Identification of thiostrepton as a novel therapeutic agent that targets human colon cancer stem cells

Accumulating evidence shows that colorectal cancer stem cells (CRSCs) are largely responsible for the metastasis and relapse of colorectal cancer (CRC) after therapy. Hence, identifying new agents that specifically target CRSCs would help improve the effectiveness of current CRC therapies. To accelerate identification of agents targeting CRSCs, the Connectivity Map (CMap) approach was used. Among the top-ranked candidates, thiostrepton, a thiazole antibiotic, was selected for further investigation because of its known tumoricidal activity. Thiostrepton could selectively induce apoptosis in CRSC subpopulations in both parental HCT-15 and HT-29 human CRC lines as well as in EMT and chemoresistant clones derived from them. Further, we investigated its inhibitory effects on the sphere- and colony-forming capabilities of the aforementioned CRC lines. The in vitro inhibition of sphere and colony formation was associated with downregulation of various modulators of the stem cell phenotype. The combination of thiostrepton and oxaliplatin eradicated both CD44+ HCT-15 and HT-29 cells more efficiently than either drug alone. FoxM1, an oncogenic transcription factor, was identified as a critical positive modulator of stemness and as the main target of thiostrepton in the CRC lines. This is the first report showing the selective killing of CRSCs by thiostrepton, which has been proposed to be a promising anti-neoplastic agent. On the basis of its synergism with oxaliplatin in killing CRSCs in vitro, if this activity is confirmed in vivo, thiostrepton may be a promising agent to be used clinically in combination with current chemotherapies to improve the efficacy of these regimens.

Colorectal cancer (CRC) is the third most common cancer and the fourth leading cause of cancer-associated deaths globally, 1 and this mortality is primarily attributed to its high incidence of metastasis. 2 CRC has a stem cell (SC) subpopulation, which is generally considered to be the source of CRC tumor metastasis. 3,4 Colorectal cancer SCs (CRSCs) were identified as a rare group of CD133 + cells present in tumors resected from patients that had high in vivo tumorigenicity in immunodeficient animals. 5,6 Other markers such as CD44, CD166, CD26, EpCAM and aldehyde dehydrogenase (ALDH) have also been found to be useful in enriching CRSCs from tumor tissues as well as established CRC cell lines. [7][8][9][10] CRSCs have also been isolated as a 'side population' by fluorescence-activated cell sorting owing to their expression of high levels of ABC-transporters which increases their ability to exclude hydrophobic dyes. 11 Additionally, CRSCs have been identified in tumor spheres formed under defined culture conditions in vitro. 12 Finally, recent studies have demonstrated a convincing link between epithelial-tomesenchymal transition (EMT) and CSCs as well as the association of these processes with CRC progression and therapeutic resistance. 13,14 As the essential roles of CRSCs in therapeutic resistance of CRC have been characterized, strategies targeting this subpopulation have been developed to improve therapeutic outcomes for CRC patients. One of the methods that can be used to identify CRSC targets is by evaluating signaling pathways specific to CSC self-renewal and/or maintenance of stemness and conduct studies to analyze the potential of diminishing these new targets for effective eradication of CRSCs. On the basis of this strategy, novel leads worthy of further investigation as potential CRC drugs have been found by targeting Wnt, Hedgehog, and Notch pathways. 15,16 For example, LGK974, a specific inhibitor for Porcupine, a membrane-bound O-acyl-transferase required for palmitoylation and secretion of Wnt ligands has been shown to suppress in vitro and in vivo growth of Wnt-dependent tumor cells. 17 Vismodegib, a Hedgehog pathway inhibitor, has been approved by the US FDA for treating certain forms of basal cell carcinoma (for a review, see Sandhiya et al. 18 ). Moreover, antitumor efficacy and prolonged disease stabilization has been observed in patients with advanced solid tumors co-treated with RO4929097, a γ-secretase/Notch signaling pathway inhibitor, and cediranib, a VEGF receptor tyrosine kinase inhibitor. 19 Strategies specifically targeting CSC surface markers such as CD44 have been explored successfully in leukemia and breast cancer in pre-clinical models. 20,21 Additionally, salinomycin, identified originally as a drug with selective toxicity against breast CSCs, 22 was shown to be effective against a variety of other CSCs. 22 Other studies have demonstrated that IGF-1 inhibition, 23 HER-2 signaling blocking, 24 IL-4 neutralization, 25 or BMP4 addition 26 inhibit in vivo tumorigenicity of CRSCs and the latter two treatments can further sensitize CRSC-derived tumors to conventional chemotherapeutic agents. Recently, mTOR suppression by rapamycin or PP242 has been shown to decrease stemness properties as well as suppress the increase in CSC populations that can be induced by 5-FU and oxaliplatin treatment of CRC cells. 27 Despite the discovery of these promising agents, more drugs (especially the ones already approved for clinical use) are still needed to expand the arsenal of agents that can combat this rare but deadly subpopulation.
