Pharmacogenomics of Scopoletin in Tumor Cells

Drug resistance and the severe side effects of chemotherapy necessitate the development of novel anticancer drugs. Natural products are a valuable source for drug development. Scopoletin is a coumarin compound, which can be found in several Artemisia species and other plant genera. Microarray-based RNA expression profiling of the NCI cell line panel showed that cellular response of scopoletin did not correlate to the expression of ATP-binding cassette (ABC) transporters as classical drug resistance mechanisms (ABCB1, ABCB5, ABCC1, ABCG2). This was also true for the expression of the oncogene EGFR and the mutational status of the tumor suppressor gene, TP53. However, mutations in the RAS oncogenes and the slow proliferative activity in terms of cell doubling times significantly correlated with scopoletin resistance. COMPARE and hierarchical cluster analyses of transcriptome-wide mRNA expression resulted in a set of 40 genes, which all harbored binding motifs in their promoter sequences for the transcription factor, NF-κB, which is known to be associated with drug resistance. RAS mutations, slow proliferative activity, and NF-κB may hamper its effectiveness. By in silico molecular docking studies, we found that scopoletin bound to NF-κB and its regulator IκB. Scopoletin activated NF-κB in a SEAP-driven NF-κB reporter cell line, indicating that NF-κB might be a resistance factor for scopoletin. In conclusion, scopoletin might serve as lead compound for drug development because of its favorable activity against tumor cells with ABC-transporter expression, although NF-κB activation may be considered as resistance factor for this compound. Further investigations are warranted to explore the full therapeutic potential of this natural product.


Introduction
Despite the breathtaking progresses in cancer biology during the past decades, a cure from the disease is still not reality for the majority of patients. A main reason for this unsatisfactory situation is the development of drug resistance, which ultimately may lead to failure of chemotherapy and a fatal outcome for patients. Unfortunately, drug resistance occurs not only with long-established cytotoxic drugs, but also with the more recent small molecules and therapeutic antibodies, which are directed against specific targets in tumor cells [1]. Hence, the urgent struggle to identify and develop novel drugs continues. Drug resistance is frequently multifactorial and several mechanisms may account for Databases [19,20].
Twelve artemisinin-based compounds and scopoletin have been included in the analysis. None of the compounds were commonly found in all 11 Artemisia species, only scopoletin was present in four species and these species clustered together (A. annua, A. abrotanum, A. dracunulus, and A. capillaris). Artemisinin derivatives were mainly found in A. annua. This dendrogram demonstrates that the considerable divergence of chemical composition in the Artemisia species.  [19,20].
Twelve artemisinin-based compounds and scopoletin have been included in the analysis. None of the compounds were commonly found in all 11 Artemisia species, only scopoletin was present in four species and these species clustered together (A. annua, A. abrotanum, A. dracunulus, and A. capillaris). Artemisinin derivatives were mainly found in A. annua. This dendrogram demonstrates that the considerable divergence of chemical composition in the Artemisia species.

Cross-Resistance of Scopoletin to Established Anticancer Drugs
We correlated the log 10 IC 50 values of the NCI cell lines to scopoletin with those of 86 standard drugs. The cellular responses of six out of 11 DNA topoisomerase I or II inhibitors drugs significantly correlated with those of scopoletin (=54.5%). Alkylating drugs were also frequently correlated to scopoletin. Four of 13 alkylating agents (=30.8%) revealed significant correlations to reserpine (p < 0.05; R > 0. 30). Comparable results were obtained for antimetabolites (4/15 drugs = 26.7%). Intermediate correlation rates were observed for antibiotics (1/4 drugs = 25%) and various other drugs (1/15 = 20%). No correlations were found for platin compounds, tubulin inhibitors, antihormones, tyrosine kinase inhibitors, mTOR inhibitors and epigenetic inhibitors ( Figure 2). The correlations to topoisomerase inhibitors, alkylating agents and antimetabolites indicate that scopoletin may interact with the DNA of tumor cells.

Cross-Resistance of Scopoletin to Established Anticancer Drugs
We correlated the log10IC50 values of the NCI cell lines to scopoletin with those of 86 standard drugs. The cellular responses of six out of 11 DNA topoisomerase I or II inhibitors drugs significantly correlated with those of scopoletin (=54.5%). Alkylating drugs were also frequently correlated to scopoletin. Four of 13 alkylating agents (=30.8%) revealed significant correlations to reserpine (p < 0.05; R > 0. 30). Comparable results were obtained for antimetabolites (4/15 drugs = 26.7%). Intermediate correlation rates were observed for antibiotics (1/4 drugs = 25%) and various other drugs (1/15 = 20%). No correlations were found for platin compounds, tubulin inhibitors, antihormones, tyrosine kinase inhibitors, mTOR inhibitors and epigenetic inhibitors ( Figure 2). The correlations to topoisomerase inhibitors, alkylating agents and antimetabolites indicate that scopoletin may interact with the DNA of tumor cells.

Tumor-Type Dependent Response towards Scopoletin
The log10IC50 values for scopoletin of all cell lines are shown in Figure 3. There was a trend whereby leukemia cell lines were more sensitive to scopoletin than cell lines from other tumor types. This result corresponds to the fact that clinically established anticancer drugs frequently show high

Tumor-Type Dependent Response towards Scopoletin
The log 10 IC 50 values for scopoletin of all cell lines are shown in Figure 3. There was a trend whereby leukemia cell lines were more sensitive to scopoletin than cell lines from other tumor types. This result corresponds to the fact that clinically established anticancer drugs frequently show high sensitivity towards leukemia, while solid cancers tend to be more resistant to cytostatic drugs. However, among the cell lines, we did not observe any significant trend regarding tumor types, which were more sensitive and resistant to scopoletin. As the tumor type seems not to reliably predict responsiveness to scopoletin, the individual gene expression might better determine sensitivity or resistance to this compound. To analyze this hypothesis in more detail, we investigated well-known mechanisms of anticancer drug resistance, e.g., ABC-transporter genes (ABCB1, ABCB5, ABCC1, ABCG2), oncogenes and tumor suppressor genes (EGFR, RAS, TP53). sensitivity towards leukemia, while solid cancers tend to be more resistant to cytostatic drugs. However, among the cell lines, we did not observe any significant trend regarding tumor types, which were more sensitive and resistant to scopoletin. As the tumor type seems not to reliably predict responsiveness to scopoletin, the individual gene expression might better determine sensitivity or resistance to this compound. To analyze this hypothesis in more detail, we investigated well-known mechanisms of anticancer drug resistance, e.g., ABC-transporter genes (ABCB1, ABCB5, ABCC1, ABCG2), oncogenes and tumor suppressor genes (EGFR, RAS, TP53).

