In Vitro Cell Toxicity and Intracellular Uptake of Doxorubicin Exposed as a Solution or Liposomes: Implications for Treatment of Hepatocellular Carcinoma

Cytostatic effects of doxorubicin in clinically applied doses are often inadequate and limited by systemic toxicity. The main objective of this in vitro study was to determine the anti-tumoral effect (IC50) and intracellular accumulation of free and liposomal doxorubicin (DOX) in four human cancer cell lines (HepG2, Huh7, SNU449 and MCF7). The results of this study showed a correlation between longer DOX exposure time and lower IC50 values, which can be attributed to an increased cellular uptake and intracellular exposure of DOX, ultimately leading to cell death. We found that the total intracellular concentrations of DOX were a median value of 230 times higher than the exposure concentrations after exposure to free DOX. The intracellular uptake of DOX from solution was at least 10 times higher than from liposomal formulation. A physiologically based pharmacokinetic model was developed to translate these novel quantitative findings to a clinical context and to simulate clinically relevant drug concentration–time curves. This showed that a liver tumor resembling the liver cancer cell line SNU449, the most resistant cell line in this study, would not reach therapeutic exposure at a standard clinical parenteral dose of doxorubicin (50 mg/m2), which is serious limitation for this drug. This study emphasizes the importance of in-vitro to in-vivo translations in the assessment of clinical consequence of experimental findings.


Introduction
The anthracycline doxorubicin (DOX) is a well-established chemotherapeutic drug, commonly used to treat solid tumors and hematologic malignancies. While DOX treatment has shown great potential in slowing down disease progression over the past decades, clinically used doses often show insufficient anti-tumor efficacy, whereas higher doses frequently result in systemic toxicity in patients. Therefore, there is a pressing need to develop drug delivery strategies with increased DOX-induced anti-tumor potency, while reducing off-target toxicity to noncancerous tissues [1].
Doxorubicin and its main metabolite doxorubicinol (DOXol) induce cytotoxicity by multiple mechanisms of action, including triggering apoptosis and cell cycle arrest by DNA intercalation and inhibition of DNA repair mechanisms, as well as generating reactive oxygen species, thereby causing further DNA damage [2]. A major factor that determines the anti-tumoral potency of chemotherapeutics is the intracellular drug exposure, as well as

Overall Study Design and Investigational Drugs
In this study, the intracellular uptake and antitumor effect of DOX was investigated following exposure of two different formulations in four human cancer lines (described in detail below). One study drug was DOX as a solution (DOX S ) and the second was pegylated liposomal DOX (DOX PL ; Doxil ® ). The micro-PK and pharmacodynamics (PD) were monitored during 24-, 48-and 72-h exposure at DOX concentrations of 0.001-1000 µM and 0.1-1000 µM for DOX S and DOX PL , respectively. The effect of lansoprazole (500 µM) on the two study formulations DOX S and DOX PL was also investigated.

Cell Culture and Culture Conditions
The three PLC cell lines (HepG2, ATCC ® HB-8065™, SNU449 ATCC ® CRL-2234™, Huh7, kind gift from Dilruba Ahmed, Karolinska Institute, Solna, Sweden) and the human breast cancer cell line MCF7 (kind gift from Johan Kreuger, Uppsala University, Uppsala, Sweden) were cultured at 37 • C with 5% CO 2 and 95% humidity within a CO 2 incubator. The HepG2, Huh7 and MCF7 cell lines were routinely cultured in DMEM supplemented with 1% antibiotic antimycotic solution and 10% FBS (cell culture media + FBS: CCM Fed ). The SNU449 cell line was cultured RPMI medium supplemented with 1% antibiotic antimycotic solution and 10% FBS (cell culture media + FBS: CCM Fed ). Standard culture medium without supplemented FBS was used during starvation (CCM SM ). Misidentification of all cell lines was checked at the Register of Misidentified Cell Lines, and none of the chosen cell lines were on the list [28]. Extracted DNA from all four cell lines were sent to Eurofins Genomics (Ebersberg, Germany) for cell line authentication using DNA and short tandem repeat-profiles. Authentication confirmed the correct identity of each cell line and each cell line tested negative for mycoplasma contamination. The choice of cell lines was based on earlier validation of their IC 50 values for DOX during normal and hypoxic conditions. SNU449 was selected as it is one of the most drug-resistant commercially available cell lines, which is not genetically engineered to be chemo-resistant [29]. A breast cancer cell line, MCF7, was also included in the study to reflect the current clinical usage of Doxil ® , which is treatment of breast cancer in Europe and Canada.

