Scientific Rationale

Cancer stem cells (CSCs), which drive tumor growth, are known to be resistant to standard chemotherapy and radiation treatment (2–6). This resistance raises a significant unmet need to find therapies that can target CSCs within tumors because these cells are proposed to be responsible for recurrence, the primary cause of patient mortality. In principle, it would be desirable to apply high-throughput screening (HTS) technologies to facilitate the identification of chemical compounds that specifically target CSCs. Since CSCs generally comprise only a small minority within cancer cell populations, standard high-throughput cell viability assays applied to the broad populations of cancer cells cannot identify agents with CSC-specific toxicity (7). Accordingly, screening for agents that preferentially kill CSCs depends, at present, on the ability to obtain and to stably propagate highly pure CSC populations in vitro. Since CSCs cannot currently be stably propagated in culture, it has not been possible to leverage HTS to identify CSC-targeting compounds.

A novel HTS assay was recently developed to identify selective inhibitors of CSCs. The approach used stable sibling cell lines that are induced into epithelial-to-mesenchymal transdifferentiation (EMT) as a method to stably propagate CSC-enriched populations (8). The availability of isogenic control cell lines for the secondary validation assay of the primary screen, which is absent in most screens, minimized the probability of advancing spurious compound hits that are not selective for CSCs but rather target genetic differences between screened cell lines.

The objective of the experiments in this probe report is to identify chemical compounds that can selectively kill breast CSCs. The chemical compounds identified by the experiments in this report would be promising new starting points for drug candidates for anticancer therapy and would also serve as useful probes to study CSC biology, which are currently lacking.

The current state of the art for probe molecules in cell culture assays against breast CSCs is salinomycin. Salinomycin has a potency of about approximately 1 μM against HMLE_sh_ECad with approximately 10-fold selectivity over the control HMLE_sh_GFP (9). Since salinomycin shows toxicity at approximately 10 μM against HMLE_sh_GFP, this may cause gross toxicity to cells in general. A detailed comparison to existing art is provided in Section 4.1 and Table A2 in Appendix G.

2.1. Assays

A summary listing of completed assays and corresponding PubChem AID numbers is provided in Table A1 (Appendix A). Refer to Appendix B for the detailed assay protocols.

2.2. Probe Chemical Characterization

After preparation as described in Section 2.3, the probe (ML245) was analyzed by UPLC, ¹H and ¹³C NMR spectroscopy, and high-resolution mass spectrometry. The data obtained from NMR and mass spectroscopy were consistent with the structure of the probe, and UPLC analysis indicated an isolated purity of >95%. Characterization data (¹H NMR spectra and UPLC chromatograms) of the probe (ML245) are provided in Appendix E.

The solubility of the probe (ML245) was experimentally determined to be 0.07 μM in Phosphate buffered saline (PBS; pH 7.4, 23 °C) solution. Plasma protein binding (PPB) studies showed that the probe was 99.7% bound in human plasma. The probe was stable in human plasma, with approximately 100% remaining after a 5-hour incubation period. The compound was found to be stable in the presence of glutathione (GSH) with 92% remaining after 48 hours.

The stability of the probe (ML245) in PBS (0.1% DMSO) was measured over 48 hours. We noticed that the concentration of the probe fluctuated between 37.5% and 131% through the course of the assay (data not shown). We believe that this fluctuation is due to the low solubility of the probe. Thus, we decided to determine the total amount of the probe present in the well after the probe was treated with PBS alone for a given length of time. We added acetonitrile at various time points to wells containing the probe in PBS and measured the total amount of the probe. This result is shown in Figure 1. From these results, the probe seems to be stable in PBS since more than 89% is still present after 48 hours of incubation. Table 1 summarizes known probe properties as displayed in PubChem.

Figure 1

Stability Data for the Probe (ML245) in PBS Buffer (pH 7.4, 23 °C) Plus Acetonitrile Over 48 Hours.

Table 1

Summary of Known Probe Properties in PubChem.

2.3. Probe Preparation

The probe (ML245) was synthesized from 5-(hydroxymethyl)furan-2-carboxylic acid 1 in three steps (see Scheme 1). Conversion of 5-(hydroxymethyl)furan-2-carboxylic acid 1 into the dichloride with thionyl chloride, followed by amidation provided furanyl chloride 2. Alkylation with 3-chlorophenol provided the probe 5-((3-chlorophenoxy) methyl)-N-(4-methylthiazol-2-yl)furan-2-carboxamide in 47% yield over three steps.

Scheme 1

Synthesis of the Probe (ML245).

Experimental procedures for the synthesis of the probe (ML245) are provided in Appendix C.

2.4. Additional Analytical Analysis

The probe (ML245) was found to be stable in human and mouse plasma with 100% remaining after 5 hours in each case. Experimental procedures for the analytical assays are provided in Appendix D.

Probe attributes

• IC₅₀ = 0.536 μM against HMLE_sh_ECad
• IC₅₀ against HMLE_sh_GFP > 14 × IC₅₀ against HMLE_sh_ECad

3.1. Summary of Screening Results

Figure 2 displays the critical path for probe development. To explore SAR, a number of analogs were synthesized and tested. Selected results are shown in Tables 2–10 in Section 3.4.

