Cellular hypoxia reprograms drug response and reveals curcumin-driven cancer stem-cell-associated marker enrichment in oral squamous cell carcinoma

Abstract

BACKGROUND

Oral squamous cell carcinoma is a deadly cancer often detected late, contributing to high mortality and treatment failure. The efficacy of first-line chemotherapeutic agents is limited by resistance and a narrow therapeutic index. Naturally derived compounds, including curcumin, are increasingly being explored for their anti-inflammatory and anticancer potential.

AIM

To determine whether curcumin retains cytotoxicity under hypoxia compared with carboplatin and assess apoptosis, reactive oxygen species (ROS), and stemness markers.

METHODS

The OECM-1 model of late-stage cancer cultured under hypoxic conditions was validated by measuring hypoxia-induced factor 1 alpha levels using enzyme-linked immunosorbent assay. Drug cytotoxicity was measured using the MTT assay. Oxidative stress was assessed using a DCFDA cellular ROS assay; apoptosis was determined using Annexin V staining; and surface cancer stem cell (CSC) markers were detected using flow cytometry with antibodies against CD44, CD133, and CD147. Dose-response curves were analyzed using nonlinear regression, and drug interactions were assessed using exploratory CI, Bliss, and Highest Single Agent analyses.

RESULTS

Hypoxia-induced factor 1 alpha levels were six-fold higher under hypoxia than normoxia (P = 0.0032). Curcumin retained cytotoxic activity under hypoxia, whereas carboplatin potency and efficacy declined. Combining carboplatin with curcumin reduced curcumin potency under hypoxia and did not show synergy by exploratory CI, Bliss, or Highest Single Agent analyses. Carboplatin did not significantly alter ROS levels, whereas curcumin alone or in combination with carboplatin reduced ROS under normoxia and hypoxia (P = 0.039 to P < 0.0001). Treatments did not induce apoptosis under normoxia but increased apoptosis under hypoxia. Carboplatin did not alter the expression of CD44, CD133, or CD147 in either model, whereas curcumin and the combined treatment increased the expression of CSC-associated markers, particularly CD44+CD133+ under hypoxia (P < 0.0001).

CONCLUSION

Curcumin retained cytotoxic activity in a hypoxic, chemoresistant late-stage model where carboplatin activity declined, but increased CSC-associated marker expression, requiring functional validation.

Key Words: Platinum drugs; Chemotherapy; Oxygen; Apoptosis; Cytotoxic

Core Tip: Curcumin retained cytotoxic activity in oral squamous cell carcinoma under hypoxic, chemoresistant conditions where carboplatin activity declined. Its effects were associated with apoptosis induction and reduced reactive oxygen species, suggesting cytotoxicity that may not depend solely on increased oxidative stress. However, curcumin also increased cancer stem cell-associated marker expression, particularly under hypoxia. This marker enrichment may reflect selective survival, phenotypic adaptation, or stress-induced reprogramming, but functional validation is required. These findings highlight the importance of hypoxia in shaping drug response and treatment-associated cellular adaptation in oral squamous cell carcinoma.

INTRODUCTION

Herbal remedies are utilized as food additives and as traditional and household medications. Naturally occurring compounds are known to be targets for several diseases. Curcumin is an organic compound sourced from Southeast Asia that belongs to the Zingiberaceae family[1]. Beyond its culinary applications as an aromatic stimulant, food preservative, and coloring agent, turmeric has been a significant component of traditional remedies. It had a long-standing history of therapeutic use, particularly in traditional medicine systems across Asia. Turmeric was used to treat common inflammatory diseases, including rheumatism and sinusitis. Curcumin, the primary active compound in turmeric, has been the subject of several recent studies reporting its pharmacological effects, based on its biological properties, including anti-inflammatory, antioxidant, anticarcinogenic, and antibacterial activities[2]. These multifaceted actions position curcumin as a promising agent for both the prevention and treatment of various diseases[3]. Such findings underscore curcumin’s potential as a highly active biological material for various modern medical treatments. Furthermore, the use of curcumin in combination with chemotherapy has been shown to improve treatment efficacy, prolong patient survival, and increase the expression of anti-metastatic markers while minimizing side effects[4]. Furthermore, many studies have shown that curcumin targets cancer-related markers[1,4,5] and reduces pathological processes associated with squamous cell carcinoma and dysplastic lesions[1,5].

