Disulfiram Sensitizes a Therapeutic-Resistant Glioblastoma to the TGF-β Receptor Inhibitor

Despite neurosurgery following radiation and chemotherapy, residual glioblastoma (GBM) cells develop therapeutic resistance (TR) leading to recurrence. The GBM heterogeneity confers TR. Therefore, an effective strategy must target cancer stem cells (CSCs) and other malignant cancer cells. TGF-β and mesenchymal transition are the indicators for poor prognoses. The activity of aldehyde dehydrogenases (ALDHs) is a functional CSC marker. However, the interplay between TGF-β and ALDHs remains unclear. We developed radiation-resistant and radiation-temozolomide-resistant GBM models to investigate the underlying mechanisms conferring TR. Galunisertib is a drug targeting TGF-β receptors. Disulfiram (DSF) is an anti-alcoholism drug which functions by inhibiting ALDHs. The anti-tumor effects of combining DSF and Galunisertib were evaluated by in vitro cell grow, wound healing, Transwell assays, and in vivo orthotopic GBM model. Mesenchymal-like phenotype was facilitated by TGF-β in TR GBM. Additionally, TR activated ALDHs. DSF inhibited TR-induced cell migration and tumor sphere formation. However, DSF did not affect the tumor growth in vivo. Spectacularly, DSF sensitized TR GBM to Galunisertib both in vitro and in vivo. ALDH activity positively correlated with TGF-β-induced mesenchymal properties in TR GBM. CSCs and mesenchymal-like GBM cells targeted together by combining DSF and Galunisertib may be a good therapeutic strategy for recurrent GBM patients.


Introduction
Glioblastoma (GBM) is the most malignant primary brain tumor in adults. The standard of care (SoC) of GBM consists of neurosurgical removal followed by radiotherapy, chemotherapy, radiochemotherapy, or a tumor treating field device, such as co-adjuvant therapy [1]. Temozolomide (TMZ) is the most common drug used in GBM chemotherapy. Radiation is a powerful approach to controlling tumor growth and significantly improves patient survival [2]. However, the benefit of radiochemotherapy varies in the case of GBM [3]. According to the World Health Organization classification of tumors in the central nervous system, in spite of receiving SoC, the median time to recurrence is about 7 months, and median survival is about 15 months [4]. Developing therapeutic resistance in residual GBM results in the failure of SoC and leads to poor prognoses [5]. GBM is derived from transformed astrocytes and other neural progenitor cells [6,7]. The heterogeneity of GBM confers therapeutic resistance.

Developing Therapeutic Resistance in Glioblastoma Promotes TGF-β-Induced Mesenchymal-Like Phenotype
In our previous finding, the development of radiation resistance in glioblastoma (GBM) enhances TGF-β secretion and reduces temozolomide (TMZ) sensitivity [24]. Therefore, we sequentially challenged radiation-resistant (RR) GBM cells with TMZ to develop a radiation-TMZ-resistant (RTR) GBM cell model. RR GBM cells responding to TMZ dosages of 50, 100, 150, 200, and 250 µM were evaluated. A single treatment of 250 µM TMZ only inhibited the viability the RR GBM cells by about 30% (72.9 ± 2.46% for 1306MG R6T1; 62.4 ± 3.52% for U87MG R4T1). However, three consecutive treatments of 250 µM TMZ were unable to suppress RR GBM cell growth (85.5 ± 1.28% for 1306MG R6T3; 88.3 ± 2.15% for U87MG R4T3). Therefore, the cells with three TMZ treatments were defined as RTR GBM cells (1306MG R6T3 and U87MG R4T3) ( Figure S1). TGF-β is a key regulator of the epithelial-mesenchymal transition (EMT) or mesenchymal differentiation. Both TGF-β and mesenchymal properties are malignant indicators in cancer progression. We hypothesized that TGF-β-induced mesenchymal transition would be enhanced in both RR and RTR GBM. We first examined the TGF-β secretions. In comparison with parental GBM cells, the secretion of TGF-β (25.7 ± 2.69 pg/mL vs. 1607.7 ± 19.16 pg/mL vs.1207.0 ± 27.33 pg/mL for 1306MG, 3.5GR6, and R6T3; 995.4 ± 10.98 pg/mL vs. 1234.1 ± 23.10 pg/mL vs. 1058.1 ± 8.56 pg/mL for U87MG, 2GR4, and R4T3) increased in both RR and RTR ( Figure 1A). In addition, the downstream effectors of TGF-β, including TGF-β receptor I (TRI), TRII, Slug, and Snail were upregulated, and the phosphorylation on TRI, TRII, and Smad2 was enhanced in both RR and RTR ( Figure 1B). RR and RTR increased the expression of the mesenchymal markers, N-cadherin and Fibronectin ( Figure 1C). Furthermore, blockade of TGF-β signaling with 200, 400, and 600 nM of the TRs inhibitor LY364947 inhibited RR-induced phosphorylation on Smad2 and reduced the expression of N-cadherin and Fibronectin. In addition, we modulated the activity of TGF-β signaling through exogenous TGF-β1 or inhibiting TRs with LY364947 and Galunisertib. The results of MTT assay showed that LY364947 did not affect the cell viability of both parental and TR 1306MG and U87MG cells ( Figure S2A). On the other hand, up to 500 µM Galunisertib was able to inhibit approximately 40% of cell viability ( Figure S2B). Accordingly, we selected 4 µM LY364947 and 100 µM Galunisertib, which inhibited 20% of cell viability, for further investigation.

