A Comparative Study of Tumor-Specificity and Neurotoxicity between 3-Styrylchromones and Anti-Cancer Drugs

Background. Many anti-cancer drugs used in clinical practice cause adverse events such as oral mucositis, neurotoxicity, and extravascular leakage. We have reported that two 3-styrylchromone derivatives, 7-methoxy-3-[(1E)-2-phenylethenyl]-4H-1-benzopyran-4-one (Compound A) and 3-[(1E)-2-(4-hydroxyphenyl)ethenyl]-7-methoxy-4H-1-benzopyran-4-one (Compound B), showed the highest tumor-specificity against human oral squamous cell carcinoma (OSCC) cell lines among 291 related compounds. After confirming their superiority by comparing their tumor specificity with newly synthesized 65 derivatives, we investigated the neurotoxicity of these compounds in comparison with four popular anti-cancer drugs. Methods: Tumor-specificity (TSM, TSE, TSN) was evaluated as the ratio of mean CC50 for human normal oral mesenchymal (gingival fibroblast, pulp cell), oral epithelial cells (gingival epithelial progenitor), and neuronal cells (PC-12, SH-SY5Y, LY-PPB6, differentiated PC-12) to OSCC cells (Ca9-22, HSC-2), respectively. Results: Compounds A and B showed one order of magnitude higher TSM than newly synthesized derivatives, confirming its prominent tumor-specificity. Docetaxel showed one order of magnitude higher TSM, but two orders of magnitude lower TSE than Compounds A and B. Compounds A and B showed higher TSM, TSE, and TSN values than doxorubicin, 5-FU, and cisplatin, damaging OSCC cells at concentrations that do not affect the viability of normal epithelial and neuronal cells. QSAR prediction based on the Tox21 database suggested that Compounds A and B may inhibit the signaling pathway of estrogen-related receptors.