Figure 2

Critical Path for Probe Development.

Table 2

SAR of the Amide.

Table 10

SAR on Furan Replacement (Phenyl).

A high-throughput screen (HTS) of 300,718 compounds (PubChem AID 2717) assessing the viability of the breast CSC-like population (HMLE_sh_ECad) was completed. HMLE_sh_ECad cells have undergone an EMT by introduction of a short hairpin RNA (shRNA) targeting the CDH1 gene, which encodes E-cadherin into human mammary epithelial cells (HMLEs). These cells provide a uniform population of breast CSC-like cells with which to conduct a large-scale screen (8). The signal was normalized to neutral (DMSO) and positive (Puromycin) control, and a 75% inhibition cutoff at a screening concentration of 3.75 μM average was used to define a hit. Next, 3,190 hits were identified as inhibitors of HMLE_sh_ECad. The list of active hits was narrowed by discarding compounds that hit in >10% of MLPCN assays or scored positive in prior mammalian cell toxicity screens. Compounds were ranked in order of their promiscuity in MLPCN assays and the top 2,500 compounds were requested. Of these, 2,244 compounds were available from the compound supplier (i.e., BioFocus).

These cherry-picked compounds were retested in 9-point dose in the primary assay using the cell lines HMLE_sh_ECad and HMLE_sh_GFP (control). A shRNA targeting of enhanced GFP (eGFP) gene, which is not endogenously expressed in the HMLE cells, was introduced to the cell line to create the control mammary epithelial cell line. These cells do not undergo EMT. From the 2,244 compounds retested at dose concentrations (range: 160 nM–19.5 μM; Table A1, Appendix A), 97% (2181 compounds) re-confirmed in this assay (AID 449748). In parallel, these 2,244 compounds were included in HMLE_sh_GFP viability assays (AID 463074) to determine if these compounds affected the viability of the control cell line. Taken together, 26 compounds were identified as 25-fold more potent against HMLE_sh_ECad versus HMLE_sh_GFP.

After completing the retests and a secondary assay at dose from DMSO stocks, 19 exact or similar compounds were ordered from commercial vendors. These 19 compounds were tested in the primary screen at dose and the secondary screen to determine if the Quality Controlled dry powder compounds retained selectivity. From this set of dry powder compounds, a furan containing compound (CID4791237, Table 2, Entry 1) retained selectivity, and this scaffold was prioritized for probe development (AID 493176 and AID 493196).

3.2. Dose Response Curves for the Probe

Figure 3Dose-dependent Activity of the Probe (ML245)

Dose-dependent activity of the probe (ML245) in Target and Counterscreen Assays: Primary screen from dry powders on HMLE_sh_ECad (IC₅₀ = 0.536 μM; AID 504535, 504667) (A); control cell line HMLE_sh_GFP (7.87 μM) Toxicity (AID 504533, 504666) (B).

3.3. Scaffold/Moiety Chemical Liabilities

A search of PubChem for the hit compound from the primary screen (CID 4791237, Table 2, Entry 1) revealed that it had been tested in 276 BioAssays and was confirmed as active only in one assay other than the assays described in this report. The assay was a screen for Niemann-Pick C1 promoter activators (1 μM).

3.4. SAR Analysis

Screening results for the primary assay are described in Tables 2–10. The results of SAR at the amide are shown in Table 2. The free acid (Table 2, Entry 2) was found to be inactive. Methylation of the amide (Table 2, Entry 3) led to a drop in potency and selectivity. Replacement of the thiazole with substituted thiazoles (Table 2, Entry 4 and Entry 5) or removal of the methyl substitution (Table 2, Entry 6) resulted in a significant drop in potency and selectivity in all cases. Replacing the methyl thiazole with benzothiazole (Table 2, Entry 7) led to loss of activity. Substitution with other heterocyclic amides (Table 2, Entry 8 and Entry 9) also resulted in inactive compounds. Replacement of the thiazole with aliphatic amides (Table 2, Entry 10 and Entry 11) also resulted in inactive compounds. Table A4 in Appendix H provides data on all other amide analogs that were tested. Based on the data in Table 2 and Table A4, we concluded that the 2-methyl thiazole amide was the optimum group for the eastern portion of the molecule for potency and selectivity against the HMLE_shEcad cell line. For all of the subsequent SAR studies we retained the 2-methyl thiazole on the eastern side.

Tables 3, 4 and 5 illustrate the effect of substitution on the phenoxy group on the western portion of the molecule. Table 3 contains data on a set of compounds that evaluates the effect on electron withdrawing and electron donating groups on the ortho-position. Substitution with a chloro (Table 3, Entry 3), a fluoro (Table 3, Entry 4), a methyl (Table 3, Entry 7) and a trifluoromethoxy (Table 3, Entry 6) led to potent compounds, however, with a loss in selectivity.

Table 3

SAR of the Phenyl Group (ortho-substitution).

Table 4

SAR of the Phenyl Group (meta-substitution).

Table 5

SAR of the Phenyl Group (para-substitution).