To identify new compounds, we used the Connectivity Map (CMap) approach, which has been shown to be effective for drug discovery and development. 28 Among several top-ranked candidates identified, thiostrepton, a thiazole antibiotic, was selected for further investigation because of its welldocumented tumoricidal activity. 29,30 In the present study, we demonstrated that thiostrepton selectively kills the SC subpopulations in HCT-15 and HT-29 human CRC lines as well as EMT and chemoresistant clones derived from them. Current work also showed that thiostrepton triggered apoptosis in HCT-15 and HT-29 CRSCs and inhibited colony and sphere formation, with downregulation of various stemness regulators. Interestingly, synergistic killing effects with thiostrepton and oxaliplatin were noted against HCT-15 and HT-29 cells as well as their SC subpopulations. Finally, we identified FoxM1 as a critical positive modulator of stemness in the aforementioned CRC lines and the main target of thiostrepton. Together, these findings warrant further investigation of the therapeutic potential of thiostrepton as an adjuvant treatment with CRC chemotherapy once its in vivo antitumor efficacy is confirmed.

Results
Thiostrepton reduces the proportion of CRSCs. To identify drug candidates with selective toxicity against SC populations present in human CRC lines, we extracted an anti-CSC gene signature from the study of Suvà et al. 31,32 and used it as the input signature to query CMap 2.0. Among the top 10 drugs that shared positive connectivity with the anti-CSC signature, thiostrepton was ranked as the top one ( Figure 1). MTT assays were then carried out to evaluate the cytotoxicity of thiostrepton on HCT-15, HT-29 and HCT-116 CRC lines. As shown in Table 1, thiostrepton killed CRC cells more efficiently than oxaliplatin, a first-line treatment for CRC. To investigate the effect of thiostrepton on human CRSCs, Figure 1 Identification of thiostrepton as a potential anti-cancer SC agent. (a) The top 10 candidates with a positive enrichment score were shown here. The 'mean' stands for the average connectivity scores of microarray profiles of various instances (e.g. HL60 cell line treated with thiostrepton) and the 'n' stands for the number of microarray profiles of a drug (e.g. thiostrepton). The 'enrichment' score was ranged from 1 to − 1 and delivers the enrichment of the instances of a drug among the 6100 instances of all drugs. (b) The colored bar with three colors represents 6100 instances in the CMap 2.0 database. Green means positive connectivity, red means negative connectivity, and gray means no connectivity. The dark lines in the green zone of the bar represent the instances of thiostrepton. It shows that thiostrepton is ranked as no. 16,114,228, and 241 compared with the other instances in the CMap 2.0 database flow cytometry was used to assess its influence on cell surface expression of the CRSC marker CD44 in HCT-15 and HT-29 cells. As shown in Figure 2a, thiostrepton treatment reduced CD44 + subpopulations in both CRC lines. In contrast, a significant increase in these cells was observed after treatment with oxaliplatin (Supplementary Figure S1). Because earlier studies have shown that CSC subpopulations could be enriched by EMT induction 14,33 or incubation in suboptimal concentrations of cytotoxic drugs, 23,34 a Snail-overexpressing HCT-15 clone (Snail OE) and an oxaliplatin-resistant HT-29 subline (r29, Supplementary Figure S2) were established. As expected, 10 and 25% increases in CD44 + subpopulations were observed, respectively, for the cell lines (Figures 2b and c). These clones were then used to evaluate the efficacy of thiostrepton against CRSCs. In agreement with the above findings, thiostrepton reduced the CD44 + subpopulation in the Snail OE clone, while oxaliplatin treatment caused an increase in CD44 + cells in similar clone (Figure 2b). A dose-dependent decrease in the CD44 + subpopulation in r29 cells was also noted after thiostrepton treatment (Figure 2c). To confirm the toxicity of thiostrepton against CRSCs, flow cytometry was used to measure the proportion of the CD44 and CD133 double positive (CD44 + /CD133 + ) HT-29 cells that remained after the parental cells were treated with oxaliplatin or thiostrepton, because these cells are considered CRSCs. 35 The CD44 + / CD133 + subpopulation in HT-29 cells was increased after oxaliplatin treatment and was decreased after thiostrepton incubation ( Figure 2d). Additionally, the freshly isolated CD44 + /CD133 + HT-29 cells were also effectively killed by thiostrepton ( Figure 2e).
