Targeting PI3K/Akt/mTOR Pathway Identifies Differential Expression and Functional Role of IL8 in Liver Cancer Stem Cell Enrichment

Hepatocellular carcinoma (HCC) has become a major global burden due to significantly increased incidence in the world; it constitutes 7% of cancer deaths and is one of the leading cause of cancer mortality globally (1). In the United States, liver cancer death rates increased 43% between 2000 and 2016 (2). Increased life expectancy and growth in old age population across developed countries have also affected cancer types previously not accounted for (3). Injured hepatocytes due to various etiologic factors (viral, alcohol abuse, toxins, obesity, and metabolic syndrome) initiate chronic inflammatory disease state which involves death and regeneration cycles of hepatocytes leading to cirrhosis and eventually to the progression toward HCC (4). In this process, hepatic progenitor cells (HPC) are damaged and transformed to liver cancer stem cells (LCSC) which are responsible for chemoresistance, tumor relapse, and metastasis (5). Sorafenib and Regorafenib are multikinase inhibitors, which are used for the treatment of advanced HCC patients. However, HCC patients acquire frequent resistance to Sorafenib (3). The underlying mechanisms of this resistance yet to be solved involve crosstalk of PI3K/Akt and JAK-STAT pathways, epithelial–mesenchymal transition events, leading to the enrichment of LCSCs (6). Therefore, identification of protein targets involved in the enrichment of LCSCs will be essential for the development of novel therapeutic approaches for HCC. The PI3K/Akt/mTOR pathway is one of the most critical pathways involved in the development and progression of many cancer types including HCC (7). Under normal circumstances, the activity of this pathway is under the control of tumor-suppressor protein PTEN. However, anomalies in PTEN function are commonly observed in HCC leading to the hyperactivation of PI3K/Akt/mTOR pathway (8). In addition, this pathway is involved in crosstalk with the JAK-STAT and MAPK pathways upon treatment with Sorafenib, which has been considered as one of the mechanisms associated with acquired Sorafenib resistance in HCC (9). Therefore, targeting the PI3K/Akt/mTOR pathway with various inhibitors has been highlighted for its therapeutic potential in combination with Sorafenib. Inflammation in chronic liver disease and cancer plays a vital role in differentiation and proliferation of HPCs. During this transformation, hepatocytes activate the resident immune cells that result in release of cytokines and chemokines into the microenvironment to recruit immune cells (10). These events cause hepatic regeneration and tumor progression to take place leading to HCC. Serum cytokine levels in HCC patients have been associated with poor prognosis and aggressive behavior of cancer cells and induce cancer stem cell–like properties (11).

Here, we report that LCSC population in HCC cell lines is affected differentially by different PI3K/AkT/mTOR pathway inhibitors, where mTOR inhibitor Rapamycin was found to inhibit the enrichment of LCSCs and suppress the upregulation of IL8 triggered by Sorafenib. Our results also revealed IL8 as a potential target for CSCs, where IL8 blockade or downregulation prevented enrichment of LCSCs significantly.

Detailed experimental procedures were given in Supplementary documentation.

HCC cell lines were obtained from the following resources: Hep3B (ATCC-HB-8064), SNU475 (ATCC-CRL-2236), Huh7 (ATCC-JCRB0403), and Mahlavu (12). All cell lines used in this study are short tandem repeat authenticated, and contamination with mycoplasma was regularly tested with mycoplasma test kit (MycoAlert, Lonza). The passaging of the cells was minimized throughout the experiments (two times a week, maximum 8–10 passages). Huh7, Hep3B, and Mahlavu cells were maintained in DMEM (Invitrogen GIBCO), supplemented with 10% FBS (Invitrogen GIBCO), and 0.1 mmol/L nonessential amino acid, whereas SNU475 cells were maintained in RPMI (Invitrogen GIBCO), supplemented 10% FBS and 2 mmol/L l-glutamine. Cells were grown in a humidified incubator at 37°C and 5% CO₂ with mediums containing 100 units/mL penicillin and streptomycin.