Table 4 describes the results from substitution at the meta-position. As with the ortho-position, substitution with a chloro (Table 4, Entry 2), a fluoro (Table 4, Entry 3), a methyl (Table 4, Entry 6) and a trifluoromethoxy (Table 4, Entry 5) resulted in compounds that were around 200–500 nM while, however, compounds substituted at the meta-position retained between 7- and 15-fold selectivity. This finding led us to further investigate the meta-position, and these results are summarized in Table A5 in Appendix H. As can be seen in Table A5, substitution at the meta-position with a variety of alkyl and alkoxy groups result in compounds that maintain potency but with decreased selectivity compared to the compounds in Table 3.

Substitution at the para-position of the phenyl group, as shown in Table 5, results mostly in compounds with a slight decrease in activity, but a significant drop in selectivity.

Having investigated the effects of a range of common substituents individually on the ortho-, meta-, and para- positions, the effects of a set of carefully designed di-substituted analogs were evaluated. Compared to the unsubstituted phenyl, the ortho-chloro analog (Table 3, Entry 3) was 10-fold more potent but less selective. The meta-chloro analog (Table 4, Entry 2) was less potent but retained the selectivity (15-fold) inherent in the parent unsubstituted analog. Hence, a collection of di-chloro substituted analogs were synthesized, and these results are summarized in Table 6.

Table 6

SAR of the Phenyl Group (di-substitution).

Several of the compounds were found to have potencies of less than 1 μM; however, most lacked reasonable selectivity. The 3,5-dichloro substituted compound (Table 6, Entry 7) was the exception. It was extremely potent with 312 nM, while also retaining selectivity against the control cell line at greater than 14-fold. Table A7 in Appendix H contains data on all other di- and tri-substituted compounds that were tested.

Table 7 contains data for compounds in which the phenyl group has been replaced with pyridyl groups. Intriguingly, replacement of the phenyl with 2-pyridyl (Table 7, Entry 2) retains some activity while being completely inactive against the control cell line and is significantly more soluble. This finding has led us to synthesize a group of compounds in which the phenyl is replaced with pyridyls in an effort to take advantage of the selectivity and solubility. However, these analogs did not improve potency. Data for these compounds can be found in Table 7 (Entries 2–4). The phenoxy group was also replaced with several alkoxy groups; however, this replacement led to mostly inactive compounds (Table A6 in Appendix H).

Table 7

SAR of Phenyl Replacement.

Modification of the western portion of the molecule was carried out, and the data is shown in Table 8. Replacement of the ether oxygen with a carbon (Table 8, Entry 2) results in similar activity but a slight loss in selectivity. Replacement with a nitrogen (Table 8, Entry 3) results in an increase in potency but a slight loss in selectivity. Conversion of the ether to a sulfone (Table 8, Entry 4) results in a loss of activity. The addition of a methylene between the phenyl group and the ether oxygen results in a loss of activity and selectivity (Table 8, Entry 5). Replacement of the phenyl ether with 2-aminopyridine (Table 8, Entry 6) results in a slight loss in activity but an increase in selectivity and solubility. This finding has led us to synthesize a group of compounds in which the phenoxy is replaced with ortho- and meta-substituted aminopyridines in an effort to take advantage of the selectivity and solubility of the aminopyridine while trying to increase potency. As in the case of hydroxypyridine, substitution did not lead to an increase in potency. Data can be found in Table A10 in Appendix H.

Table 8

SAR of the Linker.

Table 9 illustrates the effect of substitution in furan and replacement of the furan with other aromatic groups. Substitution on the furan with a methyl group (Table 9, Entry 2) results in a loss of activity and selectivity. Interestingly, replacement of the furan with a thiophene (Table 9, Entry 3) results in a compound that is much more potent towards the control cell line. Replacement with a phenyl (Table 9, Entry 4) results in a slight loss of potency and selectivity and replacement with a pyridine results in a significant loss in potency and selectivity.

Table 9

Replacement of the Furan.

Several analogs were made based on the phenyl replacement in an effort to increase potency and selectivity as seen in Table 10. Substitution with halogens and methyl at the ortho- or meta-positions results in a slight increase in potency and selectivity; however, these compounds are significantly less potent than their furan equivalents. Several aminopyridine analogs of the phenyl series were synthesized; however, while increasing solubility and selectivity these compounds were much less potent (Table A11, Appendix H).

All five of the other possible regioisomers of the furan were synthesized and tested. All of the other isomers either suffered from a loss in selectivity or potency. This data can be found in Table A9 in Appendix H.

Based on the SAR described previously, the 3-chloro-substituted analog (Table 4, Entry 2) was chosen as the probe due to its excellent potency and selectivity.

In summary, 112 synthesized compounds were tested in both primary (AID 493226, AID 493176, AID 504449, and AID 504535) and secondary assays (AID 504325, AID 493196, AID 504450, AID 504450, and AID 504533).

3.5. Cellular Activity

To confirm results obtained by the probe Screening Center, several additional experiments were conducted by the Assay Provider. The probe (CID 50910523/ML245) and five additional compounds (CID 4791237, CID 9214159, CID 50904129, CID 50904149, and CID 50944065) were investigated for their effects on viability in the HMLE_sh_ECad and HMLE_sh_GFP cell lines. Although varying absolute IC₅₀ values were obtained from the collaborators, ML245 had the greatest selectivity of all the compounds tested.