Oral squamous cell carcinoma (OSCC) is a deadly disease and has serious treatment complications. Two-thirds of patients diagnosed at an advanced stage have a 5-year survival rate of 50%[6]. OSCC can originate in any location of the oral cavity, but the sites most frequently affected are the lateral border of the tongue and the floor of the mouth[7,8]. Clinically, it presents as an ulcerated lesion with a central necrotic area surrounded by raised borders[9]. Males are at a higher risk of developing OSCC than females, most likely due to the higher tendency to practice social habits that are associated with OSCC[10]. OSCC predominantly affects men in the fifth and sixth decades of life and is rare in patients under 40, although its incidence in younger patients has increased in recent decades[11]. OSCC treatment frequently results in speech and aesthetic impairments, chewing and swallowing dysfunction, and other issues that can worsen patients’ quality of life[12].

Carboplatin is a platinum-based chemotherapeutic agent commonly used as a first-line treatment for OSCC[13]. Carboplatin’s cytotoxicity arises from its ability to promote DNA crosslinking, causing DNA damage that disrupts replication and transcription, leading to cell cycle arrest and cell death. This is especially true in rapidly growing tumor cells[14]. As with most cytotoxic agents, the primary dose-limiting side effect is myelosuppression, leading to thrombocytopenia and severe neutropenia[15]. In patients with aggressive recurrent or metastatic head and neck squamous cell carcinoma, a combination of carboplatin and paclitaxel only provided moderate efficacy[13]. The lack of efficacy is due to the activation of chemoresistant mechanisms within the tumor, which often leads to treatment failure[16,17]. The aggressiveness of the tumor is closely linked to its propensity to survive and proliferate in low-oxygen (hypoxic) microenvironments[18]. A molecular marker and a known hallmark of aggressiveness and chemoresistance in malignancies is the protein hypoxia-induced factor 1 alpha (HIF-1α), which regulates gene expression essential for cancer survival and adaptation under low-oxygen conditions[19,20].

The role of cancer stem cells (CSCs) in OSCC aggressiveness and treatment resistance has garnered significant attention. CD44, a transmembrane glycoprotein, is a prominent CSC marker in OSCC and is strongly associated with tumor initiation, metastasis, and chemoresistance[21]. High CD44 expression correlates with poor prognosis and increased recurrence rates in patients with OSCC. Similarly, CD133, also known as prominin-1, identifies a subpopulation of OSCC cells with enhanced tumorigenic potential and self-renewal capacity, thereby contributing to therapeutic resistance and disease relapse[22]. CD147 (EMMPRIN/basigin), another critical CSC marker, promotes tumor progression by inducing matrix metalloproteinase production and facilitating invasion and metastasis in OSCC. Elevated CD147 expression is associated with advanced clinical stage, lymph node metastasis, and unfavorable patient outcomes[23,24]. Importantly, emerging evidence suggests that curcumin may target these CSC populations by downregulating the expression of stemness-associated markers, including CD44 and CD133, and suppressing CSC-like properties, thereby potentially overcoming treatment resistance in OSCC[25,26].

Mechanistically, curcumin is pleiotropic, a redox modulator and an activator of DNA damage and apoptosis, and may counter hypoxia-driven platinum resistance, but direct comparative data with carboplatin under hypoxic conditions are lacking; as such, we set out to determine the pharmacological profile of curcumin in combination with, and compared to, carboplatin in cellular models of early-stage OSCC and hypoxia-induced, late-stage, chemoresistant, aggressive OSCC. Cellular apoptosis and reactive oxygen species (ROS) levels were assessed, and cell-surface CSC markers were evaluated to determine the efficacy and potency of curcumin in both models, explore mechanisms of resistance, and elucidate targets to overcome carboplatin chemoresistance.