Drug Treatment Schedule
Cells were plated at a seeding density of 1 × 10 4 cells/well in 200 µL CCM Fed onto clear black flat-bottom 96-well plates (Costar, USA), with the outermost wells filled with PBS to minimize the edge effect [30]. The cells were allowed to attach overnight. Prior to treatment, CCM Fed was removed, and the cells washed with 200 µL PBS. To allow synchronization of the cell cycle, 200 µL CCM SM was added to each well containing cells for 2 h prior treatment [30]. After cell cycle stabilization, cells were pretreated with 500 µM lansoprazole in CCM Fed for two hours (when applicable), followed by treatment with DOX S or DOX PL . For DOX S , the DOX stock solution in DMSO was used to create a range of 0.001-1000 µM in CCM Fed , and for DOX PL the product Doxil ® was used to create a range of 0.1 to 1000 µM in CCM Fed . Cells were treated with 200 µL/well of the respective DOX S and DOX PL preparations. Toxicity threshold values of DMSO were also evaluated in separate viability assays for each cell line. To determine the inert range of DMSO concentrations, the maximum dose of DMSO administered during DOX treatment (1% DMSO at 1000 µM DOX solution) times ten was set as the highest concentration. DMSO dilutions in CCM Fed in a concentration range of 0.01-10% (v/v) was administered to cells 24, 48 and 72 h.

Resazurin-Based Cell Viability Assay
Cell viability was evaluated after drug treatment for 24, 48 or 72 h by a resazurin reduction assay, as previously described [31]. A 1% solution of resazurin sodium salt (Sigma-Aldrich) was prepared according to manufacturer's recommendation and diluted 1/80 in CCM SM . Post treatment, culture medium was removed and the cells washed once with PBS. A volume of 150 µL resazurin solution was added and cells were incubated overnight. The relative fluorescence intensity was measured at excitation and emission wavelengths of 560/5 and 590/5 nm, respectively, on a Tecan Safire II plate reader. Six technical replicates were set up for each experimental group, with three biological replicates for all treatments, except for the pretreatment with lansoprazole where there was only one biological replicate.

Cell Lysate Preparation and Intracellular Determination of DOX and Major Metabolite DOXol
The cells were treated with concentrations of DOX S or DOX PL corresponding to their calculated IC 50 values, based on the pooled results of the cell viability assays in this study. The maximum concentration used for these experiments was chosen to be 200 µM. Cells were seeded at a seeding density of 4 × 10 6 cells per T75 flask (75 cm 2 , 60 mL), and allowed to attach overnight. Prior to treatment, CCM Fed was removed, and the cells washed with PBS. To allow synchronization of the cell cycle, 15 mL CCM SM was added to each flask containing cells 2 h prior to treatment. Cells were treated for 24, 48 and 72 h, respectively. Following drug treatment, culture medium was removed and kept on ice. Cells were washed twice with 5 mL PBS, which was collected and kept on ice. A volume of 3 mL trypsin-EDTA was added to each flask, after which they were incubated for 4 min. The resulting cell suspension was collected and diluted with 3 mL CCM Fed , and the cells counted using a TC20™ Automated cell counter and counting slides (Bio Rad). Using the histogram/gating option of the automated cell counter the average diameter of the cells was determined. The cell suspension was subsequently centrifuged at 140 g for 5 min, the supernatant removed, stored on ice, and the pellet was resuspended in 1000 µL ice-cold MilliQ water. Samples were lysed by two freeze-thaw cycles, consisting of snap freezing of samples in liquid nitrogen followed by thawing at 37 • C in a water bath for 10 min. After the final thaw cycle, samples were sonicated on ice for 30 s and stored at −20 • C until analysis.

Assay for Quantification of Intracellular Concentrations of DOX and Its Major and Active Metabolite Doxorubicinol (DOXol)
An ACQUITY UPLC I-Class system coupled to a single-quadrupole QDa mass detector (Waters Corporation, Milford, MA) was used to quantify intracellular concentrations of both DOX and its main metabolite DOXol. The mobile phases consisted of (A) 5.0 mM ammonium formate in water:acetonitrile (95:5) with 0.1% formic acid at pH 3.01 ± 0.07 and (B) acetonitrile. The analytes DOX and DOXol were chromatographically separated under the following LC conditions. The gradient used was: initially 5% (B), then a linear increase over 0.00-3.00 min of 5-30% (B), followed by 3.00-3.25 min linear increase to 85% (B), and then held at 85% (B) 3.25-4.25 min. Then linear decrease 4.25-4.50 min of 85-5% (B), and finally 5% (B) maintained 4.50-6.00 min. The total run time was 6 min, the flow rate was 500 µL/min, the sample injection volume was 10 µL, and the sample manager temperature was 10 • C. The column was a C18 column (ACQUITY UPLC BEH C18, 2.1 × 50 mm, particle size 1.7 µm, Waters Corporation) kept at 60 • C. In the single quadrupole QDa detector, the positive electrospray capillary voltage was set at 0.80 kV and probe and source temperatures were 600 • C and 120 • C, respectively. The quantification was performed in single-ion recording mode. The mass detection channels were set to m/z 544 (DOX), 546 (DOXol), 548 (DOX IS) and 550 (DOXol IS), each with a cone voltage of 8 V. The sampling frequency was 10 Hz.
Stock solutions of analytes and isotopically labelled internal standards were prepared in methanol (1 mg/mL). From these, working standards containing DOX and DOXol were diluted in methanol (0.25-1000 µM) and stored in amber vials at −20 • C. For quantification of DOX PL -treated samples, the commercial formulation (2 mg/mL) was diluted to working standards in methanol (5-1000 µM) and stored as mentioned above. All samples were prepared in 96-well 1.00 mL round collection plates from Waters ® . To construct calibration curves in sample relevant matrices, untreated control sample replicates were pooled and spiked (100 µL) with 25 µL of the appropriate working standard solution (0.25-1000 µM). Chilled (−20 • C) acetonitrile (375 µL) containing the isotopically labelled internal standards at matrix specific amounts (0.3-3 µM) were then added to precipitate cellular proteins. Samples were mixed (by pipetting up and down) before storage overnight at −20 • C. The following day, samples were brought to room temperature before mixing and subsequent centrifugation at 2200 rpm for 3 min at 4 • C. Portions of the supernatants (100 µL) were transferred to a new 96-well collection plate and dried under a gentle stream of nitrogen in a water bath (≈25 • C) using a 3D-printed manifold (U-PRINT) for at least 10 min. The residuals were dissolved in matrix appropriate volumes of mobile phase A (50 to 500 µL) and then injected into the instrument. Linear calibration curves (R 2 > 0.995) for both DOX and DOXol were constructed between 0.05 and 50 µM for the intracellular and washing matrices and between 1 and 200 µM for the exposure media matrix. A volume of 25 µL of methanol was added to cell portion samples (100 µL) of unknown concentration and they were subsequently treated exactly as the calibration samples. Matrix effects and extraction recoveries were evaluated according to Matuszewski et al. in 2003 [32]. The data were processed using a linear curve fit (weighting factor of 1/x) of the peak area ratio (analyte: internal standard) as a function of the analyte concentration. For both DOX and DOXol, the lowest limit of quantification (LLOQ) in matrix was 50 nM, which corresponds to the lowest point in the linear calibration curves. All the collected data were processed using TargetLynx as part of MassLynx V4.1 (Waters Corporation, Milford, MA, USA).

