Cell Culture and Reagents

The human hepatocellular carcinoma cell line HepG2 was acquired from the American Type Culture Collection (ATCC; Manassas, VA, USA; Cat. #HB-8065™). Cells were maintained at low passage numbers (passages 5–15). HepG2 cells were cultured in high-glucose Dulbecco’s Modified Eagle Medium (DMEM; Thermo Fisher Scientific, Waltham, MA, USA; Cat. #11965092) supplemented with 10% (v/v) heat-inactivated fetal bovine serum (FBS; Thermo Fisher Scientific; Cat. #26140079) and 1% (v/v) penicillin-streptomycin solution (HyClone Laboratories, Logan, UT, USA; Cat. #SV30010). Cultures were maintained at 37°C in a humidified incubator containing 5% CO and 95% air. All experimental assays were initiated when cultures reached 70–80% confluence, and routine PCR-based testing was conducted to confirm the absence of mycoplasma contamination.

Drug Preparation

Doxorubicin hydrochloride (DOX; Sigma-Aldrich, St. Louis, MO, USA; Cat. #D1515) was dissolved in sterile deionized water to yield a 1 mM stock solution. Chloroquine diphosphate (CQ; Sigma-Aldrich; Cat. #C6628), rapamycin (Sigma-Aldrich; Cat. #R0395), and cobalt(II) chloride hexahydrate (CoCl; TM Media, India) were dissolved in sterile dimethyl sulfoxide (DMSO; Sigma-Aldrich) to prepare 10 mM stock solutions. All stock solutions were divided into single-use aliquots and preserved at −20°C in the dark. Working concentrations were freshly prepared by serially diluting stock solutions into complete culture medium immediately prior to application. In all experimental conditions, the final vehicle (DMSO) concentration was maintained at ≤0.1% (v/v), a concentration validated in preliminary control assays to exert no detectable influence on cell viability or baseline apoptosis.

The fixed CQ working concentration of 25 µM was selected on the basis of preliminary dose-response cytotoxicity screening across a range of 0–400 µM under both normoxic and CoCl-induced hypoxic conditions (Supplementary Figure S4). At 25 µM, CQ exhibited minimal single-agent toxicity under normoxia while eliciting measurable growth inhibition under CoCl-induced hypoxia. Autophagy inhibition at this concentration was independently corroborated by the fluorescence microscopy experiments described below and aligns with established concentrations utilized in peer in vitro HCC autophagy inhibition studies9,14. Although therapeutic plasma CQ concentrations in patients typically reside in the low micromolar range (1–5 µM), CQ undergoes extensive lysosomal trapping wherein intracellular organellar concentrations can exceed extracellular levels by several orders of magnitude, making 25 µM a biologically and pharmacologically sound in vitro working concentration14.

Induction and Validation of Hypoxic Conditions

Chemical hypoxia was induced using CoCl, a well-characterized chemical hypoxia mimetic that blocks Fe/2-oxoglutarate-dependent PHDs, thereby stabilizing HIF-1α under normoxic oxygen levels and triggering downstream transcriptional programs7. Based on extensive dose- and time-optimization experiments, exposure to 100 µM CoCl for a total duration of 72 h was selected as the standardized hypoxic condition for all downstream drug interaction experiments, as this regimen robustly activated hypoxia-responsive genes while preserving sufficient cell viability for pharmacological combination testing (Supplementary Figure S1). This protocol aligns with standard CoCl concentrations and durations established in HCC literature7,8.