Role of Classical Drug Resistance Mechanisms for Scopoletin
ABC-transporters are well known to efflux a plethora of anticancer drugs and xenobiotic compounds protecting cancer cells from death. The clinical relevance of the major ABC-transporters ABCB1, ABCC1, and ABCG2 for failure of chemotherapy and poor survival prognosis of cancer

Role of Classical Drug Resistance Mechanisms for Scopoletin
ABC-transporters are well known to efflux a plethora of anticancer drugs and xenobiotic compounds protecting cancer cells from death. The clinical relevance of the major ABC-transporters ABCB1, ABCC1, and ABCG2 for failure of chemotherapy and poor survival prognosis of cancer patients has been shown [18]. ABCB5 represents a novel resistance-mediating ABC-transporter, which is expressed in cancer stem-like cells. Therefore, novel anticancer drugs are required that kill ABC-transporter-expressing MDR tumors. As shown in Table 1, the expression of none of the ABC transporters examined in our study correlated with the log 10 IC 50 values of the NCI cell line panel for scopoletin. For each of the transporters, suitable positive control drugs have been used, which are described in the literature as substrates of the corresponding transporters [21][22][23][24]. All of the control drugs significantly correlated with expression of the corresponding ABC transporters (Table 1). To exemplarily confirm the results obtained by the NCI cell line panel, we tested the cytotoxicity of scopoletin in a pair of sensitive CCRF-CEM and P-gp-expressing CEM/ADR5000 cell lines using the resazurin assay. As shown in Figure 4, both cell lines were inhibited by scopletin with similar efficiency. The IC 50 value for CCRF-CEM cells was 26.1˘1.6 µM and for CEM/ADR5000 was 37.5˘2.9 µM. The degree of resistance of CEM/ADR5000 cells to scopoletin was only 1.4, although this multidrug-resistant cell line exerts high degrees of resistance to many established anticancer drugs with degrees of resistance that are two to three orders of magnitudes higher [25]. This indicates that scopoletin is not a substrate of P-gp, which supports the results obtained by microarray analyses and correlation to log 10 IC 50 values for scopoletin in the NCI cell line panel (Table 1). To exemplarily confirm the results obtained by the NCI cell line panel, we tested the cytotoxicity of scopoletin in a pair of sensitive CCRF-CEM and P-gp-expressing CEM/ADR5000 cell lines using the resazurin assay. As shown in Figure 4, both cell lines were inhibited by scopletin with similar efficiency. The IC50 value for CCRF-CEM cells was 26.1 ± 1.6 µM and for CEM/ADR5000 was 37.5 ± 2.9 µM. The degree of resistance of CEM/ADR5000 cells to scopoletin was only 1.4, although this multidrug-resistant cell line exerts high degrees of resistance to many established anticancer drugs with degrees of resistance that are two to three orders of magnitudes higher [25]. This indicates that scopoletin is not a substrate of P-gp, which supports the results obtained by microarray analyses and correlation to log10IC50 values for scopoletin in the NCI cell line panel (Table 1). As tumor suppressors and activated oncogenes also contribute to drug and radiation resistance [26][27][28], we also tested whether mutations in the tumor suppressor gene TP53 or the H-RAS, K-RAS, and N-RAS genes as well as the mRNA expression of the epidermal growth factor receptor gene (EGFR) are associated with cellular response to scopoletin. It is pleasing that we did not find significant correlations between cellular responsiveness to EGFR expression or TP53 mutations (Table 1). However, the mutational status of RAS genes was significantly associated with cellular resistance to scopoletin, indicating that RAS may represent a resistance mechanism for scopoletin ( Figure 5A).
It is well known that slowly growing tumors are more resistant to chemotherapy than fast growing ones [29]. We found that this is also true for scopoletin: the cell doubling times of the tumor cell lines were significantly correlated to the log10IC50 values for scopoletin ( Figure 5B).  As tumor suppressors and activated oncogenes also contribute to drug and radiation resistance [26][27][28], we also tested whether mutations in the tumor suppressor gene TP53 or the H-RAS, K-RAS, and N-RAS genes as well as the mRNA expression of the epidermal growth factor receptor gene (EGFR) are associated with cellular response to scopoletin. It is pleasing that we did not find significant correlations between cellular responsiveness to EGFR expression or TP53 mutations (Table 1). However, the mutational status of RAS genes was significantly associated with cellular resistance to scopoletin, indicating that RAS may represent a resistance mechanism for scopoletin ( Figure 5A). To exemplarily confirm the results obtained by the NCI cell line panel, we tested the cytotoxicity of scopoletin in a pair of sensitive CCRF-CEM and P-gp-expressing CEM/ADR5000 cell lines using the resazurin assay. As shown in Figure 4, both cell lines were inhibited by scopletin with similar efficiency. The IC50 value for CCRF-CEM cells was 26.1 ± 1.6 µM and for CEM/ADR5000 was 37.5 ± 2.9 µM. The degree of resistance of CEM/ADR5000 cells to scopoletin was only 1.4, although this multidrug-resistant cell line exerts high degrees of resistance to many established anticancer drugs with degrees of resistance that are two to three orders of magnitudes higher [25]. This indicates that scopoletin is not a substrate of P-gp, which supports the results obtained by microarray analyses and correlation to log10IC50 values for scopoletin in the NCI cell line panel (Table 1). As tumor suppressors and activated oncogenes also contribute to drug and radiation resistance [26][27][28], we also tested whether mutations in the tumor suppressor gene TP53 or the H-RAS, K-RAS, and N-RAS genes as well as the mRNA expression of the epidermal growth factor receptor gene (EGFR) are associated with cellular response to scopoletin. It is pleasing that we did not find significant correlations between cellular responsiveness to EGFR expression or TP53 mutations (Table 1). However, the mutational status of RAS genes was significantly associated with cellular resistance to scopoletin, indicating that RAS may represent a resistance mechanism for scopoletin ( Figure 5A).
It is well known that slowly growing tumors are more resistant to chemotherapy than fast growing ones [29]. We found that this is also true for scopoletin: the cell doubling times of the tumor cell lines were significantly correlated to the log10IC50 values for scopoletin ( Figure 5B).