Data Analysis and Statistics
Cell viability (V), defined as the percentage of fluorescence value of treated cells compared to fluorescence value of untreated cells, according to Equation (1) [29]: where I exp is the measured fluorescence, I blank is the average fluorescence of a blank wells containing only the CCM SM+R without any cells in the plate and I cont is the average of control wells containing cells in CCM SM+R .
To accurately calculate viability values and the IC 50 values, Equation (2), referred to as a four-parameter logistic model equation, was fitted to the experimental data: where x is the concentration of DOX S or DOX PL , ∆V the viability at no DOX S or DOX PL exposure, i.e., corresponding to 100% viability, x 50 the concentration of 50% viability, and γ a parameter connected to the slope of the sigmoidal characteristic of the logistic equation.
In the fitting, V ∞ is the viability at infinite concentration of DOX S or DOX PL and was assumed to be equal to zero and consequently not fitted.
In principle, fitting of Equation (2) also gives x 50 . However, the estimation of x 50 using only Equation (2) was found to be highly dependent on the other parameters, and at viability values far from x 50 .
The value of x 50 was estimated as the linear interpolation (Equation (3)) between the x values corresponding to the V value closest above (V H ) and closest below (V L ) the V 50 = ∆V/2 value, i.e., half of the 100% viability value: In any experiment in which the cell viability did not reach less than V 50 , determination of x 50 could not be performed, and it was therefore excluded from the analysis. If more than one such experiment was detected for any replicate of a specific treatment condition, the entire treatment condition was excluded from further statistical study. The average and standard deviation of the x 50 values of the different replicate experiments were calculated, and these values are here called IC 50 .
To calculate the IC 10 values for the cells treated with DMSO, Equation (3) was used and adjusted so V H and V L were the viability values closest above and closest below the V 90 = ∆V × 0.9 value, i.e., 90% the ∆V value. From that equation, the standard error of IC 10 , here denoted s IC10 , was determined as: where s IC10 is the standard error of V 90 , s VL is the standard error of V L , s V H is the standard error of V H . For characterization of intracellular DOX and DOXol exposure there were two calculated parameters, cellular uptake ratio (IC UR ) and intracellular concentration (IC C ). These were calculated with Equations (5) and (6), and are based on the UPLC-MS measured concentration in the cell lysates.
In Equation (5), there are two main assumptions. Firstly, IC C is an average intracellular concentration that includes the higher concentration that is expected for the cell nucleus and mitochondria [2]. Secondly, it assumes that all cells are homogenous spheres with a diameter measured using a cell counter: where C lysate is the lysate concentration in µM, v lysate is the lysate solution volume, n cells the number of cells, and v cell the volume per cell. The cellular uptake ratio (IC UR ) was calculated according to Equation (6). This is an estimation of the fraction of DOX that was taken up from the surrounding media irrespective of the concentration or formulation applied. This parameter corresponds to the cellular availability of the study drug and is as such a dimensionless parameter. In this parameter, the IC C for both DOX and its main metabolite DOXol was included, as DOX is metabolized to DOXol intracellularly: where E Conc is the exposure concentration, IC C_DOX is the intracellular concentration of DOX and IC C_DOXol is the intracellular concentration of DOXol, the main metabolite of DOX (both IC C_DOX and IC C_DOXol were calculated using Equation (5)). All calculations were performed in Excel 2016 and supporting visual basic macros.
To judge the statistical significance of differences in responses for different cell lines, treatments and exposure times, mean values of IC 50 and IC UR were directly compared to respective standard deviation as well as by applying two-and three-way ANOVA on the IC 50 and IC UR data. For levels of factors containing more than two levels, the ANOVA was followed up by Tukey's post hoc test. Since there were differences of one or several orders of magnitude between values (especially when comparing DOX S and DOX PL treatments), logarithmic values were used in these comparisons and tests, which also had the advantage of higher homoscedasticy of data. The statistical software Minitab was used in these analyses.