MATERIALS AND METHODS

Cell culture and treatment with test and control drugs

OECM-1 cells are cultured with RPMI media supplemented with 100 mL/L foetal bovine serum and 10 mL/L penicillin-streptomycin, at 37 °C with 200 mL/L O₂ and 50 mL/L CO₂. Cells are subcultured at 80%-85% confluence using phosphate-buffered saline (PBS) containing 4 mmol/L EDTA, followed by centrifugation at 800 rpm for 10 minutes. The resulting cell pellets are resuspended in RPMI media, counted using a hemocytometer and trypan blue, and placed into 96-well or 6-well plates for overnight incubation. Pharmacological treatments include curcumin, carboplatin, and their combinations in addition to PBS as a negative control. The incubation duration ranges from 24 hours to 48 hours under normoxic and hypoxic conditions (humidified atmosphere with 1 mL/L O₂ and 150 mL/L CO₂ at 37 °C), the latter achieved using the MGC AnaeroPack System (JAR-70, GA-500ZE, and ANAI01ZEM, Mitsubishi Gas Chemical Company, Japan).

Cell lysis and protein determination

The OECM-1 cells were cultured in 6-well plates and subjected to drug treatments at IC₅₀ concentrations as outlined earlier. Following a 48-hour period under both normoxic and hypoxic conditions, cells were detached, collected, and lysed (Qiagen Cell Lysis buffer), with protein concentrations quantified using the BCA assay. The lysates were then stored at -90 °C for future analysis.

HIF-1α enzyme-linked immunosorbent assay

A HIF-1α Human ELISA kit (Abcam, ab171577, United Kingdom) was used to determine HIF-1α levels in OECM-1 cell lysates (1 mg/mL) from untreated cells (basal using PBS) under normoxic and hypoxic conditions over 36 hours. The procedure was carried out in accordance with the manufacturer’s guidelines. A colorimetric assessment was conducted to quantify the absorbance at 450 nm.

MTT cytotoxicity assay

A series of serial dilutions of curcumin and carboplatin were prepared in PBS, ranging from 1 mmol/L to 1 pmol/L, with PBS as the untreated control. Cells were cultured in 96-well plates at a density of 5000 cells per well in RPMI medium with supplements, then treated with increasing concentrations of curcumin or carboplatin alone or curcumin in combination with a fixed-dose carboplatin (0.1 mmol/L). The plates were incubated for 48 hours under both normoxic and hypoxic conditions. The MTT assay was conducted by adding a prepared stock solution to each well, followed by DMSO. Absorbance was measured at 570 nm after incubation.

Oxidative stress and ROS

The DCFDA/H2DCFDA cellular ROS assay (Abcam, ab113851, United Kingdom), which employs flow cytometry, was used to quantify cellular ROS levels in viable cells. Cells were cultured in 12-well plates and treated with the experimental and control agents (PBS as the negative control), as described earlier: Carboplatin (100 μmol/L) and curcumin (100 μmol/L). Following 36 hours of exposure to both normoxic and hypoxic environments, as specified earlier, cells were detached using PBS supplemented with 4 mmol/L EDTA and collected into a 1 mL PBS suspension. A total of 1 × 10⁶ cells were stained with 20 μmol/L DCFDA and incubated for 30 minutes at 37 °C. The analysis was conducted using a flow cytometer; cells were gated based on forward scatter and side scatter metrics. DCF excitation occurs at 488 nm with emission detected at 535 nm (FL1).

Apoptosis

Annexin V FITC and PI staining, used together with flow cytometry and the Annexin V-FITC/PI Apoptosis Kit (Elabscience; E-CK-A211, Wuhan, Hubei Province, China), enabled the differentiation of early apoptotic (Annexin+, PI-), late apoptotic (Annexin+, PI+), and necrotic (Annexin-, PI+) cells. After a 36-hour incubation under both normoxic and hypoxic conditions, cells were treated with carboplatin (100 μmol/L), curcumin (100 μmol/L), and their combination. PBS was used as the untreated negative control. All control and test samples were then processed for flow cytometry. The resuspended cell pellet was stained with Annexin V and PI, incubated in the dark, and analyzed for fluorescence emission at specific wavelengths.

