of interest. The Student's paired t test was used to determine statistical significance, and * denotes p Ͻ 0.05, whereas ** denotes p Ͻ 0.01. 10-fold for these cell lines. Because Wnt signaling can occur through non-canonical, ␤-catenin-independent pathways (28), it was also of interest to verify that ␤-catenin accumulation was by itself sufficient to mediate lenalidomide resistance. Adenoviruses inducing expression of GFP or ␤-catenin were therefore transduced into drug-naive KAS-6/1 and U266 cells (Fig. 8B, left and right panels, respectively), which were then treated with lenalidomide. In both lines, ␤-catenin overexpression preserved cellular viability, with an increase in IC 50 values of at least 10-fold.
We next wondered if we could reverse the lenalidomideresistant phenotype by reducing ␤-catenin expression. Lentiviral particles containing constructs for a nonspecific scrambled sequence shRNA (scr.shRNA), or two different shRNA constructs targeting ␤-catenin (␤-cat-shRNA.1 and ␤-cat-shRNA.2) were therefore used. For ANBL-6/R10R cells (Fig. 8C, left panel), both ␤-cat-shRNA constructs restored the anti-proliferative activity of lenalidomide, as judged by a reduction in viability of ϳ30 and 50% following 3-and 7-day drug exposures, respectively, compared with scr.shRNA controls. Similarly, KAS-6/R10R cells showed a reduction in cellular viability of 20 and 30% to lenalidomide, respectively, compared with controls. These findings together support the possibility that ␤-catenin activation mediates resistance to lenalidomide, and that approaches that suppress Wnt/␤-catenin may be of benefit in restoring sensitivity to this agent.

DISCUSSION
Clinical drug development in multiple myeloma usually begins by the evaluation of promising agents in patients with relapsed/refractory disease, and those that show substantial activity, such as bortezomib (29) and lenalidomide (1), are often later incorporated into front line therapy. One potential disadvantage to the earlier use of our best agents is the risk that drug-resistant disease clones will develop, thereby reducing options for treatment of patients with relapsed disease. We therefore sought to develop a model of plasma cells that were resistant to the direct anti-proliferative and pro-apoptotic effects of lenalidomide. This was with the rationale that these could ultimately help in understanding the mechanisms by which lenalidomide exerts its effects, identifying biomarkers of sensitivity, characterizing the mechanisms of lenalidomide resistance, and validating novel drugs or approaches to overcome or prevent the emergence of resistance.
In the current work, which to our knowledge is the first to present such a model, we report our initial characterization of these cells, which were tolerant of clinically relevant lenalidomide drug concentrations (6, 13) (Fig. 1). Microarray alignment of lenalidomide-resistant cells compared with their drug-naive counterparts identified the Wnt/␤-catenin signaling pathway as having multiple conserved changes (Table 1). Canonical Wnt/␤-catenin signaling starts with the extracellular binding of Wnts to their receptors and co-receptors, thereby destabilizing a destruction complex that normally FIGURE 6. Lenalidomide suppresses expression of CK1␣. Protein extracts from R10R or RR myeloma cell lines were probed to detect CK1␣, and compared with vehicle-treated drug-naive cells (WT) (A). The left panel shows a representative Western blot, and densitometry was performed to calculate the decrease in CK1␣ in each cell line after correcting for ␤-actin, which was normalized to the WT cells, which were set at 1.0. Acute exposure was studied in drug-naive cells (B), which were treated with the indicated lenalidomide concentrations for 72 (ANBL-6) or 168 h (KAS-6/1, U266). The left panels show representative Western blots, whereas the right panels show densitometry performed to evaluate CK1␣ levels corrected for ␤-actin loading, and normalized to the WT controls.

␤-Catenin Signaling Mediates Lenalidomide Resistance
targets ␤-catenin for proteasome-mediated degradation (28). This promotes ␤-catenin accumulation and its nuclear translocation, and participation as a co-activator in transcription of target genes responsible for a variety of cellular processes. A growing body of evidence has implicated Wnt/␤-catenin activation or hyperactivation through genetic (30) or epigenetic (31) mechanisms as contributing to, or even initiating oncogenesis. ␤-Catenin gain of function mutations are commonly found in colorectal and prostate carcinomas, among others (28). Overexpression of positive regulators such as secreted Wnts, the receptor protein Fzd, or signal mediators Disheveled and Frat1 occur in a variety of tumors as well, such as ovarian and breast cancers (28). Conversely, loss of function mutations in the negative regulators adenomatous polyposis C and Axin occur in several cancers, including colorectal and liver tumors (28). Moreover, emerging data also support an important role for Wnt/␤-catenin activation as a mediator of chemoresistance. For example, resistance to interferon-␣/5fluorouracil-mediated cell death in hepatocellular carcinoma required Wnt/␤-catenin activation (32). Also, studies of cisplatin resistance in laryngeal carcinoma identified a link between cell adhesion-mediated drug resistance (33) and ␤-catenin and plakoglobin levels. Activation of Wnt/␤catenin was also responsible for cell adhesion-mediated drug resistance to anthracyclines in neuroblastomas (34), acute myeloid leukemia (35), and myeloma (36). Given this background, and the important role of cell adhesion-mediated drug resistance in the natural history of myeloma (37,38), it is tempting to speculate that Wnt/␤-catenin signaling mediates lenalidomide resistance by impacting on plasma cell adhesive properties. This possibility is supported by the observation that lenalidomide-resistant cells appear to form aggregates to a greater extent than their drug-naive counterparts (data not shown), and we are currently performing further studies to identify the mechanisms behind this morphologic change. Additional studies will also be needed to examine the impact of Wnt/␤-catenin activation on the other properties of lenalidomide that contribute to its anti-myeloma activity, including the ability of this drug to act as an immune modulator, and to modify the tumor microenvironment, ideally using both murine in vivo models, as well as patient-derived clinically annotated primary samples.