Physiologically Based Modeling
Clinical evaluation of the experimental results was performed applying a physiologically based modeling approach (see Supplementary Materials for details). A physiologically based pharmacokinetic (PBPK) model developed by Hanke et al. was adopted to describe the disposition of DOX [33]. Briefly, in this model the disposition was described by unspecific tissue distribution, specific binding to DNA, unspecific metabolic hepatic clearance, unspecific elimination to bile, glomerular filtration and enterohepatic re-cycling. Furthermore, the distribution was optimized so as to accommodate for distribution to DNA containing blood cells, cell membrane translocation distribution and informed by DNA binding constants. The parameters of metabolism, bile excretion, and distribution, which could not be adequately informed by prior knowledge, were estimated by simultaneously fitting the model to clinical plasma, blood, urine and feces reference data. For the purpose of this study, an adjustment to the reported model was made in terms of reducing (1/1000) the molecular radius of DOX. This was done in order to harmonize DOX translocation over the vascular endothelial with the original model and to achieve equivalent results when using the PBPK model structure for proteins and large molecules in PK-Sim.
A model to describe the disposition of DOX PL was developed using the PBPK model structure for proteins and large molecules in PK-Sim [34]. DOX PL was modeled as an entity eliminated by DOX and limited to vascular and interstitial distribution, in accordance with the functionality and size of the drug delivery system. Model development was initiated by establishment of DOX PL distribution, where the overall distribution was described by the size of the liposome (40 nm radius assumed). The distribution was initially informed using reported information on initial (<24 h) liposomal distribution to plasma, liver, kidney, spleen and lung [35]. This signifies the liposomes distribution to the interstitial space as distribution to endosomes and intracellular space was not allowed. DOX PL elimination, i.e., release of DOX, and further optimization of the distribution (fat, muscles, bone, heart, skin, intestines, pancreas) was subsequently estimated. The release of DOX was described under the assumption that this process predominately occurs by unspecific spontaneous disintegration of the liposomes. This is supported by comparable systemic disappearance of DOX PL and unloaded liposomes [35,36]. In the model, this was parametrized as clearance via an unspecific enzyme homogenously distributed throughout the body. Optimization were performed using clinical reference data of measured total DOX (DOX and DOX PL ) in plasma (see Supplementary Materials for more details on the model development and final model) [36].
A new parameter, cellular exposure concentration (C Exp ), was established as a combination of the interstitial (C Int ) and intracellular concentration for liver cells. This parameter symbolizes the total concentration that the cells of interest will be exposed to. The established models were subsequently used, together with measured in vitro IC UR ratios for DOX, to assess whether adequate DOX concentration levels in cancer cells are achieved, i.e., within target therapeutic zone, after a clinically relevant dose (50 mg/m 2 ) DOX S or DOX PL . The therapeutic zone was defined on the basis of the cell viability investigations as the concentration range above the lowest concentration with detectable effect (~10%).

Cell Viability
Cell viability of HepG2, Huh7, SNU-449 and MCF-7 was measured after exposure to different concentrations of DOX S and DOX PL for 24, 48 and 72 h. This showed a difference in the sensitivity to both DOX S and DOX PL in the different human tumor cells, as shown by the (mean ± SD) IC 50 values (µM) (Table S1 and Figures S1-S3 in Supplementary Materials). The IC 50 values were more than one order of magnitude lower for the more sensitive cell lines HepG2, Huh-7 and MCF7, as compared to the more resistant cell line SNU449, under both DOX S and DOX PL treatment, which was statistically significant at all exposure times (p < 0.00005). In addition, Huh-7 showed a 3-fold difference to HepG2 and MCF7 when comparing Huh-7 with the other cell lines (p < 0.01). Furthermore, the relative difference between the mean IC 50 values between HepG2 and MCF7 following both DOX S and DOX PL treatments showed no clear difference, despite their different tissue origins (p > 0.5 at any time point and treatment).
As expected, there was a clear trend that longer exposure times for DOX S and DOX PL reduced the IC 50 values ( Figure 1 and Table 1). From 24 to 48 h, the IC 50 values showed a 4-18-fold decrease for all cell lines (p < 0.0005). From 48 to 72 h, the changes in IC 50 values were considerably lower for HepG2, MCF7 and SNU449 cell lines, while the increasing trend seems to continue for Huh-7 under DOX S treatment. When comparing sensitivity to DOX S versus DOX PL , it is notable that the IC 50 values in all cancer cell models were higher for the DOX PL than DOX S within the same cell model. Specifically, in the HepG2 and MCF7, we observed approximately 10 times higher IC 50 values for DOX PL compared to DOX S , as they ranged between 760 ± 120 µM (24 h in MCF7) to 110 ± 62 µM (72 h in HepG2), respectively ( Figure 1 and Table 1) (p < 0.00005). In the SNU449 cell line, and in some cases also for Huh7, the IC 50 was not precisely determinable for DOX PL . However, where it was calculable, a similar relative difference between the treatments was observed as for the other cell lines. The cell viability of the DOX-treated human tumor cells was investigated after a 2 h pretreatment with lansoprazole (500 µM), and the IC 50 data for DOX S and DOX PL are shown in Table 1. There were no clear trends in the effect of the PPI pretreatment on IC 50 values for neither DOX S nor DOX PL , and most of the pre-treatments had no appreciable effect. To assess whether the presence of DMSO in the DOX-solutions could have contributed to the toxicity, cell viability was determined after exposure to different concentrations of DMSO, as seen in Figure 2 (Table S1 in Supplementary Materials). The IC 10 values for all PLC cell lines were above 1% DMSO, which was the highest concentration of DMSO given to the cells in the cell viability assay used to determine the IC 50 values (1% DMSO at 1000 µM DOX S ). This was not the case for MCF7, which had an IC 10 of 0.23 ± 0.025% at 72 h exposure. While this is lower than the maximum dose of DMSO, the corresponding concentration of DOX (230 µM) was almost 10 times as much as was needed to kill all cells at 72 h exposure. These data suggest that DMSO was unlikely to contribute to toxicity of DOXs in any of the four different cell lines.

