Background
The expression of the RNA-binding protein Musashi-1 (MSI1) is elevated in a variety of human cancers, including glioblastoma, breast, colon and lung cancers [
1‐
10], with higher levels corresponding to poor prognosis [
3‐
5,
10‐
12]. Msi1 was first identified in
Drosophila where it plays a role in neural development and asymmetric cell division in the adult sensory organ [
13]. Subsequently, Msi1 homologs were identified in other species, with higher levels in stem and undifferentiated cells [
1,
2,
14‐
17]. Musashi-1 typically plays a role in post-transcriptional regulation of target mRNAs [
18‐
22]. Up-regulation of MSI1 in cancers appears to associate with elevated Notch/Wnt signaling, as MSI1 targets
Numb [
22,
23] and
APC (adenomatous polyposis coli) [
19] are negative regulators of Notch and Wnt signaling, respectively [
24,
25].
CDKN1A (
P21), a negative regulator of cell cycle progression, is also a direct MSI1 target [
21]. In all three cases, MSI1 blocks target mRNA translation. Knocking down MSI1 using siRNA [
3], miRNA [
26] and a small molecule inhibitor [
27] led to decreased xenograft tumor growth. Taken together, these results point to MSI1 as a potential therapeutic target.
Our previous study identified (−)-gossypol as a small molecule inhibitor of MSI1 that reduced cancer cell proliferation and xenograft growth [
27]. More recent screening in our lab using an FP assay revealed more potent and/or specific inhibitors of MSI1. One inhibitor with a Ki of 12 ± 2 nM against full length MSI1 was gossypolone (Gn), it had a higher affinity than (−)-gossypol (Ki = 476 ± 273 nM) [
27]. Gn also showed similar affinity towards Musashi-2 (MSI2) in FP assay (Ki = 7.0 ± 0.3 nM against full length MSI2). MS12 is another Musashi family member that plays both redundant and independent roles as MSI1 in neural stem cells [
28,
29]. In cancer, MSI2 expression is elevated in hematologic malignancies [
30‐
36], colorectal adenocarcinomas [
37], lung [
38], pancreatic cancers [
39‐
41], and glioblastoma [
42]. MSI1 and MSI2 share sequence and structure similarity, especially their N-terminal RNA recognition motifs (RRMs). The residues that recognize r(GUAGU) are highly conserved between MSI1 and MSI2 [
43]. Thus, Gn can potentially be used as a MSI1/2 dual inhibitor.
Gn is a major metabolite of gossypol [
44], and is oxidized in the liver by P450 enzyme [
45]. Gn shares similar biological activities as gossypol [
46‐
52], including as an inhibitor of Bcl-2 family with a Ki of 0.28 μM toward Bcl-xL [
49]. However, in colon cancer cell assays, the same concentration of Gn was less potent than (−)-gossypol [
27].
To address this problem, we introduce a new liposome-based Gn nanocarrier. Liposomes have long been used as nanocarriers for targeted cancer therapy and have demonstrated biocompatibility and controlled drug release in previous studies [
53‐
56]. Particularly, compared with unmodified liposomes, some PEGylated liposomes were reported to be less entrapped by reticuloendothelial cells and lead to enhanced drug delivery to solid tumors in vivo [
57‐
59]. In the present study, PEGylated liposomes were used to improve the bioavailability of Gn as well as to achieve tumor-targeted delivery and controlled release of Gn, which enhances its overall biocompatibility and drug efficacy in vivo.
Methods
Cell culture and reagents
Human colon cancer cell lines HCT-116, HCT-116 β/W and DLD-1, are as described by Lan et al. [
27] and tested for mycoplasma contamination [
60] before use.
Gossypolone (Gn) was prepared as previously described [
61]. (3, 4-Dimethoxyphenyl)methanimine gossypol (MP-Gr) was synthesized from gossypol [
27]. The Gn and MP-Gr powder were dissolved in DMSO at 20 mM as stock solutions. L-α-phosphatidylcholine (EPC) and 1,2-distearoyl-sn-glycero-3-phosphoethanolamine-N-[methoxy(polyethylene glycol)-2000] (PEG-DSPE) were purchased from Avati Polar Lipids, Inc. (Alabama, USA) . DiR (1,1′-dioctadecyl-3,3,3′,3′-tetramethylindotricarbocyanine iodide) was purchased from Invitrogen (Carlsbad, CA).