Two additional transformed cell lines, HMLE_sh_Twist and MDA231, were used to corroborate the findings. HMLE_sh_Twist, a second upstream model of EMT-induced breast CSC-like cells, was also significantly inhibited by the probe (ML245) (Figure 5, Table 11). Although ML245 is toxic to HMLE_Twist, the selectivity was only approximately 8-fold. This finding supports that the compound is not specific to the generated cell line HMLE_sh_ECad and can inhibit other breast CSC-like cells.

Figure 5

Dose Inhibition Assay of HMLE_sh_GFP, HMLE_sh_ECad, and HMLE_sh_Twist by the Probe (ML245). Dose curve data provided by the Gupta laboratory showing probe (ML245) inhibition of HMLE_sh_GFP (blue ), HMLE_sh_ECad (black▲), and HMLE_sh_Twist (green (more...)

Table 11

Activity Data of HMLE_sh_GFP, HMLE_sh_ECad, and HMLE_sh_Twist by ML245 and Five Analogs.

Furthermore, activity of ML245 and five analogs were tested against a breast cancer cell line (MDA231; Table 11). All of the compounds only weakly targeted the MDA231 cell line, which has been characterized as a mixed population of breast CSCs and more differentiated cells. This may explain the reduced potency since the probe would not be toxic to more differentiated cells. Additional testing of cancer cell lines will be required to see if this scaffold can inhibit other breast cancer lines. Since ML245 is selective for breast CSCs, any further development towards drug discovery should be considered in conjunction with chemotherapeutic compounds that target the differentiated cells in the tumor.

In addition, ML245 and three additional analogs (CID 4791237, CID 9214159, CID 50904129) were tested in a tumorsphere assay (AID 504623; Table 11). In this assay, Sum159 cells (a human breast cancer cell line) were grown in the presence or absence of compound for 9 days in a low adhesion environment. Under these conditions, Sum159 cells, which contain a mixture of breast CSCs and differentiated breast cancer cells, form tumorspheres. The number of tumorspheres was counted in each of the conditions. The signal (number of tumorspheres) was normalized to neutral (DMSO) and positive (Puromycin) control, and a 30% inhibition cut-off at an average screening concentration of 20 μM was used to define a hit.

Neither the probe (ML245) nor any of the analogs significantly and consistently inhibited tumorsphere formation. In previous scaffolds, nonselective analogs (CID 24816775, CID 16189574, and CID 6262187) that were toxic to all cell lines did inhibit tumorsphere formation (Table 11). In this series, CID 50904129, which is similarly toxic/nonselective, did not similarly inhibit tumorsphere formation. It is possible that the viscous methylcellulose/media used to keep tumorspheres from growing together also inhibited this compound series from inhibiting tumorsphere formation. Another possibility is that these compounds are not effective against SUM159 cells. Further work needs to be done to determine if these compounds can inhibit tumor formation.

Although the probe did not inhibit tumor formation, previous probes ML239 and ML243 also did not inhibit tumor formation. Salinomycin, the only other small molecule that is known to selectively inhibit HMLE_sh_ECad also does not inhibit SUM159 tumorsphere formation(8). Additional assays and cell lines will be required to address these issues. Taken together, the probe (ML245) clearly meets the established probe criteria in whole cells specified for this project (Figure 2, Figure 4, and Table 12); experimental details are provided in Section 2.1.2.

Figure 4

Dose-dependent Activity of Select Analogs in the Primary Assay. Four curves of analogs tested in the primary screen at dose: CID4791237 (A; Table 2, Entry 1), CID51003702 (B; Table 6, Entry 6), CID9214159 (C; Table 2, Entry 10), CID50944077 (D; Table (more...)

Table 12

Comparison of the Probe (ML245) to Project Criteria.

3.6. Profiling Assays

To gain insight into the possible mechanism or signaling pathways that the probe (ML245) could be targeting, gene expression studies were conducted. HMLE_sh_ECad and HMLE_sh_GFP cells were treated with the probe (ML245), analog CID50904149, or salinomycin for 24 hours at the IC₅₀ identified after 3-day treatment in the primary assay (0.5 μM, 6 μM, and 1 μM, respectively). In addition, control samples were included where DMSO was added for 24 hours. cRNA synthesis from the total RNA samples were analyzed on a Human HT-12 Expression BeadChip (Illumina). All samples passed QC standards for gene expression and hybridization.

The samples were normalized, and gene expression of the triplicate samples was compared using the Comparative Marker Selection Module in GenePattern (Open source: http://genepattern.broadinstitute.org/). Gene expression that had p ≤ 0.005 and a change at least 150% (± 50%) change in expression was selected. First, the control conditions between the cell lines were compared and 624 genes were differentially expressed. Comparable results were observed previously (11). Moreover, many of the genes confirmed the results observed by the Weinberg laboratory using these cell lines (12). Specifically, of the 19 genes indicated by Onder et al. as statistically significantly different, 14 repeated in our assay (11, 12). Moreover, Onder et al. identified several proteins with altered expression, including vimentin, beta-catenin, gamma-catenin, CK-8, and fibronectin (12). We observed similar changes in our gene expression assay, except for fibronectin, where we did not observe any change in gene expression. However, gene expression does not always correlate with alterations in protein expression. Further characterization is required to determine if those results varied.