Intracellular Concentration and Uptake Ratio of DOX and DOXol Determined with a Bioanalytical Method Using UPLC-MS
In this study, a UPLC-MS method was developed to simultaneously quantify DOX and its primary metabolite DOXol in cell lysate (intracellular compartment) and in extracellular media. Matrix effects were most pronounced in the intracellular sample portions, suggesting ionization enhancement effects. Interestingly, extraction recoveries were higher for DOXol than DOX, but the isotopically labelled internal standards corrected for these differences (Table S2). The selectivity was demonstrated using matrix blanks injected before each calibration curve and sample. No peaks were detectable at the retention times corresponding to the analytes and internal standards of interest. After high-concentration calibrators or samples, blanks were injected and no carry-over peaks were observed. Quality control (QC) samples at low (0.1 µM), medium (1 µM) and high (10 µM) concentration levels, within the linear range, were prepared in triplicates for both analytes in each matrix and for each sample run. The relative bias and relative standard deviation for DOX QC samples were 3.0 and 11.7% (low), 5.4 and 10.4% (medium) and 0.5 and 6.0% (high), re-spectively. For DOXol QC samples, the numbers were −0.5 and 20.0% (low), 1.7 and 8.6% (medium), and −0.4 and 7.6% (high). In Table 2, the measured extra-and intracellular amounts of DOX and DOXol are shown. Cell lines were treated with extracellular concentrations of DOX S or DOX PL corresponding to their calculated IC 50 values ( Table 2). The maximum concentration of DOX S or DOX PL used was 200 µM. The relationship between intracellularly and extracellularly quantified DOX and DOXol is demonstrated in Figures 3 and 4, respectively. The measured median diameter of the cells was 14 µm for SNU449, Huh7, MCF7, and 12 µm for HepG2. The number of tumor cells left after the end of the drug exposure period for each tested cell line is given in Table 3. Table 2. The calculated amounts (mean± SD) intracellular DOX (IC A DOX) and DOXol (IC A DOXol), as well as the extracellular DOX (EC A DOX) and DOXol (EC A DOXol) of each exposure amount (E A ). When the quantified amount was below the LLOQ, it is symbolized with "n/a". Each E A was based on the calculated IC 50 of that treatment with a maximum E A of 3000 nmol. Please note that IC A DOXol is reported in pmol, while the other amounts are reported in nmol.   The intracellular uptake ratios (IC UR ) for DOX S and DOX PL into the human cell lines during the three different treatment periods are given in Table 3. In all cell lines and treatment periods, the IC UR for DOX S and DOX PL ranged from 4.5 to 1500 and 0.12 to 5.2, respectively. In general, the ratio IC UR DOX S /IC UR DOX PL had a median value of 87 across all cell lines and time points (p < 0.00005). The lowest IC UR and IC C of DOX for both formulations was observed for SNU449, which is in accordance with this cell line having the highest IC 50 -value (Tables 1 and 3). The range of IC UR and intracellular concentration did not show any trend among these cell lines Huh7, MCF7, and HepG2. Table 3. The number of cells that were quantified, calculated mean (±SD) intracellular concentration (IC C ) of DOX and DOXol and intracellular uptake ratio (IC UR ) for DOX in the different cell lines at different exposure concentrations (E Conc ). When the quantified amount was below the LLOQ, it is symbolized with "n/a". Each E Conc was based on the calculated IC 50 of that treatment with a maximum E Conc of 200 µM, see Table 1 for more details. DOXol was formed in all cell lines to a similar extent after 24 h exposure of DOX S (Figure 4a). There was no difference between SNU449 and the three more sensitive cell lines, which suggests that metabolic formation of DOXol mediated by cytosolic enzymes, such as carbonyl reductases and/or aldo-keto reductases, does not explain the lower sensitivity of SNU449. The extracellular concentration of DOXol was higher at all exposure times for SNU449 and HepG2 when exposed for DOX S which is most likely mediated by efflux proteins expressed in these cell lines. The efflux of DOXol was most pronounced in HepG2 and SNU449 (Figure 4a,c,e). In general, the formation of DOXol was far lower after treatment of DOX PL , which is in accordance with its lower IC UR of DOX as carbonyl reductases and/or aldo-keto reductases are located intracellularly (Figure 4b,d,f).