Stem cell surface marker detection

The expression levels of CSC markers were evaluated using flow cytometry. CD44 (FITC-Mouse Anti-human-CD44v6 IgG), CD133 (PE-Anti-human CD133 IgG), and CD147 (perCPCy5.5-Mouse Anti-human-CD147 IgG). Cells were cultured in 6-well plates and treated with test and control agents at IC₅₀ concentrations and their combinations for 48 hours under either normoxic or hypoxic conditions. PBS was used as a negative, untreated control. Following treatment, the cells (1 × 10⁶) were washed in PBS and then incubated with the respective antibody in the dark. They were washed again, and flow cytometric analysis was performed using FL1, FL2, and FL3 detection channels.

Dose-response and drug-interaction assessment

Dose-response data were analyzed in GraphPad Prism Software version 9.5.1 (San Diego, CA, United States) using nonlinear regression with a variable-slope four-parameter logistic model. Cell viability values were expressed as normalized percentages relative to untreated controls. The bottom plateau was constrained to ≥ 0% viability, while the remaining parameters were fitted freely. IC₅₀ values were calculated from the fitted dose-response curves and expressed in μM. Maximal cytotoxic efficacy was estimated from the fitted lower plateau of the viability curve.

For drug-interaction assessment, normalized viability values were converted to fraction affected values using the formula: Fa = 1 - (% viability/100). An exploratory Chou-Talalay-based combination index (CI) analysis was performed using Python (version 3.12.17). Since the combination design used increasing concentrations of curcumin while keeping carboplatin at a fixed concentration, CI values were interpreted as dose-specific estimates. CI was calculated as:

CI = dcurcumin/dxcurcumin + dcarboplatin/dxcarboplatin

D represents the actual dose of each drug used in the combination, and Dx represents the estimated dose of each drug alone required to produce the same fractional effect as the combination. CI values < 1, = 1, and > 1 were classified as synergistic, additive, and antagonistic, respectively. Expected-effect modeling was also performed in Python using Bliss independence and Highest Single Agent (HSA) analyses. Bliss expected effect was calculated as EBliss = EA + EB - EAEB, where EA and EB represent the fractional effects of curcumin and carboplatin alone. HSA analysis compared the observed combination effect with the strongest single-agent effect at the corresponding dose. Analyses were performed separately under normoxic and hypoxic conditions.

Statistical analysis

Statistical analysis were performed using GraphPad Prism. For HIF-1α, ROS, apoptosis, and CSC marker experiments, three technical replicates were performed in each of three biological replicates. Graphs show the mean ± SD. For analysis, comparisons between two groups were performed using an unpaired t-test. For comparisons involving more than two groups, one-way analysis of variance (ANOVA) was followed by appropriate post hoc multiple-comparison testing. A P value < 0.05 was considered statistically significant.

RESULTS

HIF-1α expression under normoxic and hypoxic conditions

HIF1-α expression levels were assessed in OECM-1 cells grown under normoxic and hypoxic conditions. Results highlight a significant increase (6-fold) in HIF1-α in the hypoxic condition (Figure 1).

Figure 1 Expression of hypoxia-induced factor 1 alpha in OECM-1 cells grown under normoxic and hypoxic conditions. Statistical analysis was performed using an unpaired t-test. ^aP = 0.0032, comparison between normoxic and hypoxic groups.

Dose-dependent cytotoxicity of curcumin, carboplatin, and combined treatment

To characterize the potency and efficacy of the drugs, the MTT cytotoxicity assay was performed. Both curcumin and carboplatin induced dose-dependent cytotoxicity in OECM-1 cells, with near-equal efficacy (approximately 75%) under normoxic conditions, although carboplatin was more potent (Figure 2A, Table 1). Under hypoxic conditions, carboplatin’s potency and efficacy were markedly reduced and far less than curcumin (Figure 2B, Table 1). Interestingly, the potency of curcumin was maintained regardless of oxygen conditions, whereas carboplatin’s potency was significantly reduced under hypoxic conditions (Table 1). Combining carboplatin with curcumin had no significant effect on the curcumin potency or efficacy in normoxia (Figure 2C, Table 1). However, under hypoxic conditions, curcumin’s potency was significantly decreased compared to curcumin alone (Figure 2D, Table 1).