Introduction
Many anti-cancer drugs have been reported to cause side effects such as oral mucositis, neurotoxicity, and extravasation (leak of intravenously injected drugs and fluids out of the blood vessels) in clinical practice [1][2][3][4][5][6][7][8][9]. Oral mucositis not only reduces the patient's quality of life due to accompanying pain, but lowers oral intake, therefore causing dehydration, and can trigger systemic bacterial infection and invasion. Reduced administration doses of chemotherapeutic drugs and delayed treatment schedules should reduce the therapeutic effects and survival of patients. Furthermore, there are currently no established prevention or treatment methods for oral mucositis [1][2][3]. Even if the drugs were withdrawn, the induced peripheral neuropathy may leave rare recovery and lifelong disability [4][5][6]. Extravasation is more likely to occur when the peripheral venous wall is fragile or when phlebitis occurs, leading to the constriction and clogging of blood vessels, and then inflammation and necrosis [7][8][9]. Therefore, it was necessary to find novel drugs with high tumor-specificity and low toxicity toward normal cells of the oral cavity and neuronal cells.
Sakagami et al. have reported previously that many anti-cancer drugs show strong toxicity to human normal oral epithelial cells (human oral keratinocyte (HOK), human gingival epithelial progenitor cell (HGEP)) and induce apoptosis (loss of microvilli on the cell membrane surface and activation of caspases) in in vitro experiments [10]. Sakagami et al. have established in vitro quantification method of anti-tumor activity, using four human oral squamous cell carcinoma (OSCC) cell lines (gingival-derived Ca9-22, tongue-derived HSC-2, HSC-3, HSC-4), three human mesenchymal oral normal cells (gingival fibroblast (HGF), periodontal ligament fibroblast (HPLF), and pulp cell (HPC)) and two human epithelial oral normal cells (HOK and HGEP) [11,12]. Tumor-specificity (TS) was calculated as the ratio of the mean CC 50 against normal cells to the mean of CC 50 against OSCC cells. When mesenchymal cells (M) or epithelial cells (E) are used, TS M and TS E are obtained, respectively. Since most anti-cancer drugs exhibit strong cytotoxicity against epithelial cells, TS M values were adopted at the initial stage of a comprehensive search for new anti-tumor substances. Using this method, Sakagami et al. found that (i) most anti-cancer drugs show one to three orders of magnitude higher tumor-specificity against OSCC cell lines as compared with human oral mesenchymal normal cells yielding the TS M value of 10-1000, (ii) three major polyphenols, namely lignin glycosides, tannins, and flavonoids, show a much lower TS M value of 1-10, and (iii) among 12 series of chromone derivatives and 5 series of esters and amides (a total of 291 compounds), 7-methoxy-3-[(1E)-2-phenylethenyl]-4H-1-benzopyran-4-one (Compound A) and 3-[(1E)-2-(4-hydroxyphenyl)ethenyl]-7-methoxy-4H-1-benzopyran-4-one (Compound B) showed the greatest tumor selectivity (TS M = 301 and 182, respectively) [13,14]. However, Compounds A and B induced G 2 /M phase accumulation [14], suggesting the possible induction of strong neurotoxicity.
Sugita and Takao et al. manufacture four or five newly synthesized chromone derivatives every year, expecting to find more active and less adverse compounds than Compounds A and B. This time, Sugita and Takao et al. newly synthesized a total of 65 compounds: 2-indolylchromones (9 compounds: Series A) [15], indole-auron derivatives (10 compounds: Series B) [16], capsaicin derivatives (23 compounds: Series C) [17,18], 6,7styrylchromone derivatives (12 compounds, Series D) [19] and benzylidene chromanones (11 compounds: Series E) [20] (Figure 1). In the present study, whether the TS M values of these newly synthesized compounds exceed those of Compounds A and B was first investigated. Secondly, since this was not the case, we further confirmed the superiority of Compounds A and B over newly synthesized materials, and investigated the TS M , TS E, and neurotoxicity of Compounds A and B in comparison with four positive controls-5-FU, cisplatin (CDDP) and docetaxel (DTX)-used clinically for the treatment of OSCC [21], and doxorubicin (DOX). Thirdly, the possible signaling pathway of Compounds A and B was investigated using QSAR prediction based on the Tox21 database.  -22). D/B is the ratio of mean CC50 (f two or three normal oral cells)/mean CC50 (for two or four OSCC cell lines), while C/A is the ratio mean CC50 (HGF)/CC50 (Ca9-22) ( Table S1). The CC50 of Compounds A and B were cited from [14 Red in the figure indicates the most potent compound(s) in each series of compounds.  -22). D/B is the ratio of mean CC 50 (for two or three normal oral cells)/mean CC 50 (for two or four OSCC cell lines), while C/A is the ratio of mean CC 50 (HGF)/CC 50 (Ca9-22) ( Table S1). The CC 50 of Compounds A and B were cited from [14]. Red in the figure indicates the most potent compound(s) in each series of compounds.
Capsaicin derivatives 23 compounds (Series C) were synthesized by the condensation of various fatty acid chlorides with vanillylamine derivatives [17,18]. 6,7-styrylchromone derivatives 12 compounds (Series D) were synthesized by the coupling of bromochromones with various styrene derivatives using the Heck reaction [19].
We have established a simple method for preparing differentiated PC-12 cells by omitting the toxic coating step, using nerve growth factor (NGF) in a serum-free medium without a medium change [24]. We confirmed that differentiated PC-12 cells stop growing and express neurites tangles like the primary culture of neurons. On the other hand, we failed to differentiate SH-SY5Y cells even if we added sequentially all-trans retinoic acid (10 µM) and brain-derived neurotrophic factor (100 ng/mL) treatment for 9 days (data not shown) as recommended by previous report [25]. We did not further differentiate LY-PPB5 cells, since this cell already expresses differentiated phenotypes such as vimentin, S-100 protein, neuron-specific enolase, myelin basic protein, and glial fibrillary acidic protein in varying degrees, indicating neurogenic derivation [26].
Gingival epithelial progenitor cells HGEP were purchased from CELLnTEC and cultured in a Cnt-PR medium [10].

Cytotoxicity Assay
After detachment of cells with 0.25% trypsin solution (containing EDTA), 0.1 mL of cell suspension (2 × 10 4 /mL) was seeded in a 96-microwell plate with triplicate, and cultured in a 5% CO 2 incubator at 37 • C for 48 h to achieve complete attachment to the plate. After the replacement of fresh medium containing samples of various concentrations, cells were further cultured for 48 h to measure the absorbance at 560 nm (that reflects the viable cell number) by the MTT method. All samples were dissolved in DMSO. The toxicity of DMSO alone was calculated and subtracted. From the dose-response curve of test samples, 50% cytotoxic concentration (CC 50 ) was determined.

