Anthelmintic Niclosamide Disrupts the Interplay of p65 and FOXM1/β-catenin and Eradicates Leukemia Stem Cells in Chronic Myelogenous Leukemia

Chronic myelogenous leukemia (CML) arises from hematopoietic stem cells (HSC) malignantly transformed by the BCR-ABL oncogene. CML generally progresses from a chronic phase (CP) to an accelerated phase (AP), then a stage of blast crisis (BC; refs. 1, 2). A 10-year clinical follow-up demonstrated that treatment with a tyrosine kinase inhibitor (TKI) imatinib mesylate achieved remission in more than 80% of patients with CP CML and significantly prolonged the event-free survival of CML patients harboring the wild-type BCR-ABL (3). However, acquired resistance to imatinib is a challenge in CML treatment. Point mutations (e.g., T315I, G250E, and E255K/V) in BCR-ABL are major causes of imatinib-resistance (4). Nilotinib and dasatinib, the second-generation TKIs, can achieve good clinical response in most CML patients harboring most mutant isoforms of BCR-ABL except T315I (5). Ponatinib, a third-generation TKI, has been approved for treatment in imatinib-resistant CML patients harboring T315I BCR-ABL despite its potential cardiotoxicity (6, 7). Thus, acquired mutation resistance to imatinib is becoming a manageable clinical issue with these novel TKIs.

The other mechanisms of resistance to imatinib may include leukemia stem cells (LSC) or leukemia-initiating cells (8), and BCR-ABL–independent clones (9). LSCs, characterized by their capacity for self-renewal, and insensitivity to TKIs, may confer clinical resistance to imatinib and lead to CML relapse. Current evidence supports that LSCs are retained in CML patients with remission induced by TKI treatment (10), for a potential source of CML recurrence (11). Identification of novel agents capable of eradicating LSCs may be a critical strategy to cure CML.

The regulation of self-renewal in LSCs has not been fully understood. The intrinsic regulators include developmental signaling pathways (e.g., WNT/β-catenin, Hedgehog) and transcription factors [e.g., forkhead box M1 (FOXM1), NF-κB; ref. 12]. Evidence supports that loss of β-catenin impairs LSCs renewal in vivo, and genetic deletion and pharmacologic inhibition of β-catenin targets LSCs in CML (13, 14). Deregulated inflammatory cytokines in leukemic bone marrow may also affect intrinsic regulators of LSCs; for instance, high levels of TNFα in bone marrow promotes CML LSC survival by activating the NF-κB pathway (15). However, whether β-catenin and NF-κB, two common intrinsic regulators of LSCs, have interplay, remains unclear.

The current study aimed to examine the interplay between β-catenin (with its partner FOXM1; ref. 16) and NF-κB, and the effect of its chemical intervention on CML LSCs. We discovered a positive-feedback loop regulation between NF-κB p65 and FOXM1/β-catenin in CML LSCs. Disrupting this loop with niclosamide inhibited survival and self-renewal of CML LSCs in vitro and in vivo.

Niclosamide, Annexin V-FITC, cycloheximide, DMSO, anti-Flag, anti-HA, anti-actin, and anti-tubulin were from Sigma-Aldrich. p-Niclosamide, a water-soluble derivative of niclosamide, was designed by adding a phosphate group to niclosamide with diethyl phosphate (17). Imatinib was from Novartis Pharmaceuticals. MG-132 was from Calbiochem. Recombinant human TNFα was a product of Peprotech. Recombinant human WNT3A was from R&D Systems. Protein A/G agarose beads and antibodies against FOXM1 (C-19, K-20), p65 (c-20, F-6), proliferating cell nuclear antigen (PCNA), and IκBα were from Santa Cruz Biotechnology. Anti-β-catenin and FITC conjunct IgG1κ isotype control were from BD Biosciences. Phospho-IκBα (S32) was from Cell Signaling Technology. Anti-active-β-catenin (clone 8E7) was from Upstate Tech. Anti-mouse immunoglobulin G and anti-rabbit immunoglobulin G horseradish peroxidase–conjugated antibodies were from Pierce Biotechnology.