Figure 2 Treatment induced cytotoxicity in OECM-1 cells under normoxic and hypoxic conditions. A: Single treatments of increasing concentrations of carboplatin and curcumin under normoxia; B: Single treatments of increasing concentrations of carboplatin and curcumin under hypoxia; C: Combination treatment with increasing doses of curcumin and fixed 0.1 mmol/L carboplatin under normoxia; D: Combination treatment with increasing doses of curcumin and fixed 0.1 mmol/L carboplatin under hypoxia.

Table 1 Cytotoxic potency and maximal efficacy of curcumin, carboplatin, and combined treatment in oral squamous carcinoma cells under normoxic and hypoxic conditions, mean ± SD.

	IC50 (μM)		Maximal cytotoxic efficacy (%)
	Normoxic	Hypoxic	Normoxic	Hypoxic
Curcumin	137 ± 22	110 ± 3.1	77.4 ± 3.1	76.1 ± 4.0
Carboplatin	81.9 ± 5.0	670 ± 489	73.9 ± 1.0	42.4 ± 6.6
Curcumin + carboplatin	129 ± 48	410 ± 205	77.89 ± 8.4	82.7 ± 4.9

Values were derived from nonlinear regression using a variable-slope four-parameter model with the bottom plateau constrained to ≥ 0% viability. OECM-1, oral squamous cell carcinoma cell line used in this study. IC₅₀: Half-maximal inhibitory concentration.

Expected-effect modeling was performed using Bliss independence and HSA analysis to determine whether fixed-dose carboplatin enhanced the cytotoxic effect of curcumin. Under normoxic conditions, both models showed predominantly negative interaction scores, with mean HSA scores ranging from approximately -0.59 to -0.03 and mean Bliss scores from approximately -0.66 to -0.05, indicating a lower-than-expected combination effect. Under hypoxic conditions, HSA scores were mostly close to zero, ranging from approximately -0.05 to +0.07, whereas Bliss scores remained generally negative, ranging from approximately -0.17 to -0.05. Exploratory Chou-Talalay-based CI analysis similarly showed predominantly CI values > 1, supporting a non-synergistic interaction. Although one hypoxic dose point showed a mean CI of 0.86, this effect was not consistent across the dose range. Overall, the interaction analyses indicate that fixed-dose carboplatin did not synergistically enhance curcumin cytotoxicity under either normoxic or hypoxic conditions.

Effects of treatment on ROS levels

Total ROS levels were determined in OECM-1 cells treated with the drugs under normoxic and hypoxic conditions. Carboplatin treatment had no significant effect on ROS levels compared to untreated cells in either normoxic or hypoxic environments (Figure 3). Curcumin, however, significantly decreased ROS levels by approximately 50% in cells under both conditions. Combining carboplatin with curcumin did not affect the curcumin antioxidant effect.

Figure 3 Total reactive oxygen species levels in live OECM-1 cells after treatment with carboplatin, curcumin, and curcumin combined with carboplatin under normoxic and hypoxic conditions. A: Normoxia; B: Hypoxia. Statistical analysis was performed using one-way ANOVA with Tukey’s multiple comparison test. ^aP < 0.0001, ^bP = 0.039, ^cP = 0.004, ^dP < 0.0001. ROS: Reactive oxygen species.

Treatment-induced apoptosis under normoxic and hypoxic conditions

Drug-induced apoptosis was assessed in the OECM-1 cell line grown under normoxic and hypoxic conditions. In normoxia, neither single nor combined treatments led to any significant change in the proportion of cells undergoing apoptosis relative to the norm, as most of the population remained viable (Figure 4A). On the other hand, under hypoxic conditions, a significant change in the proportion of apoptotic cells was evident. Approximately 75% of the cells were in apoptosis after carboplatin treatment (Figure 4B), and approximately 50% in both the curcumin treatment alone group and the combined treatment group (Figure 4B).