Hypoxic induction was validated at the transcriptional level via reverse transcription-quantitative PCR (RT-qPCR). The RT-qPCR data referenced herein were previously acquired, analyzed, and published as part of our initial characterization of this HepG2/CoCl experimental system13, and are referenced to confirm that the identical validated hypoxia-mimetic conditions were applied in the present study. In that work, HepG2 cells were exposed to 50 µM or 100 µM CoCl for 72 h, with vehicle-matched DMSO controls included across all plates. Total RNA was isolated using the easy-BLUE™ Total RNA Extraction Kit (iNtRON Biotechnology, Seongnam, Republic of Korea). One-step RT-qPCR was conducted using the Luna® Universal One-Step RT-qPCR Kit (New England Biolabs, Ipswich, MA, USA) on a real-time PCR detection system. Primer sequences (synthesized by Phusa Genomics, Ho Chi Minh City, Vietnam) designed using Primer-BLAST were as follows:

• HIF-1α: Forward 5′-TGCTTGCCAAAAGAGGTGGA-3′, Reverse 5′-GGGGCCAGCAAAGTTAAAGC-3′

• VEGF: Forward 5′-GGACATTGCTGTGCTTTGGG-3′, Reverse 5′-ATGGGCTGCTTCTTCCAACA-3′

• GLUT1: Forward 5′-CTGTCGTGTCGCTGTTTGTG-3′, Reverse 5′-AAAGATGGCCACGATGCTCA-3′

• β-actin (reference gene): Forward 5′-GACTTAGTTGCGTTACACCCTTTCT-3′, Reverse 5′-GAACGGTGAAGGTGACAGCAGT-3′

Target transcript levels were normalized to β-actin and quantified using the comparative 2 method (n = 3). As documented in Dao Minh, Pham Anh, & Nguyen (2025), treatment with 100 µM CoCl for 72 h elicited statistically significant upregulation of VEGF and GLUT1 mRNA relative to untreated normoxic controls (p < 0.05). HIF-1α mRNA was modestly elevated but did not attain statistical significance, a finding fully consistent with the primary post-translational mechanism of CoCl-mediated PHD inhibition, wherein HIF-1α is stabilized at the protein level via degradation blockade without requiring transcriptional upregulation. Consequently, induction of the downstream transcriptional targets VEGF and GLUT1 serves as the functional indicator of hypoxic pathway activation in this cell model.

Cell Viability Assay

Cell viability was quantified using the resazurin-based Alamar Blue fluorometric assay (Invitrogen, Thermo Fisher Scientific; Cat. #DAL1025). HepG2 cells were seeded into 96-well culture plates at a density of 2,000 cells per well in 100 µL of complete growth medium and allowed to adhere for 24 h at 37°C. Following drug treatments, Alamar Blue reagent was added to each well at a final concentration of 10% (v/v) and incubated for 4 h at 37°C in the dark. Fluorescence emission intensity was recorded on a DTX880 multimode microplate reader (Beckman Coulter, Brea, CA, USA) with excitation at 560 nm and emission at 590 nm. The fraction of cells affected (Fa) was calculated as Fa = 1 − (% Cell Viability / 100).

Drug Combinations and Synergy Analysis

Two primary drug-treatment paradigms were investigated. Under chemically induced hypoxia, cells were pre-treated with 100 µM CoCl for 48 h to establish a stable hypoxic transcriptional state, followed by simultaneous co-treatment with DOX (ranging from 0.016 to 0.5 µM) and CQ (fixed at 25 µM) for an additional 24 h (cumulative CoCl exposure: 72 h; designated as the DCC combination). Under normoxia, cells were co-treated simultaneously with DOX (0.016–0.5 µM) and CQ (fixed at 25 µM) for 24 h (designated as the DOX:CQ combination). Because CoCl was incorporated into the cell culture medium as a hypoxia-mimicking experimental variable rather than as a therapeutic candidate, CI and DRI values calculated for the DCC group reflect the pharmacodynamic interaction within this three-component experimental model and should not be construed as clinical dose-reduction targets.

Single-agent and pairwise dose-response profiles utilized to fit median-effect reference curves (DOX alone, CQ alone, CoCl alone, CQ:CoCl, and DOX:CoCl) were generated using a 24 h CoCl pre-treatment protocol matching the dose-optimization assays in Supplementary Figure S4, rather than the 48 h pre-treatment plus 24 h co-treatment protocol (72 h total) utilized for the DCC combination and flow cytometry assays. This technical distinction in pre-incubation duration is noted for clarity in interpreting the pairwise reference curves.

