The Influence of Matrix-Induced Dormancy on Metastatic Breast Cancer Chemoresistance

Abstract

Metastasis remains the leading cause of cancer-associated death worldwide. Disseminated tumor cells can undergo dormancy upon infiltration of secondary organs, and chemotherapeutics fail to effectively eliminate dormant populations. Mechanistic understanding of dormancy-associated chemoresistance could lead to development of targeted therapeutic strategies. Toward this goal, we implemented three poly(ethylene glycol) (PEG)-based hydrogel formulations fabricated from proteolytically degradable PEG (PEG-PQ), integrin ligating PEG-RGDS, and the non-degradable cross-linker N-vinylpyrrolidone (NVP) to induce three distinct phenotypes in triple negative MDA-MB-231 breast cancer cells. With constant 5% w/v PEG-PQ, PEG-RGDS and NVP concentrations were tuned to induce (i) a growth state characterized by high proliferation, high metabolic activity, significant temporally increased cell density, and an invasive morphology; (ii) a balanced dormancy state characterized by a temporal balance (~1:1 ratio) in new live and dead cell density and a non-invasive morphology; and (iii) a cellular dormancy state characterized by rounded, solitary quiescent cells with low viability, proliferation, and metabolic activity. The cellular responses to doxorubicin (DOX), paclitaxel (PAC), and 5-fluorouracil (5-FU) in the three phenotypic states were quantified. Under DOX treatment, cells in dormant states demonstrated increased chemoresistance with a 1.4- to 1.8-fold increase in half maximal effective concentration (EC₅₀) and 1.3- to 1.8-fold increase in half maximal inhibitory concentration (IC₅₀) compared to cells in the growth state. PAC and 5-FU treatment led to similar results. To mechanistically investigate the role of dormancy in conferring DOX resistance, cytoplasmic and nuclear accumulation of DOX was measured. The results indicated comparable DOX accumulation between all three phenotypic states; however, the intracellular to intranuclear distribution indicated a ~1.5 fold increase in DOX nuclear accumulation in cells in the growth state compared to the two dormant states. These results further validate the utility of implementing engineered hydrogels as in vitro platforms of breast cancer dormancy for the development of anti-dormancy therapeutic strategies.

Graphical Abstract

1. INTRODUCTION

Advances in cancer detection, diagnosis, and treatment over the past few decades have significantly improved patient outcomes.¹ Despite these significant advances, metastasis (dissemination, invasion, and proliferation of tumor cells in secondary organs) is still responsible for 90% of cancer-associated mortality.² Currently, metastatic breast cancer patients have a dismal 5 year survival rate of 27%.¹ One significant challenge in metastasis treatment and prevention is a poor understanding of the underlying mechanisms that regulate both metastatic relapse and the latency period commonly observed prior to relapse.

The latency period is determined by cellular dormancy, in which residual disseminated tumor cells (DTCs) reside in a dormant state that retards the growth of clinically detectable tumors.³ Dormant cells may undergo cellular quiescence ranging from months to decades⁴ before returning to an actively proliferative state resulting in metastatic relapse. DTCs originating from the primary tumor may also extravasate into secondary organs where <2% of the cells survive.^5,6 Surviving DTCs undergo one of three possible fates: cellular dormancy,^3,7-9 tumor mass dormancy,^3,7,10-12 or growth.^13,14 In cellular dormancy, surviving cells employ anti-apoptotic survival mechanisms to remain in G₀-G₁ cell cycle arrest. Cells undergoing tumor mass dormancy maintain a delicate balance between proliferation and apoptosis, which prevents tumors (or multi-cellular clusters) from growing to a clinically detectable size. A multitude of microenvironmental factors including immunosuppression, lack of angiogenic signaling, extracellular matrix (ECM) properties, and environmental factors regulate cancer dormancy.^12,15

In particular, the ECM has key biochemical and mechanical components, which influence how a tumor cell interacts with its microenvironment. For instance, in a permissive microenvironment, the ECM provides matrix components (proteins and proteoglycans) that support cell adhesion via integrin-mediated adhesions and cell surface receptors.¹⁶ Subsequently, these adhesions trigger downstream intracellular pathways that modulate cancer cell fate including cell cycle progression and migration. Alternatively, when the ECM is not as permissive to integrin ligation or ligates different integrins, cancer cells may enter a state of dormancy. Remodeling of the dormant niche over time may permit engagement of dormant tumor cells with altered ECM and enable their escape from dormancy. However, there is still a poor understanding of the underlying mechanisms that induce and maintain cancer dormancy as imaging resolution limitations that prevent dormant cells from being easily detected in vivo make it difficult to use animal models for development of therapeutics that target dormant micrometastases.