The Role of ALDH in Therapeutic Resistance
Cancer stem cell properties play a role in cancer progression. The activity and expression of ALDHs are the indicators of cancer stemness. ALDH activity was measured using an ALDEFLUOR assay. We found that developing RR and RTR activated ALDHs ( Figure 3A,B). Also, the expression of ALDH1 was upregulated in the RR and RTR GBM cells (1 ± 0 vs. 1.05 ± 0.04 and 1.10 ± 0.34, p = 0.0153 and 0.0017 for U87MG vs. 2GR4 and R4T3) ( Figure 3C). Disulfiram (DSF), the pan-ALDHs inhibitor, is an old drug used for alcoholism. DSF as a single or co-adjuvant treatment for cancers increases anti-tumor efficacy [25,26]. Additionally, DSF is able to enhance radiation responses and chemotherapy sensitivity in cancers, but its targets and the underlying mechanism are not known [27,28]. Therefore, we evaluated the cell response to different doses of DSF ( Figure S3A). A 20% dose of growth inhibition (GI 20 ) was selected for further investigation (8 µM for 1306MG, 3.5GR6, and R6T3; 150 nM for U87MG, 2GR4, and R4T3). NCT-501, a specific inhibitor of ALDH1A1, was selected to elucidate the role of ALDH1 in therapeutic resistance. We first evaluated the effects of NCT-501 on cell growth. As shown in Figure S3B, NCT-501 did not affect the cell viability.

Disulfiram Sensitizes Resistant GBM to Galunisertib
In order to examine the anti-tumor effects and toxicity of DSF and Galunisertib (G) alone, and that when combining DSF and Galunisertib (D + G), we developed orthotopic xenograft GBM model by injecting luciferase-expressing therapeutic-resistant GBM cells into the striatum of NOD/SCID mice. At day 10 post-transplantation, the mice were