Drug interaction dynamics were formally modeled according to the Chou–Talalay median-effect equation (Chou, 2010) using CompuSyn software (ComboSyn, Inc., Paramus, NJ, USA). The Combination Index (CI) was calculated to define the nature of the pharmacological interaction: CI < 1 denotes synergism; CI = 1 denotes an additive effect; and CI > 1 denotes antagonism. The Dose Reduction Index (DRI) was computed to define the theoretical fold-reduction in the concentration of each therapeutic agent required within the combination to achieve a given fraction affected (Fa) compared with the concentration required for single-agent monotherapy.

Apoptosis and Cell Death Assay by Flow Cytometry

Apoptotic and necrotic cell death fractions were quantitatively resolved using an Annexin V/7-AAD dual-staining apoptosis detection kit (BD Biosciences, San Jose, CA, USA). For hypoxic groups, HepG2 cells were pre-incubated with 100 µM CoCl for 48 h followed by the addition of DOX (62.5 nM) and/or CQ (25 µM) for an additional 24 h (total CoCl exposure: 72 h). DOX at 62.5 nM was selected as a standardized sub-IC concentration based on preliminary viability dose-response assays indicating a normoxic DOX IC of 250–500 nM in this cell line, placing 62.5 nM at approximately 12–25% of the IC. This sub-IC concentration was specifically chosen to evaluate resistance and sensitization mechanisms operating under conditions of partial cytotoxic pressure, reflecting physiological tumor niches where incomplete drug penetration occurs. For normoxic groups, cells were treated with DOX and/or CQ for 24 h without CoCl.

Following incubation, floating and adherent cells were harvested, washed twice with ice-cold phosphate-buffered saline (PBS), and resuspended in 1× Annexin V Binding Buffer. Cells were stained with allophycocyanin (APC)-conjugated Annexin V and peridinin-chlorophyll-protein (PerCP)-conjugated 7-AAD for 15 min at room temperature (25°C) in the dark. Flow cytometric acquisition was conducted on a BD FACS Melody™ cell sorter (BD Biosciences), recording a minimum of 10,000 gated singlet events per sample. Flow cytometric data were analyzed using FlowJo software (Tree Star Inc., Ashland, OR, USA; BD Biosciences). Quadrant gates were set to resolve four distinct subpopulations: viable cells (Annexin V/7-AAD), early apoptotic cells (Annexin V/7-AAD), late apoptotic or secondary necrotic cells (Annexin V/7-AAD), and primary necrotic cells (Annexin V/7-AAD). Total cell death was defined as the aggregate percentage of early apoptotic, late apoptotic, and necrotic populations.

Flow Cytometric Analysis of LC3 Expression

To establish baseline autophagic induction protocols and calibrate flow cytometric detection of microtubule-associated protein 1A/1B-light chain 3 (LC3), LC3 expression was evaluated by flow cytometry. HepG2 cells were seeded into 6-well plates at a density of 1 × 10 cells per well. After 24 h, cells were exposed to rapamycin (0, 1, 2, or 4 µM) for 24 h in a dose-calibration study to determine the lowest effective concentration yielding reproducible LC3 accumulation. Based on these assays, 1 µM rapamycin was selected as the positive control for autophagy induction in subsequent fluorescence microscopy (viability curve in Supplementary Figure S2; flow cytometric validation in Supplementary Figure S3, demonstrating a shift in LC3A/B positivity from 0.13% to 74.4%). Because 1 µM rapamycin yielded near-saturating LC3A/B positivity, higher concentrations (2 and 4 µM) provided no additional analytical resolution and were not evaluated further.