Figure 4 Apoptotic induction of OECM-1 cells after treatment with carboplatin, curcumin, and curcumin combined with carboplatin under normoxic and hypoxic conditions. A: Normoxia; B: Hypoxia.

Treatment-associated changes in CSC marker expression

Cell-surface CSC markers were assessed to determine treatment-induced phenotypic changes and selection pressure in the OECM-1 cell line grown under normoxic and hypoxic conditions. Untreated cells had more than 90% of the population lacking any of the assessed markers (other), and carboplatin treatment did not alter the cell marker population relative to untreated cells under both normoxic and hypoxic conditions (Figure 5).

Figure 5 Stem cell surface marker expression in OECM-1 cells treated with carboplatin, curcumin, and curcumin combined with carboplatin under normoxic and hypoxic conditions. A: Normoxia; B: Hypoxia. Statistical analysis was performed using two-way ANOVA with Tukey’s multiple comparison test.

Under normoxic conditions (Figure 5A), curcumin treatment increased the CD44+ population (approximately 20%), but this difference was not statistically significant. It also significantly increased the CD44+CD133+ population (approximately 60%) compared with untreated controls or carboplatin treatment (P < 0.0001). In the combined treatment (curcumin and carboplatin), approximately 75% of the population was CD133+ (P < 0.0001), and approximately 20% were CD44+CD133+.

In the hypoxic model (Figure 5B), curcumin treatment resulted in a CD44+CD133+ population of approximately 70%, which was not significantly different from that in the normoxic model but was a striking change relative to the untreated or carboplatin-treated groups under hypoxia (P < 0.0001). The combination treatment under hypoxic conditions resulted in more than 25% of the population being CD133+, which was significantly lower than in the normoxic model (77%) (P < 0.0001) and similar to the curcumin-alone treatment in the hypoxic model. The combined treatment yielded a double-positive CD44+CD133+ population of approximately 55%, which was not significantly different from that of the curcumin-alone treatment but was substantially lower than the double-phenotype in the combined treatment under normoxia (22%) (P = 0.0018).

DISCUSSION

OSCC is characterized by a high recurrence rate and significant management challenges. In this study, we investigated how curcumin, a well-known anti-inflammatory herbal compound, influenced both the early and late stages of OSCC using the OECM-1 cell line, which harbors a p53 missense mutation and low epidermal growth factor receptor expression. Our focus was on elucidating how the tumor microenvironment, mainly oxygenation (hypoxia) and curcumin’s presence as an edible herbal metabolite, impacts the efficacy of carboplatin, the first-line chemotherapy for OSCC. Additionally, we explored the potential of curcumin as a primary therapeutic option in late-stage, aggressive, chemoresistant cancers. We also examined how microenvironmental variations affect the expression of key surface stem cell markers and cellular stress responses, providing further insight into mechanisms underlying treatment resistance and disease progression. Hypoxia is a central microenvironmental driver of resistance, CSC maintenance, and metabolic and epigenetic reprogramming, thereby undermining platinum efficacy and shaping responses to phytochemicals[27]. Hypoxia reduces drug uptake, enhances DNA repair, activates HIF-1α survival programs, and triggers the unfolded protein response and endoplasmic reticulum stress responses that promote platinum resistance in head-and-neck subsites[27]. On the other hand, curcumin inhibits multiple pro-survival pathways, including nuclear factor kappa B, phosphatidylinositol 3-kinase/protein kinase B, and HIF-1α[28]. The clinical implication is that tumor regions with low oxygen are less responsive to DNA-damaging agents that rely on oxidative processes and proliferative kinetics. In the current study, we found that HIF-1α levels were significantly higher in OECM-1 cells under hypoxia than under normoxia, which correlated with carboplatin chemoresistance, as evidenced by reduced carboplatin efficacy and potency. Whereas the potency and efficacy of curcumin were maintained in the late-stage hypoxic model.