Thiostrepton induces apoptosis of CRSCs. Because thiostrepton has been reported to inhibit growth and induce apoptosis in various human cancer cells, 30,36,37 the effect of this drug on regular as well as SC-enriched CRC cells was next investigated. As can be seen, a dose-dependent increase in the sub-G1 fraction in HT-29 and r29 cells was observed after thiostrepton treatment ( Figure 3a). Annexin V and 7-AAD double-positive apoptotic populations in the two CRC lines were also increased in a dose-dependent manner by this drug (Figure 3b). Furthermore, apparent cleavage of PARP in HCT-15 and its Snail-overexpressing clone as well as in HT-29 and r29 cells was detected after thiostrepton treatment ( Figure 3c). These results clearly demonstrated that thiostrepton induces apoptosis of CRSC-enriched clones derived from both HCT-15 and HT-29 cells.
Thiostrepton suppresses sphere and colony formation, as well as the expression of some stemness-related proteins in human CRSCs. Because thiostrepton reduced CD44 + subpopulations present in human CRC lines, we assessed whether other relevant phenotypes for CRSCs were also suppressed by this antibiotic. Sphere formation of HCT-15 and HT-29 cells in the continuous presence of thiostrepton or oxaliplatin was examined. The degree of tumor spheres formed by both CRC lines was decreased by thiostrepton treatment. A marked reduction in HCT-15 tumor sphere formation after oxaliplatin treatment was also noted ( Figure 4a). The influence of short exposure of thiostrepton on CRSCs was evaluated by colony formation of HCT-15 and HT-29 cells after they were treated with oxaliplatin and/or thiostrepton for 6 h prior to seeding. As shown in Figure 4b, while oxaliplatin treatment of both CRC lines only reduced the size of colonies, short exposure to thiostrepton drastically reduced their colony numbers. These results suggested that thiostrepton, unlike oxaliplatin, was able to eradicate clonogenic subpopulations present in these human CRC cell lines. To elucidate the molecular mechanisms underlying the suppressive effects of thiostrepton on CRSC self-renewal, immunoblotting was used to analyze the protein levels of various markers of stemness, including CD44, ALDH1, Oct4, Nanog, Bmi1, Snail, and Twist in Snail OE and r29 clones after thiostrepton treatment. To no surprise, protein levels of all the aforementioned factors were decreased by thiostrepton (Figures 4c and d). Because earlier studies have reported that some of the anticancer effects of thiostrepton could be attributed to its proteasome inhibitory activity, 38 we assessed the contribution of this activity to the stemnesssuppressing effects of thiostrepton by evaluating the stemness properties of HCT-15 and HT-29 cells after treatment with the proteasome inhibitors MG132 and Carfilzomib (Kyprolis). Strikingly, neither inhibitor decreased the CD133 + CD44 + subpopulations in HCT-15 and HT-29 cells (Supplementary Figure S3). Moreover, despite the reduction in protein levels for stemness markers such as Oct4, Nanog, and Bmi1 by both inhibitors, neither agent decreased SC marker protein levels in the CRC lines as effectively as thiostrepton (Supplementary Figure S4). Collectively, the above findings suggested that the proteasome inhibitory activity of thiostrepton is not a key contributor to its stemnessreducing effects on CRSCs.
