Spatially resolved quantification of drug metabolism and efficacy in 3D paper-based tumor mimics

Abstract

Paper-based cultures are an emerging platform for preparing three-dimensional (3D), tissue- and tumor-like structures. The ability to stack individual sheets of cell-containing paper affords a modular means of assembling structures with defined cellular compositions and microenvironments. These layered stacks are easily separated at the end of an experiment, providing spatially resolved populations of live cells for further analysis. Here we describe a workflow in which cell viability, drug penetration, and drug metabolism can be quantified in a spatially resolved manner. Specifically, we mapped the distribution of the drug irinotecan and its bioactive metabolite SN38 in a colorectal cancer cells-containing stacked structure with liquid chromatography-mass spectrometry (LC-MS). This paper provides the first example of a 3D culture platform that quantifies viability and drug metabolism in a spatially resolved manner. Our data show that this stacked structure mimics observations in solid tumors. Cells at the bottom of the stack are more drug-resistant than layers in contact with the culture medium, similar to cells in the nutrient-poor center of a proliferating tumor being more drug-resistant than rapidly dividing cells at a tumor's periphery. The powerful combination of quantitative viability and drug metabolism measurements will enable future studies to determine the exact mechanism(s) of drug resistance in different regions of a tumor.

GraphicalAbstract

Introduction

The tumor microenvironment (TME) is defined by the cellular and extracellular components surrounding a rapidly proliferating tumor mass. The architecture and chemical composition of this dynamically evolving environment influence cancer progression, including the increased resistance to chemotherapies.^1-3 The chemical cues that promote drug resistance arise from paracrine signaling with the stroma and gradients of abiotic factors such as oxygen, nutrients, and pH that span the tumor mass. Current drug screening platforms, which aim to predict the efficacy and selectivity of potential cancer therapies, rely on monolayer cultures of cells grown directly on the surface of culture-compatible plasticware. While easily prepared and maintained, these monolayer structures lack some critical components of the tumor microenvironment. The planar surfaces that support monolayer cultures limit cell-cell and cell-extracellular matrix (ECM) interactions. They also eliminate the heterogeneous chemical environment within the tumor because the cells are exposed to a uniform concentration of abiotic factors and drugs. The replacement of this currently relied upon culture format with tissue-representative three-dimensional (3D) environments is necessary to better estimate chemotherapeutic efficacy, toxicity, and metabolism in vivo.^4-6

Incorporating aspects of the tissue microenvironment into in vitro cultures can improve current drug discovery and preclinical drug testing, using 3D structures to generate more representative cellular responses and phenotypes than monolayer cultures.^7-10 These 3D models can be classified by their method of preparation:¹¹ placing cells in contact with a naturally occurring or synthetically prepared ECM, loading cells into a preformed scaffold, or placing the cells in an environment that promotes the formation of multicellular aggregates or spheroids. Spheroids contain physiologically relevant abiotic and proliferation gradients¹², resulting from rates of cellular consumption outpacing diffusion. These structures are well-established tools to evaluate the distribution, efficacy, and metabolism of chemotherapies.^13-15

We and others have described an alternative 3D model for preparing the tumor-like structures found in spheroids.^16-19 These structures are prepared by stacking sheets of paper infused with cell-laden hydrogels. The number of sheets in the stack defines its overall thickness, the identity and density of cells in each layer of the stack define its composition. Using molecular biology-based readouts and oxygen sensing optodes, our group demonstrated that playing the stacks a holder that restricts exchange with culture medium to only the top results in chemical and proliferation gradients similar to those found in spheroids.^20-23 The paper-based culture platform overcomes three experimental limitations of spheroids. First, paper-based cultures are compatible with a greater number of cell types, as the scaffolds do not require tight cell-cell interactions. Second, preparation and stacking of the cell-laden scaffolds reduce the overall culture times needed to obtain a 3D structure of the desired thickness. Lastly, the separation of each cell-containing scaffold provides spatially resolved datasets of cellular viability after exposure to a drug without the need for fixation or histological slicing.