Cells were fixed with 4% paraformaldehyde for 20 minutes and stained with anti-CD133/1 (AC133)-Biotin, anti-biotin-PE, anti–EpCAM-FITC, or anti–CD90-FITC antibodies according to the manufacturer's protocol (Miltenyi). For stemness protein detection, cells were permeabilized with 90% ice-cold methanol 10 minutes prior to blocking, and NANOG and OCT4 antibodies, and anti-rabbit IgG-Alexa 647 were used for staining. Analyses were done using BD Accuri C6 Flow Cytometer (BD Biosciences) or Novocyte flow cytometer (ACEA Biosciences). CD133, EpCAM-positive, and -negative cells were sorted with FACSMelody Cell Sorter (BD Biosciences), and their purity was assessed using flow cytometry.

Cells were inoculated into 96-well plates (1,000–2,000 cells/well). After 24 hours, cells were treated with the inhibitors and DMSO control in 40 to 0.1 μmol/L concentrations. Cells were fixed after 72 hours by cold 10% (w/v) trichloroacetic acid (MERCK) for an hour. Cells were washed 3 times with ddH₂O and then 50 μL of 0.4% sulforhodamine B (SRB) dye (Sigma-Aldrich) was applied to each well and incubated at room temperature for 10 minutes. Wells were washed with 1% acetic acid for 4 times and left for air-drying. SRB dye was solubilized in 100 μL/well 10 mmol/L Tris-Base solution, and absorbance was measured at 515 nm (ELx800, BioTek).

Cells formed spheres in ultralow attachment 96-well plates for 6 to 12 days containing DMEM/F12 serum-free medium (Gibco, Invitrogen) supplemented with EGF (20 ng/mL; Thermo Fisher), basic fibroblast growth factor (10 ng/mL; Sigma), B27 supplement (1:50; Invitrogen), Heparin (2 μg/mL; Sigma), insulin (5 μg/mL; Sigma), hydrocortisone (0.5 μg/mL; Sigma), and 100 units/mL penicillin and streptomycin (Gibco, Invitrogen).

Huh7 cells were treated with the optimum mutual inhibitory concentration (OMIC) values of the inhibitors and DMSO controls for 72 hours, then total RNA was extracted (RNeasy Mini Kit Qiagen), and nCounter PanCancer Panel gene expression analysis was performed as instructed by the manufacturer. Quality control, normalization, and differential expression analysis were done with nSolver 3.0.

Differential expression values (Supplementary Tables S1 and S2) were ranked, and the lists were uploaded to Cytoscape 3.5 software platform and analyzed with the pathway scoring application on “Pathways in cancer (hsa05200)” KEGG pathway. The enrichment scores assigned by the algorithm reflect the biological response of a specific process in the pathway.

Three different IL8 siRNAs (Hs_IL8_6 FlexiTube siRNA_validated, Hs_IL8_5 FlexiTube siRNA_validated, and Hs_IL8_4 FlexiTube siRNA) were purchased from Qiagen (Qiagen Inc.). Hs_IL8_6 FlexiTube siRNA was used for IL8 knockdown. Cells were seeded onto 6-well plates (105 cells/well). After 24 hours, at 50% to 70% confluency, cells were treated with Hs_IL8_6 FlexiTube siRNA_validated (50 nmol/L) or negative control siRNA (si-neg) using HiPerFect transfection reagent according to the manufacturer's instructions. Media were replaced with 10% complete DMEM after 16 hours, and cells were collected after 24, 48, or 72 hours for further experiments.

RevertAid First Strand cDNA synthesis kit (Thermo Scientific) was used for cDNA synthesis from Total RNA. Quantitative PCR was performed using Light Cycler 96 Real-Time PCR System with LightCycler 480 SYBR Green I Master (Roche). The experimental Ct (cycle threshold) was calibrated against GAPDH. The ΔΔCt method was used to determine the amount of target gene in inhibitor treated cells relative to DMSO-treated cells.

Anti-human IL8 ELISA studies were conducted using Nunc MaxiSorp ELISA plates (Millipore Sigma) and human IL8 (CXCL8) ELISA development kit (Mabtech, Sweden) as recommended by the manufacturer. Optical density values were detected on an ELISA reader (Spectramax, Molecular Devices) at 405 nm.