Plasmid encoding FOXM1 gene promoter–driven luciferase (Luc) reporter was described previously (18). The pNF-κB-Luc plasmid was from Stratagene. The TCF/LEF reporter plasmid, pTOPflash, and its mutant control, pFOPflash, were from EMD Millipore. The Renilla luciferase reporter construct, pEFRenilla-Luc, was from Promega. pcDNA3-β-catenin was from Addgene. FOXM1-flag (16), p65-HA, p65 1–286, and p65 286–551 (19), HA-β-catenin (20), and (His)₆-ubiquitin (21) were described previously.

siRNA duplexes against p65 (#1, sc-29410; #2, sc-44212) and FOXM1 (#1, sc-270048; #2, sc-43769) were from Santa Cruz Biotechnology. siRNA duplexes against β-catenin (#1, 6225; #2, 6238) were from Cell Signaling Technology; ON-TARGET plus Non-Targeting Pool siRNA control was from Dhamacon RNA Tech (22).

K562 and KBM5-T315I cells were grown as reported previously (23). 293T and Plat-E cells were cultured in DMEM supplemented with 10% FBS. MEF p65^+/+ and p65^−/− cells were grown in IMDM (Invitrogen) supplemented with 10% FBS (24). All the cell lines were tested and authenticated by using short tandem repeat (STR) matching analysis of cells last month. No cross-contamination of other human cells was found in all six lines of cells.

For K562 and human primary CD34⁺ cells, siRNA duplexes were transduced into cells with the Cell Line Nucleofector Kit T (Amaxa) and program O-17 (22). For 293T cells, transfection involved use of Lipofectamine 2000 (Invitrogen).

Whole-cell lysates were prepared in RIPA buffer. Cytoplasmic, mitochondrial, and nuclear extracts were prepared as described previously (17, 22).

Immunoprecipitation (IP) and immunoblotting (IB) were performed as reported previously (16).

An amount of 1 × 10⁷ K562 cells was prepared with a ChIP kit (EMD Millipore). The precipitated DNA complexes were analyzed by real-time quantitative PCR (22). The primers were listed in Supplementary Table S1.

Dual luciferase assay was followed as reported previously (22). Briefly, cells were transfected with plasmids encoding FOXM1-Luc (0.5 μg), NF-κB-Luc (0.5 μg), pTOPflash (0.5 μg), pFOPflash (0.5 μg), and pEFRenilla-Luc (10 ng) with Lipofectamine 2000. Luciferase activity was measured with Dual Luciferase Assay Kit (Promega) as described previously (22). The luciferase activity was normalized to Renilla luciferase activity.

Cells treated with or without niclosamide for 24 hours followed by stimulation of TNFα or WNT3A were collected and cytospun to slides. Immunofluorescence staining was performed as described previously (22).

KBM5-T315I cells were transfected with the indicated constructs in the presence of DMSO or 2.0 μmol/L niclosamide for 24 hours and then treated with MG-132 (20 μmol/L) for the last 6 hours. In vivo ubiquitination assay was performed as reported previously (21).

Normal bone marrow samples (n = 4) were obtained from healthy donors at the first affiliated hospital of Sun Yat-sen University (Guangzhou, China). CML samples (n = 9) were obtained from the First Affiliated Hospital of Sun Yat-sen University and Guangdong General Hospital. Nucleated cells were isolated by Ficoll separation. CD34⁺ cells were fractionated by use of a positive magnetic bead selection protocol (Miltenyi Biotec). All patients and healthy donors gave their signed informed consent. The study approved by our institute followed the Declaration of Helsinki principles. The clinical information of CML patients is described in Supplementary Table S2.

The staining procedure was described previously (22). Samples were analyzed by Accuri C6 (BD Biosciences) flow cytometry.

The apoptosis of quiescent cells was analyzed according to the previous report (25). Briefly, carboxyfluorescein succinimidyl amino ester (CFSE)-stained CD34⁺ cells (CellTrace CFSE Cell Proliferation Kit, Invitrogen) were incubated with different treatments for 96 hours and then stained with Annexin V–PE and analyzed by Accuri C6 flow cytometry. CFSE^maxCD34⁺Annexin V⁺ cells were determined as apoptotic quiescent cells (25).