Drug distribution and metabolism in spheroids have been quantified with both optical and mass spectrometric techniques. Optical techniques rely on drug molecules that contain a unique spectroscopic handle, such as the intrinsic fluorescence of doxorubicin.^24,25 Mass spectrometry provides a more universal means of quantifying drug distribution and metabolism because the parent drug molecule and metabolites can be distinguished by their masses alone, alleviating the need for attaching a label. When coupled with the serial trypsinization of spheroids, liquid chromatography-mass spectrometry (LC-MS) quantified drug metabolism, as well as global metabolic changes, in different regions of the spheroid.^26,27 A microfabricated single-probe collection device also quantified extracellular metabolites and chemotherapeutics in spheroids without the need for serial trypsinization or slicing.²⁸

Previously, we showed that the antineoplastic agent SN38 had the most significant effect on HCT-116 colorectal cancer (CRC) cells in the top layers of the paper-based stacks, where the cells were proliferating.²⁰ This study relied on fluorescence-based readouts to assess viability, proliferation and determine the oxygen concentration in each layer of the stack. A limitation of this study was the inability to determine if the regions at the bottom of the stack, which were least affected by the presence of SN38, were due to decreased drug uptake or a consequence of a microenvironment-promoted phenotype. Using the same CRC cell line, we mapped cell viability and drug metabolism in these tumor-mimicking stacked structures composed of five cell-containing layers, dosed with irinotecan (CPT-11) for 24-96 h. Using the LC-MS method described, we show the intracellular concentration of CPT-11 and its bioactive metabolite (SN38) increased over the 96-h period. We show that cellular accumulation of SN38 is position-dependent, with CPT-11 metabolism occurring in regions of the stack that were in contact with the culture medium and corresponding to the proliferative region (i.e., the outermost part) of a tumor or spheroid. We also show that increased concentrations of SN38 correlate with decreased cell viability. This new workflow allowed us to confirm our hypothesis: that the microenvironment in lower regions of the tumor stacks (possibly hypoxia, low pH, decreased proliferation, or drug efflux) is causing cells to evade drugs. The method outlined here with any chemotherapy or cell type will enable future studies that determine which microenvironmental influences reduce drug efficacy.

Materials and Methods

Reagents.

All reagents were used as received unless otherwise noted. Acetonitrile (ACN), formic acid, methanol (MeOH), and water (Optima™ Grade) were purchased from Fisher Scientific. Anhydrous dimethyl sulfoxide (DMSO) was purchased from Sigma-Aldrich. Irinotecan (CPT-11), SN38, and camptothecin (CPT) were purchased from TargetMol. All cell culture medium and supplements were purchased from Gibco, except for McCoy's 5A medium (Corning) and fetal bovine serum (FBS, VWR).

Cell Culture, Seeding, and Tissue Stack Assembly.

The HCT116 colorectal carcinoma cell line was purchased from the American Type Culture Collection (ATCC) and validated by short tandem repeat sequencing. Cells were maintained as monolayers at 37 °C and 5% CO₂ in McCoy's 5A culture medium with L-glutamine, supplemented with FBS (10% v/v), HEPES (12.5 mM), and penicillin-streptomycin (1% v/v). Culture medium was exchanged every 2-3 days, and the cells were passed at 90% confluency with TrypLE, using standard procedures.

HCT116 cells suspended in Matrigel (Corning) were cultured in wax-patterned paper scaffolds, whose preparation and sterilization we detailed previously.²⁹ Sheets of Whatman 105 lens tissue paper were wax-patterned with a Xerox ColorQube 8750 printer, cut out with a 13" Silver Bullet Professional Series automated cutter, and sterilized overnight under ultraviolet light. Different grades of Whatman paper have been used as scaffolds to generate 3D environments in which cells and extracellular matrix readily penetrate to form a culture of defined thickness. Whatman 105 lens tissue is the thinnest paper made by Whatman (40μm), allowing for high spatial resolution of viability and drug metabolism in the paper stack. This grade of paper has a high porosity, allowing cells and extracellular matrix to penetrate the scaffold to generate a 3D culture environment. Large single zone (LSZ) scaffolds were used for all mass spectrometry experiments to detect low abundance metabolites. These scaffolds contained a 12.4 mm diameter cell-containing region, which was seeded with 2 × 10⁶ cells suspended in 12.5 μL of Matrigel. Small single zone (SSZ) scaffolds had a 3 mm cell-containing region seeded with 20,000 cells suspended in 0.5 μL of Matrigel. The same density of cells (cells/cm³) was maintained in the LSZ and SSZ scaffolds across all experiments. Schematics of each scaffold are in the Supporting Information (SI).