For staining, harvested cells were fixed with 4% paraformaldehyde for 15 min and permeabilized with 90% ice-cold methanol at 4°C for 20 min. Non-specific binding was blocked with 5% bovine serum albumin (BSA) in PBS for 30 min. Cells were stained with 1 µL of DyLight™ 488-conjugated LC3A/LC3B polyclonal antibody (Thermo Fisher Scientific; Cat. #PA1-16930-DYLT488) for 25 min at 4°C in the dark, washed with PBS, resuspended in 1% BSA/PBS, and analyzed via flow cytometry. This assay measures total cellular LC3A/B protein accumulation in permeabilized cells rather than dynamic flux in real time.

Fluorescence Microscopy for Autophagy Assessment

Autophagic vesicle morphology was characterized by sequential Acridine Orange (AO) vital staining and LC3A/B immunofluorescence on identical cell preparations. HepG2 cells were seeded at 1 × 10 cells per well into 6-well plates containing glass coverslips. For CoCl-containing cohorts (CoCl alone and CoCl + CQ), cells received 100 µM CoCl for 48 h followed by 25 µM CQ for 24 h (72 h total CoCl exposure). Normoxic control cohorts (vehicle control, 1 µM rapamycin, and 25 µM CQ alone) were treated for 24 h under standard culture conditions.

Following incubation, live cells were first stained with 0.2 µg/mL AO (Sigma-Aldrich) in complete medium for 15 min at 37°C in the dark to label acidic vesicular organelles (AVOs, including autolysosomes and lysosomes) in their native, unperturbed physiological state. Cells were rinsed with PBS, immediately fixed with 4% paraformaldehyde for 15 min, and permeabilized with 90% ice-cold methanol at 4°C for 20 min. Following blocking with 5% BSA/PBS for 30 min, cells were incubated with DyLight™ 488-conjugated anti-LC3A/B polyclonal antibody (Thermo Fisher Scientific; Cat. #PA1-16930-DYLT488; green channel) for 25 min at 4°C in the dark. Fluorescent signals for AO (red-orange channel, Texas Red/long-pass filter set) and LC3 (green channel, FITC filter set) were captured in a single imaging session per field of view using a Zeiss Axio Observer A1 inverted fluorescence microscope (Carl Zeiss, Jena, Germany). Puncta quantification and individual cell area measurements were performed using ImageJ software (NIH, Bethesda, MD, USA). Three independent replicate preparations were quantified per experimental condition (one representative field per replicate; individual raw counts documented in Supplementary Table S6). Because AO staining was performed prior to cell fixation, the red-orange signal reflects the distribution of acidic compartments at the moment of vital staining; while aldehyde fixation can partially attenuate AO fluorescence intensity, it preserves relative inter-group staining differences. The green LC3 signal, detected post-permeabilization, reflects aggregated LC3A/B puncta corresponding to autophagosome structures. The coordinated occurrence of elevated green LC3 puncta and depleted red-orange AO fluorescence was interpreted as indicative of autophagosome accumulation secondary to blocked autolysosome maturation, in accordance with established autophagy guidelines15,16,17.

Statistical Analysis

Data are presented as mean ± standard deviation (SD). Viability assays were conducted with three technical replicates per condition and repeated across two independent biological experiments (N = 2) using distinct cell passages and independently prepared reagents. The limitation of N = 2 relative to standard statistical requirements is acknowledged in the Limitations section. The RT-qPCR data were statistically analyzed in the original publication13 and were not re-evaluated here. Flow cytometric apoptosis percentages are reported descriptively; formal non-parametric hypothesis testing was not performed on these data because the small sample size (2–3 technical replicates from a single flow experiment) cannot yield a two-tailed Mann–Whitney Up-value below 0.10.