Protein electrophoresis (Mini-PROTEAN Tetra Cell Systems and TGX precast gels) and transfer system (Trans-Blot Turbo Transfer System) were used according to the manufacturer's protocol for Western blot analysis. Proteins were transferred to a PVDF-LF membrane and visualized using C-Digit or Odyssey Clx imaging system (Ll-COR) after incubation with primary and secondary antibodies (provided in Supplementary Information).

Huh7 cells were inoculated into 16-well E-plates (8,000 cell/well). After 16 hours, cells were either transfected with IL8 siRNA (experimental group) or left inside culture media (control group) for 16 hours. Then, cells were treated with Sorafenib to be monitored for at least 96 hours by real-time cell electronic sensing system (RT-CES), xCELLigence SP system (ACEA Biosciences), as recommended by the manufacturer. Cell index values were used to plot growth curves of cells.

All data in this study were obtained from three independent experiments and n ≥ 3 biological replicates. Statistical analysis was performed using a Student t test (Prism, Graphpad, or Microsoft Excel). Mean florescence intensity values of all flow cytometry analysis results were provided in Supplementary Fig. S11. Statistically significant results were represented as follows: *, P < 0.05; **, P < 0.01; and ***, P < 0.001.

We recently reported that CD133 and EpCAM were suitable to detect LCSCs of epithelial-like cells, whereas CD90 was efficient to detect LCSCs of mesenchymal-like cells (13, 14). CD133⁺/EpCAM⁺ and CD133⁻/EpCAM⁻ populations from Huh7 and Hep3B cells were separated by cell sorter with a purity and sorting efficiency of approximately 95% (Supplementary Fig. S1A). To characterize stemness features of the isolated cells, sphere formation and stemness factors' (NANOG and OCT4) expression were compared in CD133⁺/EpCAM⁺ and CD33⁻/EpCAM⁻ populations. CD133⁺/EpCAM⁺ cells were able to form vivid and dense spheres, whereas CD133⁻/EpCAM⁻ cells were unable to form such spheres (Supplementary Fig. S1B). In addition, NANOG and OCT4 expressions were higher in CD133⁺/EpCAM⁺ cells (Supplementary Fig. S1C). Thus, these markers were suitable for studying cells that have LCSC characters.

Moreover, the activity of the PI3K/Akt/mTOR pathway was analyzed in the presence of Sorafenib (Selleck chemicals) in Huh7 and Mahlavu cells. After 72 hours of treatment with Sorafenib, the PI3K/Akt/mTOR pathway proteins were shown to remain active (Supplementary Fig. S2), which supported the findings reported in previous studies (6, 9).

Ten PI3K/Akt/mTOR pathway inhibitors, DNA intercalators [Camptothecin (15; MERCK), and Doxorubicin (16; SABA)], and Sorafenib were used to determine their effects on LCSCs. Initially, the inhibitory concentrations of these inhibitors on HCC cells (Huh7, Hep3B, Mahlavu, and SNU475) were identified. The IC₅₀ values showed small differences between cell lines; therefore, an OMIC value was set for all the inhibitors targeting HCC cells throughout the following experiments in this study (Supplementary Fig. S3A).

To observe the changes in the expression of stem cell markers with each inhibitor, cells were treated for 72 hours with their OMIC values and compared with DMSO controls. Treatment of HCC cells with Sorafenib and DNA intercalators enriched the CD133⁺/EpCAM⁺ cells. PI3K/Akt pathway inhibitors led either to the enrichment of CD133⁺/EpCAM⁺ cells [ZSTK474 (17), NVP-BEZ235 (18)] or did not have any effect [Akti-1,2 (19), Akti-2 (20), PI3Ki-α (21), PI3Ki-β (22)]. However, mTOR inhibitor-Rapamycin (ref. 23; Calbiochem), PI3K inhibitor-LY294002 (ref. 24; Calbiochem), and Notch inhibitor-DAPT (ref. 25; Selleck chemicals) significantly reduced the CD133⁺/EpCAM⁺ cells (Fig. 1; Supplementary Fig. S3B). Yet, because LY294002 displayed its effect in lower significance, Rapamycin was selected for combination experiments with Sorafenib.