Long-term culture-initiating cell (LTC-IC) assay and limiting dilution LTC-IC assay were performed following the manufacturer's instructions (9). CML-nucleated cells (2 × 10⁶) were cultured with irradiated (80 Gy) M2-10B4 murine fibroblasts in MyeloCult H5100 (StemCell Technologies) supplemented with 10⁻⁶ mol/L hydrocortisone (long-term culture medium) in the presence of drugs for the first week of culture. Medium was replaced by half drug-free medium change weekly. After 6 weeks, all cells were harvested and plated into MethoCult H4435. LTC-IC–derived colonies were counted after 14 days. For limiting dilution assay, pretreated mononuclear CML cells were cultured with irradiated (80 Gy) M2-10B4 murine fibroblasts in MyeloCult H5100 supplemented with 10⁻⁶ mol/L hydrocortisone at serial dilutions (10⁴, 3 × 10³, 10³, 300, 100). Half of the medium was refreshed weekly. After 5 weeks, cells were harvested and seeded in MethoCult H4435, and then the colonies were counted. LT-HSC frequency was analyzed by Poisson statistics online by using the Bioinformatics facility of The Walter & Eliza Hall Institute of Medical Research (Melbourne, Australia; ref. 26).

CD34⁺ cells (5,000/well) were seeded in the H4434 MethoCult with niclosamide for the first round. Colonies were counted 7–10 days after culture, and then 5,000 cells from the colonies were sequentially replated in the H4434 MethoCult for another two rounds, respectively (9).

CML CD34⁺ cells were cultured with or without 2.5 μmol/L niclosamide for 48 hours. Cells (1–2 × 10⁶ cells/mouse) were then collected, washed, and transplanted into 8-week-old NOD.Cg-Prkdc^scidII2rg^tm1Sug/JicCrl mice (NOG mice, CIEA) via the tail vein (27). Cells were allowed to grow for 10 weeks. Mice were sacrificed, and fractionated mononuclear cells from bone marrow and spleen were labeled with antibodies for flow cytometry analysis (BD LSRFortessa). Antibodies were from BD Biosciences: CD45-APC Cy^TM7, CD34-FITC, CD33-PE Cy^TM7, CD14-PerCP-Cy5.5, CD11B-PE, CD19-APC, and CD3-Alexa Fluor 700 (27, 28).

The retroviral construct MSCR-IRES-BCR-ABL-WT-EGFP was used to generate high-titer helper-free retrovirus by transient transfection of Plat-E cells as reported previously (27). Bone marrow cells from 5-fluorouracil–treated (200 mg/kg) 6- to 8-week-old C57BL/6 male donor mice were transduced twice with BCR-ABL retrovirus by centrifugation in the presence of IL3, IL6, and stem cell factor (SCF). Cells (0.5 × 10⁶) were then transplanted by tail vein into irradiated (5.50 Gy) receipt female mice. Mice were treated with placebo, imatinib, p-niclosamide, and imatinib combined with p-niclosamide for 2 weeks.

Cells were obtained from bone marrow (both femurs and tibias) or spleen. Antibodies were as follows: Lin-APC, Sca-1-PE-CF594, c-Kit-PE, Flt3-PE-Cyanine 5, CD150-PE-Cyanine 7, and CD48-APC-Cyanine7. Analysis and LSK cell sorting were conducted using flow cytometer (BD FACSAria II, BD Biosciences).

The qPCR experiments were carried out as described previously (22). The primers were listed in Supplementary Table S1.

Bone marrow and splenic cells from 3–5 CML mice with different treatments were harvested and injected into secondary receipt mice (irradiated at 5.50 Gy) at serial concentrations of cells (2 × 10⁶, 1 × 10⁶, 5 × 10⁵). GFP⁺ cells in peripheral blood were monitored by cytometry every week. GFP⁺ cells >0.5% were considered as positive transplantation. LSC frequency was determined 16 weeks after the secondary transplantation using Poisson statistics online by using the Bioinformatics facility of The Walter & Eliza Hall Institute of Medical Research (Parkville, Victoria, Australia; ref. 26).

GraphPad Prism 5.0 (GraphPad Prism Software) was used for statistical analysis. All experiments were carried out at least three times, and results were presented as mean ± SEM unless otherwise stated. Comparison between two groups was analyzed by t test and between more than two groups by one-way ANOVA with post hoc comparison by Tukey test. P < 0.05 was considered statistically significant.

To examine the potential interaction between NF-κB and FOXM1, 293T cells were cotransfected with HA-tagged p65 and Flag-tagged FOXM1 and subjected to IP analysis. IB analysis revealed FOXM1 presented in anti-p65 IP pellets (Fig. 1A, left) and p65 in anti-FOXM1 IP pellets (Fig. 1A, right). Moreover, reciprocal co-IP experiments revealed physical interaction of endogenous p65 and FOXM1 proteins in CML cells (Fig. 1B). Our results suggest that NF-κB protein physically interacts with FOXM1 protein.