Physiologically Based Pharmacokinetic Modeling
The model predicted extensive intracellular accumulation of DOX, mediated by the parametrization of DNA binding, where a maximum exposure to an interstitial DOX concentration ratio of~10,000 was simulated for the liver (Figure 5a,b). Measured IC UR values for investigated cancer cell lines were adopted as a surrogate for intracellular accumulation in tumor tissue for different variants of HCC indicated possibilities for large differences in intracellular DOX exposure with potential consequences of susceptibility to DOX. The predicted average of 168 h cellular exposure, i.e., the sum of interstitial and intracellular concentration, was 0.5-1 µM for HepG2, Huh-7 and MCF7, while it was 40 times lower (0.02 µM) for SNU449 (Figure 5c). In relation to the in vitro viability investigation, this indicates that a SNU449-like tumor would not reach therapeutic intracellular DOX levels at a parenteral dose of 50 mg/m 2 . Similar levels, as well as differences, in cellular exposure between cell lines, were predicted for a simulation of DOX PL (50 mg/m 2 ) (Figure 5d). Notably, the total cellular exposure (AUC) was comparable between the two formulations (2500 and 1800 µM·h, respectively), despite a 500-fold higher plasma exposure after administered as DOX PL compared to DOX S (3.0 and 1600 µM·h, respectively).

Figure 5.
The results from the physiologically based simulation. (a,b) Simulated cellular exposure (C Exp , a combination of the interstitial and intracellular concentration for liver cells), interstitial concentration in the liver (C Int ), and plasma concentration (C P ) of DOX when given as DOX PL and DOX S . Treatment was set to a dose of 50 mg/m 2 for both treatments, with C Exp in red, C Int in green and plasma concentration in teal, for both DOX S (full lines) and DOX PL (dotted lines) as well as the non-liposomal DOX found in plasma after DOX PL treatment (C P.nl , circles). (c,d) C Exp is modelled based on experimental IC UR for the PLC cells, with healthy liver cells (red), SNU449 (purple), HepG2 (blue) and Huh7 (orange). This is shown for DOX S treatment (a,c) and DOX PL treatment (b,d) at a dose of 50 mg/m 2 . The dotted line symbolizes the therapeutic zone, which was defined from the cell viability investigations as the concentration range above the lowest concentration with detectable effect (~10%). Simulations were performed with PBPK models for DOX S and DOX PL , as described in Section 2.9, adopting a typical male individual (30 years old, 73 kg).