Interestingly, activation of Wnt/␤-catenin was seen both in drug-naive cells acutely exposed to lenalidomide (Fig. 3,  B-D), and in drug-resistant cells chronically treated with this agent (Fig. 3A). Because the former were undergoing lenalidomide-induced apoptosis, whereas the latter proliferated in the presence of drug, and under some conditions to an even greater extent than with vehicle alone (Fig. 1, C and E), this may at first seem to undermine the role of Wnt/␤-catenin in drug resistance. However, a direct link between pathway activation and lenalidomide resistance was convincingly supported by studies in which Wnt/␤-catenin was manipulated using targeted, molecular approaches. These showed that pathway activation with recombinant Wnt (Fig. 8A), or overexpression of ␤-catenin (Fig. 8B), reduced lenalidomide sensi-tivity in drug-naive cells, whereas shRNA-mediated suppression in lenalidomide-resistant cells enhanced drug activity (Fig. 8C). It is also notable in this regard that time course experiments with ANBL-6 ( Fig. 1A) and U266 cells (Fig. 1B) showed that the majority of the anti-proliferative and proapoptotic effects of lenalidomide were fully expressed by day 5 of exposure. Thus, by day 7, which was the time point used in our studies, the vast majority of the drug-sensitive cells have already been eliminated, and the remaining cells are already functionally drug-resistant, but may have not yet acquired this property in a stable phenotype, which takes a longer period of exposure. This rapid emergence of drug-resistant cells also suggests the possibility that these cell lines, and possibly primary tumors as well, consist of a mixture of inherently lenalidomide-sensitive and -resistant populations, and that it is the ability of the latter to better activate Wnt/␤catenin and survive drug treatment that leads to disease persistence and subsequent relapse. Studies are underway to de-termine by immunophenotyping if two such populations can indeed be identified. If this could be done, it would be interesting to compare these with genomic and proteomic techniques to provide further insights into the mechanisms of drug resistance, as well as to possibly provide markers of drug sensitivity that could be useful clinically to predict which patients would most benefit from therapy.
Our current findings suggest a role for two mechanisms in induction of the Wnt/␤-catenin pathway by lenalidomide.
One of these appears to be enhanced ␤-catenin transcription, because we observed increased transcript levels in chronically treated cells (Table 1), and in acutely exposed cells (Fig. 2). These findings are interesting considering that ␤-catenin has been felt to be primarily regulated at the protein level, whereas transcriptional control was under a constitutive promoter. However, Bandapalli et al. (39) recently showed that ␤-catenin, in conjunction with TCF4, bound to and regulated its own promoter. Moreover, previous ␤-catenin promoter FIGURE 8. Modulation of Wnt/␤-catenin pathway alters lenalidomide sensitivity. Drug-naive KAS-6.wt and ANBL-6.wt cells (A) were exposed to vehicle or 25 ng/ml of human recombinant Wnt-3a (rWnt-3a) prior to exposure to lenalidomide for 168 h. Cellular viability measurements were performed using the WST-1 assay, and all data points were normalized to the vehicle control (0.01% dimethyl sulfoxide), which was set at 100%. Mean viability values are provided from three independent experiments, along with the S.D. KAS-6.wt and U266.wt were then infected with adenoviral particles (B) containing CMV promoter-driven expression vectors for GFP (Ad-GFP) or ␤-catenin (Ad-␤-cat) for 24 h. They were then exposed to lenalidomide, analyzed, and are presented as described in panel A. Finally, ANBL-6/R10R and KAS-6/R10R cells were infected with lentiviral particles containing shRNA constructs targeted to a scrambled sequence (scr.shRNA), or an shRNA targeting ␤-catenin (␤-cat-shRNA-1 and Ϫ2). Twenty-four h later they were exposed to 10 M lenalidomide for 72 or 168 h, and analyzed for viability. All values are normalized to the vehicle control, and presented as the average of triplicate measurements Ϯ S.D. The Student's paired t test was used to determine statistical significance, and ** denotes p Ͻ 0.005, whereas *** denotes p Ͻ 0.001 for C. mapping studies showed an LEF/TCF binding site, and high affinity sites for E2F1 and NF-B (40). Because the former is regulated by cyclin D1, a downstream Wnt/␤-catenin target, it is possible that induction of ␤-catenin transcription is a secondary event stimulated by the accumulation of ␤-catenin protein itself (Fig. 3), and activation of the pathway as a whole (Fig. 4), in a positive feedback loop. However, our experiments have not probed for the possibility that lenalidomide may have a direct epigenetic effect on the ␤-catenin promoter, as it does on p21 (19), or that it may directly or indirectly influence other transcriptional activators and/or repressors that control ␤-catenin transcription. Further studies will be needed to determine the contribution of these various possible mechanisms to the actions of lenalidomide.