COMPARE and Hierarchical Cluster Analyses of mRNA Microarray Data
In a hypothesis-generating bioinformatic approach, we mined the NCI database and correlated the microarray-based transcriptome-wide mRNA expression of tumor cell lines with the log10IC50 values, in order to identify novel putative molecular determinants of cellular response to scopoletin. Based on Pearson's rank correlation tests, scale rankings of genes were obtained by COMPARE analysis. The top 20 genes with positive and the top 20 with negative correlations are shown in Table 2. The identified genes belong to different functional groups such as cell cycle and proliferation (CCNB1, CDKN3, KIF11, MAPRE2, MKI67, SMC3), signal transduction processes (ARRB2, ARHGAP12, CDC42BPA, MAPRE2, RHEB, TSC22D4) and others.
Then, these genes were subjected to hierarchical cluster analysis. Only the mRNA expression data, but not the log10IC50 values of scopoletin for the cell lines were included into the cluster analysis. Three main cluster branches appeared ( Figure 6). As the log10IC50values of scopoletin were previously included into the cluster analysis, we now analyzed whether or not the obtained gene expression profile predicted sensitivity or resistance of cell lines to scopoletin. Indeed, a significant relationship was obtained (p = 2.05 × 10 −5 ; Table 3).  It is well known that slowly growing tumors are more resistant to chemotherapy than fast growing ones [29]. We found that this is also true for scopoletin: the cell doubling times of the tumor cell lines were significantly correlated to the log 10 IC 50 values for scopoletin ( Figure 5B).

COMPARE and Hierarchical Cluster Analyses of mRNA Microarray Data
In a hypothesis-generating bioinformatic approach, we mined the NCI database and correlated the microarray-based transcriptome-wide mRNA expression of tumor cell lines with the log 10 IC 50 values, in order to identify novel putative molecular determinants of cellular response to scopoletin. Based on Pearson's rank correlation tests, scale rankings of genes were obtained by COMPARE analysis. The top 20 genes with positive and the top 20 with negative correlations are shown in Table 2. The identified genes belong to different functional groups such as cell cycle and proliferation (CCNB1, CDKN3, KIF11, MAPRE2, MKI67, SMC3), signal transduction processes (ARRB2, ARHGAP12, CDC42BPA, MAPRE2, RHEB, TSC22D4) and others.
Then, these genes were subjected to hierarchical cluster analysis. Only the mRNA expression data, but not the log 10 IC 50 values of scopoletin for the cell lines were included into the cluster analysis. Three main cluster branches appeared ( Figure 6). As the log 10 IC 50 values of scopoletin were previously included into the cluster analysis, we now analyzed whether or not the obtained gene expression profile predicted sensitivity or resistance of cell lines to scopoletin. Indeed, a significant relationship was obtained (p = 2.05ˆ10´5; Table 3).

COMPARE and Hierarchical Cluster Analyses of mRNA Microarray Data
In a hypothesis-generating bioinformatic approach, we mined the NCI database and correlated the microarray-based transcriptome-wide mRNA expression of tumor cell lines with the log10IC50 values, in order to identify novel putative molecular determinants of cellular response to scopoletin. Based on Pearson's rank correlation tests, scale rankings of genes were obtained by COMPARE analysis. The top 20 genes with positive and the top 20 with negative correlations are shown in Table 2. The identified genes belong to different functional groups such as cell cycle and proliferation (CCNB1, CDKN3, KIF11, MAPRE2, MKI67, SMC3), signal transduction processes (ARRB2, ARHGAP12, CDC42BPA, MAPRE2, RHEB, TSC22D4) and others.
Then, these genes were subjected to hierarchical cluster analysis. Only the mRNA expression data, but not the log10IC50 values of scopoletin for the cell lines were included into the cluster analysis. Three main cluster branches appeared ( Figure 6). As the log10IC50values of scopoletin were previously included into the cluster analysis, we now analyzed whether or not the obtained gene expression profile predicted sensitivity or resistance of cell lines to scopoletin. Indeed, a significant relationship was obtained (p = 2.05 × 10 −5 ; Table 3).

NF-κB Motif Analysis
We hypothesized that the expression profile identified by COMPARE analysis shown in Table 2 might be commonly regulated at the transcriptional level. The idea was that a wide array of functionally diverse genes might be commonly up-or down-regulated by a transcription factor that is involved in resistance to cytotoxic compounds. We screened the gene promoter regions of this set of genes for binding motifs of the transcription factor, NF-κB to investigate this hypothesis. NF-κB mediates resistance towards diverse cancer therapeutics by inhibition of apoptosis. This broad spectrum of resistances represents another multiple drug resistance profile, which is different from the one mediated by ABC transporters, oncogenes and tumor suppressors.
Interestingly, all 40 genes identified by COMPARE analysis (Table 2) harbored binding motifs for NF-κB. We performed motif analysis for upstream promoter sequences of 40 genes that correlated with the cytotoxic activity of scopoletin. With 329 hits, the NF-κB DNA binding motif (Rel) was observed to be significantly enriched (with a log p-value of´8.2) in regions of 25 kb upstream promoter of all genes (Figure 7), indicating that NF-κB may play a considerable role for the regulation of genes associated with cellular response to scopoletin.