We next explored whether transcription factor p65 promotes FOXM1 expression. IB analysis indicated increased level of endogenous FOXM1 protein with increasing amount of plasmids encoding full-length p65 (Fig. 1C, top). In a parallel set of experiments cotransfecting a construct containing FOXM1 promoter-driven luciferase reporter (18), p65 dose dependently promoted FOXM1 promoter–driven transcription activity (Fig. 1C, bottom). In contrast, transfection of constructs of truncate p65 (a.a. 1–286) lacking the TAD domain (19), and truncate p65 (a.a. 286–551) containing NLS and C-terminal domains did not elicit an increase in endogenous FOXM1 protein (Fig. 1D), as well as in transcriptional activity of the Luc-FOXM1 promoter (Fig. 1E). These data suggest that the C-terminal portion including the NLS appears to be required for the promotion of FOXM1 transcription.

We next examined whether cellular endogenous p65 directly bound to the promoter sequence of FOXM1 gene by using chromatin immunoprecipitation (ChIP) assay. Fragmented chromatin of K562 cell lysates was immunoprecipitated with anti-p65 antibody, and specific primers were used to amplify p65-binding sites in the FOXM1 promoter region. The results showed recruitment of endogenous p65 to the FOXM1 gene promoter but not CDS and intron regions of FOXM1 or irrelevant gene GAPDH promoter in CML cells (Fig. 1F).

We next investigated whether NF-κB was required for FOXM1 expression. K562 cells were transfected with siRNA duplexes against RELA gene, and cell lysates were then subjected to IB and luciferase activity assay. Silencing p65 in K562 cells by transfection of siRNA duplexes against RELA gene led to downregulation of FOXM1 expression as detected by IB (Supplementary Fig. S1A, left). Parallelly, p65 knockdown decreased the transcriptional activity of Luc-FOXM1 promoter which was cotransfected in K562 cells (Supplementary Fig. S1A, middle). These results suggest that NF-κB is required for FOXM1 expression. In addition, parallel experiments showed that p65 knockdown in K562 cells reduced the levels of β-catenin and its downstream TCF/LEF–dependent transcription activity (Supplementary Fig. S1A, right), which further indicates the regulation of β-catenin function by NF-κB.

To further delineate the regulation of FOXM1 by canonical activation of NF-κB, K562 cells with p65 silenced by siRNA were exposed to TNFα for different durations. Control siRNA–treated cells showed TNFα-upregulated FOXM1 expression, which was abrogated in p65 siRNA–treated cells (Supplementary Fig. S1B). These data suggest that TNFα upregulates FOXM1 in a p65-dependent manner. Furthermore, p65-deficient MEF cells showed higher turnover rate of FOXM1 as compared with MEF p65^+/+ cells (Supplementary Fig. S1C).

To better clarify the relationship of p65 and FOXM1, MEF cells harboring p65^+/+ or p65^−/− were stimulated by TNFα for various durations, nuclear and cytosolic fractions were examined by IB. The purity of nuclear and cytosol fractions was first verified (Fig. 1G). In MEF p65^+/+ cells, TNFα stimulation triggered FOXM1 nuclear translocation, coupled with p65 nuclear translocation (Fig. 1H). In stark contrast, TNFα stimulation did not trigger nuclear translocation of FOXM1 in MEF p65^−/− cells (Fig. 1H). A similar effect of TNFα treatment on nuclear relocation of FOXM1 and p65 was observed by confocal fluorescence microscopy (Fig. 1I). Moreover, WNT3A stimulation triggered concomitant nuclear translocation of β-catenin and FOXM1 protein in MEF p65^+/+ cells but not MEF p65^−/− cells (Supplementary Fig. S1D). These results suggest that p65 is required for FOXM1 nuclear translocation.

Because niclosamide is capable of blocking the canonical activation of NF-κB pathway (17), K562 cells pretreated with or without niclosamide were exposed to TNFα. IB examination showed that in the absence of niclosamide, IκBα was phosphorylated shortly after TNFα stimulation (Fig. 1J, left). Accordingly, p65 level was decreased in the cytosolic fraction and increased in the nuclear fraction (Fig. 1J, left). Of note, FOXM1 and β-catenin levels were concomitantly decreased in the cytosolic fraction and increased in the nuclear fraction with the change in p65 level (Fig. 1J, left). Niclosamide completely abolished TNFα-induced nuclear translocation of p65, FOXM1, and β-catenin (Fig. 1J, right).