Disulfiram Sensitizes Resistant GBM to Galunisertib
In order to examine the anti-tumor effects and toxicity of DSF and Galunisertib (G) alone, and that when combining DSF and Galunisertib (D + G), we developed orthotopic xenograft GBM model by injecting luciferase-expressing therapeutic-resistant GBM cells into the striatum of NOD/SCID mice. At day 10 post-transplantation, the mice were treated twice weekly with DSF, G, D + G, or DMSO as a vehicle (Veh). As shown in Figure 8A, in comparison with Veh, DSF and G did not affect the in vivo growth of RR U87MG 2GR4 cells, while D + G significantly suppressed the in vivo growth of RR U87MG 2GR4 cells ( Figure 8A) Figure 8C). In addition, a Kaplan-Meier analysis of the survival of the Veh-, DSF-, G-, and D + G-treated mice is presented in Figure 8D. The median survival of Veh-treated mice was 54 days while the median survival of DSF-, G-, and D + G-treated mice were not defined. In order to exclude the cell line-specific and resistance-specific effects, the luciferase-expressing RR 1306MG 3.5GR6 and RTR 1306MG R6T3 cells were injected intracranially into the NOD/SCID mice. The same procedures of IVIS observation and the treatments of Veh, DSF, G, and D + G were applied to the RR 1306MG 3.5GR6 and RTR 1306MG R6T3 cells in vivo growth. As shown in Figure 8E, compared with the Veh-treated groups, DSF and G did not affect tumor growth while D + G significantly delayed the in vivo tumor growth of RR 1306MG 3.5GR6. Furthermore, D + G showed better anti-tumor effects on RR 1306MG 3.5GR6 in vivo than DSF or Galunisertib alone ( Figure 8E). A two-way ANOVA revealed a main effect of DSF, G, D + G vs. Veh (F (3, 443) = 2.843, p = 0.0375), interaction (F (39, 443) = 0.8691, p = 0.6968), and days after treatment (F (13, 443) = 11.89, p < 0.0001) ( Figure 8F). In comparison with the Veh-treated groups, DSF, G, or D + G did not affect the body weight of the mice ( Figure 8G). However, the survival curve of DSF, G, D + G, and Veh did not show the differences ( Figure 8H). The similar phenomena were found in Figure 8I, only D + G was able to inhibit the in vivo growth of RTR 1306MG R6T3 cells while Veh, DSF, G did not affect the tumor growth. A two-way ANOVA revealed a main effect of DSF, G, D + G vs. Veh (F (3, 418) = 5.795, p < 0.0001), interaction (F (39, 418) = 0.7455, p = 0.8694), and days after treatment (F (13, 418) = 4.255, p < 0.0001) ( Figure 8J). Also, DSF, G, or D + G did not affect the body weight of the mice ( Figure 8K).
Conversely, we evaluated whether ALDH inhibition by DSF was able to inhibit therapeutic resistance-induced TGF-β signaling. As shown in Figure 9D,E, DSF suppressed the secretion of TGF-β in the RR and RTR GBM cells. DSF primarily reduced the phosphorylation of TRI. Also, it downregulated the expression of TRI in the RR and RTR GBM cells. In the RTR U87MG cells, DSF also inhibited the phosphorylation of TRII. In terms of intracellular signal transduction, DSF not only inhibited the phosphorylation of Smad2 in the RTR U87MG cells, but directly reduced the expression of Smad2 in the RR and RTR U87MG cells. In addition, the dose of DSF GI 20 downregulated the transcription factor Slug and reversed therapeutic resistance-induced MD by decreasing N-cadherin and Fibronectin ( Figure 9F,G). Furthermore, DSF and NCT-501 were able to de-sensitize RR and RTR GBM cells to exogenous TGF-β. Combining the inhibitory effects of the ALDHs and TGF-β receptors efficiently decreased the expression of mesenchymal N-cadherin and Fibronectin ( Figure 9H).