NF-κB Motif Analysis
We hypothesized that the expression profile identified by COMPARE analysis shown in Table 2 might be commonly regulated at the transcriptional level. The idea was that a wide array of functionally diverse genes might be commonly up-or down-regulated by a transcription factor that is involved in resistance to cytotoxic compounds. We screened the gene promoter regions of this set of genes for binding motifs of the transcription factor, NF-κB to investigate this hypothesis. NF-κB mediates resistance towards diverse cancer therapeutics by inhibition of apoptosis. This broad spectrum of resistances represents another multiple drug resistance profile, which is different from the one mediated by ABC transporters, oncogenes and tumor suppressors.
Interestingly, all 40 genes identified by COMPARE analysis (Table 2) harbored binding motifs for NF-κB. We performed motif analysis for upstream promoter sequences of 40 genes that correlated with the cytotoxic activity of scopoletin. With 329 hits, the NF-κB DNA binding motif (Rel) was observed to be significantly enriched (with a log p-value of −8.2) in regions of 25 kb upstream promoter of all genes (Figure 7), indicating that NF-κB may play a considerable role for the regulation of genes associated with cellular response to scopoletin.

Molecular Docking
As a next step to investigate the interaction of scopoletin with the NF-κB system in more detail, we performed molecular docking analyses to study binding to NF-κB, I-κB kinase β, I-κB kinase β-NEMO complex and NF-κB-DNA complex in silico. Our results indeed showed that scopoletin strongly bound to the pharmacophores of two of these target proteins. Scopoletin bound with calculated binding energies of less than −7 kcal/mol to the NF-κB DNA complex and I-κB kinase (Figure 8 and Table 4),

Molecular Docking
As a next step to investigate the interaction of scopoletin with the NF-κB system in more detail, we performed molecular docking analyses to study binding to NF-κB, I-κB kinase β, I-κB kinase β-NEMO complex and NF-κB-DNA complex in silico. Our results indeed showed that scopoletin strongly bound to the pharmacophores of two of these target proteins. Scopoletin bound with calculated binding energies of less than´7 kcal/mol to the NF-κB DNA complex and I-κB kinase (Figure 8 and Table 4), but not to free NF-κB and the I-κB kinase-NEMO complex (data not shown). Hence, scopoletin may target different steps of the NF-κB pathway to modulate NF-κB-mediated functions. Table 4. Molecular docking of scopoletin to NF-κB pathway proteins (details see Figure 8).

Protein
Binding but not to free NF-κB and the I-κB kinase-NEMO complex (data not shown). Hence, scopoletin may target different steps of the NF-κB pathway to modulate NF-κB-mediated functions.  Table 4. Molecular docking of scopoletin to NF-κB pathway proteins (details see Figure 8).

NF-κB Reporter Assay
The effects of scopoletin on NF-κB are controversially discussed in the literature. This compound can either activate or inhibit the activity of this transcription factor [32,33]. From our molecular docking studies we obtained only clues for binding, which implies an interaction of scopoletin with the transcription factor machinery. However, it cannot be concluded, whether this interaction leads to activation or inhibition. Therefore, we used a SEAP-driven NF-κB reporter cell line to address this question. Preliminary experiments with this cell line showed that scopoletin did not inhibit TNF-α-induced NF-κB activation. Therefore, we investigated, whether scopoletin would rather activate NF-κB. For this reason, we partially inhibited NF-κB by a low concentration (1.6 nM) of a well-known NF-κB inhibitor, triptolide [34,35]. Indeed, scopoletin (20 µM or 40 µM) significantly induced NF-κB activity (Figure 9). Activation of NF-κB by scopoletin was observed at two different treated conditions: incubation with triptolide for 1 h followed by incubation with scopoletin for another hour and co-incubation with both triptolide and scopoletin for 24 h (Figure 9).

NF-κB Reporter Assay
The effects of scopoletin on NF-κB are controversially discussed in the literature. This compound can either activate or inhibit the activity of this transcription factor [32,33]. From our molecular docking studies we obtained only clues for binding, which implies an interaction of scopoletin with the transcription factor machinery. However, it cannot be concluded, whether this interaction leads to activation or inhibition. Therefore, we used a SEAP-driven NF-κB reporter cell line to address this question. Preliminary experiments with this cell line showed that scopoletin did not inhibit TNF-α-induced NF-κB activation. Therefore, we investigated, whether scopoletin would rather activate NF-κB. For this reason, we partially inhibited NF-κB by a low concentration (1.6 nM) of a well-known NF-κB inhibitor, triptolide [34,35]. Indeed, scopoletin (20 µM or 40 µM) significantly induced NF-κB activity (Figure 9). Activation of NF-κB by scopoletin was observed at two different treated conditions: incubation with triptolide for 1 h followed by incubation with scopoletin for another hour and co-incubation with both triptolide and scopoletin for 24 h (Figure 9).

Chemoprofiling
As a part of our ongoing research on the anticancer activity of A. annua, in the present study we focused on scopoletin. While artemisinin-based compounds are mainly found in A. annua alone, scopoletin can be found in other Artemisia species and also in other genera (Brunfelsia, Kleinhovia, Mallotus, Scopolia, Solanum, Urticaria, Viburnum). Although the use of chemoprofiles has been proposed as a taxonomical parameter, our results with 11 different Artemisia species do not support this concept. This observation is in accordance with data obtained for other plant genera [36].
Artemisinin is the most well-known constituent of A. annua and it is overseen sometimes that other bioactive compounds are also present in the plant [37]. In two independent laboratories in Germany and China, two samples analyzed by TLC and another three samples by UHPLC showed that scopoletin was much more abundant than artemisinin in A. annua. Previously, other authors also reported that A. annua contains scopoletin [38,39]. In addition to scopoletin and artemisinin, it is well possible that some more phytochemicals of A. annua reveal cytotoxicity towards cancer cells [15]. A. annua is frequently used in Africa outside the approved clinical practice by indigenous people as herbal tea to treat malaria. Although hot water decoctions indeed show some activity against malaria, WHO warns against the use of herbal teas instead of approved artemisinin combination therapies (ACT). The apprehension is that suboptimal artemisinin concentrations in the tea may or may not treat malaria on the short run, but would foster the development resistance on the long run. Of course, that would be a fatal consequence that has to be avoided. However, several aspects have to be considered in this context.
Artemisinin is a rather water insoluble molecule and it cannot be assumed that herbal teas contain considerable amounts of artemisinin. The observed anti-malaria effects of A. annua decoctions might be due to other water-soluble compounds in the plant. The problem of poor water-solubility of artemisinin is known since the detection of artemisinin. Youyou Tu, who obtained the Nobel Price