To study dormant micrometastases, engineered systems that induce tumor dormancy in vitro have been developed. Specifically, the role that microenvironmental properties play in regulating dormancy has garnered significant interest.⁷ Studies have demonstrated that the ECM can be tuned to induce cancer cell dormancy in vitro using both natural and synthetic biomaterials.^7,17-20 Amongst several factors, physical confinement, cell–matrix interactions, and matrix strain/deformation have been manipulated to induce dormancy.⁷ Toward this effort, our group recently developed a suite of 16 poly(ethylene glycol) (PEG)-based hydrogels with varying biochemical (ligand density and degradability) and mechanical (elasticity, swelling, mesh size) properties to induce three distinct phenotypes (growth, balanced dormancy, and cellular dormancy) in triple negative breast cancer cells, MDA-MB-231s (231s).^21-23

One major roadblock in treating dormant cancer cells and preventing cancer recurrence is that current chemotherapeutics often fail to effectively treat dormant, quiescent cells. Several patient studies indicate that dormant DTCs resist common forms of chemotherapy.^24-29 For example, patients treated with adjuvant therapy that had DTCs present in bone marrow aspirates displayed a significantly decreased 5 year survival rate.²⁶ In vitro platforms that induce cancer dormancy can be implemented to characterize chemoresistance.^18,20,30 For example, ECM-induced dormant cells displayed a 0.9- to 10.6-fold increase in the half maximal inhibitory concentration (IC₅₀) compared to control conditions.²⁰ Several in vitro studies have confirmed these findings and have even used chemotherapy as a method of inducing dormancy.^{7,18,20,30,31}

One study seeded 231s in an ex vivo system and used standard chemotherapeutics to selectively target and eliminate proliferating 231s while dormant subpopulations remained viable.³² While dormant cancer cells can display increased chemoresistance, the underlying mechanisms dictating this resistance remain elusive and platforms available for such studies are lacking. Current hypotheses state that cytotoxic therapies meant to target proliferative cells fail to target quiescent, dormant DTCs,^18,33 while others suggest dormant cells may have evolved mechanisms (such as increased expression of efflux pumps) to increase resistance to therapy.^32-34 An understanding of how ECM properties induce dormancy-associated chemoresistance could potentially lead to the development of therapeutics to prevent metastasis or relapse.

In this study, we implemented three hydrogel formulations that induce growth, cellular dormancy, or balanced dormancy to characterize the chemoresistant behavior of dormant 231s compared to cells in a growth state. 231s were encapsulated and maintained in three-dimensional (3D) culture in each of the three hydrogel formulations: growth (10 mM PEG-RGDS, 0 mM NVP), balanced dormancy (1 mM PEG-RGDS, 9.4 mM NVP), or cellular dormancy (0 mM PEG-RGDS, 0 mM NVP). Tuning the PEG-RGDS concentration altered adhesive ligand density of the hydrogel matrix required for integrin-mediated cell adhesion and survival and NVP incorporation increased the number on non-degradable cross-links of the hydrogel matrix, inhibiting cell-mediated matrix degradation leading to physical confinement of encapsulated cells in their local niche. Fifteen days post encapsulation, 231s were exposed to doxorubicin (DOX), paclitaxel (PAC), or 5-fluorouracil (5-FU) for 48 h followed by quantification of cell viability to assess the efficacy of each drug and to quantify differences in cellular responses among the three phenotypic states. For all three drugs tested, 231s undergoing either cellular or balanced dormancy demonstrated significantly increased chemoresistance compared to 231s in the growth state. For DOX, dormant cells had a 1.4- to 1.8-fold increase in half maximal effective concentration (EC₅₀) and a 1.3- to 1.8-fold increase in IC₅₀ values compared to 231s in the growth state. Dormant 231s exposed to PAC displayed statistically significant higher cell viability compared to those in the growth state, with >72% of dormant cells viable compared to 39 ± 4% when exposed to 1 mM PAC. Similarly, dormant cells exposed to 5-FU displayed significant increases in cell viability, with >93% cell viability in 231s residing in either dormant state, compared to 75 ± 5% for cells in the growth state at 10 mM 5-FU. To understand how chemoresistance to DOX was exhibited by dormant cells, the intracellular concentration of DOX was quantified via intensity measurements of DOX autofluorescence. No significant differences in the intracellular intensity (concentration) of DOX were observed, but the intracellular distribution, as quantified by the nuclear to cytoplasmic (N/C) drug ratio, indicated that dormant cells had a ~1.5-fold decrease in relative DOX accumulation in the nuclei compared to 231s in the growth state, suggesting that dormancy-associated chemoresistance may be due to decreased drug nuclear localization, at least for DOX. These results demonstrate how tuning the matrix properties alter 231 phenotypes and subsequent chemoresistance, lend additional validation that this hydrogel platform can be used to induce breast cancer dormancy, and support its use for future development and testing of new anti-dormancy therapeutic approaches. While this platform does not include other microenvironmental stimuli that influence dormancy such as immune responses, hypoxia, nutrient deprivation, soluble factors, and secondary cell signaling, the results highlight the contributions of hydrogel properties in inducing and maintaining cancer dormancy and the influence of matrix properties on dormant cancer chemoresistance.