Tumor-Specificity and Neurotoxicity
Tumor-specificity (TS) was calculated as the ratio of the mean of CC 50 for normal cells to that for OSCC cells. When mesenchymal cells (M) (HGF, HPLF, HPC) or epithelial cells (E) (HGEP) were used, TS M and TS E were obtained, respectively. TS M = mean CC 50 against two to three normal human oral mesenchymal cells/mean CC 50 against two to four OSCC cell lines, as shown by D/B in Table S1, Figure 1, and Table 1. Since both Ca9-22 and HGF cells were derived from gingival tissue, the relative sensitivity of these cells was also compared (as shown by C/A in Tables 1 and S1). TS E = CC 50 (HGEP)/CC 50 (OSSC).
The ratio of mean CC 50 for three popular neuronal cells (PC-12, SH-SY5Y, LY-PPB6) or differentiated dPC-12 to that for OSCC (defined as TS N and TS DN , respectively) indicates the safety of neurotoxicity. When the TS N or TS DN is higher, the ratio of anti-cancer activity to neurotoxicity is higher. The compounds having the highest TS M , TS E , TS N , and TS DN are the best compounds.

Calculation of Chemical Descriptors
The activities against 59 signaling pathways [27], agonist and antagonist activities of the nuclear receptor, and stress response pathway were calculated by the chemical structures. In other words, all chromone derivatives were classified as positive or negative based on the calculated probabilities in Tox21 activity scores of 1 or higher for each signaling pathway using the Toxicity Predictor, a QSAR based on machine learning models trained on the Tox21 10K compound library [27]. Through Toxicity Predictor, we can wash and standardize the SMILES strings (salt, counterions, fragment removal, and adjustment of protonation state (neutralization)) to determine the optimal 3D conformer. The optimized molecular structures were confirmed using MarvinView (ChemAxon Kft., Budapest, Hungary).

Statistical Processing
All experiments were conducted in triplicate, and the average value was represented as the mean ± standard deviations (SD) of triplicate determinations. The significance of values was examined by one-way analysis of variance (ANOVA) and appropriate Bonferroni's post-test. A value of p < 0.05 was considered to indicate statistically significant differences.

Continuous Search for New Chromone Derivatives with Higher TS M -Confirmation of Prominent Tumor-Specificity of Compounds A and B
We are continuing to search for new chromone derivatives having higher TS M than Compounds A and B. In the present study, five series of chromones derivatives were investigated for this purpose ( Figure 1) (CC 50 values described in Supplementary data Table S1).

Rapid Decay of Cell Growth by Chromone Derivatives
Ca9-22 cells were treated with Compound A or selected seven compounds at various times, replaced with fresh drug-free medium, and incubated for 48 h after the start of drug administration, and then viable cell numbers were then determined by the MTT method ( Figure 2). Cytotoxicity of chromone derivatives (Series A, C, D, and E compounds) reached the maximum after a 24 h treatment. On the other hand, series B, which belongs to indole-auron derivatives (closed by blue circles), showed rather cytostatic growth inhibition even after 48 h incubation. This suggests that cytotoxic, rather cytostatic growth inhibition, is characteristic of the active chromone derivatives.

Higher Tumor-Specificities and Less Neurotoxicities of Compounds A and B Than Those of Popular Anti-cancer Drugs
The 50% cytotoxic concentration (CC50) of Compound A to OSCC (Ca9-22, HSC-2) was below 1 µM, as little as 1/400 of CC50 for normal oral epithelial system (HGEP) and

Higher Tumor-Specificities and Less Neurotoxicities of Compounds A and B than Those of Popular Anti-Cancer Drugs
The 50% cytotoxic concentration (CC 50 ) of Compound A to OSCC (Ca9-22, HSC-2) was below 1 µM, as little as 1/400 of CC 50 for normal oral epithelial system (HGEP) and mesenchymal cells (HGF, HPC). The CC 50 values for neuronal cells (PC-12, SH-SY5Y, LY-PPB6) were distributed between OSCC and normal oral cells ( Figure 3A). Compound B showed a similar distribution pattern ( Figure 3B). When PC-12 cells were cultured in a medium containing 50 ng/mL NGF in a serum-free medium for 2, 5, and 7 days, differentiated dPC-12 cells with elongated neurites gradually increased (Supplementary Figure S1). Differentiated dPC-12 cells with much reducing proliferating activity (closed green) showed similar sensitivity to Compounds A and B with parent PC-12 cells that rapidly grow (open green) ( Figure 3A,B, closed green). This indicated that the higher sensitivity of neuronal cells does not depend on their growth potential.  5-FU showed cytostatic growth inhibition of cancer cells and epithelial normal cells, without completely killing them ( Figure 3C). Cisplatin damaged neuronal cells more potently than OSCC cells ( Figure 3D). DOX damaged neuronal cells, normal mesenchymal and epithelial cells ( Figure 3E). Docetaxel (DTX) at as little as 39 nM showed cytostatic growth inhibition against epithelial normal cells ( Figure 3F). In contrast, Compounds A and  Figure 3C). Cisplatin damaged neuronal cells more potently than OSCC cells ( Figure 3D). DOX damaged neuronal cells, normal mesenchymal and epithelial cells ( Figure 3E). Docetaxel (DTX) at as little as 39 nM showed cytostatic growth inhibition against epithelial normal cells ( Figure 3F). In contrast, Compounds A and B showed no keratinocyte toxicity up to 400 µM ( Figure 3A,B).