Chemoprofiling
As a part of our ongoing research on the anticancer activity of A. annua, in the present study we focused on scopoletin. While artemisinin-based compounds are mainly found in A. annua alone, scopoletin can be found in other Artemisia species and also in other genera (Brunfelsia, Kleinhovia, Mallotus, Scopolia, Solanum, Urticaria, Viburnum). Although the use of chemoprofiles has been proposed as a taxonomical parameter, our results with 11 different Artemisia species do not support this concept. This observation is in accordance with data obtained for other plant genera [36].
Artemisinin is the most well-known constituent of A. annua and it is overseen sometimes that other bioactive compounds are also present in the plant [37]. In two independent laboratories in Germany and China, two samples analyzed by TLC and another three samples by UHPLC showed that scopoletin was much more abundant than artemisinin in A. annua. Previously, other authors also reported that A. annua contains scopoletin [38,39]. In addition to scopoletin and artemisinin, it is well possible that some more phytochemicals of A. annua reveal cytotoxicity towards cancer cells [15]. A. annua is frequently used in Africa outside the approved clinical practice by indigenous people as herbal tea to treat malaria. Although hot water decoctions indeed show some activity against malaria, WHO warns against the use of herbal teas instead of approved artemisinin combination therapies (ACT). The apprehension is that suboptimal artemisinin concentrations in the tea may or may not treat malaria on the short run, but would foster the development resistance on the long run. Of course, that would be a fatal consequence that has to be avoided. However, several aspects have to be considered in this context.
Artemisinin is a rather water insoluble molecule and it cannot be assumed that herbal teas contain considerable amounts of artemisinin. The observed anti-malaria effects of A. annua decoctions might be due to other water-soluble compounds in the plant. The problem of poor water-solubility of artemisinin is known since the detection of artemisinin. Youyou Tu, who obtained the Nobel Price for Physiology or Medicine 2015, obtained inconsistent antimalarial effects at the beginning. Only as she went back to the original historical records on the traditional use in Chinese medicine (Handbook of Prescriptions for Emergency Treatment, Hou Bei Ji Fang, by Hong Ge, 281-340 B.C.), she recognized that the press juice rather than hot water decoction was recommended [8]. Furthermore, plant extracts prepared by water decoction, press juice or other forms contain always more than one single phytochemical and are combination therapies. Hence, plant extracts act against the development of resistance, since different compounds reveal different modes of action. Indeed, this has recently been shown in animal experiments [40]. Whole plant preparations of A. annua overcame existing resistance to pure artemisinin in the rodent malaria Plasmodium yoelii. In a long-term artificial selection for resistance in Plasmodium chabaudi and stable resistance to A. annua whole plant preparations was achieved three times more slowly than stable resistance to pure artemisinin. The authors concluded that the resilience of the whole plant may be attributable to a multicomponent defense system, which is more effective than isolated single compounds.

Cross-resistance and Drug Resistance Mechanisms
The correlation analysis of scopoletin with a panel of standard anticancer agents revealed that scopoletin tends to be cross-resistant to drugs targeting DNA, e.g., DNA toposiomerase I or II inhibitors, alkylatin agents, or antimetabolites, but not to mitotic spindle poisons or antihormones. Given that scopoletin might be developed as anticancer drug in the future, combination therapy with classical DNA-damaging anticancer drugs might not lead to improved tumor cell killing because of resistance to drugs affecting tumor DNA. It is interesting that not only cellular response to scopoletin was not associated with anticancer drugs addressing other cellular targets, but also that some of them belong to novel categories of targeted chemotherapeutics such as tyrosine kinase inhibitors, mTOR inhibitors or epigenetic drugs. This specific cross-resistance profiles that spares these targeted drugs may open a therapeutic window for novel combination therapies of scopoletin. The fact that scopoletin derivatives have been developed that are active against human xenograft tumors in vivo [41,42] can be taken as a clue that this natural product may be subjected to drug development to find its way into the clinic.
A central topic of our study was to investigate whether scopoletin may be active against otherwise drug-resistant tumor cells. Because of resistance to a broad spectrum of anticancer drugs [28], novel drugs with improved features are desperately required in clinical oncology. Out of the 48 human ABC-transporters, P-gp/MDR1/ABCB1, ABCB5, MRP1/ABCC1 and BCRP/ABCG2 have been extensively shown to be associated with MDR phenomena in human tumors [18]. Therefore, we analyzed the role of these ABC transporters to scopoletin. We found that the expression and/or function of these efflux pumps were not related to the cytotoxic activity of scopoletin, indicating that MDR caused by these ABC transporters may not hamper the cellular responsiveness towards scopoletin. It has been recently shown that P-glycoprotein in the blood-brain barrier did not affect scopoletin transport [40], a result which is in agreement with our data. The role of other ABC transporters for cytotoxicity of scopoletin against cancer cells has not been investigated.
We also found that the expression of EGFR and mutation of TP53 did not affect scopoletin's activity in cancer cells, while mutations in RAS oncogenes were significantly correlated with scopoletin resistance. EGFR represents an important oncogene that triggers malignant growth [43,44]. EGFR-overexpression is associated with unfavorable response to chemotherapy and short survival times of cancer patients. Mutations in the EGFR gene confer drug resistance. Again, scopoletin inhibited cancer cell lines independent of their EGFR expression, indicating that this oncogene does not represent a resistance factor for scopoletin. These results fit to data published recently, which showed that Morinda citrifolia extract rich in scopoletin (and epicatechin) inhibited lung tumor growth in vivo and down-regulated EGFR and other tumor markers [45].
RAS mutations are known for quite a long time to cause resistance to DNA-damaging drugs and irradiation [46,47]. We found that resistance to scopoletin was also associated with the mutational status of RAS oncogenes. This result fits to the correlation of scopoletin's cellular responsiveness to those of DNA-damaging agents (see above), which supports the assumption that scopoletin might also act against tumor cells by damaging DNA.
There is a long-lasting debate in the literature about the clinical relevance of the tumor suppressor TP53 for therapy against radioresistance [27]. DNA lesions are recognized by p53 leading to cell cycle arrest and DNA repair. If damage exceeds the cellular repair capability, TP53 may sense the induction of apoptosis. Both DNA repair and apoptosis ultimately maintain health of the organism. TP53 mutations lead to p53 protein which lost its function. This in turn causes carcinogenesis and drug resistance. It was pleasing to observe that scopoletin activity is not related to the TP53 mutational status, indicating that TP53-mutated otherwise resistant tumors may still be responsive to scopoletin.