2. EXPERIMENTAL SECTION

2.1. Cell Culture.

The triple negative breast cancer cell line, MDA-MB-231 (American Type Culture Collection (ATCC)), was cultured in fibronectin coated (10 μg/mL) T75 tissue culture flasks in Dulbecco’s modified Eagle’s medium (DMEM; Thermo Fisher) supplemented with 10% (v/v) fetal bovine serum (FBS; Thermo Fisher) and 1% (v/v) penicillin–streptomycin (Lonza). Cells (passages 35–40) were cultured to 80% confluence at 37 °C and 5% CO₂ prior to passaging. Cells used for encapsulation were serum starved in serum-free DMEM for 48 h prior to encapsulation to synchronize the cell cycle.

2.2. PEG Macromer Synthesis and Characterization.

Acryl-PEG-SVA (PEG-SVA, MW: 3400 Da, Laysan Bio) was reacted with a proteolytically degradable peptide sequence GGGPQG↓IWGQGK (PQ, MW: 1141.24 Da, Layson Bio., ↓ denotes the cleavage site by matrix metalloproteinase-2 and matrix metalloproteinase-9) at a 2.1:1 molar (M) ratio (PEG-SVA:peptide) in DMSO with N,N-diisopropylethylamine (DIPEA) at a 4:1 M ratio (DIPEA:PQ) at room temperature for 48 h to form the PEG-diacrylated PEG-PQ-PEG (PEG-PQ) macromer. PEG-SVA was similarly reacted to the integrin-ligating peptide sequence RGDS (MW: 433.42 Da, American Peptide) at a 1.1:1 M ratio (PEG-SVA:peptide) in DMSO with DIPEA at a molar ratio of 2:1 (DIPEA:RGDS) to form the PEG-monoacrylated PEG-RGDS macromer. Reacted products were dialyzed against deionized (DI) water for 24 h with four water changes (MWCO 3500 units, Regenerated Cellulose, Spectrum Laboratories), frozen, lyophilized, and stored at −80 °C under argon. PEG-peptide conjugation was verified via gel permeation chromatography (GPC; Waters, aqueous phase).