To generate a 200 μm-thick stack with the diffusion-dominated environment found in solid tumors, five cell-containing scaffolds were stacked and sandwiched between a custom-made acrylic holder. The bottom and top acrylic pieces were laser-cut from clear cast acrylic (McMaster-Carr). The base plate was a solid piece. The top plate contained an opening whose diameter matched the size of the cell-containing regions. This opening allowed a single point of medium exchange with the cell stack. The dimensions and patterns for both pieces are detailed in the SI. The holders were secured with four stainless steel screws, tightened to 25 cNm using a digital torque-measuring screwdriver. Assembled stacks were incubated in culture medium at 37 °C and 5% CO₂. Tissue stacks of LSZ scaffolds were placed in a sterilized beaker containing 35 mL of medium; stacks of SSZ scaffolds were placed in a 12-well plate containing 2.5 mL of medium.

Drug Treatment and LC-MS Sample Preparation.

Individual scaffolds or tissue stacks were incubated in culture medium containing 586.7 ng/mL (1 μM) CPT-11 or 2.5 ng/mL (6.4 nM) SN38 for either 24, 48, 72, or 96 h. Dosing concentrations of each drug were prepared by diluting stock solutions (1000x, in DMSO) in cell culture medium. The relative viability of the dosed cells was determined against vehicle controls containing the same concentration of DMSO (0.1% v/v). After drug treatment, each scaffold was washed for 5 min in 1X PBS to remove excess culture medium and drugs and then placed in a 1.5 mL microcentrifuge tube containing 1 mL of ice-cold MeOH (−20 °C) and 5 ng/mL CPT. To ensure complete lysis of the cells, the tubes were placed in a −20 °C freezer for 30 min and vortexed every 10 min. After 30 min, the scaffolds were removed, and the solutions centrifuged (20,000 xg, 4 °C, 15min) to pellet protein and remove any remaining paper debris. The supernatant was then decanted, dried under vacuum, and reconstituted in 100 μL 50/50 (v/v) H₂O/MeOH for analysis by LC-MS. Figure 1 is an overview of this workflow.

Figure 1.

Workflow for the assembly and analysis of a paper-based tumor stack model comprised of five scaffolds. (A) Individual scaffolds were seeded with HCT116 cells suspended in Matrigel at a density of 2x10⁶ cells/zone. (B) Five scaffolds were stacked and placed in a holder that limited the exchange with the culture medium. The stacked structures mimic aspects of the tumor microenvironment, forming consumption-generated gradients as indicated by the blue triangle. After a 24 – 96 h dose with CPT-11 or a no-drug control, (C) the layers of the tumor stack were peeled apart, the cells lysed with ice-cold methanol, and the small molecules extracted. (D) Prior to targeted quantification of drug and metabolite with LC-MS, the supernatant was concentrated under vacuum.

Cellular Viability.

Cell-containing SSZ scaffolds were exposed to either 586.7 ng/mL of CPT-11 or 2.5 ng/mL SN38 and viability was determined with the CellTiter-Glo (CTG) assay. These scaffolds were compatible with the CTG assay, whose linear range is 0-80,000 cells per scaffold.

LC-MS Method Development.

A Waters Acquity UPLC system was interfaced to a ThermoFisher LTQ XL linear ion trap mass spectrometer. Samples were separated on a Waters C8 Sunfire column (100 Å, 3.5 μm, 2.1 x 100 mm) with a solvent profile of 0.00–1.00 min of 10% B, 1.00–6.00 min of 90% B (gradient), 6.00–6.50 min of 90% B, and 6.60–8.60 min of 10% B. Solvent A was 0.1% formic acid in water. Solvent B was 0.1% formic acid in acetonitrile. The ionization source parameters were: source voltage, 4.8 kV; capillary voltage, 8.00 V; capillary temperature, 300 °C; HESI temperature, 300 °C; sheath gas flow, 50.00; auxiliary gas flow, 15.00; tube lens voltage, 110.00 V. Targeted analysis for each analyte was performed with the following MS/MS transitions: CPT-11 from m/z 587 to 543, normalized collision energy (NCE) 25, retention time (t_R) 3.00 min; SN38 from m/z 393 to 349, NCE 25, t_R 3.95 min; CPT from m/z 349 to 305, NCE 26, t_R 4.10 min. The automatic gain control was set to 3 × 10⁴ for MS/MS with a maximum allowed injection time of 100 ms. Data were analyzed with the Xcalibur Software package (ThermoFisher).