COMPARE and Hierarchical Cluster Analyses of mRNA Microarray Data
Many-if not all-natural products act in a multifactorial manner [41,42,44,45,[48][49][50][51][52][53][54]. Therefore, we performed COMPARE and hierarchical cluster analyses of transcriptome-wide microarray-based mRNA expressions. We identified a gene expression profile, which was significantly correlated with the log 10 IC 50 values of the NCI cell lines for scopoletin. Genes of diverse functional groups and signaling routes appeared. The identified genes belong to many different functional groups, although genes regulating the cell cycle and cell proliferation (CCNB1, CDKN3, KIF11, MAPRE2, MKI67, SMC3) as well as genes involved in signal transduction (ARRB2, ARHGAP12, CDC42BPA, MAPRE2, RHEB, TSC22D4) appeared more frequently than genes from other functional classes. This may indicate that these two cellular processes may be of specific relevance for the responsiveness of tumor cells towards scopoletin. Although the role of the specific genes identified here for cellular responsiveness towards scopoletin is still unknown, cell cycle regulating genes [46,47,49,50] and signal transduction pathways in general play a tremendous role for cancer drug resistance [51][52][53]. Therefore, genes involved in cell cycle and proliferation as well as signal transducing processes may contribute to sensitivity or resistance of tumor cells to scopoletin.
The fact that all 40 genes identified by COMPARE analysis harbored binding motifs in their gene promoter sequences and that scopoletin binds in silico to the NF-κB-DNA complex as well as the APT binding site of IKK speaks for a role of NF-κB in regulating cellular response to scopoletin. By using a SEAP-driven NF-κB reporter cell line, we found that scopoletin activated NF-κB. We suggest that NF-κB activation might confer drug resistance. This implies that NF-κB may be a resistance mechanism of cancer cells towards scopoletin and other drugs, although main resistance mechanisms such as ABC-transporters, EGFR, and TP53 do not affect cellular resistance to scopoletin.

Mutagenic and Genotoxic Potential of Scopletin
Scopoletin as well as 8-methoxypsoralen (8-MOP) are coumarin derivatives. 8-MOP is an approved drug, which is used together with UV-A light for photochemotherapy (PUVA therapy) of vitilago and for photophoresis of Sezary syndrome. There is controversy about the long-term cancer risk of PUVA therapy [60,61]. Therefore, the mutagenic and genotoxic potential of coumarin has been investigated, since coumarin is present in foods and used in cosmetic fragrances. The majority of investigations suggest that coumarin is not a genotoxic agent [62]. The contrasting data in animal and human experimental models may be explained by different metabolization routes. In human beings the prevalent detoxification pathway is the 7-hydroxylation pathway, whereas the 3,4-epoxidation pathway predominant in rats and mice leads to the formation of toxic, harmful metabolites [63]. Hence, coumarin in food and cosmetics bears no health risk for human subjects. The mutagenic potential of scopoletin is not well investigated as of yet. Scopoletin can even act in an anti-mutagenic manner by inhibiting the mutagenicity of cigarette smoke condensate [64]. Further safety assessment is necessary, if scopoletin would be further developed as anticancer drug.
The prospect that scopoletin might serve as lead compound for drug development and would find its way into the clinics, is supported by the favorable activity against tumors expressing well-known drug resistance mechanisms, although RAS mutations and NF-κB may hamper the effectiveness of scopoletin. Further investigations are warranted to explore the full therapeutic potential of this natural product.

Chemicals and Extracts
Artemisia samples were harvested in the years 1999 and 2000 and kindly provided by the TCM Hospital Bad Kötzting (Bad Kötzting, Germany). Artemisinin and scopoletin were used as standards. Artemisinin was isolated from A. annua, whereas scopoletin was obtained from Sigma-Aldrich (Taufkirchen, Germany). Three different batches of A. annua were purchased from Anguo Ruiqi Chinese Herbal Medicines Co., Ltd. (Anguo, China).

Thin Layer Chromatography
The thin layer chromatopgraphy (TLC) method was performed as described [64].

Preparation of Standard Solutions and the A. annua Methanol Extract
Artemisinin and scopoletin standard solutions were prepared by dissolving them with DMSO to a concentration of 50 mM, then diluted into 50 µM with methanol, and 1 µL of 50 µM standard solution was injected for analysis. The methanol extracts of A. annua was prepared as follows: 5 g A. annua powder was refluxed with 200 mL methanol for 1 h. Then, this procedure was repeated with 150 mL methanol for another 1 h. The solution was set to 350 mL, 1 mL of the extraction solution was filtered and 1µL was injected for analysis.

Cell Lines
A panel of 56 human tumor cell lines of the Developmental Therapeutics Program of the NCI (Bethesda, MD, USA) consisted of leukemia, melanoma, non-small cell lung cancer, colon cancer, renal cancer, ovarian cancer, breast cancer, and prostate carcinoma cells as well as tumor cells of the central nervous system [65]. Cells treated with scopoletin for 48 h were assayed by means of a sulforhodamine B assay [66].