Moreover, Sorafenib and Doxorubicin significantly increased CD90 positivity in mesenchymal-like HCC cells (Mahlavu and SNU475), whereas Rapamycin and DAPT decreased the amount of CD90-positive cells (Supplementary Fig. S4). Hence, rapamycin was able to prevent enrichment of CSCs in both epithelial- and mesenchymal-like HCC cells similar to the effect of DAPT.

Recent studies indicate that mTOR plays critical roles in CSCs through stemness-related functions, and mTOR inhibition significantly leads to sensitization of CSCs (26). However, the potential of combinatorial treatment using PI3K/Akt/mTOR inhibitors and Sorafenib on LCSC population remains unclear. Therefore, two different sequential treatment strategies were followed: treatment of cells with strategy 1; Sorafenib or Doxorubicin before either mTOR inhibitor Rapamycin or PI3K inhibitor LY294002 or DAPT, and strategy 2; treatment of cells with Rapamycin or DAPT before Sorafenib. The first strategy failed to inhibit enrichment of LCSCs and even enhanced this process (Supplementary Fig. S5). However, the second strategy decreased the CD133⁺/EpCAM⁺ cells significantly in both Huh7 and Hep3B cells (Fig. 1B) and resulted in significantly smaller and lower number of spheres in cells treated with Rapamycin prior to Sorafenib (Fig. 1C). Consequently, strategy 2 was promising for the inhibition of LCSCs' enrichment compared with Sorafenib alone.

In order to investigate the genes altered in cancer pathways and stemness-related molecular mechanisms in the presence of Sorafenib, Rapamycin, and DAPT, “PanCancer pathways” gene expression panel (770 genes) was analyzed with nCounter Technology. Differentially expressed gene (DEG) analysis revealed that upon Rapamycin or DAPT treatment, stemness pathway genes (WNT and NOTCH) were downregulated (Fig. 2A), whereas Sorafenib treatment caused upregulation of JAG1 which encodes for Jagged-1, a Notch receptor ligand involved in cancer development and angiogenesis including HCC (ref. 27; Supplementary Fig. S6 and Supplementary DEG). When the top 15 DEGs were clustered with K-means-3 approach, the grouping of drug treatments was in correlation with our data given in Fig. 3 with regard to LCSC enrichment. In addition, expression of FLNA (Filamin A), FLNC genes (potential markers for the advanced HCC) were upregulated upon Sorafenib treatment (Fig. 2B; ref. 28).

DEG data were then used to perform “Score Flow” gene enrichment tool using Cytoscape software (enrichment strategy described in detail in Supplementary Fig. S7A; ref. 29) in order to evaluate the activity of cancer signaling pathways. IL8 gene gained significantly enriched scores in Sorafenib-treated cells compared with DAPT or Rapamycin treatment (Fig. 2C). Parallel with the score flow analysis, when top 15 DEGs were clustered, IL8 expression was strongly upregulated (∼5-fold, log₂) upon Sorafenib treatment. In addition, Rapamycin treatment before Sorafenib resulted in the suppression of IL8 along with FLNA and FLNC upregulation triggered by Sorafenib (Fig. 2B). Furthermore, clustering of the DEG data with the combination treatments supported our results with LCSC enrichment (Fig. 2B). Once the cells were treated with Rapamycin, DAPT, or Rapamycin and then Sorafenib, the expression of the above-mentioned genes was found to be regulated inversely. Altogether, DEG and score flow analysis results emphasized the potential of IL8 as a target in the inhibition of LCSC enrichment.