Cell growth, MTT, colony formation, western blot analysis, Caspase-3 activation assay, RT-PCR and quantitative real-time PCR were carried out according to our previous publications [
27,
62‐
66]. Protein expression and purification, FP competition assay, SPR, NMR, and Wnt luciferase reporter assay were carried out as previously described [
27]. The primer sequences, the primary and the secondary antibodies used were from Lan et al. [
27]. Live cell imaging was carried out using EVOS FL Auto Cell Imaging System (Invitrogen, Thermo Fisher Scientific) and images were cropped and processed using ImageJ (NIH).
For all cell based studies, the DMSO concentration was 0.1% except where indicated below (for CETSA).
Computational modeling
The AutoDock4.2.6 program [
67] was used for docking calculations. The three-dimensional structure of Musashi1’s RBD1 in complex with RNA (PDB: 2RS2) was used to dock the gossypolone compound at the MSI1 RBD1 - RNA interface. A grid box of size 40*44*56 Å with 0.375 Å spacing centered around residue F23 was used for docking. A total of 200 docking runs were carried out using the Lamarckian genetic algorithm. The docked conformation with lowest energy was selected as the final predicted binding mode.
Cellular thermal shift assay (CETSA)
CETSA was carried out according to Molina et al. [
68]. For Gn dose CETSA, the HCT-116 β/W cell lysates with different concentrations of Gn were incubated for 30 min and heated individually at 52 °C for 3 min (StepOnePlus™ Real-Time PCR System, Applied Biosystems/Life Technologies) followed by cooling for 3 min at 25 °C. The soluble fractions were analyzed by western blot. The concentration of DMSO in each sample is 3.3%. Musashi-1 antibody used for CETSA was anti-MSI1 (01–1041, Millipore, Billerica, MA). Western band intensities were measured using Image Studio Ver 4.0 (LI-COR Bioscience, Lincoln, NE), and normalized to α-Tubulin.
Preparation and characterization of gossypolone-encapsulated liposomes (Gn-lip)
Gn-lip was formed using a mixture of Gn, EPC, PEG-DSPE, and cholesterol in chloroform, at a molar ratio of 30/85/6/9. The solution was dried under vacuum to form a thin film of Gn/carrier mixture, which was then dissolved in DPBS to produce Gn-encapsulated liposomes. Blank liposomes were prepared similarly without the addition of Gn. To prepare the samples for TEM image, both Gn-lip and blank liposomes were diluted in DI water, respectively. The suspensions were applied to a grid and negatively stained by 4% uranyl acetate. Images of liposomes were acquired using FEI Tecnai G2 Polara 200 kV TEM (FEI Company, OR, USA). The size distribution and zeta potential of liposomes in DI water were measured at 25̊ C using a Malvern instrument (Nano-ZS90, Malvern, UK). The size stability of Gn-lip for 3 months was investigated at 4 °C. The drug loading efficiency (DLE%) and drug loading content (DLC%) of Gn were determined using filtration method. Gn-lip solution was filtered using an ultra centrifugal filter unit (MWCO 3000 Da, Amicon®, Merck KGaA, Germany). The concentration of free drug in the filtrate was determined using a UV-vis spectrophotometer. The DLE% and DLC% of Gn were calculated as follows: DLE% = (weight of loaded Gn ÷ total weight of input Gn) × 100%; DLC% = (weight of loaded Gn ÷ total weight of Gn-lip) × 100%.
The viabilities of HCT-116 and DLD-1 cells in the presence of free Gn or Gn-lip were determined using MTT-based assay, as described previously.