Discussion
GBM is the most common and lethal primary brain tumor in adults. Despite the neurosurgical intervention following standard radiation and TMZ, GBM easily recurs and results in dismal outcomes. These cells resist both radiation and TMZ. Due to the high cellular heterogeneity of GBM, an effective therapeutic strategy has to target both cancer stem cells and malignant-resistant cells [29]. In our experimental resistant cell model, the mesenchymal-like and cancer stem properties were enhanced in therapeutic-resistant GBM cells, which was attributed to activated TGF-β signaling and facilitation of the activity of ALDHs. TGF-β signaling and ALDHs are highly correlated to cancer progression [18,30], and EMT promotes cell motility, invasiveness, and even resistance to cancer therapies [31,32]. However, the interplay between TGF-β signaling and ALDH remains unclear. In certain cancers, blockade of TGF-β signaling can decrease ALDH activity [33]. In parallel, TGFβ-induced EMT promotes ALDH-modulated cancer stemness [34]. In contrast, other researchers have discovered that activating TGF-β signaling reduces the ALDH-positive cancer population [35], and inhibiting TGF-β signaling by LY364947 increases the ALDHpositive cell population [36]. In our resistant GBM model, inhibiting ALDHs did not affect the cellular behavior of parental GBM cells but reduced resistance-induced cell migration. Blockade of ALDHs inhibited the formation of tumor spheres. Additionally, inhibiting TGF-β signaling by LY364947 and Galunisertib reversed therapeutic-resistance-induced mesenchymal-like phenotype and inhibited cell motility. Exogenous TGF-β did not affect the activity of ALDHs, while LY364947 and Galunisertib downregulated ALDH activity. Conversely, inhibiting ALDHs suppressed the secretion of TGF-β of radiation-resistant and radiation-TMZ-resistant GBM cells. Furthermore, inhibiting ALDHs directly reduced the expression of TRI, Smad2, and Slug to inhibit TGF-β-induced mesenchymal-like phenotype. However, our study focused on the correlation between ADLH and TGF-β signaling in comprehensive population after experimental radiation alone or radiation followed with TMZ. In addition, although DSF was developed decades prior as an anti-alcoholism drug by inhibiting pan-ALDHs, the current studies reported that the effects of DSF may occur in an ALDH-independent manner [37]. Li et al. indicated that DSF can induce tumor cell death in an ALDH-independent manner [38]. In addition, DSF is reported as an inhibitor of NF-κB to interfere TGF-β-induced epithelial-mesenchymal transition in cancers [39]. In order to clarify the intact interaction between ADLH and TGF-β signaling, the selection of ALDH-positive subgroups and the major ALDH isoform(s) consisting of ALDEFLUOR activity need to be figured out. Also, for mechanistic investigation, gene modification, such as specific ALDH isoform knockdown/overexpression and functional mutation, is necessary. Therefore, in this study, we summarized that ALDH activity may be potentially involved in the regulation of TGF-β signaling and TGF-β-induced mesenchymal-like phenotype in this experimental therapeutic-resistant GBM model.
Otherwise, based on the above findings, DSF and Galunisertib were introduced to evaluate the therapeutic efficacy of inhibiting TGF-β signaling and ALDHs. DSF is an anti-alcoholism drug. In 2017, Karamanakos' groups treated a GBM patient with DSF in combination with the standard of care and improved progression-free and overall survival [28]. Although DSF has been suggested as an anti-tumor agent [40][41][42], and a phase II clinical trial indicated that DSF is safe and is well-tolerated, combining TMZ and DSF did not significantly improve the prognosis of recurrent GBM patients [43]. In addition, various studies have indicated that DSF may be a good sensitizer for radioand chemotherapy [44]. A combination of DSF and copper enhances radiosensitivity via inducing cell death or interfering with DNA repair [45,46]. Also, DSF can overcome drug resistance and potentiates anti-tumor efficacy [47,48]. On the other hand, Galunisertib is an approval drug targeting TGF-β receptors. In 2015, a clinical study indicated that Galunisertib was safe for patients with glioma [49]. However, Galunisertib in combination with standard TMZ-based radiochemotherapy or alone with lomustine did not significantly improve the survival of newly diagnosed or recurrent GBM patients [50,51]. Accordingly, combining DSF and Galunisertib may enhance the anti-tumor activity by dual targeting of cancer stem cells and malignant mesenchymal-like GBM cells. We found that combining DSF and Galunisertib showed better inhibitory effects on cell proliferation, mobility, and the growth of tumor spheres. Furthermore, due to the activation of TGF-β signaling and ALDHs by therapeutic resistance, the combined treatment had better anti-tumor effects in RR and RTR GBM cells than in parental GBM cells. In vivo, sole DSF and Galunisertib did not affect the growth of RR GBM. However, combining DSF and Galunisertib led to efficient suppression of resistant GBM growth. In spite of the in vitro and in vivo results showing the anti-tumor potential combining DSF and Galunisertib by targeting both cancer stem cells and mesenchymal-like cells, however, we only examined the inhibitory effects of combining DSF and Galunisertib by using tumor sphere formation assay. The effects of combination of DSF and Galunisertib on the self-renewal ability, the expression of cancer stem cell markers (such as CD133, SOX2, and NANOG), and the ability for differentiation into multiple cell types are necessary. Various cell-based assays have been developed for evaluating cancer stemness. For instance, secondary tumor sphere formation assay can be used to evaluating the ability of self-renewal [52]. Also, sorting out subpopulation by multiple markers is one of common manner to investigate the role of cancer stem cells in both basic and clinical cancer biology [53]. Furthermore, xenotransplantation with extreme limiting dilution assay is able to evaluate tumor-propagating and tumor-initiating abilities [54]. Therefore, investigating the effects of combining DSF and Galunisertib on the cancer stem cell properties is needed for elucidating the mechanism of combination action.