Working solutions of CPT-11, SN38, and CPT in 50/50 (v/v) H₂O/MeOH were prepared from a single dilution of 1000x stock solutions in anhydrous DMSO. Working solutions of CPT-11 (5000 and 1000 ng/mL) and SN38 (500 and 100 ng/mL) were used to prepare calibration curves. Calibration standards were prepared by spiking 100 μL of each drug standard into cell lysate extracted from the LSZ scaffolds. The limit of detection (LOD) and limit of quantitation (LOQ) values were calculated using 3s and 5s of the y-intercept for calibration curves analyzed on separate days. Quality control (QC) samples were prepared separately from 50, 500, and 800 ng/mL CPT-11 or 0.5, 10, and 20 ng/mL SN-38. These samples were prepared using the same procedure as calibration standards.

Extraction Efficiency.

Extraction efficiency was evaluated for both cell-containing and Matrigel-only LSZ scaffolds. To evaluate efficiency, 10μL of a 5887 ng/mL (10μM) solution of CPT-11 was spiked onto either cell-containing or Matrigel-only LSZ scaffolds. The drug spiked LSZ was placed in ice cold MeOH containing 5 ng/mL CPT and extracted using the procedure described above. For comparison, the same volume of CPT-11 was spiked into a "blank" extraction solution containing lysate from the LSZ. For the blank extract, CPT-11 was spiked into the supernatant post-centrifugation. Both sets of samples were dried under vacuum and resuspended in 100μL of 50/50 H₂O/MeOH prior to quantification by LC-MS.

Data Analysis.

Unless otherwise noted, values are reported as the average and standard deviation of at least three individual cell-containing samples prepared from the same passage of cells. Statistical comparisons and the fitting of the CPT-11 dose-response datasets were performed in GraphPad Prism 7. Statistically significant differences correspond to a p-value of ≤ 0.05.

Results and Discussion

Here we describe a targeted reversed-phase LC-MS/MS method to quantify the parent drug CPT-11 and its active metabolite SN38 in the tumor-like environment afforded by the paper-based culture platform. Specifically, we generated a diffusion-limited configuration that resembles the cross-section of a 400 μm-diameter spheroid by stacking five scaffolds containing HCT116 cells. We previously showed that cells at increasing distances from the culture medium experienced increased oxygen stress and hypoxia.^20,21 Hypoxia is known to confer drug resistance in cancer through several pathways³⁰ and has been correlated with poor outcomes in CRC. CRC is the third most lethal cancer in men and women in the United States,^31,32 with drug resistance accounting for 90% of chemotherapeutic failures in patients.³³ CPT-11 was chosen for this study because it is one of the most commonly used chemotherapeutics for CRC.³⁴ The distribution and metabolism of CPT-11 have been well- characterized in HCT116 spheroids,^26,35 making this combination an ideal benchmark for these paper-based models.

Method Development

Selectivity.

To quantify CPT-11 and SN38, CPT was included in all samples as an internal standard to account for matrix effects and sample loss during preparation. No-drug control samples confirmed there were no co-eluting interferents in the cell lysate. The LC run (8.6 min total) contained a 5 min gradient, which separated CPT-11 (t_R = 3.01 min), SN38 (t_R = 3.96 min), and CPT (t_R = 4.09 min). Representative chromatograms of all three compounds in cell extract are shown in Fig. S1. Fragmentation of CPT-11 resulted in multiple product ions, with the most abundant transition being m/z 587 to 543. Both SN38 and CPT had a single product ion, with transitions of m/z 393 to 349 and 349 to 305, respectively.

Linearity and Sensitivity.