Biodistribution of DiR-loaded liposomes in tumor-bearing mice
NOD.CB17-Prkdcscid (SCID) mice were purchased from Harlan laboratory (Indianapolis, IN) and bred at the University of Kansas Animal Care Unit. The in vivo tumor-specific distribution of liposomes was studied using DiR, a near-infrared (NIR) fluorescent dye. DiR-loaded liposome was formed using a mixture of DiR, EPC, PEG-DSPE, and cholesterol in chloroform, at a molar ratio of 1/85/6/9. The solution was dried under vacuum to form a thin film of DiR/carrier mixture, which was then dissolved in DPBS to produce DiR-loaded liposomes. Two DLD-1 tumor-bearing SCID mice were used for in vivo fluorescence imaging according to our previous studies with modifications [
69,
70]. Briefly, 10 nmol DiR-loaded liposome in 200 μL was intravenously (
i.v.) injected into one mouse; 200 μL 10 nmol DiR ethanol/water (1:4
v/v) mixed solvent as the control was
i.v. injected into another mouse. At different time points, the biodistributions of DiR in both mice were observed using a Carestream Molecular Imaging System (Carestream Health, Rochester, NY), with excitation at 750 nm and emission at 830 nm using an exposure time of 60 s. Mice were euthanized at 72 h post-injection by CO2 overdose and confirmed by cervical dislocation as recommended by the Panel on Euthanasia of the American Veterinary Medical Association. Organs and tumors of mice were obtained for further ex vivo fluorescence imaging. The fluorescence intensities of tumors at different time point in vivo, and tumors and livers ex vivo, were quantified using the ‘Image Math’ function of Carestream Molecular Imaging Software (Carestream Health, Inc). To produce calibration curves for DiR-lip and free DiR, 50 μL DPBS containing different amount of DiR-lip or free DiR was added in each well of a 96-well plate, followed by in vitro imaging using the same settings with that of the in vivo imaging. The calibration curves were produced using the fluorescence intensity of each well. The amount of dye in each tissue was calculated using its fluorescence intensity and the corresponding calibration curve. The fluorescence percentage of injected dose per gram (%ID/g) of each tissue was calculated using the following formula:
$$ \%\mathrm{ID}/\mathrm{g}=\frac{{\mathrm{M}}_{\mathrm{DiR}}\kern0em }{\mathrm{ID}\times {\mathrm{W}}_{\mathrm{Tissue}}}\times 100\% $$
in which M
DiR is the amount (nmol) of DiR in the tissue, ID is the injected amount (nmol) of DiR, and W
Tissue is the weight (g) of tissue.
In vivo drug efficacy of Gn in DLD-1 tumor-bearing nude mice
The in vivo experiments were carried out with 5 to 6-week-old female athymic NCr-nu/nu nude mice purchased from the Harlan laboratory (Indianapolis, IN). After alcohol preparation of the skin, mice were inoculated subcutaneously with 200 μL DLD-1 cell suspension (1 × 106 cells) in plain DMEM on both flanks using a sterile 23-gauge needle. When tumors reached 40 mm3 on average, the mice were randomized into 2 groups. Group 1 (10 mice, 20 tumors) was given vehicle as the control; group 2 (5 mice, 10 tumors) was given 10 mg/kg Gn-lip. Gn-lip was administrated intravenously 2 times weekly for 3.5 weeks. Tumor size and body weight of each mouse were measured twice a week, and tumor volumes were determined as a × b2/2, in which a and b represent the longest and shortest diameter of the tumors, respectively. All animal experiments were carried out according to the protocol approved by the Institutional Committee for the Use and Care of Animals of University of Kansas.
Statistical analysis
Using Prism 5.0 software (GraphPad Prism), one-way ANOVA and t-Test were used to analyze the in vitro data, two-way ANOVA was used to analyze the in vivo data. A threshold of P < 0.05 was defined as statistically significant.
Discussion
In this study, we sought to identify additional small molecule inhibitors of Musashi-1(MSI1). Such inhibitors can downregulate Notch/Wnt signaling and block cell cycle progression through MSI1. To this end, we used FP-based screening and identified Gn as a potential MSI1 inhibitor. We further confirmed Gn binding to MSI1 using SPR and using NMR, identified the amino acids in the RNA-Recognition Motif 1 that were involved in the binding. Using colon cancer cell lines, we showed that Gn inhibited cell growth, induced autophagy, inhibited Notch/Wnt signaling in these cells and led to apoptotic cell death. However, compared to 10 μM (−)-gossypol [
27], the same concentration of Gn was less active in cell assays. This result could be due to poor water-solubility of Gn. Therefore, we used a liposomal carrier to deliver Gn in animals. The liposome could efficiently increase the apparent solubility of Gn in water. We showed that Gn loaded liposomes induced apoptosis, inhibited tumor growth and Notch/Wnt signaling in DLD-1 xenograft model. Our study identified a new target for Gn and provided a new delivery method for this poorly bioavailable compound.