Similarly, immunofluorescence staining revealed coupled nuclear translocation of p65 and FOXM1, and p65 and β-catenin in K562 cells with TNFα treatment (Supplementary Fig. S1E and S1F), which suggests that pharmacologic inactivation of NF-κB by niclosamide abrogates nuclear translocation of FOXM1 and β-catenin in CML cells.

We next examined whether FOXM1 and β-catenin affected NF-κB level. 293T cells were transfected with plasmids encoding human FOXM1 or β-catenin, and p65 level was assessed by IB. Ectopic expression of FOXM1 and β-catenin increased endogenous p65 level (Fig. 2A).

We determined the effect of WNT/β-catenin activation by WNT3A on p65 level in K562 cells. IB analysis of K562 cells exposed to WNT3A revealed that β-catenin and FOXM1 levels were progressively decreased over time in the cytosolic fraction and progressively increased in the nuclear fraction (Fig. 2B), which supports nuclear translocation of β-catenin during canonical activation of WNT/β-catenin pathway. IB results showed a coupled and concurrent change of p65 with β-catenin in the cytosolic and nuclear fractions in WNT3A-treated cells (Fig. 2B). WNT/β-catenin activation promoted the nuclear translocation of p65.

In K562 cells transfected by two independent siRNA duplexes against human FOXM1, the protein levels of p65 and β-catenin were reduced along with FOXM1 knockdown (Fig. 2C). Accordingly, NF-κB– and β-catenin–dependent reporter gene transcription was reduced (Fig. 2C). Similarly, silencing β-catenin by siRNA in K562 cells reduced p65 protein level and NF-κB–dependent reporter gene transcription (Fig. 2D), lowered the FOXM1 protein level and the transcription activity of FOXM1, as evaluated by IB and FOXM1 gene promoter–dependent luciferase activity, respectively (Fig. 2D). Collectively, these data suggest a positive-feedback regulation in expression between NF-κB and FOXM1/β-catenin.

Because niclosamide is capable of blocking canonical activation pathways of NF-κB as well as WNT/β-catenin (17), we determined whether niclosamide disturbed the regulatory loop between NF-κB and FOXM1/β-catenin. Niclosamide dose dependently reduced β-catenin protein level in CML cells (Fig. 2E). Furthermore, in vivo ubiquitination assay showed that the ubiquitination of β-catenin was increased with niclosamide treatment (Supplementary Fig. S2A), which suggests that niclosamide induces ubiquitin–proteosome–dependent degradation of β-catenin.

Furthermore, immunofluorescence staining experiments revealed that niclosamide completely abrogated β-catenin nuclear translocation in K562 cells stimulated with WNT3A (Fig. 2F and Supplementary Fig. S2B). Concomitantly, the nuclear translocation of p65 was abolished in K562 cells treated with WNT3A combined with niclosamide. Collectively, niclosamide blocked NF-κB and FOXM1/β-catenin and disrupted the regulatory loop.

Because TNFα supports the survival of CML stem/progenitor cells by promoting NF-κB pathway activity (15), and β-catenin is essential for stemness maintenance of CML LSCs (14), we examined whether the regulatory loop between NF-κB and FOXM1/β-catenin existed in the primary CD34⁺ cell populations from CML patients. Purified CD34⁺ cells from CML patients were labeled with antibodies against p65, FOXM1, and active-β-catenin, then underwent flow cytometry analysis. The results showed that the levels of intracellular p65, FOXM1, and active-β-catenin were significantly higher in CML CD34⁺ cells than those in normal bone marrow CD34⁺ cells (Fig. 3A). Immunofluorescence detection revealed that p65, FOXM1, and β-catenin appeared predominantly distributed in nucleus of the CML CD34⁺ cells versus NBM counterparts (Supplementary Fig. S3). In addition, overlay between p65 and β-catenin under immunofluorescence observation implied interplay of NF-κB and FOXM1/β-catenin pathways in the primary CML CD34⁺ cells (Supplementary Fig. S3).

To validate the above implications, primary CD34⁺ cells from CML patients were cotransfected with constructs of p65 or β-catenin with constructs of Luc-FOXM1 promoter and pEFRenilla-Luc. The results showed that either p65 or β-catenin significantly increased the transcription of FOXM1 in primary CML CD34⁺ cells (Fig. 3B, left). Moreover, cotransfection of constructs of pTOPflash (or pFOPflash) and pEFRenilla-Luc in combination with p65 or FOXM1 demonstrated that either p65 or β-catenin remarkably increased the TCF/LEF–dependent transcription in primary CML CD34⁺ cells (Fig. 3B, middle). In separate cotransfection experiments, either β-catenin or FOXM1 significantly increased NF-κB–dependent reporter gene transcription in primary CML CD34⁺ cells (Fig. 3B, right).