The dramatic increase in carboplatin IC₅₀ under hypoxia (from 81.9 μmol/L to 670 μmol/L) compared to curcumin’s more stable IC₅₀ (137 μmol/L to 110 μmol/L) suggests that oxygenation status strongly influences platinum sensitivity in this model. This is consistent with previous reports showing that HIF-1α stabilization under hypoxia can contribute to platinum resistance through coordinated effects on drug efflux, DNA repair capacity, and metabolic adaptation[29,30]. Specifically, Li et al[29] demonstrated that extracellular signal-regulated kinase-mediated phosphorylation of PHD2 prevents HIF-1α degradation, leading to HIF-1α accumulation that quantitatively correlates with platinum resistance in ovarian cancer models. Pharmacologic HIF-1α inhibition (YC-1) restored cisplatin sensitivity both in vitro and in vivo. In contrast, the preserved cytotoxic activity of curcumin under hypoxia may reflect its broader multi-target profile, including modulation of nuclear factor kappa B, signal transducer and activator of transcription 3, phosphatidylinositol 3-kinase/protein kinase B, and apoptotic signaling pathways[31,32]. The pathways may remain targetable regardless of oxygenation status. The current study assessed total ROS levels at a fixed time endpoint following treatment, in both normoxic and hypoxic conditions. Under normoxic and hypoxic conditions, curcumin treatment reduced ROS levels, whereas apoptosis was detected only under hypoxia, suggesting a key difference in metabolic responses between the two conditions. It is important to highlight that these are endpoint assays; delayed apoptosis could have occurred under normoxia; hence, time-course assays would provide further insights. Curcumin reduced total ROS in both models, whereas apoptosis was detected primarily under hypoxic conditions. This suggests that curcumin-induced cytotoxicity in OECM-1 cells may not depend solely on increased total ROS levels. Previous reports have also described ROS-independent DNA damage, cell-cycle arrest, and caspase-dependent apoptosis following curcumin treatment in other cancer models[33,34]. However, the DCFDA assay measures total ROS and does not distinguish mitochondrial or compartment-specific ROS; ROS responses are highly time-dependent. Therefore, the ROS findings should be interpreted cautiously. Future studies should include time-course analyses, mitochondrial ROS assays, and assessments of downstream HIF-1α targets, such as vascular endothelial growth factor, glucose transporter type 1, 3-phosphoinositide dependent protein kinase-1, CA9, and multidrug resistance 1, to further clarify the relationship among hypoxia, redox signaling, and treatment response in OSCC.

The combination of curcumin with fixed-dose carboplatin did not improve curcumin activity under hypoxic conditions. Instead, curcumin potency was reduced in the combination treatment, and additional drug-interaction analyses did not support synergy. Expected-effect modeling using Bliss independence and HSA analyses showed predominantly additive to sub-additive effects, while exploratory Chou-Talalay-based CI analysis also indicated a largely non-synergistic interaction. These findings suggest that the benefit of curcumin-platinum combinations may depend on tumor oxygenation status and formulation strategy, as previous reports of additive or synergistic effects in oral cancer models were largely conducted under normoxic conditions or used nanoformulated delivery systems[35-38].

Several mechanisms may contribute to the lack of synergy under hypoxia, although these remain speculative. Hypoxia can activate adaptive survival pathways, including autophagy, unfolded protein response signaling, and HIF-1α-mediated stress responses, which may reduce sensitivity to platinum-induced cytotoxicity[39,40]. In addition, carboplatin and curcumin may exert opposing effects on redox balance, as platinum agents are commonly associated with ROS-mediated cytotoxicity[41], whereas curcumin reduced ROS levels in the present study and has been reported to have antioxidant/redox-modulating properties[42]. This potential redox conflict may partly explain the lack of synergy observed under hypoxia. Further studies in additional OSCC models and in vivo systems are required to determine whether hypoxic, late-stage tumors are less responsive to the combination of curcumin and carboplatin than to curcumin alone. The current findings should be interpreted within the context of the OECM-1 model and require validation in additional OSCC models and in vivo systems. A key limitation is that OECM-1 carries a p53 missense mutation; therefore, treatment responses may not fully represent all OSCC subtypes. Since p53 status can influence platinum sensitivity and hypoxia-associated treatment resistance, further validation with additional p53-mutant and p53-wild-type OSCC models is also required[43].