Thiostrepton synergizes with oxaliplatin in killing CRSCs. Having demonstrated that thiostrepton alone could reduce the SC subpopulations in HCT-15 and HT-29 cells, we asked what effect the combination of thiostrepton with oxaliplatin would have on CRC cell killing, especially the CD44 + subgroups. As can be seen, while at least 45% of surviving cells were CD44 + after oxaliplatin treatment, CD44 + ones accounted for less than 15% and 30% of the viable HCT-15 and HT-29 cells, respectively, after being incubated with thiostrepton. Even greater reductions in CD44 + cells in HCT-15 and HT-29 lines were detected when the two drugs were combined (Figure 5a). Also, established tumor cell spheres generated from HCT-15 cells were reduced by treatment with this drug combination even though sphere growth was enhanced by oxaliplatin alone (Figure 5b).  In agreement with the earlier observations made by others, mRNA levels of FoxM1 in these cells drastically decreased after a short exposure (6 h) to thiostrepton (Figure 6a), subsequently, reduced protein level could be observed after 48 h treatment ( Figure 6b). These results support FoxM1 as the main target of this antibiotic in human CRC cells because direct transcription suppression on FoxM1 is the most likely explanation for the effects observed after short-term treatment with this thiazole drug. We examined whether HCT-15 SC subpopulations (CD44 + ) expressed elevated levels of FoxM1, possibly affecting their response to thiostrepton. Higher FoxM1 protein levels were found in CD44 + HCT-15 cells (Figure 7a), and this subpopulation was more susceptible to thiostrepton (Figure 7b). After short-term (6 h) treatment with this drug, CD44 + HCT-15 cells formed colonies at a slower rate and they were smaller (Figure 7c), demonstrating the selective cytotoxicity of thiostrepton against CRSCs. These data indicate that elevated expression of FoxM1 in HCT-15 CRSCs may account for their increased sensitivity to thiostrepton.
FoxM1 is required for maintenance of the CS subpopulation and the susceptibility of CRSCs to thiostrepton in HT-29 and HCT-116 cells. Having found that FoxM1 may positively modulate the stemness in HCT-15 cells and be the primary target mediating their sensitivity to thiostrepton, we dissected the role of this transcription factor in the maintenance of CD44 + subpopulations in two other CRC lines as well as evaluating their response to thiostrepton. Significant decreases in CD44 expression levels were found in the FoxM1-knockdown clones derived from HT-29 and HCT-116 cells (Figure 8a). Additionally, the CD44 + subpopulation in the above-mentioned FoxM1-knockdown clones was also reduced (Figure 8b). Consistent with these results, the sensitivity of these knockdown clones to thiostrepton was reduced in comparison with vector control cells (Figure 8c). These results suggested that FoxM1 is not only responsible for maintaining key stemness phenotypes of the CRC lines evaluated but also is the primary target for thiostrepton.

Discussion
To improve the efficacy of current treatments for CRC, novel agents that can eradicate the tenacious CRSCs present in patients need to be identified and developed. In this study, our clear demonstration that thiostrepton could synergize with oxaliplatin, a widely used CRC chemotherapeutic agent, in  killing CRSCs, strengthens the rationale for its clinical usage in CRC treatment.
To characterize the potential target of thiostrepton in CRSCs, we focused on Forkhead box M1 (FoxM1), an oncogenic transcription factor critical for various cellular functions including self-renewal of normal SCs, 39,[46][47][48] because the inhibition of this factor was responsible for the apoptotic effects of thiostrepton and another thiazole antibiotic, siomycin A. 30,49 In earlier studies, FoxM1 was proposed to be a promising therapeutic target for cancers because upregulated expression of this protein was found in most human solid tumors including CRC. 50 Subsequently, crucial roles played by FoxM1 in promoting tumor initiation, 51 invasion, 52 and metastasis 53,54 have also been reported. More interestingly, the involvement of this factor in modulating signaling pathways crucial for SC development has recently been reported by others. For example, FoxM1 is not only required for pluripotency maintenance of P19 embryonal (b) Clonogenic assays were performed using cells exposed previously to DMSO (Ctrl), oxaliplatin, or thiostrepton for 6 h as described in 'Materials and Methods'. Ten days after seeding, colonies were stained by crystal violet and their numbers were counted using Colony 1.1 software. *Po0.05 when compared with the DMSO-treated cells by Student's t-test. Total lysates prepared from (c) Snail OE and (d) r29 cells after being treated with the indicated doses of thiostrepton for 48 h were subjected to western blot analyses using antibodies against CD44, ALDH1, Oct4, Nanog, Bmi1, and Twist as probes. Tubulin signal served as a loading control carcinoma cells by regulating Oct4, 43 but also enhances tumorigenesis of glioblastoma cells by promoting the nuclear translocation of β-catenin, a critical mediator of Wnt signaling, via a direct protein-protein interaction. 