In order to characterize the association of IL8 with LCSC enrichment features, we carried out functional assays in the presence of the inhibitors. Initial confirmation of nCounter DEG data was done with qRT-PCR experiment. NANOG expression was in parallel with IL8 expression in Huh7 cells and further decreased upon Rapamycin treatment (Fig. 3A). We then assessed the autocrine secretion of IL8 with ELISA from tissue culture supernatants. IL8 protein levels increased approximately 2-fold upon treatment with Sorafenib and did not change significantly upon DAPT or Rapamycin treatment (Fig. 3B). Time- and dose-dependent ELISA experiments showed that, as the cell viability decreases upon Sorafenib treatment, IL8 concentrations increase (Fig. 3C; Supplementary Fig. S8). When we analyzed and compared the data in the presence of the inhibitors and their combinations from Fig. 1A in terms of LCSC ratio with IL8 expression originated from two different sources (direct hybridization with nCounter platform and qRT-PCR), there was a strong correlation between the ratio of LCSCs and the expression of IL8. As the LCSC population was enriched, the expression of IL8 gene increased (Fig. 3D; Supplementary Fig. S9). As opposed to Sorafenib alone, Rapamycin treatment before Sorafenib prevented IL8 gene expression upregulation, hence impeded the stemness switch of HCC cells. IL8 has been widely studied in many cancers including breast (30), lung (31), pancreatic (32), colorectal (33), and HCC (34), but not yet been associated with drug-resistant CSC-like properties under drug stress in HCC.

Our results have led us to concentrate on the possibility that HCC cells could be sensitized to Sorafenib treatment through blockade of IL8 signaling as we observed with the effect of Rapamycin. Reparixin (ref. 35; Cayman Chemical), IL8 receptor inhibitor (IC₅₀: 50 μmol/L), was used to inhibit IL8 signaling in Huh7 and Hep3B cells. Reparixin alone and in combination with Sorafenib resulted in significant reduction in positivity of LCSC markers (Fig. 3E), as well as their capacity to form spheres (Fig. 3F). In addition, both NANOG and SOX2 protein levels were reduced in spheres formed by Huh7 cells treated with DAPT and Reparixin, compared with Sorafenib treatment. Moreover, the combination of Sorafenib and Reparixin resulted in lower levels of SOX2 and NANOG proteins (Fig. 3G).

Knockdown experiments were performed using three different siRNAs targeting IL8 gene. Hs_IL8_6 FlexiTube siRNA was selected for further experiments due to its effective knockdown of IL8 in Huh7 cells (Supplementary Fig. S10). siRNA treatment lead to reduced mRNA expression levels of IL8 as well as NANOG after 48 hours (Fig. 4A). Autocrine secretion of IL8 also decreased with respect to untreated or negative siRNA-treated Huh7 cells (Fig. 4B). CD133/EpCAM-positive population as well as the sphere formation capacity of Huh7 cells decreased significantly 72 hours after siRNA transfection (Fig. 4C). Simultaneously, the expression of epithelial marker E-cadherin and stemness marker SOX2 decreased significantly, indicating that knockdown of IL8 increases both epithelial features of Huh7 cells and reduces stemness-related protein expression (Fig. 4D).

Moreover, the effects of IL8 silencing on the sensitivity of cells to Sorafenib were investigated using an RT-CES. siRNA-transfected cells grew more slowly and were more sensitive to Sorafenib compared with untransfected controls, such that 10 μmol/L Sorafenib resulted in twice the amount of growth inhibition in these cells (Fig. 4E). We concluded that both inhibition of IL8 signaling as well as knockdown of IL8 sensitized HCC cells toward treatment and reduced LCSC population in HCC cells.

We also evaluated LCSC enrichment in parallel with IL8 expression on Huh7 cells in the presence of Regorafenib (another multikinase inhibitor approved for advanced HCC treatment, IC₅₀ 1.5 μmol/L; Selleck chemicals). Regorafenib enriched LCSC population (Fig. 5A) and upregulated not only IL8, but also other stemness-promoting interleukins such as IL11 and IL1β, as well as STAT expression (Fig. 5B; Supplementary Fig. S6B). Moreover, Sorafenib and Regorafenib treatment resulted in denser and larger spheres, whereas IL8 signaling inhibition with Reparixin or DAPT treatment inhibited the number and the size of the spheres with respect to DMSO control (Fig. 5C). Furthermore, with the Score Flow algorithm applied to the DEG results of Regorafenib-treated cells, MAPK and PI3K gene expressions were significantly enriched (Fig. 5D). Altogether, Regorafenib and Sorafenib lead to the enrichment of CD133⁺/EpCAM⁺ population and upregulate cytokines associated with stemness switch which can be prevented with adjuvant therapies.