In contrast, cotransfection experiments with p65 siRNA duplexes in combination with constructs of Luc-FOXM1 promoter and pEFRenilla-Luc showed that p65 knockdown significantly decreased the transcription of FOXM1 in primary CML CD34⁺ cells (Fig. 3C, left). Cotransfection of constructs of pTOPflash (or pFOPflash) and pEFRenilla-Luc in combination with p65 siRNA duplexes revealed that p65 remarkably decreased the TCF/LEF–dependent transcription in primary CML CD34⁺ cells (Fig. 3C, right). In addition, FOXM1 knockdown significantly reduced NF-κB– and TCF/LEF–dependent reporter gene transcription in primary CML CD34⁺ cells (Fig. 3D). Similarly, β-catenin knockdown significantly decreased NF-κB–dependent reporter gene transcription and Luc-FOXM1 promoter transcriptional activity in primary CML CD34⁺ cells (Fig. 3E).

Next, we evaluated the effect of niclosamide on NF-κB and β-catenin in primary CML CD34⁺ cells. The purified CML CD34⁺ cells were incubated with control or niclosamide ± imatinib for 24 hours, levels of p65, FOXM1, and β-catenin were detected by IB. The results showed that niclosamide alone or combined with imatinib elicited a robust suppression in FOXM1 and p65 levels (Fig. 3F). Furthermore, niclosamide decreased levels of p65, FOXM1, and β-catenin by immunofluorescence staining examination (Fig. 3G).

Taken together, these results suggest that the positive feedback loop formed by p65 and FOXM1/β-catenin is active in human CML LSCs, and that niclosamide may abrogate the interplay.

We next ascertained the effect of niclosamide on survival and growth of CML primitive stem/progenitor cells. Purified CML CD34⁺ cells were labeled with CFSE (25), cultured with niclosamide, then stained with Annexin V–PE for flow cytometry analysis. Niclosamide greatly induced apoptosis in the CFSE^max CML CD34⁺ cell population but not in the CFSE^max NBM CD34⁺ cell population (Fig. 4A). Niclosamide may preferentially kill quiescent CML stem/progenitor cells, sparing quiescent NBM HSCs.

We employed CFC/replating and LTC-IC assay to assess whether niclosamide affected the self-renewal capacity of CML LSCs. The results showed that niclosamide alone or combined with imatinib treatment decreased CFC/replating ability and the number of LTC-IC–derived colonies in CML CD34⁺ cells but not NBM CD34⁺ cells (Fig. 4B and C).

We further quantify the frequency of CML LSCs with limiting dilution LTC-IC assay. Niclosamide reduced the frequency from 1/104 (control) to 1/457 (niclosamide). Combination treatment of niclosamide and imatinib substantially reduced the LT-HSC frequency to a greater degree (Fig. 4D; Supplementary Table S3). Therefore, niclosamide alone or combinational treatments of niclosamide and imatinib inhibited the self-renewal capacity of CML LSCs and reduced the CML LT-HSC frequency.

To evaluate the long-term ex vivo effect of niclosamide on survival of CML CD34⁺ cells, purified human CML CD34⁺ cells were exposed to niclosamide, then intravenously injected into NOG mice (Fig. 5A). The proportion of human CD45⁺ cells in the bone marrow and spleens of NOG mice was analyzed by flow cytometry 10 weeks after transplantation. Niclosamide significantly reduced the engraftment of human CML CD45⁺ cells in murine bone marrow (Fig. 5B) and spleens (Fig. 5C). In addition, niclosamide lowered the human diverse myeloid lineages in the NOG murine bone marrow (Fig. 5D) and spleens (Fig. 5E) as detected by flow cytometry after staining with antibodies of CD34, CD33, CD11b, CD14, CD19, and CD3, respectively. These results suggest that ex vivo niclosamide treatment inhibits the long-term engraftment of primary human CML CD34⁺ cells.