For fluorescence microscopy data, puncta counts were quantified across three independent preparations per group and are presented descriptively. Morphological cell area measurements were obtained from individual cells across multiple fields, providing sufficient statistical power for inferential analysis via the non-parametric Kruskal–Wallis test followed by Dunn’s multiple comparisons post hoc test, with p < 0.05 established as the threshold for statistical significance. As an exploratory proof-of-concept investigation, prospective power calculations were not conducted. Statistical evaluations and graphical plotting were executed using GraphPad Prism software (version 9.0; GraphPad Software Inc., San Diego, CA, USA).

Validation of Chemical Hypoxia Induction by CoCl

The establishment of chemical hypoxia via CoCl in this HepG2 model was validated in our previous study13, which demonstrated significant transcriptional upregulation of downstream hypoxic markers (VEGF and GLUT1) and heightened DOX chemoresistance. In the present study, all hypoxia-mimetic drug combination experiments were performed using the identical validated CoCl regimen (100 µM, 72 h total exposure). In that validated dataset, exposure to 100 µM CoCl for 72 h elicited statistically significant increases in VEGF and GLUT1 mRNA expression compared with untreated controls (p < 0.05). In contrast, HIF-1α mRNA exhibited a minor, non-significant elevation. This differential pattern aligns with the established mechanism of CoCl, which stabilizes HIF-1α post-translationally by suppressing PHD-dependent enzymatic degradation, thereby driving the transcription of downstream effector genes (VEGF, GLUT1) without necessarily altering HIF-1α transcript abundance6,7. Because this validation was conducted at the mRNA level without direct chamber oxygen tension measurements or Western blot confirmation of HIF-1α protein accumulation, the experimental condition represents a chemically induced hypoxia-mimetic state rather than full physiological hypoxia.

CQ Inhibits Autophagic Flux in HepG2 Cells under Basal and Hypoxic Conditions

To characterize autophagic alterations, HepG2 cells were analyzed by sequential live-cell AO vital staining and fixed-cell LC3A/B immunofluorescence microscopy (Figure 1, Figure 2; full replicate imaging panels in Supplementary Figure S6). Treatment cohorts were structured to isolate the specific contributions of CQ and CoCl to autophagic machinery; DOX was deliberately omitted from imaging assays to prevent topoisomerase II-mediated DNA damage and off-target cytotoxic artifacts from confounding autophagic morphological readouts. Under basal normoxic conditions, vehicle-treated HepG2 cells exhibited low baseline intensities of both green LC3 fluorescence and red-orange AO signal, reflecting low basal autophagic turnover (Figure 1A). Treatment with the positive control rapamycin (1 µM, 24 h) induced a pronounced increase in both green LC3 puncta (autophagosomes) and red-orange AO staining (acidic autolysosomes), validating assay sensitivity for detecting robust autophagic pathway induction (Figure 1B).

Treatment with CQ alone (25 µM, 24 h) produced a distinct morphological pattern characterized by prominent accumulation of green LC3 punctate fluorescence alongside a markedly attenuated red-orange AO signal (Figure 1C). This phenotype—elevation of early autophagosome markers coinciding with loss of acidic vesicle fluorescence—is characteristic of impaired autophagosome–lysosome fusion and lysosomal deacidification, consistent with the pharmacological mechanism of CQ14. Under CoCl-induced chemical hypoxia (Figure 1D), both LC3-positive autophagosomes and AO-positive acidic vesicles increased compared with normoxic controls, indicating that hypoxic stress independently activates cytoprotective autophagy. Co-treatment with CoCl and CQ (Figure 1E) resulted in marked accumulation of green LC3 puncta alongside a substantial loss of red-orange AO signal compared with CoCl monotherapy. This indicates that CQ effectively impedes autolysosome maturation under hypoxic stress, causing autophagosome accumulation despite heightened upstream autophagic drive.