Calibration curves of CPT-11 and SN38 were prepared from frozen drug stocks on the day of analysis. CPT-11 concentrations of 20.00, 50.00, 100.00, 200.00, 300.00, 500.00 and 1000.00 ng/mL were analyzed. SN38 concentrations of 0.25 0.50, 1.00, 3.00, 5.00, 10.00, and 25.00 ng/mL were analyzed. A 1/X weighting factor accounted for heteroscedastic errors across the concentration range; this fit resulted in linearities of R² > 0.99 (Fig. 2) and residual plots that confirmed no bias at low drug concentrations.

Figure 2.

Representative calibration curves for (A) CPT-11 and (B) SN38 with a 1/X weighting. The peak area ratio between CPT-11 or SN38 and the CPT internal standard is plotted on the y-axis of each curve. Both curves display duplicate injections as separate points. The first injection of the curve was done at the beginning of analysis and the second after all samples were run.

The SN38 curve had a sensitivity of approximately 0.03, a calculated LOD of 0.21 ng/mL, and a calculated LOQ of 0.36 ng/mL. The CPT-11 curve had a sensitivity of approximately 0.009, a calculated LOD of 10.51 ng/mL, and a calculated LOQ of 22.86 ng/mL. These values are summarized in Table 1. The concentrations of CPT-11 and SN38 in each dosed sample were above the LOQ values.

Table 1.

Calculated limit of detection (LOD) and limit of quantitation (LOQ) values for CPT-11 and SN38.^a

	CPT-11 (ng/mL)	SN38 (ng/mL)
LOD	10.51	0.21
LOQ	22.86	0.36

^aThe LOD and LOQ values were calculated using 3s and 5s of the y-intercept for calibration curves prepared on four separate days.

Accuracy and Precision.

We evaluated the accuracy and precision of CPT-11 and SN38 quantification with quality control (QC) samples made from stock solutions on the day of analysis. Three concentrations of CPT-11 and SN38 were evaluated: 50.0, 500.0 and 800.0 ng/mL and 0.5, 10.0, and 20.0 ng/mL, respectively. Inter-day accuracy and precision values, obtained over four different days and analyzed in conjunction with the calibration curves, are summarized in Table 2. Accuracy is defined as the percent error between the nominal and calculated concentration values of QC samples; these values are within 15% of the nominal concentration for both compounds. Precision is defined as the %RSD values calculated from the concentration of QC samples between days; these precision values are <10% for both compounds.

Table 2.

The accuracy and precision values for QC samples containing known concentrations of CPT-11 and SN38 in cell lysate collected from a LSZ scaffold containing 2 × 10⁶ HCT116 cells suspended in Matrigel.^a,b

CPT-11		Average (ng/mL)	Std. Dev (ng/mL)	%RSD	Accuracy (%)
High	800.0 ng/mL	690.19	8.67	1.26	86.28
Medium	500.0 ng/mL	448.75	9.74	2.17	89.75
Low	50.0 ng/mL	44.05	2.48	5.64	88.10
SN38
High	20.0 ng/mL	18.36	0.59	3.20	91.81
Medium	10.0 ng/mL	9.73	0.42	4.27	97.30
Low	0.5 ng/mL	0.44	0.02	4.78	98.75

^aPrecision is defined as the relative standard deviation (%RSD) obtained from analyzing multiple quality control samples.

^bValues represent quality control samples prepared and analyzed on four separate days. Calibration curves of known concentrations of CPT-11 and SN38 were prepared separately from the QC samples and were also collected each day.

Extraction efficiency from cell-containing and Matrigel-only scaffolds

To quantify the extraction efficiency of our method, we determined the concentration of a known amount of CPT-11 added to LSZ scaffolds containing 2 × 10⁶ cells immediately before extraction. Control samples prepared by adding the same amount of CPT-11 to post-extraction lysate and analyzed in parallel. The extraction efficiency of CPT-11 from the cell-containing scaffolds was 85.4 ± 5.8%; the efficiency from samples containing only Matrigel was 85.5% ± 15.8%. These data are summarized in Fig. S5. The concentration of CPT-11 was lower in the extracted samples for both the cell-containing and Matrigel-only scaffolds. These data show that sample loss cannot be attributed to the cells but rather to non-specific adsorption to the paper fibers or Matrigel proteins. The removal of both the scaffold and the proteins are necessary components of the extraction procedure, and thus can be best estimated with the addition of the CPT internal standard.