Discussion
In this study, cytotoxic potency and its relation to intracellular exposure of DOX and its active metabolite doxorubicinol (DOXol) was investigated in four human cancer cell lines, following treatment of DOX as a solution (DOX S ) or as the nano-sized pegylated liposome Doxil ® (DOX PL ). Three human liver cancer cell lines (HepG2, Huh-7 and SNU449) and one human breast cancer cell line (MCF7), used as a reference, were selected based on previously published data and on their response to DOX [29]. The cell viability was determined by calculating IC 50 values for DOX S and DOX PL at wide extracellular concentration ranges using a novel and more accurate approach, prioritizing the crucial concentrations in the mid part of the slope in the viability curves (Equations (2)-(4)). The effects on cell viability between the different cell lines and exposure times were related to the cellular uptake and intracellular concentrations of DOX and DOXol. This comparison was also extended into a theoretical model where extra-and intracellular concentration-time profiles of DOX were simulated to assess the significance of experimental IC UR values in terms of antitumor effect in a clinical situation. These findings drive the field forward in increasing the understanding how drug response between is affected by intracellular uptake of DOX, both in the context of preclinical research using cell lines, as well as the potential translation to the clinic through theoretical modelling.
HepG2 and MCF7 were the most sensitive cell lines to both treatments in this study [37,38]. HepG2 cells have been reported to have limited tumorigenic potential in in vivo xenograft models, and overall express less oncogenic proteins than other, more resistant, cell lines, such as Huh7 [29]. In this study, the Huh-7 cell line has been shown to be more resistant to DOX PL than HepG2 and MCF7 at all-time points, while this difference is not as clear for DOX S . One potential explanation for the difference in cell viability is that HepG2 cells carry wild-type p53, while Huh-7 cells are characterized by a point-mutation in the p53gene [39,40]. Inactivation of p53 has been shown to mediate resistance to DOX treatment in breast cancer cell lines and could thus play a similar role in determining sensitivity to DOX in the different PLC cell lines [39,41].
In this study, it was clearly demonstrated that SNU449 was the most resistant cell line against both treatments, which is in line with earlier reports [29,42]. A possible explanation for this increased resistance is the lower intracellular uptake ratio of DOX from both DOX S and DOX PL in the SNU449 cells, compared to the other cell lines. Membrane efflux, as well as other Fascin-1-mediated mechanisms that affect cell adhesion, cell-cell interactions and motility, could also be linked to the observed increased resistance [35]. The lower IC C could be a result of lower uptake and more extensive metabolism. An argument against the higher metabolism in this cell line is the results shown in Figure 4, where the lower intracellular uptake ratio in SNU449 did not correspond with a higher amount of DOXol formed, neither inside nor outside the cells, as compared to the more sensitive cell lines. Based on previous research, it is known that poorly differentiated cell lines, such as SNU449, provide better study models for human HCC profiles, and may contribute to understanding drug-resistance observed in clinical practice [43]. Our theoretical model simulations reveal that the level of resistance observed in SNU449 cells could lead to sub-therapeutic DOX exposure in a clinical context (50 mg/m 2 ) (Figure 5c,d). This was irrespective of whether DOX was administered as a solution or as DOX PL , despite the large difference in total DOX plasma exposure, which is in accordance with DOX treatment being ineffective in some patients. Although a certain level of resistance was experimentally determined for the other PLC cell lines investigated (HepG2 and Huh-7), the cellular exposure was predicted to reach the therapeutic zone (>0.1 µM) after a 50 mg/m 2 dose (Figure 5c,d). This further supports the importance of tumor characterization before selecting a strategy for therapeutic intervention, and emphasizes the potential of personalized medicine [44]. In addition, these data confirm SNU449 cells as a valuable study model for studying resistance to DOX and DOX PL in HCC and for using multiple cell lines in cancer research, in order to draw conclusions that are relevant for different tumor phenotypes.
The lower IC 50 values at longer exposure times observed in this study could be attributed to an overall increased cellular uptake of DOX and a subsequent accumulation inside the cell nucleus, where it induces cell death through different mechanisms, including ferroptosis [45,46]. However, the effect of further exposure time is significantly decreased or ceases between the 48 to 72 h time interval for the HepG2, SNU449 and MCF7 cell lines, while it might be maintained for Huh-7 [29,45,47,48]. The clinical benefit of an extended drug extracellular exposure time, which is a localized formulation strategy for TACE, can lead to a potential synergistic increase in intracellular uptake ratio of DOX and a prolonged intracellular retention time and effect duration. However, our results suggest that this synergistic effect might depend on the tumor phenotype and intercellular differences, as we did not observe this in all cell lines. The in vitro exposure times in our study were selected based on their clinical relevance for the local pharmacokinetics of DOX and DOXol following TACE injections with an emulsion-based formulation (Lipiodol ® ) [24]. The DOX plasma concentrations in HCC patients have been reported to be approximately in the range 0.2-30 µM for several days in central and peripheral veins, with corresponding DOXol plasma concentrations between 0.01-0.03 µM [14,49,50]. The total intracellular concentrations of DOX in the cell lines used in our study, when exposed to DOX S or DOX PL at their IC 50 values, were several fold higher than their exposure concentrations (Table 3). With TACE, DOX is delivered directly to the hepatic tumor tissue, and the blood flow is then often embolized to maintain a high DOX concentration around the tumor, as well as to block the tumor's blood supply [1,23]. In these cases, the local concentration in or around the tumor tissue in patients is expected to be substantially higher. Studies using VX2rabbits have shown that total intra-tumoral DOX concentrations following conventional TACE of 2 mg/kg DOX were in the range 8-17 µM up to 3 days after dosing. In the same study, DOX was also given at the same dose in microspheres, with DOX intra-tumoral concentrations ranging between 20 to 65 µM [49]. In addition, multi-sampling of plasma in a domestic pig model (10-12 kg), showed that two repeated intravenous doses of 64 mg DOX resulted in resected liver tissue concentrations (corrected for blood contribution) of approximately 25 µM [50]. This concentration range corresponds with the PBPK model in this study ( Figure 5).