Studies focusing on the mechanisms underlying the heterogeneity of HCC and the acquired resistance to chemotherapeutics reported that the activation of compensatory pathways such as PI3K/Akt/mTOR, JAK-STAT signaling, and the existence of LCSC population are involved in drug resistance (36). In this study, we initially characterized CD133 and EpCAM marker positivity as LCSC markers in Huh7 and Hep3B cells and further validated with sphere formation and expression of NANOG and OCT4 as reported previously (Supplementary Fig. S2; ref. 37). Literature has also defined the association between PI3K/Akt/mTOR pathway and LCSC markers CD133 and EpCAM through analysis of tumor and nontumor tissues from HCC patients and found that p-Akt and p-mTOR expressions were positively correlated with stemness marker expression (37). Although Sorafenib and other small-molecule kinase inhibitors targeting PI3K/Akt or Ras/Raf pathways are widely studied for their anticancer activity, it is a great concern that their efficacy is unpredictable in HCC patients due to high-level tumor heterogeneity. Because Sorafenib inhibits Raf/(MAPK) signaling but activates PI3K/Akt pathway (ref. 38; Supplementary Fig. S2), it is suggested that combinations targeting PI3K/Akt/mTOR or Ras/MAK pathways could be promising in ensuring better clinical outcomes. PKI-587, an inhibitor of PI3K/Akt/mTOR pathway together with Sorafenib, is reported to be advantageous in HCC (39). However, a recently completed phase II clinical trial (40) which compares everolimus (PI3K/Akt/mTOR pathway inhibitor) combined with Sorafenib and a phase III trial of Sorafenib and erlotinib (EGFR inhibitor; ref. 41) showed no improved survival rates compared with Sorafenib alone (42). Given the fact that not all PI3K/Akt/mTOR pathway inhibitors exhibit promising effect on HCC, in this study a group of small-molecule inhibitors targeting PI3K/Akt/mTOR pathway, Sorafenib, and clinically used DNA intercalators (Camptothecin, Doxorubicin) were investigated for their efficacy against LCSCs. As expected, and in parallel with the literature, each inhibitor was found to have differential action on LCSCs where Sorafenib and DNA intercalators and PI3K/Akt pathway inhibitors ZSTK and NVP-BEZ235 enriched LCSC population, whereas Rapamycin and LY294002 decreased this population significantly (Fig. 1). The potent inhibitors against LCSCs were then further tested in combination with Sorafenib to propose a treatment strategy that would inhibit LCSC enrichment and sensitize cells to Sorafenib. Sequential treatment with Rapamycin prior to Sorafenib resulted in reduced levels of LCSCs (Fig. 1). When the order of inhibitors was changed, LCSCs were likely to enrich even more compared with Sorafenib treatment alone (Supplementary Fig. S5). We concluded that pretreatment of cells with Rapamycin regulates stemness-related factors and sensitizes LCSCs toward sorafenib. The potential of mTOR pathway inhibitors has already been shown in stemness-related pathways indicating that these inhibitors act on the proliferation of CSCs and non-CSCs in Huh7 cells equally (43).

To understand the alterations in the expression patterns of cancer pathway–related genes and to identify novel molecular targets that could have a significant role in survival mechanisms of LCSCs, a panel consisting of 770 genes were studied and analyzed using nCounter technology based on direct hybridization of cellular RNA (Fig. 2). Upon DAPT or Rapamycin treatment, genes related with Notch pathway were downregulated as expected. Differential expression analysis revealed that expression of Ras and MAPK pathway–related genes (PDGFRA, EFNA1, EFNA3, VEGFA, and RASA4) were downregulated, whereas genes belonging to Wnt and Notch pathway (DKK1 and JAG1) were upregulated upon treatment with Sorafenib. The differential expression data in the presence of selected inhibitors were further analyzed with the score flow algorithm (Cytoscape gene prioritization application) to evaluate the activity of KEGG cancer signaling pathway (Fig. 2C). This gene prioritization application does not only enrich the DEGs but also depends on the network topology. The score flow algorithm, using the expression data, calculates weights for the protein nodes considering protein–protein interaction edge connectivity information such as feedbacks and loops within the pathway. This method allowed us to single out IL8 as a significant differential target during Sorafenib, DAPT, Rapamycin, and combination treatments when compared with simple DEG analysis. Autocrine IL8 secretion by liver cancer cells was previously reported, along with its possible role in tumor growth and angiogenesis, suggesting IL8 as a potential target for liver cancer therapy (44). However, there is not any study indicating the involvement of IL8 in drug resistance–related liver cancer stemness (34).