Next, we compared differences in gene expression after 24-hour treatment with ML245, analog CID 50904149 (Table 4, Entry 10), or salinomycin (CID 3085092) in HMLE_sh_ECad or HMLE_sh_GFP (Figure 6 and Figure 7) to yield a better understanding of genes and/or pathways regulated by the breast CSC-like selective compounds. Using the same criteria as above, 349 genes were identified as changing when comparing RNA-isolated, DMSO-treated versus ML245-treated HMLE_sh_ECad cells (Figures 6A, 6C, 6D). However, only two genes (CYP1B1 and DHRS3) had altered expression by the same criteria after ML245 treatment in the control line, HMLE_sh_GFP (Figure 6D). This strongly suggests that ML245 is highly selective for the HMLE_sh_ECad cells.

Figure 6

Heat Map Comparisons of Differential Gene Expression by ML245 and CID50904149. Genes that were significantly differentially expressed (p <0.005) and had greater than 1.5-fold difference in expression were identified. Heat map representations of (more...)

Figure 7

Heat Map Comparisons of Differential Gene Expression by Salinomycin. Gene expression profile of prior art, salinomycin. HMLE_sh_ECad (shECad) (A) and HMLE_sh_GFP (sh_GFP) (B) were treated with DMSO or salinomycin (Sal). Genes that had significantly different (more...)

To gain a better insight to which genes and/or pathways are affected by this scaffold, both the CSC-like cell line (HMLE_sh_ECad) and control (HMLE_sh_GFP) cell lines were treated with analog CID 50904149. After treatment with CID 50904149, 238 genes were alternately expressed in HMLE_sh_ECad, while 160 genes were altered in HMLE_sh_GFP (Figures 6B, 6C, 6D). The higher number of genes affected in HMLE_sh_GFP cell line may be expected since CID 50904149 is less selective than ML245 (3-fold versus 14.7-fold, respectively). More critically, when comparing the genes altered by ML245 and CID 50904149 in the HMLE_sh_ECad line, 84 genes overlapped in both groups (Figure 6A–D).

To compare our compounds with the prior art (i.e., salinomycin), we conducted parallel studies with ML245 (Figure 7). Again, we compared salinomycin-treated cells to DMSO-treated cells using the same criteria as above. Salinomycin altered the expression of 683 genes in the CSC-like cell line. Interestingly, salinomycin had a greater impact on the control line (HMLE_sh_GFP) with 1021 genes having differential gene expression (Figure 6). There were 84 genes with altered expression by salinomycin that were also affected by ML245 (Figure 6D). Interestingly, all of the 84 genes identified when comparing the effects of ML245 and CID5094149 were identified again as having altered gene expression with treatment of salinomycin. were the same 84 genes found when comparing ML245 and CID 5094149 (Figure 6 and Figure 7). These 84 genes suggest ML245, CID 50904149, and salinomycin could be killing CSCs via a similar mechanism.

These 84 genes were, therefore, prioritized as genes of interest. A portion of the genes were transcription factors or DNA-modifying enzymes. At least 12 of the 84 fell into this category, including five zinc finger-containing genes (EGR1, ZCCHC7, ZNF295, CGRRF1), leucine zipper proteins (LZTFL1, TSC22D3), several additional transcription factors (ATF4, LMO4, CEBPG, NUPR1, TAF1D), and DNA-modifying enzymes such as GADD45A. Taken together, these 12 genes do not represent a clear signaling pathway; however, it may be possible that the compound directly targets nuclear protein gene set.

One particular gene of interest is the RING finger protein CGRRF1, which is a protein that determines cell-cycle arrest. An increased expression of CGRRF1 correlated with a decreased division of Burkitt lymphoma cell lines (1), and a potential way to stop the proliferation of lymphomas. The probe (ML245), CID 5094149, and salinomycin increased the expression of this particular gene in our gene expression studies. Therefore, this gene will be a prioritized gene of interest.

Furthermore, several genes included in the 84 prioritized genes are involved in apoptotic signaling and mitochondrial maintenance. These include caspase recruitment domain family member 10 (CARD10), Activating Transcription Factor 4 (ATF4) (13), ATPase subunit C targeting peptide (ATP5G1) (14), cytochrome P450, family 20 (CYP20A1) (15), mitochondrial ribosomal protein (MRPL12) (16). Since the expression of these five genes was altered after treatment with CSC-selective compounds, the mechanism of selective death may be increasing pro-apoptotic signaling. Future studies will be conducted to address the role of the probe (ML245) role in DNA regulation and apoptotic signaling.

The probe (ML245) was screened against a panel of 68 targets in radioligand binding assays that are commonly used in drug discovery for lead profiling. The assays were done by Ricerca Biosciences, LLC and include targets from various areas such as, GPCRs, ion channels, and transporters. The probe (ML245) was active in eight of the 68 assays and displayed inhibition of adenosine A₁, A_2A, and A₃ receptors at 79%, 88%, and 94% at 10 μM, respectively. The probe (ML245) also displayed inhibition against two transporters, Norepinephrine (NET) and Dopamine (DAT) at 98% and 101%, respectively. The other three targets that the probe (ML245) inhibited were the GABA_A (93%), Opiate κ (87%), and serotonin (5-Hydroxytryptamine) 5-HT_2B (78%) receptors. These assays only assess binding to the above mentioned targets and the functional consequence of that binding will need to be assessed.