There was a clear difference in cell viability between DOX S or DOX PL in all investigated cell lines, with DOX S several-fold more potent than DOX PL . These ratios are in accordance with earlier reports, in which DOX S was shown to be 10 to 1000 times as potent as DOX PL during different exposure conditions [15,48]. Other in vitro studies have also reported that encapsulated DOX might have lower IC 50 values than DOX S . For instance, Li et al. in 2020 reported DOX-loaded micelles in the range of 0.35-0.50 µM in three different cell lines (for HCT-116, HT-29 and SW480 cells) [51]. Their IC 50 for free DOX was reported to be 2.6-3.1 µM [51]. These findings were also supported by comparing free DOX and DOX loaded in dendrimers, where IC 50 s were 1.4 µM and 0.50 µM, respectively [52]. In our study, we found that the IC C for DOX PL was lower than for DOX S , showing a 1 to 3 log value difference for each cell line and time point. For MCF7, the IC UR value increased with exposure time, while no clear trends can be noted in the other cell lines. A plausible hypothesis for the lower in vitro potency for DOX PL is a low, variable and incomplete release of DOX extracellularly from the liposome (i.e., prior to DOX cellular uptake) [53]. Another explanation could also be a low cellular uptake of this particular nano-sized DOX PL and/or low intracellular DOX release. Some reports propose that the main mode of action for DOX PL is attributed to DOX that is released extracellularly and subsequently taken up by the cells [15,54]. There is currently no consensus regarding the extent of transport of the pegylated liposome (Doxil ® ) into cancer cell lines in vitro. Taking together our findings supporting a decreased potency of DOX PL compared to DOX S and the existing controversy in published literature, further research of the structure-activity relationships for cellular targeting and uptake for various types of TNP is encouraged, including pegylated liposomes.
DOX passes across cell membranes by passive lipoidal diffusion, as well as by carriermediated influx and efflux processes [3]. Cellular pharmacokinetics, cellular uptake and intracellular distribution have all been shown to be pH-dependent in accordance with the pH-partitioning model [10]. In this study, the extracellular pH ranged from 7.33-8.81 after treatment (Supplementary Material Table S3), which is expected to facilitate the passive membrane diffusion of DOX. In our in vitro study, the 2-h pretreatment of lansoprazole had no consistent effect on the observed IC 50 values for DOXs or DOX PL in the examined cell lines. One plausible explanation is that slightly acidic conditions (pH 6) are required in order to have a PPI-effect on DOX potency [55].
The high IC UR of DOX and the lack of distinct correlation with the extracellular concentration in the four cell lines suggests that passive diffusion is a key mechanism for the overall intracellular delivery of DOX into these cell lines, which is supported by previous studies [4,10]. The high intracellular concentration of DOX is possible due to the high binding and retention capacity of DOX to nuclear and mitochondrial DNA, which has been clearly reported to be main intracellular binding sites of DOX [26,56]. The high cellular accumulation of DOX after treatment of DOX S might contribute to reduced anti-tumor effect in avascular tumor regions in patients, as it reduces the intercellular diffusion of deeper tumor regions.
The metabolite DOXol can be quantified in vivo to measure how efficiently DOX is taken up and metabolized by cells; however, this strategy is seldom replicated for in vitro cell assays [57]. This can possibly be related to failure in detecting the often low-abundant DOXol in various cellular compartments. Our approach to quantify DOXol in extra-and intracellular samples proved to be successful and revealed that the extracellular amount of DOXol exceeded the intracellular amount approximately two-fold, especially in HepG2 and SNU449, when exposed to DOX S . However, the concentration of DOXol was still overall higher intracellularly, as the liquid volume measured extracellularly was consistently higher compared to the intracellular volume. The difference in concentration might be due to the fact that DOXol was formed inside the cells and is still mostly bound to different intracellular compartments, reducing the diffusion and/or carrier-mediated transport of DOXol out of the cells. The quantification of DOX and DOXol is assumed to reflect one single compartment, which is a simplification as DOX most likely accumulates in the cell nucleus, as previously reported. Instead, this experimental setup gives an average intracellular concentration that encompasses both the "free" DOX in the cytoplasm as well as the DOX intercalated in the nuclear and mitochondrial DNA.
In all stages of drug discovery, there is a clear need to distinguish between the observed effects of the agent's concentration-dependent potency and its potential active metabolites at the site of action. Improving the understanding of drug exposure at the site of action is especially relevant in pharmacological cellular or tissue models, as the fraction of the drug that reaches the site of action in the cell or some sub-cellular space is usually unknown. In this study, the IC 50 values were calculated based on the determined cell viability data at various exposure concentrations using a novel and more accurate approach, which prioritizes crucial concentrations in the mid part of the slope in viability curves. As the cellular uptake is several folds higher for the sensitive cell lines (HepG2, MCF7 and Huh-7), the difference in the calculated in vitro IC 50 could be attributed to cellular net uptake of DOX and intracellular disposition (such as metabolism, efflux and lysosometropic activity). These differences between cell lines, formulations and exposure time which affect the in vitro cellular uptake, intracellular retention and exposure, will affect cell viability after DOX-treatment and warrants further investigations in order to fully optimize the concept of localized tumor drug treatment.

Conclusions
Overall, the results of this study show DOX S to be a more potent chemotherapeutic formulation compared to the nano-sized liposomal DOX PL (Doxil ® ) during the applied in vitro conditions. This is most likely due to a slower and incomplete uptake of DOX from the liposomal formulation, which might be a result of lower extracellular availability of DOX when dosed as DOX PL . While preclinical studies have reported on the potential of TACE and TNPs combination therapy in HCC treatment, the data in this study demonstrate the need to design TNPs with a high drug load and a tumor triggered release of DOX [58]. We found that the total intracellular concentrations of DOX had a median value of 230 times higher than the exposure concentrations. This high concentration-independent cellular uptake and accumulation further supports the hypothesis that passive diffusion is a key transport mechanism for DOX in these cell lines. The potent anti-tumoral effect of DOX from both DOX S and DOX PL maintained over time (lower IC 50 ), suggests that extending high local DOX concentration in the liver tumor during TACE treatment could be an efficient therapeutic strategy for HCC. The PBPK model used in this study provided a more accurate description of the DOX concentration-effect relationship. The performed simulations exemplify the relevance of in vitro to in vivo translations in the assessment of clinical consequence of experimental findings, and highlight the importance of tumor characterization before selecting a strategy for therapeutic intervention.