Resazurin Assay
The cytotoxicity of scopletin has been assayed by resazurin reduction [24]. Resazurin as indicator dye is reduced to highly fluorescent resorufinin viable cells, but not in dead cells, which cannot metabolize resazurin anymore. We previously described the protocol for the resazurin assay and the cell lines (drug-sensitive CCRF-CEM leukemia cells and their P-gp-expressing multidrug-resistant CEM/ADR5000 subline) in detail [26,27]. The cells were treated with scopoletin for 72 h. Experiments were performed three times with at least six replicates per experiment. The 50% inhibition concentrations (IC 50 ) were calculated from dose response curves of each cell using Excel 2013 software (Microsoft, Redmond, WA, USA).

DNA Analyses with the NCI Cell Line Panel
DNA analyses focused on the measurement of gene copy numbers and point mutations. Gene copy numbers relative to normal female DNA were determined using array comparative genomic hybridization (ABCB1, ABCC1, EGFR). DNA samples were labeled with Cy3-dUTP and Cy5-dUTP were hybridized to OncoBAC DNA microarrays and quantitated. Normal human female DNA was used for the control.
Point mutations in the TP53 gene were detected by cDNA sequencing. PolyA mRNA was extracted from cells using a MicroFast Track mRNA isolation kit (Invitrogen, Waltham, MA, USA) and cDNA was prepared using a First-Strand cDNA synthesis kit (Pharmacia, Berlin, Germany) according to the manufacturer's recommendations. Three overlapping TP53-specific PCR primers covering the entire open-reading frame were used to bi-directionally sequence the TP53 cDNA using Taq dideoxy-sequencing methodology on an Applied Biosystems model 373A automated sequencer (Applied Biosystems, Waltham, MA, USA).
The transcriptional activity of TP53 cDNA from individual cell lines was measured in a yeast survival assay. The assay uses 3 centromeric plasmids. The 1st plasmid, pLS76 (positive control), has the LEU2 gene as a selectable marker and also full length wild-type human TP53 cDNA under the control of the ADH1 promoter. The 2nd plasmid, pSS16, is identical to pLS76 except that the wild-type TP53 sequence from codons 68 through to 347 is replaced by the selectable marker URA3. The 3rd plasmid, pSS1, contains a HIS3 gene under the control of a single TP53 binding site derived from the ribosomal gene cluster. Input TP53 was generated from polyA mRNA extracted from cells. TP53 cDNA was PCR amplified using Pfu polymerase (Stratagene, La Jolla, CA, USA) and yeast were co-transformed with the PCR generated TP53 and HindIII/StuI restricted plasmid pSS16 (which releases the URA3 gene segment). Repair of the gapped pSS16 plasmid with the TP53 derived from the input TP53 PCR product occurs in vivo through homologous recombination. Transformants which have successfully repaired pSS16 were selected on media lacking leucine and such replicates lack growth on uracil-minus media. If the yeast has been also transfected with pSS1 then growth on media lacking histidine identifies colonies that contain transcriptionally active TP53. Leu+ colonies containing only wild-type TP53 sequence grew successfully on plates lacking histidine. Leu+ colonies containing only TP53 loss-of-function mutations failed to grow on plates lacking histidine. Leu+ colonies derived from TP53 cDNA prepared from cancer cells containing heterozygous TP53 gene status yield approximately 50% His+ colonies.
To determine the mutational status in H-, Kand N-RAS genes, Genomic DNAs were isolated using a Cell Culture DNA kit (Qiagen, Hilden, Germany). Exons 1 and 2 of each RAS gene were amplified by PCR from the genomic DNA's using primer pairs specific for each exan/intron junction. PCR amplifications were performed in a GeneAmp PCR System 9600 (Perkin-Elmer, Rodgau, Germany) and were all preceded with one cycle of denaturation at 94˝C for 3 min and finished with the incubation at Products were purified from a 6% polyacrylamide gel. DNA sequence was determined for both strands directly from the purified PCR products using the same primers as those for PCR amplification. DNA sequencing reactions were carried out with the PRISM Ready Reaction DyeDeoxy Terminator cycle sequencing kit (Applied Biosystems, Waltham, MA, USA), and the products were analyzed on an ABI 373A DNA sequencer.

mRNA Analyses with the NCI Cell Line Panel
The transcriptome-wide mRNA expression measured by microarray hybridizations as well as data from diverse assays to validate the microarray results have been deposited at the website of the Developmental Therapeutics Program of the NCI (http://dtp.nci.nih.gov). The expression of ABCB1, ABCB5, and EGFR was measured utilizing Affymetrix Human Genome U95Av2 A-E arrays and the expression of ABCC1 and ABCG2 using Affymetrix U133A and U133B chips. Many of these probe sets can be linked to known genes (based on clustering information from UniGene and GeneCards, where information on possible function can be found). All gene assignments should be considered tentative. Sequence information for a probe set can be found through GenBank or UniGene.
For validation of microarray-based mRNA data, different methods have been used. For measurement of ABCB1 or ABCB5, total RNA was purified with the RNeasy kit. Levels of ABC transporters were measured by real-time quantitative RT-PCR using the LightCycler RNA Amplification SYBR Green kit and a Roche LightCycler machine. The gene-specific data was normalized to actin RNA levels, and are the average of two experiments.
Messenger RNA expression of ABCC1 was measured relative to SW620 cells. Total RNA was extracted and subjected to quantitative PCR using serially diluted samples. RNA isolated from SW620 was used as a reference standard. The level in the SW 620 cell line was arbitrarily assigned a value of 10, and all other values were determined relative to this. All quantifications were performed by densitometry.
Expression of EGFR mRNA expression was determined by RNAse protection assay. A 141bp fragment of EGFR cDNA, a 470-bp fragment of c-erbB2 cDNA, and a 603-bp fragment of TGFA DNA were used to generate 32 P-labeled antisense riboprobes. Thirty milligrams of total RNA were hybridized with 3ˆ10 5 cpm-labeled riboprobe and digested for 30 min at 25˝C with RNases A and T1. Following extraction, samples were separated by PAGE and autoradiographed. The relative amount of mRNA was quantified using a Phosphor Imager.