Quantitative image analysis of autophagic puncta and cell area is summarized in Figure 2 (individual counts in Supplementary Table S6). Rapamycin induced the highest total puncta count (mean 7,961 puncta/field across categories), followed by CoCl alone (2,636), CoCl + CQ (1,128), vehicle control (340), and CQ alone (137) (Figure 2A). In the CoCl + CQ group, autolysosome-associated puncta dropped from a mean of 973 (CoCl alone) to 165 puncta/field, whereas autophagosome-associated puncta remained elevated (650 puncta/field). Morphometric analysis revealed significant increases in mean cell area in the rapamycin (p < 0.0001), CoCl alone (p < 0.001), and CoCl + CQ (p < 0.05) groups relative to vehicle controls (Figure 2B), whereas CQ monotherapy produced no significant change. This cellular enlargement may reflect cytoplasmic vacuolation and swollen autophagosomal structures secondary to disrupted catabolic clearance.

CQ Sensitizes HepG2 Cells to Cell Death under Normoxic and Hypoxic Conditions

Cell death and apoptosis were quantified via Annexin V/7-AAD flow cytometry across all treatment groups (Figure 3, Table 1; complete replicate dot plots in Supplementary Figure S5, raw replicate data in Supplementary Table S5). Under normoxia, DOX monotherapy (62.5 nM, 24 h) induced minimal cell death (4.50% ± 0.28%). Co-administration of CQ (25 µM) markedly elevated total cell death to 98.32% ± 0.61%. Under CoCl-induced hypoxia, single-agent DOX elicited slightly reduced cytotoxicity (3.43% ± 0.45%), consistent with hypoxia-mediated chemoresistance, although this modest difference was not evaluated for statistical significance due to sample size constraints. The DCC triple combination (DOX + CQ + CoCl) produced 95.76% ± 0.88% total cell death.

Notably, the hypoxic control group receiving CoCl + CQ without DOX exhibited 96.43% ± 0.74% cell death, which is statistically indistinguishable from the DCC triple combination (95.76%). This demonstrates that under prolonged hypoxic stress (72 h CoCl), HepG2 cells become exceptionally dependent on basal lysosomal autophagic clearance for survival, such that late-stage autophagy inhibition via 25 µM CQ alone induces near-complete cell death. Consequently, the fixed-dose apoptosis assay displays a ceiling effect at 25 µM CQ, obscuring DOX-specific synergistic increments in this specific assay format. The full dose-ranging Chou–Talalay viability analysis described below resolves this interaction across dynamic effect levels.

Chou–Talalay Pharmacological Synergy Analysis

Effect of the DCC Triple Combination under CoCl-Induced Hypoxia

Under CoCl-induced hypoxia, the DCC combination demonstrated potent, dose-dependent inhibition of cell growth across the entire DOX concentration range (0.016–0.5 µM), with the fraction affected (Fa) ranging from 0.690 to 0.790 (Table 2, Figure 4). This elevated baseline Fa reflects the underlying growth-inhibitory contribution of 100 µM CoCl. Calculated CI values remained consistently below 1 across all tested concentrations (CI = 0.354–0.548), indicating stable drug synergism (Table 2, Figure 4D). Decomposition of the three-component CI equation (CI = 1/DRI + 1/DRI + 1/DRI) indicated that the CoCl term contributed 87% of the total index at 0.016 µM DOX and 47% at 0.5 µM DOX. Thus, the DCC values reflect a three-drug mathematical model and are not directly comparable to two-drug normoxic indices.

DOX Concentration (µM)	Fraction Affected (Fa)	Combination Index (CI)	DOX Dose Reduction Index (DRIDOX)
0.016	0.690	0.354	48.68
0.031	0.712	0.355	28.96
0.063	0.732	0.367	17.04
0.125	0.716	0.451	7.48
0.250	0.784	0.422	6.84
0.500	0.790	0.548	3.64

Dose Reduction Index (DRI) calculations for DOX revealed substantial dose-sparing potential, ranging from 3.64-fold at 0.5 µM DOX to a maximum of 48.68-fold at 0.016 µM DOX (Table 2, Figure 4C). DRI values for CQ and CoCl were omitted because both agents were maintained at fixed concentrations rather than varied at constant equipotent ratios with DOX, precluding rigorous pharmacological interpretation of their individual DRI values. In Figure 4B, the pairwise CQ:CoCl reference curve displayed minimal growth inhibition (Fa < 0.01) because it was generated using a shorter 24 h CoCl pre-treatment protocol during initial curve fitting, whereas the DCC and apoptosis assays utilized a 48 h pre-treatment plus 24 h co-treatment protocol (72 h total CoCl).