5-FU showed cytostatic growth inhibition of cancer cells and epithelial normal cells, without completely killing them (
Based on these results, the tumor-specificity and neurotoxicity of Compound A and Compound B were compared with those of four typical anti-cancer drugs in newly performed experiments (Table 1) All six drugs used in this study showed strong neurotoxicity. Among these, Compound A (D/A = 11.1) showed the weakest neurotoxicity, followed by DTX (D/A = 7.9) and Compound B (D/A = 3.1). DOX, cisplatin, and 5-FU were found to damage neuronal cells at the concentrations that damage OSCC (D/A = 1.2, 1.3, 0.3) ( Table 1). Table 1. Higher tumor-specificities and lesser neurotoxicities of Compounds A and B than those of popular anti-cancer drugs. The CC 50 values were calculated using the data in Figure 3.

Rational of Using Human Normal Oral Cells as a First Stage of Screening of High TS M Cells
Since many anti-cancer drugs are known to induce oral mucosal diseases and neurotoxicity, it is important to manufacture compounds that have higher tumor-specificity and lower adverse effects. In addition, considering future clinical applications, it is desirable to use human-derived cancer cells and normal cells as target cells. In this study, anti-cancer drugs (DTX, cisplatin, 5-FU, DOX) used as positive controls showed high cytotoxicity against normal epithelial cells (Table 1); therefore, human oral squamous cells (OSCCs) and human mesenchymal oral normal cells were used to quantify tumor selectivity at the initial screening stage.

Failure of 65 Newly Synthesized Chromones Derivatives to Exceed the TS M of Chromones A and B
Previous studies have reported that among 291 chromone derivatives, Compounds A and B showed the greatest tumor selectivity [13,14]. In this study, we challenged ourselves to investigate a total of 65 new compounds (classified into five groups), hoping we may find more potent anti-tumor substances. Contrary to our expectation, the TS M of any of these compounds did not exceed that of Compounds A and B (Figure 1). This shows that Compounds A and B are the most potent chromone derivatives so far investigated. Based on these results, the side effects of Compounds A and B and four anti-cancer drugs were next compared at the same time. cisplatin (6.5) > 5-FU (4.5), confirming our previous findings [10,13,14,28]. Since 5-FU showed only cytostatic growth inhibition rather than cytotoxic action ( Figure 3C), concomitant use of other anti-cancer drugs may be a good choice. Actually, 5-FU combined with oxaliplatin significantly increased the survival of cancer patients [29].

Keratinocyte Toxicity
When tumor-specificity was evaluated with OSCC and human normal epithelial cell lines, TS E values of Compounds A and B (661.8 and 479.4, respectively) were approximately two orders higher than those of DTX and DOX (5.1 and 2.3, respectively) ( Table 1). Sakagami et al. also reported previously that (E)-3-(4-hydroxystyryl)-6-methoxy-4H-chromen-4-one showed one to two-fold higher TS E values than DOX and 5-FU, and induced the mitochondrial vacuolization, autophagy suppression followed by apoptosis induction, and changes in the metabolites involved in amino acid and glycerophospholipid metabolisms [30].