Protein Analyses with the NCI Cell Line Panel
To determine the protein expression of ABCG2, western blotting was performed. Cell pellets were solubilized in a lysis buffer, resolved by SDS-PAGE, transferred onto nitrocellulose membranes and incubated with an anti-BCRP polyclonal antibody. Blots were incubated with a peroxidase-conjugated donkey anti-rabbit secondary antibody and visualized using ECL Plus chemiluminescence detection kit (GE Healthcare, Chicago, IL, USA).
The EGFR protein expression was measured using Rev Prot Lysate Array. Protein lysates were diluted to a uniform protein concentration, denatured, and dilution series printed onto slides. These arrays were probed with antibody specific for the indicated protein, using the DAKO signal amplification system. P-glycoprotein encoded by ABCB1 acts as drug efflux pump. Rhodamine 123 is a fluorescent substrate for P-glycoprotein, which can be used to measure the functional activity of this drug transporter. Suspensions of logarithmic phase cells were obtained from tissue culture plates by trypsinization. During the accumulation period, four aliquots of cells were re-suspended in rhodamine-containing medium (improved minimum essential medium with 10% fetal calf serum and 0.5 µg/mL rhodamine 123) and incubated with or without 3.0 µg/mL cyclosporin A at 37˝C in 5% CO 2 for 30 min. After the accumulation period, efflux was initiated by sedimentation at 600ˆg and resuspension in rhodamine-free medium (improved minimum essential medium with 10% fetal calf serum), with or without 3.0 µg/mL cyclosporin A. The efflux was carried out at 37˝C in 5% CO 2 for 90 min. At the end of both the accumulation and efflux periods, cells were sedimented, washed in ice-cold Hank's buffered salt solution, placed in Hank's buffered salt solution with 10% fetal calf serum on ice, and kept in the dark until flow cytometric analysis. Samples were analyzed on a FACScan flow cytometer equipped with a 488 nm argon laser. The green fluorescence of rhodamine 123 was collected after a 530 nm band pass filter. Samples were gated on forward scatter versus side scatter to exclude cell debris and clumps. A minimum of 6000 events were collected for each sample. Histograms were generated for each cell line and the median rhodamine fluorescence obtained for each peak were identified by channel number, and this number (denoting the logarithm of fluorescence intensity) was used to indicate the amount of rhodamine in the cells in each incubation condition. A difference of 256 channel numbers represents a 10-fold change in fluorescence intensity.

Cell Doubling Times
The cell lines in the panel grow at different rates, which may affect sensitivity to certain compounds. Doubling time (in hours) was obtained by determining the change in SRB signal over time at the standard conditions and cell densities used for the NCI anticancer drug screen. These data are averages from 8 years of screening.

Statistical Analysis
Data on cytotoxicity (log 10 IC 50 values) and mRNA microarray data of the NCI tumor cell line panel [67] are deposited at the NCI website (http://dtp.nci.nih.gov). For hierarchical cluster analysis, objects were classified into dendrograms by calculating distance according to the closeness of between-individual distances by means of the Ward method (WinSTAT program, Kalmia, Cambridge, MA, USA). Cluster models have been previously validated for gene expression profiling and for approaching molecular pharmacology of cancer [68,69]. The application of this method for pharmacogenomics of phytochemicals has been described in detail [15].
COMPARE analyses were performed to produce rank-ordered lists of genes expressed in the NCI cell lines as previously described [70,71]. Briefly, every gene of the NCI microarray database was ranked for similarity of its mRNA expression to the log 10 IC 50 values for scopoletin. To derive COMPARE rankings, a scale index of correlation coefficients (R-values) was created.
Pearson's correlation test was used to calculate significance values and rank correlation coefficients as a relative measure of the linear dependency of two variables (WinSTAT, Kalmia). The chi-squared test was applied to bivariate frequency distributions of pairs of nominal scaled variables (WinSTAT, Kalmia). It was used to calculate significance values (p-values) and rank correlation coefficients (R-values) as a relative measure of the linear dependency of two variables.

Binding Motif Analysis
COMPARE analysis showed 40 genes that either directly or inversely correlated with log 10 IC 50 values of a panel of 60 NCI cell lines upon scopoletin treatment. Twenty five kilobases long upstream promoter sequences were retrieved from UCSC Genome Browser Gene Sorter (http://genome.ucsc.edu) and BED (Browser Extensible Data) files were created to screen for the possible binding motifs for each gene. Promoter sequences were screened using SeqPos tool implemented in Galaxy, Cistrome software [72].

Molecular Docking
Molecular docking was performed to calculate the interaction energy and geometry of scopoletin with I-κB kinase β, I-κB kinase β-NEMO (NF-κB essential modulator) complex, NF-κB, and NF-κB-DNA complex. The protocol for molecular docking was previously reported by us [33]. X-ray crystallography-based protein structures were obtained from Protein Data Bank [73,74]; I-κB kinase β (PDB ID:3RZF), I-κB kinase β-NEMO complex (PDB ID: 3BRT), NF-κB (p52/RelB heterodimer, PDB ID: 3DO7), NF-κB-DNA complex (p50/p65 heterodimer bound to DNA, PDB ID: 1VKX) and set as rigid receptor molecules. A grid box was then constructed to define docking spaces in each protein according to its pharmacophores. Docking parameters were set to 250 runs and 2,500,000 energy evaluations for each cycle. Docking was performed three times independently by Autodock4 and with AutodockTools-1.5.7rc1 [34] using the Lamarckian Algorithm. The corresponding lowest binding energies and predicted inhibition constants were obtained from the docking log files (dlg). Mean˘SD of binding energies were calculated from three independent dockings. Visual Molecular Dynamics (VMD) was used to depict the docking poses of scopoletin.

NF-κB Reporter Assay
HEK-Blue-Null1 was derived from the human embryonic kidney-293 (HEK-293) cell line and obtained from Invivogen (San Diego, CA, USA). These cells stably express an optimized secreted embryonic alkaline phosphatase (SEAP) reporter gene preceded by the NF-κB promoter.