Effect of the DOX:CQ Double Combination under Normoxia

Under normoxic conditions, the two-agent DOX:CQ combination achieved a lower maximum growth inhibition (Fa ≈ 0.571) compared with the hypoxic DCC combination (Fa ≈ 0.790) (Table 3, Figure 5). In contrast to the stable synergism observed under hypoxia, the normoxic DOX:CQ combination exhibited a pronounced dose-dependent transition: robust synergism was observed at low DOX concentrations (CI = 0.345 at 0.016 µM; CI = 0.297 at 0.031 µM), which progressively shifted toward additivity (CI = 0.810 at 0.25 µM) and frank antagonism at the highest tested concentration (CI = 1.581 at 0.5 µM DOX) (Table 3, Figure 5C).

DOX Concentration (µM)	Fraction Affected (Fa)	Combination Index (CI)	DOX Dose Reduction Index (DRIDOX)
0.016	0.320	0.345	3.68
0.031	0.432	0.297	4.10
0.063	0.434	0.535	2.07
0.125	0.521	0.581	1.85
0.250	0.571	0.810	1.29
0.500	0.571	1.581	0.65

DRI values for DOX under normoxia ranged from 0.65 to 4.10, peaking at 0.031 µM DOX (Table 3, Figure 5B). Growth inhibition reached a plateau at 0.25–0.50 µM DOX (Fa = 0.571). Median-effect goodness-of-fit coefficients were r = 0.983 (r = 0.966) for DOX and r = 0.995 (r = 0.991) for CoCl, but r = 0.918 (r = 0.843) for CQ, introducing a degree of mathematical uncertainty that propagates into derived CI and DRI values.

Study Limitations

Several methodological limitations must be considered when interpreting these findings:

• Single Cell Line Model: All experiments were performed exclusively in HepG2 cells; validation in additional HCC lines (e.g., Huh7, PLC/PRF/5) and patient-derived primary models is essential to account for HCC heterogeneity.

• Chemical Hypoxia vs. Physical Hypoxia: CoCl₂ was utilized to chemically mimic hypoxia. Although validated by VEGF and GLUT1 mRNA upregulation13, CoCl₂ does not fully replicate physiological low-oxygen tension and may exert pleiotropic heavy-metal effects. Furthermore, direct Western blot confirmation of HIF-1α protein stabilization was not performed in this study.

• Autophagic Flux Validation: Autophagy was assessed via sequential AO/LC3 immunofluorescence rather than gold-standard biochemical assays (Western blotting for LC3 lipidation and p62/SQSTM1 degradation).

• Sample Size and Replicates: Viability assays were repeated across two independent biological experiments (N = 2), below the conventional standard of N ≥ 3 required for robust inferential statistics, and apoptosis assays comprised 2–3 technical replicates from a single experiment.

• Mathematical Modeling Constraints: The hypoxic Chou–Talalay model incorporated CoCl₂ as a third component drug, preventing direct numerical comparison with normoxic two-agent CI values. In addition, the median-effect fit for CQ (r² = 0.843) was lower than for DOX (r² = 0.966) and CoCl₂ (r² = 0.991).

• Lack of Normal Hepatocyte Controls: Non-malignant hepatocytes (e.g., primary human hepatocytes or LO2 cells) were not evaluated, leaving the therapeutic index and cancer cell selectivity uncharacterized.