Tumor cells produce factors to attract and regulate a variety of cell types in their microenvironment. The inflammatory cytokines IL1β, IL6, and IL8 play essential roles in regulating the interaction between LCSCs and their microenvironment by activating Stat3/NF-κB pathways in both tumor and stromal cells. Due to the activation of these pathways, further cytokine production causes positive feedback loops that stimulate self-renewal of LCSC (10). In this study, differential expression of IL8 under various drug treatments draws our attention because IL8 signaling has been studied in other cancers, especially in breast cancer, and its knockout resulted in poor survival of BCSCs (30). In fact, it has been reported previously that IL8 levels increase during progression of HCC and play role in development of distant metastasis (45). Another study has described that IL8 plays an important role in driving HCC tumor growth, self-renewal, and angiogenesis in CD133⁺ liver cancer cells (34). In our study, treatment of cells with Rapamycin prior to Sorafenib prevented the upregulation of IL8 unlike other combination strategies, which further suggested that Rapamycin could be inducing cellular events that do not allow IL8 expression to elevate, yet triggers a mechanism which inhibits survival of LCSCs. IL8 expression was significantly correlated with NANOG expression and LCSC ratio in Huh7 cells (Fig. 3A). In fact, inhibition of IL8 through treatment with its receptor inhibitor Reparixin resulted in the reduction of CD133⁺/EpCAM⁺ population, their sphere formation capacity, and the levels of stemness-related proteins (Fig. 3E–G). Comparably, identical results were obtained from knockdown experiments of IL8 gene in Huh7 cells (Fig. 4C and D), and sensitivity of Huh7 cells to Sorafenib was increased when cells were transfected with IL8 siRNA beforehand (Fig. 4E). Altogether, both IL8 signaling inhibition and IL8 knockdown enhance sensitivity toward Sorafenib treatment in HCC cells and display inhibitory effects on survival of stem cell–like cancer cells, correlated with the data represented in previous studies on other cancer types (30, 31, 46, 47).

Apart from Sorafenib, a recently FDA-approved multikinase inhibitor Regorafenib also enriched LCSC population and upregulated several inflammatory cytokines (IL8, IL11, and IL1β; Fig. 5). The upregulation and secretion of these inflammatory cytokines from cancer cells were associated with the enrichment of side populations, which further emphasizes the effect of autocrine signaling on LCSCs and their regulation of survival under chemotherapeutic stress.

In summary (Fig. 6), the data represented herein elucidate the mechanisms associated with the survival of LCSC population under drug stress. We provide a better understanding of alterations in distribution of stem/nonstem cells with different Akt/mTOR pathway inhibitors along with clinically administered chemotherapeutic agents Sorafenib and Regorafenib. Finally, expression and functional analysis of the cellular pathways under drug stress allowed us to identify Rapamycin as one of the drugs capable of suppressing the overexpression of inflammatory cytokines, particularly IL8 pathway as a target involved in the regulation of LCSC survival. The data address the necessity for new approaches against HCC in clinics because Rapamycin is already in use for various liver diseases and Reparixin is exploited in clinical trials for other cancers.

No potential conflicts of interest were disclosed.

Conception and design: D.C. Kahraman, T. Kahraman, R. Cetin-Atalay

Development of methodology: D.C. Kahraman, T. Kahraman, R. Cetin-Atalay

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D.C. Kahraman

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D.C. Kahraman, T. Kahraman, R. Cetin-Atalay

Writing, review, and/or revision of the manuscript: D.C. Kahraman, T. Kahraman, R. Cetin-Atalay

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R. Cetin-Atalay

Study supervision: R. Cetin-Atalay

The authors acknowledge the Turkish Ministry of Strategy and Budget grant (Ref: 2016K121540) and the Scientific and Technological Research Council of Turkey (TUBITAK) grant (Ref: 113S540) received by R. Cetin-Atalay. The authors would also like to thank Professor Dr. Ihsan Gursel for his valuable comments during this study.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.