Increasing intracellular SN38 leads to decreased viability in the paper scaffolds

We quantified the conversion CPT-11 to SN38 in paper scaffolds seeded with 2 × 10⁶ HCT116 cells suspended in Matrigel over a 96 h incubation period. To account for unintended hydrolysis of CPT-11 to SN38 caused by the experimental setup, we also analyzed scaffolds containing only Matrigel. Each scaffold was placed in culture medium containing 586.7 ng/mL of CPT-11. The CPT-11 concentration in the cell-containing zones (Fig. 3A) increased 1.3-fold over the 96 h period, from 470.2 ng/mL to 631.4 ng/mL; the Matrigel-only scaffolds contained significantly less CPT-11 across the entire dosing period (165.6 ng/mL). The CPT-11 detected in the Matrigel-only zones is attributed to the drug's non-specific adsorption to the Matrigel and paper fibers.

Figure 3.

Time course for individual scaffolds containing either 2 × 10⁶ HCT-116 cells suspended in Matrigel or cell-free Matrigel incubated in 586.7 ng/mL (1 μM) CPT-11 for 96 h. At each time point, the amount of (A) CPT-11 and (B) SN38 in cell-containing zones (■) and Matrigel-only zones (▼) was determined following extraction and LC-MS/MS quantification. Each point represents the average and standard deviation of at least 10 individual cell-containing scaffolds obtained over three separate cell passages (N=3). Statistical significance (p < 0.05) to the 24h time point is indicated. The SI (Table S1) lists the numerical values and significance for CPT-11 and SN38 concentrations at each time point.

The intracellular concentration of SN38 increased 3.7-fold over the 96 h period (Fig. 3B). These concentrations did not increase monotonically with time but rather in a stepwise fashion: from 0.64 ng/mL at 24 h to 1.31 ng/mL at 48 h, displaying no discernible change between 48 and 72 h, and to 2.38 ng/mL at 96 h. The Matrigel-only controls contained no SN38, indicating the cells were converting CPT-11. The low conversion rate of CPT-11 to SN38 is expected as the expression of the carboxylesterase 1 and 2 enzymes that catalyze the reaction is low in HCT-116 cells.³⁶ When delivered intravenously, CPT-11 is predominately metabolized in the liver at an efficiency of 2-5% in vivo, and then delivered to the tumor.³⁷

To complement the LC-MS datasets, we determined cellular viability in each layer of the stack at each time point during the 96 h period (Fig. 4A). Increasing concentrations of SN38 correlated with decreased viability. These data support previous studies, which showed that SN38 is 1000-2000 times more cytotoxic than CPT-11.³⁸ The largest decrease in viability (between 48 and 72 h) is attributed to the replication of genomic DNA that is necessary in cell division. While this period is significantly longer than the experimentally determined doubling time of HCT116 in the paper scaffolds (31.4 h, Fig. S4), it likely is due to either decreased proliferation caused by CPT-11 or the time needed to accumulate a lethal dose of SN38. Previous studies showed that CPT-11 inhibits cell growth in monolayer cultures.³⁹ To determine if these results were due to time-dependent accumulation of SN38, we dosed cells with the concentration of SN38 present after the 96 h exposure to CPT-11: 2.5 ng/mL (6.4 nM) SN38. Direct exposure to SN38 show the largest decrease in cell number occurred at 48 h (Fig. 4B), supporting an increased doubling time in the presence of CPT-11.

Figure 4.

Cell viability as a function of exposure of HCT116-containing paper scaffolds to CPT-11 or SN38 exposure. Each graph plots viability relative to a no-drug control exposed to 0.1% DMSO. The red dashed lines mark the doubling time of the HCT116 cells in the paper scaffolds, in the absence of CPT-11. (A) A plot of viability (■) and SN38 concentration (■) as a function of exposure to 586.7 ng/mL (1 μM) CPT-11. Each point is the average and standard error of the mean of at least seven separate cell-containing scaffolds, obtained from two cell passages. (B) Plot of viability (●) as a function of exposure to 2.5 ng/mL (6.4nM) SN38. Each point represents the average and SEM eight separate cell-containing scaffolds, obtained from two cell passages.