To investigate the in vivo effect of niclosamide on CML LSCs, we used a human BCR-ABL gene–driven CML mouse model (27). The CML mice were treated with placebo, parenteral p-niclosamide, imatinib, and the combination for 2 weeks (Fig. 6A). p-Niclosamide alone or in combination with imatinib significantly prolonged the overall survival of CML mice as compared with the placebo group (Fig. 6B). The populations of leukemic BCR-ABL-GFP⁺ WBC and bone marrow myeloid cells (Gr-1⁺Mac1⁺) were greatly reduced in the mice receiving p-niclosamide or imatinib, and was further reduced in the mice receiving combinational treatment (Fig. 6C). CML mice receiving p-niclosamide, imatinib, or the combination showed reduction in size and nodule in spleen than those in placebo group (Supplementary Fig. S4A). Pathology of lungs revealed decreased infiltration of leukemic cells in the CML mice receiving p-niclosamide, imatinib, or the combination (Supplementary Fig. S4B).

Flow cytometry analysis indicated that p-niclosamide alone or with imatinib significantly lowered the number of primitive Lin⁻Sca1⁺Kit⁺ (LSK) cells, LT-HSC (Flt3⁻CD150⁺CD48⁻), ST-HSC (Flt3⁻CD150⁻CD48⁻) in the bone marrow (Fig. 6D and E) and spleens (Supplementary Fig. S4C and S4D) in CML mice.

Sorted LSK cells from bone marrow displayed significant reduction in mRNA levels of Rela, Foxm1, and Ctnnb1 in the mice receiving p-niclosamide alone or with imatinib (Fig. 6F). In vivo limiting dilution analysis of LSCs further showed that p-niclosamide alone prevented the engraftment of GFP⁺ leukemia cells in the secondary recipient mice at 16 weeks (Fig. 6G and H). p-Niclosamide alone or combined with imatinib significantly decreased CML LSC frequencies (Fig. 6I; Supplementary Table S4).

Elucidation of survival and the self-renewal regulation mechanism of CSCs is of importance for targeting CSCs. In this study, we documented that NF-κB physically interacted with FOXM1, and was recruited to the FOXM1 gene promoter to increase the transcription of FOXM1. Reciprocally, FOXM1 and β-catenin positively regulated NF-κB expression and its transcriptional activity in CML LSCs. Disrupting the interplay of NF-κB and FOXM1/β-catenin by niclosamide treatment significantly induced apoptosis and reduced in vitro LSC self-renewal capacity in human CML CD34⁺ cells. Niclosamide prolonged survival in a BCR-ABL–driven CML mouse model and decreased the in vivo burden of CML LSCs.

Oncogene activation in LSCs may cause NF-κB pathway activation (15, 29). LSCs may also produce autocrine TNFα to support their own survival via the NF-κB pathway because increased TNFα was detected in the serum of CML patients (15). We discovered aberrant overexpression of intracellular FOXM1, β-catenin, and NF-κB in human CML primitive CD34⁺ cells as compared with NBM CD34⁺ cells. The differential expression may offer a rationale to target LSCs.

FOXM1 is widely overexpressed in human tumors, and plays a critical role in cell cycle, DNA replication, mitosis, and genomic stability (16). The association of FOXM1 and β-catenin hints that FOXM1 may be involved in the regulation of self-renewal of CSCs. FOXM1 is essential for the maintenance of HSCs (30). Mechanistically, loss of FOXM1 downregulates cyclin-dependent kinase inhibitors (e.g., p21, p27) by directly suppressing the expression of the gene encoding NURR1 (30). Given that LSCs share properties with normal HSCs, FOXM1 might control LSCs in CML.

Why FOXM1 is overexpressed in malignant hematologic cells is not fully understood. Our results revealed that NF-κB activation might execute three layers of regulation of FOXM1: increasing FOXM1 nuclear translocation, direct binding to the FOXM1 gene promoter to increase transcription, and facilitating the stability of FOXM1.

Although the interplay of p65 and β-catenin is controversial, more studies have supported a cooperative contribution of NF-κB and β-catenin in tumorigenesis and metastasis (31, 32). Simultaneous activation of both β-catenin and NF-κB signaling pathways but neither alone is required for the enhanced CSC phenotypes (33). These findings support a connection between NF-κB and β-catenin. Concordantly, reports have shown the complex crosstalk between NF-κB and WNT/β-catenin in a mouse model of smoke-induced inflammation with lung cancer growth (32). Mice lacking myeloid RelA/p65 displayed tumor growth delay on inhibition of the WNT/β-catenin pathway (34). Our findings provide a mechanistic explanation for such a connection. We found that NF-κB also positively regulates β-catenin by increasing its protein level. NF-κB may increase β-catenin protein stability because we found that pharmacologic inhibition of NF-κB by niclosamide increased ubiquitin-mediated degradation of β-catenin.