Developing Therapeutic-Resistant GBM Cell Lines
GBM cells were subjected to γ-ray irradiation by the accelerator in the Department of Radiation Oncology of Kuo General Hospital in Tainan. The median effective dose of irradiation was selected to develop the consecutive irradiation-induced resistance. Therefore, 6 times of 3.5 Gy irradiation were required for 1306MG (1306MG 3.5GR6) and 4 times of 2 Gy irradiation were required for U87MG (U87MG 2GR4). The indicators involved in radiation resistance were examined in the previos study [24]. In order to develop radiation-temozolomide (TMZ) resistance, the radiation-resistant 1306MG 3.5GR6 and U87MG 2GR4 cells were treated with 0 µM, 50 µM, 100 µM, 150 µM, 200 µM, and 250 µM TMZ to determine the dose of TMZ. After TMZ treatment, the cells with the selected dose of TMZ were amplified and consecutively treated sequential concentration of TMZ until the viability was significantly higher than the viability of the first treatment. To develop radiation-TMZ-resistant GBM cell lines, 250 µM TMZ treatments for 3 times to 1306MG 3.5GR6 (1306MG R6T3) and U87MG 2GR4 (U87MG R4T3) were required.

MTT Assay
4000 cells were incubated in 96-well plate for 24 h. The cultured media were replaced with media with different doses of LY364947 or Galunisertib for 24 h. The media of post-treated cells were replaced with 0.5 mg tetrazolium salt (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide or MTT) in 1 mL regular media for incubating for 3 h at 37°C with 5% CO 2 . The insoluble crystals were dissolved in DMSO. The colored solution was quantified by measuring absorbance at 575 nm using The FlexStation ® 3 Multi-Mode Microplate Reader (Molecular Devices, San Jose, CA, USA).

ALDEFLUOR Assay
Cells were incubated in 6-well plates and were subjected to single or combined treatments. The activity of ALDH was measured using an ALDEFLOUR Kit (StemCell Technologies, Vancouver, BC, Canada). The ALDH + cells were determined using CytoFlex-6 colors (Beckman Coulter, Brea, CA, USA).

Trypan Blue Exclusion Assay for Cell Viability
The post-treated cells were suspended with trypsin. The cell suspension was mixed with trypan blue at a ratio of 1:3. The cell concentration was counted with a cytometer, and the viability was calculated using the following equation: [(cell number of treatment)/(cell number of control)] × 100%.

Combination Index Calculation and Determination of Combined Effects
The combination index (CI) of combining Disulfiram and Galunisertib was calculated by using free CompuSyn software (ComboSyn, Inc., Paramus, NJ, USA). According to the cell viability curves of Disulfiram and Galunisertib, the cytotoxicity of combining Disulfiram and Galunisertib of 20% inhibiting cell viability, and half-dose of above combination, the CI values were calculated. The combined effect of combining Disulfiram and Galunisertib was determined by Chou's group reported criteria [55].

Wound Healing Assay
The cells were plated in 2-chamber culture inserts (ibidi). After the cells formed a monolayer, the inserts were removed to create a wound. The cells were then subjected to different treatments in serum-free media. The wound area was observed from 0 h to different timepoints. 18 h were required for 1306MG, and 24 h were required for U87MG. The area of wound was measured using ImageJ software. The migration rate was calculated using the following equation: the percentage of wound area = [(wound area at timepoint)/(wound area at 0 h)] × 100%.

Transwell Assay
The cell monolayers formed in the upper chambers of the Transwell inserts (CORN-ING). The cells were subjected to different treatments in serum-free media for 9 h, while the 10% FBS-containing media was added in the lower chamber. The cells in the lower Transwell chamber were fixed with paraformaldehyde, and the nuclei was stained with hematoxylin (Sigma, Merck, Darmstadt, Germany), as previously described [24].

Tumor Sphere Formation Assay
1200 cells were suspended in the media for the purpose of forming tumor spheres, as previously described [24]. Treatments were repeated ten times with DMEM/F12 containing 10 ng/mL EGF, 10 ng/mL bFGF, and B27 was supplemented twice weekly. The cells were subjected to different treatments for 21 days. The diameters of the spheres were measured using ImageJ software.

Animals
NOD.CB17-Prkdcscid/NcrCrl (NOD/SCID) mice were purchased from the Laboratory Animal Center, College of Medicine, National Cheng Kung University (NCKU). Five mice were housed in a cage at a controlled temperature (22 ± 2 • C), humidity (55 ± 5%), and a 12 h light/dark cycle. The mice were given free access to water and food. All procedures were approved by the Institutional Animal Care and Use Committee of the College of Medicine, NCKU, with project approval number (#107106 and #110013).