4.1. Comparison to Existing Art and How the New Probe is an Improvement

The current state-of-the-art for the probe molecule salinomycin exhibits a potency of about 1 μM against breast CSCs (8); however, it should be noted that salinomycin shows toxicity against the control cell line (HMLE_sh_GFP) at 10 μM. Whereas the new probe (ML245) exhibits a similar potency at 0.536 μM, this probe is more selective and more attractive structurally allowing for rapid analog synthesis.

In the investigation of prior art relevant to the breast CSC project, the following databases were used: SciFinder, Entrez, PubMed, PubChem, and PatentLens. The search terms applied and hit statistics for the prior art search are provided in Table A2 (Appendix G). Abstracts were obtained for all references returned and were analyzed for relevance to the current project. The searches were performed on, and are current as of, September 22, 2011.

Literature and patent searches revealed nine small molecules capable of selectively inhibiting breast CSCs (Figure 9). It is important to note that salinomycin is the only small molecule shown to be effective in the cell line HMLE_sh_ECad used in the primary assay. The effectiveness of salinomycin in HMLE_sh_ECad cells was determined by the Assay Provider, in which they developed an HTS screen to discover compounds that were selectively toxic for breast CSCs. Salinomycin was shown to provide a reduction in the proportion of CSCs with an IC₅₀ value of approximately 1 μM against HMLE_sh_ECad with approximately 10-fold selectivity over the control HMLE_sh_GFP (8). In addition, ex vivo and in vivo treatment of mice with intraperitoneal injection of salinomycin inhibited mammary tumor growth and increased differentiation of epithelial tumor cells. To date, this is the only report of selectively active compounds against the HMLE_sh_ECad cell line.

Figure 9

Known Small Molecules Inhibitors of Breast Cancer Stem Cells.

There has been a recent increase in the research of selective killing of breast CSCs. Though interesting, the cells used in this report are not susceptible to these agents. In one approach, researchers targeted the breast CSC pathways involved in self-renewal and survival. These pathways include NF-κB, Notch, Hedgehog, and Wnt. The NF-κB inhibitors parthenolide and 8-quinolinol (8Q) were reported by Zhou et al. to inhibit breast CSC-dependent MCF7 sphere cell growth selectively over bulk MCF7 cell growth (17). They reported that parthenolide and pyrrolidinedithiocarbamate inhibited sphere MCF7 cell growth by 33% and 51%, respectively, and showed no obvious growth inhibition of bulk MCF7 cells. This work was followed by a 2009 publication, in which Zhou et al. reported that 8Q inhibited the sphere cell proliferation by 86% at 5 μM and inhibited the bulk cells by 30% (18). 8Q also showed some antitumor activity in both MCF7 and MDAMB-435 (breast cancer cell line) xenograft models, but it had a much better effect when combined with paclitaxel than either agent administered alone. Kakarala et al. reported that curcumin and piperine, respectively, inhibited Wnt signaling, ALDH+ cells (breast CSC biomarker) and mammosphere formation by 50% at 5 μM and completely at 10 μM (19). In 2010, Li et al. showed that sulforaphane down-regulated the Wnt-pathway and could inhibit ALDH+ cells in vitro by 65%–80% at 1–5 μM, but had minimal effect on the population of bulk breast cancer cell lines. In addition, sulforaphane inhibited mammosphere formation and the cancer cells from sulforaphane-treated mice failed to produce any tumors in recipient mice up to 33 days after implantations (20).

Metformin, a known oral hypoglycemic agent, was reported to selectively kill breast CSCs through an unknown mechanism. In this study, the combination of metformin and doxorubicin had a synergistic effect in reducing tumor mass and preventing relapse in xenograft mice (20). In addition, Jiralerspong et al. reported a study of 2529 patients undergoing neoadjuvant chemotherapy, in which the pathologic complete response rate following surgery was 24% for diabetic patients taking metformin, 8% for diabetic patients not taking metformin, and 16% for nondiabetic patients (21). Furthermore, using both CXCR1-blocking antibodies and the small molecule inhibitor, repertaxin, demonstrated that CXCR1 selectively decreased breast CSCs in vitro and in NOD/SCID xenograft models, respectively (20).

Berberine was investigated to determine its effects on the enriched stem cell-like population of MCF-7 cells. After 72 hours, the viability of MCF-7 cells was reduced to 96%, 84%, 79%, 65%, and 56%, and the stem cell-like population was reduced to 1.4% (1.5% of MCF7 cells were the stem cell-like population cells), 1.1%, 0.9%, 0.5%, and 0.2% by 10 nM, 100 nM, 1 μM, 10 μM, and 20 μM of berberine, respectively (22).

All nine compounds identified through this project are registered with the NIH Molecular Libraries, and the respective compound information is provided in Table A3 (Appendix G).