Neurotoxicity
The ratio of OSCC toxicity/neurotoxicity (equivalent to TS N ) was investigated. Compound A showed the highest TS N value (TS N = 11.1, TS ND = 12.4), followed by DTX (7.9) > Compound B (TS N = 3.1, TS ND = 5.6) > cisplatin (TS N = 1.3) > DOX (TS N = 1.2) > 5-FU (TS N = 0.3). These data suggest that Compounds A and B, and DTX shows higher cytotoxicity against OSCC than neuronal cells. On the other hand, DOX, cisplatin, and 5-FU damaged both OSCC and neurons to a comparable extent. Paclitaxel-induced neurotoxicity has been reported to be suppressed by the addition of antioxidants (such as docosahexaenoic acid, acetyl-L-carnitine hydrochloride, N-acetyl-L-cysteine, and sodium ascorbate) in cultured cells [31]. Also, in animal studies in rats, DOX-induced neurotoxicity and behavior were potentially protected by coenzyme Q10 [32], and DOX-induced cardiotoxicity is significantly suppressed with candesartan and quercetin [33]. For Compound A, we plan to search for drugs that alleviate the neurotoxicity and enhance the toxicity to OSCC.

Search for Target Molecules
Both Compound A and docetaxel accumulated Ca9-22 cells in the G 2 /M phase, but the former has a cytocidal action, while the latter exhibits a cytostatic effect ( Figure 3A,F), confirming previous report with Compound A [14] and docetaxel [28,31], suggesting the different site of action. The possibility of microtubule inhibition in neurotoxicity should also be investigated [34,35].
At present, the target site of Compounds A and B is not identified. Nuclear receptors and stress response pathways that are possibly involved in the inhibition of OSCC growth by Compounds A and B were searched using Toxicity Predictor [23]. The specific cytotoxicity of 14 chromone derivatives, including Compounds A and B against OSCC cells, were correlated with estrogen-related receptor inhibitory activity (ERRPGC_ant) in the presence of PPARγ activators (Figure 4), but not with other 58 signaling pathways (Tables 2 and  S2). These data suggest that Compounds A and B may inhibit the signaling pathway of estrogen-related receptors.
Medicines 2023, 10, x FOR PEER REVIEW 2 of 5 by Compounds A and B were searched using Toxicity Predictor [23]. The specific cytotoxicity of 14 chromone derivatives, including Compounds A and B against OSCC cells, were correlated with estrogen-related receptor inhibitory activity (ERRPGC_ant) in the presence of PPARγ activators (Figure 4), but not with other 58 signaling pathways (Tables 2  and S2). These data suggest that Compounds A and B may inhibit the signaling pathway of estrogen-related receptors. In silico study suggests the inhibition of Compound A against the estrogen-related receptor-alpha signaling pathway, which is identified as an adverse marker for breast cancer progression and poor prognosis. Data on CC50 and TSM values were derived from a previous study [14].      Table S2.
There was a possibility that the higher drug sensitivity of neuronal cells may be due to their high proliferative capacity. However, this possibility seems to be low, since both undifferentiated PC-12 cells (rapidly growing) and differentiated PC-12 cells (growth retarded) [24] showed comparable sensitivity to Compounds A and B (Figure 3A,B). Experiments using primary neurons that have stopped dividing and maintained their nerve function may support this point.

Conclusions
Chromone ring is widely distributed into flavonoids such as flavanol, flavone, flavanone, and isoflavone. Compound A has a styryl series attached to a chromone ring ( Figure 2) derived from natural products and, therefore, can be easily adaptable to living organisms. Compound A has higher tumor selectivity against OSCC cells than the low-molecular polyphenols and anti-cancer drugs 5-FU, cisplatin, and DOX, which many researchers have worked on, and has weaker keratinocyte toxicity and neurotoxicity. It is thus expected to be a potential lead compound for the discovery of new oral cancer therapeutic drugs.
For the clinical application, the production rate of Compounds A and B per month (at present, 50~60 mg each) should be scaled up. The present study suggests that Compounds A and B may destroy oral cancer cells that express estrogen receptors [37,38] without affecting normal cells in the oral cavity and the nervous system. Potential inhibition of estrogen-related receptor signaling pathways opens up possibilities for further research and application of these compounds in the field of hormone-dependent cancer and other diseases.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/medicines10070043/s1, Figure S1: Induction of neurite extension in PC-12 cells by NGF, Table S1: Tumor-specificity of newly synthesized 65 chromone derivatives, Table S2: Search for signaling pathway involved in 3-styrylchromone induced-induced selective tumor-specificity against human OSCC cell lines. Institutional Review Board Statement: Ethical review and approval were waived for this study since we used only cultured cells that had been purchased from the RIKEN Cell Bank and the human normal oral cells established in our laboratory 11 years ago according to the guidelines of the intramural Ethics Committee (No. A0808).

Informed Consent Statement:
Not applicable since we used only cultured cells.

Data Availability Statement:
The data that support the findings of this study are available from the corresponding author upon reasonable request.