Conversely, activation of a canonical WNT/β-catenin pathway resulted in the simultaneous translocation of p65 and β-catenin. Ectopic expression of β-catenin increased p65 protein level, but silencing β-catenin decreased p65 protein level and transcriptional activity. Crosstalk between β-catenin and NF-κB may involve other molecules besides FOXM1. For instance, Schön and colleagues (35) demonstrated that β-catenin regulates NF-κB probably via TNFRSF19, a β-catenin target gene in colorectal cancer cells. Likewise, in our study, forced expression of β-catenin increased p65 protein level, whereas silencing β-catenin decreased p65 protein level and transcriptional activity.

The interplay of FOXM1 and β-catenin in glioma cells was demonstrated by Zhang and colleagues (16). Consistent with that report (16), we confirmed the interdependence of FOXM1 and β-catenin in CML cells. WNT3A stimulation triggered coupled nuclear translocation of β-catenin and FOXM1. Silencing β-catenin lowered FOXM1 expression, while forced expression of β-catenin increased FOXM1 level. Reciprocally, manipulation of FOXM1 positively affected β-catenin levels.

Taken together, we discovered the positive feedback loop of p65/FOXM1/β-catenin in CML LSCs. Therefore, targeting this p65/FOXM1/β-catenin regulatory loop may be a promising approach to eliminate CML LSCs (a proposed model, Fig. 6J).

Niclosamide has been an oral anthelmintic drug used to treat tapeworms for approximately 50 years (17). Recent studies from different groups including us have independently shown that niclosamide is effective against CSCs in acute myeloid leukemia, ovarian carcinoma, breast cancer, glioblastoma, and colon cancer (36–39). The underlying mechanisms may involve inactivation of pathways of NF-κB and WNT/β-catenin and increased ROS production in CSCs (17, 37, 40, 41). Niclosamide lowers the levels of Dishevelled-2 (DVL2) and LRP6, components of the WNT pathway (40, 41). Inhibition of NF-κB niclosamide involves Tak1 and IKK in the NF-κB pathway (17). In the current study, we discovered that niclosamide and its water-soluble derivative downregulated, and thereby disrupted the regulatory loop of p65/FOXM1/β-catenin in human CML CD34⁺ cells and in LSCs of CML mice, respectively. Niclosamide treatment decreased the survival and self-renewal capacity of human CML LSCs in vitro and in long-term engraftment in vivo. In addition, parenteral p-niclosamide administration significantly prolonged the survival of human BCR-ABL gene–driven CML mice and decreased the in vivo LSC frequency in the bone marrow and spleen of CML mice. Our findings in CML are consistent with previous reports in diverse types of cancer such as ovarian cancer and acute myeloid leukemia (17, 42), and shed new light on its antineoplastic mechanism.

In short, our findings provide new insight into the self-renewal regulatory network of CML LSCs and offer a rational approach to eliminate LSCs in CML. These results warrant further clinical study of parenteral p-niclosamide in CML.

No potential conflicts of interest were disclosed.

Conception and design: B. Jin, J. Pan

Development of methodology: B. Jin, J. Pan

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): B. Jin, C. Wang, J. Pan

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): B. Jin, C. Wang, J. Pan

Writing, review, and/or revision of the manuscript: B. Jin, J. Pan

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K. Ding

Study supervision: J. Pan

Other (provide specimens of CML patients): J. Li, X. Du

The authors thank Dr. Hans Clever (Department of Immunology, University Hospital Utrecht, the Netherlands) for generously providing plasmid encoding FOXM1 gene promoter–driven luciferase (LUC) reporter. The authors also thank Dr. Sai-Ching J. Yeung (The University of Texas MD Anderson Cancer Center, Houston, TX) for critical reading of the manuscript.

This study was supported by grants from National Natural Science Funds (81025021, U1301226, 81373434, and 91213304; to J. Pan), the National Basic Research Program of China (973 program grant 2009CB825506; to J. Pan), the Research Foundation of Education Bureau of Guangdong Province, China (grant cxzd1103; to J. Pan), the Research Foundation of Guangzhou Bureau of Science and Technology, the Fundamental Research Funds for the Central Universities (to J. Pan), the Natural Science Foundation of Guangdong province (grant 2015A030312014; to J. Pan), and the Fundamental Research Funds of the State Key Laboratory of Ophthalmology (grant 2015QN07; to B. Jin).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.