Expression of CSC-associated markers such as CD44, CD133, and CD147 is associated with OSCC aggressiveness, recurrence, treatment resistance, and immune evasion[21-24]. In the present study, carboplatin did not significantly alter CD44, CD133, or CD147 expression, whereas curcumin and curcumin combined with carboplatin increased CSC-associated marker expression, particularly the CD44+CD133+ population, with a more pronounced effect under hypoxic conditions. This finding appears paradoxical, as curcumin also reduced cell viability and induced apoptosis under hypoxia. Importantly, apoptosis and CSC-associated marker expression were assessed in parallel assays rather than by single-cell co-localization; therefore, the present data cannot determine whether apoptotic and marker-positive cells represent the same or distinct cellular subpopulations. Accordingly, the observed increase in marker-positive cells should be interpreted as enrichment for CSC-associated markers rather than definitive evidence of functional CSC expansion. Several non-exclusive explanations may account for this marker shift. Curcumin may impose treatment-associated selection pressure, allowing survival of a marker-positive subpopulation, or may induce stress-associated phenotypic adaptation in surviving cells. Curcumin’s redox-modulating and broad pathway effects may also contribute to transient changes in CSC-associated marker expression. The literature supports curcumin’s dual redox activity and broad pathway modulation, both of which can favor short-term marker upregulation or the selective survival of marker-positive cells[44,45]. In the current study, the paradoxical phenotypic shift is likely to be a consequence of selection pressure, and although plant antioxidants and polyphenols can modulate epigenetic networks, lowering ROS levels, favoring cellular reprogramming[45] and CSC marker expression[44], the observed large proportion of cells expressing CD44+CD133+ markers may not correlate with functional stemness, self-renewal, and tumorigenicity. Rather, functional assays are needed to confirm these presumed properties[21,46-48]. Contrary to our results, Jiang et al[49] reported that cisplatin increases CD133+CD44+ in head and neck squamous cell carcinoma, whereas curcumin suppressed cisplatin-induced CSC phenotypes. Therefore, the present CSC-related findings should be considered hypothesis-generating. Functional assays, including tumorsphere formation, ALDH activity, and in vivo limiting-dilution tumourigenicity assays, are required to determine whether the marker-positive cells observed after curcumin treatment represent functionally active cancer stem-like cells.

Translation of curcumin-based therapy into clinical practice remains challenging, primarily due to limited systemic bioavailability. Standard oral doses of 8-12 g/day have been reported to achieve peak plasma concentrations of only 0.05-0.5 μmol/L[50], which are considerably lower than the IC₅₀ values observed in the present study. This suggests that conventional oral curcumin formulations may be insufficient to achieve therapeutically relevant systemic concentrations. However, tumor tissue accumulation may differ from plasma exposure, and direct measurements in OSCC remain limited. Improved delivery systems, including liposomal and poly(lactic-co-glycolic acid)-encapsulated curcumin, may help overcome these bioavailability limitations[51,52]. In addition, nano-curcumin combined with photodynamic therapy has shown beneficial effects in an in vivo rat OSCC model, including modulation of BCL2 and caspase-3 activity, as well as improved histological outcomes[53]. Future studies should therefore evaluate optimized curcumin formulations, local tumor drug accumulation, HIF-1α dynamics, treatment-resistant cancer cell phenotypes, CD44/CD133 expression, and functional stemness. In addition, p53 mutation status, platinum resistance, and tumor oxygenation should be considered when assessing curcumin-based therapeutic strategies.

CONCLUSION

Overall, these findings highlight the complex, context-dependent effects of curcumin in OSCC, where cytotoxic activity may coexist with treatment-associated cellular adaptation. This may reflect selective survival/enrichment, stress-induced reprogramming, selection pressure, and redox modulation; however, functional assays are required to confirm whether the observed marker changes represent true CSC activity.

ACKNOWLEDGEMENTS

The authors acknowledge, Princess Nourah bint Abdulrahman University Researchers Supporting Project (number: PNURSP2026R362), Princess Nourah bint Abdulrahman University, Riyadh, Saudi Arabia. This acknowledgment refers to an institutional post-publication recognition scheme and not to direct funding for the conduct of this study or for publication costs.