To date, we have identified 3 probe compounds, ML239 (Probe 1), ML245 (Probe 2), and ML243 (Probe 3). The three probes are potent inhibitors (0.5–2uM) of the Breast Cancer Stem Cell-like cells, sufficiently selective (14.7–32 fold) against HMLE_sh_ECad, and meet the probe criteria. ML245 is the most potent against the primary target cell line, while ML243 has the greatest selectivity compared to HMLE_sh_ECad (see Table 13).

Table 13

Comparison of Breast Cancer Stem Cells: Probes 1, 2, and 3.

4.2. Mechanism of Action Studies

We have included gene expression assays that indicate pathways that are regulated by addition of the probe (ML245). Future studies will be required to determine the target of the probe (ML245).

4.3. Planned Future Studies

To date, the probe (ML245) has been used in gene expression profiling studies, where HMLE_sh_ECad and HMLE_sh_GFP have been treated for 24 hours (see Figure 6). Treatment of the breast CSC-like cell line (HMLE_sh_ECad) by ML245 yielded interesting results identifying genes in pro-apoptotic signaling pathways as well as numerous transcription factors as potential targets. Furthermore, numerous transcription factor genes were regulated in these studies, suggesting that ML245 either directly or indirectly regulates transcription.

Although these initial gene profiling experiments suggest that these pathways are critical to the probe’s selective killing of breast CSCs, follow-up studies are required to confirm these results. Since we have identified two additional probes (Table 13) that confer selective killing of the HMLE_sh_ECad cell line, we plan to revisit the gene expression profiling studies to determine if the combined data sets provides greater insight into critical signaling pathways and potential targets within breast CSCs.

Primary HTS Using CellTiter-Glo (AID 2717)

Primary Retest (AID 449748)

Repeat of the primary screen at dose using CellTiter-Glo.

Primary Cell Viability Protocol with HMLE_sh_ECad Cells Using CellTiter-Glo (AID 493226, AID 493176, AID 504449, AID 504535)

HMLE_sh_ECad cells were prepared as previously described (8).

1. Propagate HMLE cells expressing either shRNA targeting E-cadherin (sh_ECad), control shRNA targeting eGFP (sh_GFP), or shRNA targeting Twist (sh_Twist) in a 1:1 mixture of 10% fetal bovine serum (FBS; HyClone), 1% Penicillin/Streptomycin (Pen/Strep; Cellgro), 1% Glutamax-1 (Invitrogen), 70 nM Hydrocortisone (Sigma), 12 μg/ml Insulin (Sigma), 50 μg/ml Gentamicin (Sigma), 12.5 μg/ml Plasmocin (InVivogen), 10 ng/ml EGF in DMEM (Cellgro) with Mammary Epithelial Cell Growth Medium (MEGM complete medium; Lonza, Basel, Switzerland).
2. Incubate at 37 °C, 5% CO₂.
3. For screening, count the cells and resuspend in complete media without serum.
4. Plate 2,000 cells in 50 μl per well in white, tissue culture-treated, 384-well plates (Corning).
5. Incubate the cells at 37 °C, 5% CO₂ for at least 4 hours and pin with 100 nl of compounds.
6. Incubate the cells approximately 72 hours, then add 30 μl of CellTiter-Glo (Promega) diluted 1:3 with PBS to the well.
7. Read the plates using the EnVision (PerkinElmer; Luminescence 0.1 sec/well) after 12 minutes.

Secondary Cell Viability Protocol with HMLE_sh_GFP Cells Using CellTiter-Glo (AID 504325, AID 493196, AID 504450, AID 504533)

HMLE_sh_GFP cells were prepared as previously described (8). Refer to the procedures as described in the Primary Cell Viability Protocol.

Assay Provider Cell Viability Protocol with CellTiter-Glo

HMLE_sh_Twist cells were prepared as previously described (8). Refer to the procedures as described in the Primary Cell Viability Protocol.

Secondary Orthogonal Assay for In Vitro Inhibition of Tumorspheres with Sum159 Cells (AID 504623)

Sum159 cells were provided by Dr. Gupta as previously described (8).

1. Perform tumorsphere assays were performed as previously described (10) with minor changes.
2. Propagate Sum159 cells in 5% FBS (HyClone), 1 % Pen/Strep, 1% Glutamax-1, 12 μg/mllInsulin, 50 μg/ml Gentamicin in F12/DMEM (Cellgro).
3. Plate 100 μl media respective to cell type in 96-well, ultra-low adhesion plates (Costar).
4. Pin 400 nl of compounds into the media.
5. Harvest the cells, count, and resuspend in their propagation media with 1% methylcellulose (ES-CultM3120, Stem Cell Technologies).
6. Add 100 μl of resuspended cells to the plates containing media with compound for a final count of 2000 cells/well in 200 μl with 0.5% methylcellulose.
7. Allow tumorspheres to form for 9 days incubated at 37 °C, 5% CO₂.
8. Image tumorspheres using a 2× objective on the ImageXpress Micro (Molecular Devices). Identify cell clusters greater than 100 μM in diameter were identified using MetaXpress software (version 3.1; Molecular Devices, Sunnyvale, CA).