Orthotopic Xenograft Animal Model and Bioluminescence Imaging
GBM cells were stable labeled with fluorescent protein (GFP) and firefly luciferase (Luc) genes through a lentiviral infection. The cells were suspended by trypsin and PBS. After 600× g centrifuge for 5 min at room temperature, the suspension was removed. The cell pellet was re-suspended in PBS. Before adjusting the cell concentration, the percentage of viable cells were examined by trypan blue staining. The percentage of viable cells must be higher than 90%. 5 × 10 5 GBM cells in 2 µL PBS were inoculated orthotopically into 8 to 10-week-old male NOD/SCID mice. The cells were injected into the right side of brain, 0.5 mm anterioposterior and 2.0 mm mediolateral to the bregma, and at a depth of 3.0 mm. DMSO served as a vehicle control, and DSF (50 mg/kg BW) (D), Galunisertib (75 mg/kg BW) (G), and DSF + Galunisertib (D + G) treatments started from post-transplantation day 10 to day 55. Tumor growth was monitored with the IVIS spectrum Live Imaging System (IVIS-200, Xenogen, PerkinElmer, Inc., Waltham, MA, USA) twice weekly, as described previously [24]. The luciferase radiance was quantitated using Live Imaging Software (Xenogen, PerkinElmer, Inc., Waltham, MA, USA) and analyzed with GraphPad Prism software.

Statistics
All results are presented as means ± standard error of the mean (SEM). Independent in vitro experiments were conducted using an unpaired two-tailed Student's t-test, and a two-way ANOVA was used to analyze the differences in in vivo growth of GBM at different time points during each treatment. Significance was defined as a p-value less than 0.05.
Microsoft Excel 2013 and GraphPad Prism6 software (GraphPad Software, San Diego, CA, USA) were used for the statistical calculation and analyses.

Conclusions
In conclusion, we developed radiation resistance and radiation-TMZ resistance in GBM cells. Dual targeting on TGF-β and ALDH showed the potential of reversing resistance-induced malignancies. Furthermore, DSF was able to desensitize GBM cells to TGF-β stimulation and was also able to sensitize therapeutic-resistant GBM cells to Galunisertib ( Figure 10). Therefore, combining DSF and Galunisertib may be an effective therapeutic strategy by which to improve the prognosis of recurrent GBM patients. software.

Statistics
All results are presented as means ± standard error of the mean (SEM). Independent in vitro experiments were conducted using an unpaired two-tailed Student's t-test, and a two-way ANOVA was used to analyze the differences in in vivo growth of GBM at different time points during each treatment. Significance was defined as a p-value less than 0.05. Microsoft Excel 2013 and GraphPad Prism6 software (GraphPad Software, San Diego, CA, USA) were used for the statistical calculation and analyses.

Conclusions
In conclusion, we developed radiation resistance and radiation-TMZ resistance in GBM cells. Dual targeting on TGF-β and ALDH showed the potential of reversing resistance-induced malignancies. Furthermore, DSF was able to desensitize GBM cells to TGF-β stimulation and was also able to sensitize therapeutic-resistant GBM cells to Galunisertib ( Figure 10). Therefore, combining DSF and Galunisertib may be an effective therapeutic strategy by which to improve the prognosis of recurrent GBM patients. Figure 10. Summary of the activation of TGF-β-induced mesenchymal-like phenotype and ALDH in therapeutic resistance in GBM. Dotted red arrow means the potential interplay between ALDH activity and TGF-β/Smad2 signaling; the black solid up arrow means the development of therapeutic resistance Figure 10. Summary of the activation of TGF-β-induced mesenchymal-like phenotype and ALDH in therapeutic resistance in GBM. Dotted red arrow means the potential interplay between ALDH activity and TGF-β/Smad2 signaling; the black solid up arrow means the development of therapeutic resistance or TGF-β/Smad2 signaling and ALDH activity-induced phenotypes; the red solid up arrow means therapeutic resistance-increased ALDH activity.  Informed Consent Statement: Not applicable.

Data Availability Statement:
The data supporting the findings and conclusion of the study are included in this manuscript. The used software is given in the Materials and Methods section.