Drug concentration and efficacy are position-dependent in the tumor stack

To determine if CPT-11 metabolism varied across a 200 μm-thick stack structure, we applied the workflow shown in Figure 1 to quantify CPT-11, SN38, and viability in each layer of the five-scaffold stack after a 96 h dose with 586.7 ng/mL CPT-11. Figure 5 summarizes this data. In agreement with our previous studies, we found that regions furthest from the source of nutrients and oxygen were less susceptible to drug exposure. The LC-MS data suggests cells in each layer uptake CPT-11 (Fig. 5A); the extent of this uptake decreases from the top (layer 1) to the bottom (layer 5) of the stack. The concentration of SN38 in each layer of the cell-containing stack (Fig. 5B) mirror the trend observed for CPT-11, with the highest amount of drug metabolite in the top layer. Layers 1 and 2 of the stack are the most susceptible to the drug, with the most cell death occurring in those layers relative to a no-drug control. No-cell control stacks contained significantly less CPT-11 in each layer when compared to the cell-containing stacks. The concentration of CPT-11 in each layer of the no-cell control stacks were also equivalent after a wash step.

Figure 5:

Graphical summary of the HCT-116 cell-containing stacks in which the x-axis refers to position in the relative to the source of fresh culture medium. Layer 1 is at the top of the stack. Layer 5 is at the bottom of the stack. Quantification of (A) CPT-11 and (B) SN38 in cell-containing (■) and Matrigel-only (▼) zones was determined with LC-MS. Each black square point is the average of four separate tumor stacks, prepared from a single cell passage. (C) Cellular viability in each layer of 200 μm-thick HCT116 tumor stacks after a 96 h dose with a no-drug control (■) or 586.7 ng/mL CPT-11 (■) was determined with the CTG assay. (D) Relative viability in each layer of the drug-dosed stack compared to the no-drug control. Each point in C and D is the average of five separate stacks, prepared over two passages of cells.

Figure 5C plots the luminescence intensity of each scaffold of the cell-containing stacks exposed to CPT-11 or a no-drug control. These luminescence values were determined with the CTG assay and are proportional to the cell number for conditions used. Figure 5D plots the ratio of the luminescence intensities for each layer of the CPT-11 stacks to the no-drug control stacks, highlighting changes in cell viability resulting from drug exposure and metabolism. Both plots show that the drug did not affect cellular viability in layers 3–5. This lack of effectiveness in these regions could be microenvironment-mediated decreases in cellular proliferation, increases in efflux pump expression, decreased pH leading to decreased drug uptake, or exposure to sub-lethal doses of SN38. Detailed studies of the tumor stack's cellular and microenvironmental aspects are needed to determine the exact mechanism(s) of drug resistance.

Conclusion

Leveraging the unique ability of the paper-based culture platform, we devised a quantitative method to map the distribution and metabolism of the drug CPT-11 and its effects on cellular viability in a multilayer structure that mimics the diffusion-dominated environments found in solid tumors. The LC-MS analysis of cell lysates obtained from the paper scaffolds accurately (<15% error) and precisely (<10% RSD) determined the concentration CPT-11 and SN38. We found that decreased cellular viability in the tumor stacks correlated with increased intracellular SN38 concentration. Despite distributing throughout the cell-containing stacks, CPT-11 and SN38 were less effective in regions furthest from the source of fresh culture medium. These findings suggest the microenvironment conferred drug resistance; the exact mechanism of this resistance cannot be determined from the current study.

The success of this approach provides a means of identifying regions of drug resistance within a tumor-like environment. It also promises to improve drug screening efforts with an easily set up and analyzed tumor model that can spatially resolve cellular responses to drugs in a tumor-like environment. An advantage of paper-based cultures is their modularity, allowing for cell lines that do not form spheroids to be evaluated in a 3D structure. The ability to incorporate different cell types into the stacks will allow us to determine how CRC tumors evade CPT-11 and other chemotherapeutics.

Highlights.

• LC-MS quantification of drug metabolism and viability in paper-based tumor models

• Intracellular SN38 accumulation over 96 h correlates with decreased cell viability

• Drug efficacy was highest in stack layers that support proliferating cells

• Drug concentration and metabolism decreases in hypoxic layers

• Quantitative analysis of tumor model indicates microenvironment-mediated resistance