Gene Expression Assays with HMLE_sh_ECad and HMLE_sh_GFP Cells

1. ML245, CID50904149, Salinomycin/CID23682228 are selectively toxic to HMLE_shECad cell lines (over HMLE_shGFP cell lines) in 3-day toxicity assays. The IC₅₀ concentration of toxicity towards HMLE_shECad cell line in these 3-day toxicity assays is the same concentration used for experiments below (i.e., 0.54 μM, 6.7 μM, and 1 μM, respectively). 300,000 cells HMLE_shECad or HMLE_shGFP cells were plated in a 6-well dish and allowed to adhere for 4 hours.
2. Replicate samples (3–4) of HMLE_shECad and HMLE_shGFP cells were treated with vehicle, ML245, CID50904149, or Salinomycin at the above concentrations for 24 hours prior to isolation of RNA. 30
3. Isolation of total RNA was conducted using the protocols for the RNeasy Protect Mini Kit (Qiagen).
4. Genome Analysis Platform (GAP) at the Broad Institute analyzed RNA samples for quality using Aglient Bioanalyzer Chips, and prepared cRNA synthesis from the total RNA samples passing QC for analysis on a HumanHT-12 Expression BeadChip (Illumina, San Diego, CA) according to the manufacturer’s instructions. Results were confirmed using GenomeStudio (version 2010.3; Illumina, San Diego, CA).
5. All of the samples were run in replicates (3–4) and the entire data set was normalized using the quantile module available in open source program GenePattern (http://genepattern.broadinstitute.org/).
6. Gene expression was compared between the DMSO-treated and compound-treated using the ComparitiveSelection Module in GenePattern.
7. Those genes with a significant difference (p>0.005) and had 1.50-fold difference in expression were selected.

Chemistry Experimental Methods

General details. All oxygen and/or moisture-sensitive reactions were carried out under nitrogen (N₂) atmosphere in glassware that had been flame-dried under vacuum (approximately 0.5 mm Hg) and purged with N₂ prior to use. All reagents and solvents were purchased from commercial vendors and used as received, or synthesized according to methods already reported. NMR spectra were recorded on a Bruker 300 (300 MHz ¹H, 75 MHz ¹³C) or Varian UNITY INOVA 500 (500 MHz ¹H, 125 MHz ¹³C) spectrometer. Proton and carbon chemical shifts are reported in ppm (δ) referenced to the NMR solvent. Data are reported as follows: chemical shifts, multiplicity (br = broad, s = singlet, d = doublet, t = triplet, q = quartet, m = multiplet; coupling constant(s) in Hz).

Unless otherwise indicated, NMR data were collected at 25 °C. Flash chromatography was performed using 40–60 μm Silica Gel (60 Å mesh) on a Teledyne Isco Combiflash R_f. Tandem Liquid Chromatography/Mass Spectrometry (LC/MS) was performed on a Waters 2795 separations module and 3100 mass detector. Analytical thin layer chromatography (TLC) was performed on EM Reagent 0.25 mm silica gel 60-F plates. Visualization was accomplished with ultraviolet (UV) light and aqueous potassium permanganate (KMnO₄) stain followed by heating. High-resolution mass spectra were obtained at the MIT Mass Spectrometry Facility (Bruker Daltonics APEXIV 4.7 Tesla Fourier Transform Ion Cyclotron Resonance Mass Spectrometer).

5-(hydroxymethyl)furan-2-carboxylic acid (1.15 g, 8.09 mmol) was diluted with Benzene (81 ml). Thionyl chloride (3.54 ml, 48.6 mmol) was added and then the reaction is heated to reflux. The reaction was stirred overnight. The reaction was concentrated and carried on directly to the next step. The dichloride was dissolved in dichloromethane (112 ml) and 4-methylthiazol-2-amine (1.302 g, 11.17 mmol) was added followed by 4-Dimethylaminopyridine (DMAP;0.138 g, 1.117 mmol) and triethylamine (3.89 ml, 27.9 mmol). The reaction was stirred at room temperature until complete by LC-MS analysis. The reaction was concentrated and purified by silica gel chromatography to yield the furanyl chloride (1.61 g) in 56% yield over two steps. The furanyl chloride (63.1 mg, 0.246 mmol) was dissolved in dimethylformamide (2.5 ml). Sodium iodide (3.68 mg, 0.025 mmol), potassium carbonate (51.0 mg, 0.369 mmol), and 3-chlorophenol (31.6 mg, 0.246 mmol) were added and the reaction was stirred at room temperature until complete by LC-MS analysis. The reaction was concentrated under vacuum and purified by silica gel chromatography to provide 5-((3-chlorophenoxy)methyl)-N-(4-methylthiazol-2-yl)furan-2-carboxamide (56 mg) (CID50910523/ML243) in 65% yield.

¹H NMR (300 MHz, CDCl3) δ 7.19 (s, 2H), 6.91 (s, 1H), 6.85 – 6.74 (m, 2H), 6.50 (s, 2H), 5.28 – 5.19 (m, 2H), 2.30 (s, 3H),; ¹³C NMR (75 MHz, CDCl3) δ 157.38, 155.83, 153.78, 151.82, 146.26, 144.93, 133.81, 129.33, 120.72, 116.44, 114.20, 112.05, 111.37, 107.34, 61.33, 15.77. calculated [M⁺H] 349.0408 experimental [M⁺H] 349.0408.