55 Taken together, these findings support our data that the cytotoxic effect of thiostrepton on CRSCs is due to higher expression of FoxM1 in these cells and explain the decreases in both SC subpopulations and the susceptibility to this antibiotic observed in FoxM1-knockdown clones. However, chemosensitization resulting from FoxM1 silencing in other tumor cell types has also been reported. 56,57 More work is needed to resolve these discrepancies. Because enrichment of CSCs from various types of cancers after treatment with conventional therapeutic agents has been demonstrated, 33,34,58 strategies that reduce the expression of stemness-related genes were proposed as a way to overcome drug resistance. In this regard, the nuclear factor-erythroid 2 p45-related factor 2 (Nrf2)/antioxidant response element (ARE) pathway has been shown to be the plausible cause for the resistance of HT-29 cells to 5-FU treatment. Accordingly, silencing of Nrf2, a newly identified pluripotency gene, 54 not only inhibited the expression of its target genes involved in cytoprotection but also increased 5-FU cytotoxicity. 59 Moreover, the chemosensitization of HCT-116 cells to 5-FU by curcumin, a plant polyphenol, has recently been attributed to its ability in suppressing CSC pools. 59 Consistent with these findings, the CD44 + subpopulations in HCT-15 and HT-29 cells were dramatically increased after oxaliplatin treatment (Supplementary Figure S1), so was the sphere-forming ability of the former (Figure 5b). On the contrary, combined treatment with oxaliplatin and thiostrepton markedly diminished not only the CD44 + subpopulations in two CRC lines but also HCT-15 sphere formation ( Figure 5). Collectively, our data suggest that thiostrepton, via uncharacterized mechanisms, reverses the SC-enriching effect of oxaliplatin and sensitizes CRSCs to this commonly used anti-CRC drug.
In conclusion, thiostrepton was identified as a novel agent with selective in vitro cytotoxicity against CRSCs present in HCT-15, HT-29, and HCT-116 cell lines. However, to enable the clinical use of this thiazole antibiotic as an adjunct in CRC treatment, animal studies are required to assess the in vivo efficacies of thiostrepton alone or in combination with oxaliplatin and/or other clinically available agents in suppressing tumor growth and tumor relapse after treatments are stopped as well as the safety of this drug. Continued research to dissect the cytotoxic mechanisms of thiostrepton on CRSCs may allow us not only to identify more novel target(s) in this rare subpopulation but also to design better treatments for CRC.

Materials and Methods
Anti-CSC signature. Differentially expressed probe sets whose adjusted P-value ≤ 0.05 and abs (fold change) ≥ 2 were selected as the anti-CSC signature by using GEO2R. 60 The up and down probe sets were used as the input signature and were uploaded to CMap 2.0 website (https://www. broadinstitute.org/cmap/). 31 Cell culture. Chemical. Thiostrepton obtained from Sigma-Aldrich (Taufkirchen, Germany) was dissolved in 100% DMSO at a final concentration of 10 mM as a stock solution. The working concentration of thiostrepton was either 5 or 10 μM with a less than 0.1% final concentration of DMSO throughout this study.
Virus preparation and generation of stable clones. The Snail expression vector was a kind gift from Dr Chiou S.H. (Institute of Pharmacology, National Yang-Ming University, Taipei, Taiwan, R.O.C.). FoxM1-knockdown stable clones were generated according to a previously described protocol. 61 Briefly, two pLKO.1 lentiviral vectors containing different shRNA sequences (5′-CGCTAC TTGACATTGGACCAA-3′ designated as sh1; 5′-GCCAATCGTTCT CTGACAGAA-3′ designated as sh2) targeting FoxM1 mRNA were purchased from the National RNAi Core Facility Platform (Nankang, Taipei) and were used to generate recombinant virus for infection. A Snail-overexpressing HCT-15 clone (Snail OE) was established by selecting virus-infected cells in a medium containing 100 μg/ml hygromycin. FoxM1-knockdown clones from HCT-116 (sh1 and sh2) and HT-29 (sh1 and sh2) cells were generated by a similar protocol except that infected cells were selected in 1.5 and 1 μg/ml of puromycin, respectively.
Isolation of CD44 + cells by fluorescence-activated cell sorting.