The second mechanism of ␤-catenin activation suggested by our findings is through suppression of expression of CK1␣ (Fig. 6) and activity of GSK3␣/␤ (Fig. 7). CK1␣, a serine/threonine kinase, is responsible for the initial phosphorylation priming step at Ser 45 of ␤-catenin. Casein kinases are ubiquitous (41), but less is understood about their regulation or turnover. Unlike other casein kinases that are regulated by phosphorylation events, CK1␣ lacks the consensus sequences needed to submit to this type of regulation. Potential mechanisms for a reduction of CK1␣ could therefore be similar to those discussed above for ␤-catenin, including a suppression of transcription, or enhanced protein turnover. Interestingly, our microarray data indicated that there was a consistent increase in CK1␣ mRNA levels (not shown), suggesting that lenalidomide both activated CK1␣ transcription, and either suppressed its translation, or enhanced its turnover.
GSK3␣/␤ is also a common threonine/serine kinase made up of a heterodimer of the ␣ and ␤ subunits. It is unphosphorylated when active, and is responsible for the phosphorylation of ␤-catenin at Thr 41 /Ser 37 /Ser 33 . This generates the recognition signal that allows the E3 ubiquitin ligase ␤-TRCP to bind and polyubiquitinate ␤-catenin, leading to its proteasomemediated degradation. Following an array of stimulatory responses, GSK3␣/␤ is phosphorylated at Ser 21 of the ␣ subunit, and Ser 9 of the ␤ subunit, by either protein kinase A (PKA) (42) or protein kinase B (PKB)/Akt (43). This phosphorylation leads to inactivation of GSK3␣/␤, and is associated with a pro-survival cellular response. Our data indicate that phosphorylated GSK3␣/␤ levels are increased upon acute and chronic lenalidomide exposure. Consistent with an inhibition of GSK3␣/␤ activity, this resulted in decreased abundance of ␤-catenin phosphorylated at Thr 41 /Ser 37 /Ser 33 (not shown). This could be occurring through a Wnt-dependent autocrine or paracrine process resulting in dissolution of the destruction complex, given that our microarray data showed increased transcription of several Wnt isoforms. However, we were unable to confirm the presence of increased Wnt production and secretion into the medium (not shown), providing an interesting parallel with the data for CK1␣, above. It is also possible that lenalidomide directly stimulates PKA or PKB/Akt, resulting in enhanced GSK3␣/␤ phosphorylation, although this is not supported by previous studies suggesting that lenalidomide suppressed Akt (20). Once again, additional studies will be needed to define all of the mechanisms by which this compound exerts such pleiotropic effects.
Taken together, our experimental efforts have identified activation of the Wnt/␤-catenin pathway as a conserved effect of acute and chronic exposure of plasma cells to lenalidomide. Activation is mediated by decreased ␤-catenin destruction and increased transcription, and results in downstream activation of targets such as LEF/TCF, cyclin D1, and c-Myc. Additionally, we have shown that selective modulation of this pathway can alter the anti-proliferative potency of this drug against myeloma cells. These drug-resistant models may also allow us in the future to examine whether Wnt/␤-catenin activation can be a biomarker for lenalidomide sensitivity, and whether it can be targeted through pharmacological means to induce chemosensitization and reverse resistance. If validated, these approaches will also need to be considered in the context of the role of the Wnt/␤-catenin pathway in myeloma bone disease. Overproduction of Dkk, a secreted LRP inhibitor produced by myeloma cells, induces bone resorption through its effects on marrow osteoblasts (44). Binding of Dkk to LRP extinguishes Wnt signaling, and Dkk antibodies, which would enhance Wnt/␤-catenin activity, are being evaluated as bone anabolic agents (45). Thus, there is the risk that approaches that would inhibit Wnt/␤-catenin could enhance the efficacy of lenalidomide, but also enhance bone resorption, unless they could somehow be targeted only to myeloma cells.