MSI1 is an RNA-binding protein that promotes cell proliferation and survival through Notch/Wnt signaling (Fig.
7d). Inhibiting MSI1 is a promising therapeutic strategy for preventing cancer cell proliferation and progression. Here we identified Gn as a more potent inhibitor of MSI1 compared to (−)-gossypol [
27] in our binding assays. Our NMR studies showed that Gn bound to the same residues as the cognate RNA. The NMR peaks of the RNA-binding residues (K93, F23 and W29) showed the most changes upon titration of Gn (Fig.
1c). This suggests that these residues are involved in the binding to Gn, consistent with our docked model. The majority of non-RNA binding residues were unaffected when titrated with Gn, suggesting that the interaction between Gn and MSI1-RBD1 tends to be more localized near the RNA binding pocket as compared to the binding event between (−)-gossypol and MSI1-RBD1 [
27]. This might explain the improved potency of Gn compared to (−)-gossypol. Additional X-ray crystallography studies will be helpful in determining the high resolution structure of the complex between MSI1-RBD1 and Gn.
Based on the sequence identity of RBDs/RRMs of MSI1 and MSI2, we tested the binding of Gn towards MSI2 and showed that Gn can disrupt the binding of MSI2 to
Numb RNA. Like MSI1, MSI2 is also a member of the Musashi family of RNA-binding protein and shares similar roles as MSI1 in stem cells [
28,
29]. In colorectal cancer initiation and maintenance, a recent study demonstrated the functional redundancy between MSI1 and MSI2 [
78]. Thus using a MSI1/MSI2 dual inhibitor such as Gn in patients with MSI overexpression suggests an improved therapeutic outcome.
The introduction of PEGylated liposomes improved the dispersion of Gn in aqueous environment, thus making it possible to produce injectable drug solutions, which is essential for a better bioavailability of Gn. As shown in the results of MTT assay (Fig.
5c), the encapsulation of Gn using liposomes did not compromise the cytotoxicity of Gn in vitro. Finally, the Gn-lip exhibited a significant tumor inhibition efficacy and a low systemic toxicity in the mice. Our work is important in that our study provides a proof-of-concept to develop the Gn-lip as a novel molecular therapy for colon cancer with MSI overexpression.
The limitation of our current inhibitors is that they are not specific to MSI1/MSI2, because Gn and the previously reported (−)-gossypol [
27] are both Bcl-2 family inhibitors as well [
46‐
52,
62,
63,
76,
79]. With Gn treatment, we saw apoptosis/autophagy induction via Bcl-2 and cell proliferation inhibition via Wnt and/or Notch signaling pathways. Our goal is to develop potent and specific MSI1/MSI2 inhibitors, and ultimately move these new inhibitors into clinical applications in the treatment of cancers with MSI overexpression. Towards this goal, our future efforts will focus on utilizing computer modeling and medicinal chemistry for identifying new chemical scaffolds that selectively inhibit MSI1/2.
Conclusions
Gn was identified as a MSI1/2 duo inhibitor in this study. It disrupted binding of MSI1 to its target mRNAs by binding to the RBD1 of MSI1. Gn inhibited colon cancer cell growth, induced autophagy, down-regulated Notch/Wnt signaling and led to apoptotic cell death. The introduction of tumor-targeted liposomes significantly improved the bioavailability of Gn, meanwhile maintaining its drug efficacy. Gn-lip has promising antitumor effects and biocompatibility in vivo, warranting further study to determine its suitability for cancer treatment.
Acknowledgements
We are grateful to OpenEye Scientific Software (Santa Fe, NM) for providing an academic license for the use of OMEGA, ROCS, MolProp, and QuacPac. We thank the NCI/DTP Open Chemical Repository (
http://dtp.cancer.gov) libraries.