2.3. Cell Encapsulation.

Three prepolymer solutions were prepared to induce either (i) growth, (ii) balanced dormancy, or (iii) cellular dormancy in 231s as previously described.^21-23 In all solutions, PEG-PQ (MW: 7900 Da) was held constant and reconstituted in phosphate-buffered saline (PBS) to a final concentration of 5% w/v (6.3 mM). PEG-RGDS (MW: 3800 Da) was reconstituted in PBS to final concentrations of 0 mM (cellular dormancy), 1 mM (balanced dormancy), or 10 mM (growth). For the balanced dormancy hydrogel formulation, NVP was mixed with PEG-PQ and PEG-RGDS precursor solutions to a final concentration of 1.0 μL/mL (9.4 mM) NVP. The UV-sensitive photo-cross-linker, lithium phenyl-2,4,6-trimethylbenzoylphosphinate (LAP), was added to all solutions at a final concentration of 3.0 mg/mL (10 mM). 231s were trypsinized, counted, and resuspended in the desired prepolymer solution at 10 million cells/mL. To control the hydrogel height, 500 μm-thick poly(dimethyl siloxane) (PDMS) spacers were placed on a perfluoroalkoxy alkane (McMaster-Carr) coated Petri dish. Droplets (3 μL) of the cell-laden precursor were pipetted on the Petri dish, and a glass slide was gently placed over the droplets to create 500 μm-thick disks (~2.8 mm in diameter). The droplets were exposed to broad spectrum UV (Blak-Ray flood UV lamp, wavelength: 365 nm, intensity: 10 mW/cm²) and photo-cross-linked for 1 min. The polymerized cell-laden hydrogels were transferred to a well plate and cultured in the medium for 15 days, with medium changes every 4 days.

2.4. Hydrogel Characterization.

To characterize the properties of the three hydrogel formulations the compressive modulus, degradation rate, PEG-RGDS incorporation, mesh size, and swelling ratio were quantified. For mechanical characterization, cell-laden PEG-PQ hydrogels with either 0 or 9.4 mM NVP were polymerized in 3 mm diameter, 1 mm-tall cylindrical PDMS molds using 15 μL of prepolymer solution. Unconfined compression testing was performed immediately after encapsulation and after 15 days in culture. A Universal Testing System 3340 Series (Instron) using platens for unconfined compression testing was applied to measure compressive modulus. Samples (n = 3) were compressed with a 10 N load cell with an initial load of 0.02 N at 2 μm/s for 100 s. Compressive modulus was determined as the slope of the linear regime of the resulting stress–strain curve, within the first 20% of strain compression.

To calculate the degradation rate, PEG-PQ hydrogels containing 1 mM methacryloxyethyl thiocarbamoyl rhodamine B (Polysciences) and 0 or 9.4 mM NVP were prepared and allowed to swell overnight in PBS (n = 3). Hydrogels were incubated in pre-warmed (37 °C) 100 μg/mL collagenase IV (Worthington, 250 U/mg) in PBS. Hydrogels were imaged every 15 min over 3 h with a Zeiss AxioObserver Z1 inverted fluorescence microscope. Using FIJI software (NIH, 1.51r), fluorescence intensity was measured over time. The degradation rate was determined by fitting the intensity versus time data and determining the slope of the linear portion. A control group of hydrogels in PBS without collagenase was included to account for potential photobleaching.

Fluorescence analysis was used to determine the concentration of PEG-RGDS that was photocoupled in PEG-PQ hydrogels containing 0 NVP and 10 mM PEG-RGDS or 9.4 mM NVP and 1 mM PEG-RGDS (n = 4).^21-23 Prepolymer solutions containing 0.5 mM PEG-RGDS-488 (a PEG-RGDS macromer fluorescently labeled with Alexa Fluor 488) were prepared for each formulation. PEG-RGDS concentrations were adjusted to 9.5 and 0.5 mM. To obtain a baseline fluorescence reading, prepolymer solutions were imaged prior to polymerization. After photopolymerization, hydrogels were imaged again using the same settings. FIJI software (NIH) was used to measure fluorescence intensities and calculate how much the fluorescence intensity reduced due to photobleaching. Photobleached hydrogels were allowed to swell in PBS overnight, allowing any unconjugated PEG-RGDS and PEG-RGDS-488 to rinse away. Using the same imaging parameters, rinsed hydrogels were imaged again. Calculations comparing intensities immediately after cross-linking and after rinsing were used to determine PEG-RGDS conjugation efficiency as previously described.^21-23

To determine theoretical mesh size, PEG-PQ hydrogels with either 0 or 9.4 mM NVP were polymerized and allowed to swell in DI water overnight (n = 4). Hydrogels were transferred to solutions of 1 mg/mL 150 kDa fluorescein isothiocyanate (FITC)-labeled dextran (Thermo Fisher) and incubated at 4 °C for 48 h. Hydrogels containing 150 kDa FITC-dextran were blotted clean and placed into well plates with fresh DI water. Dextran that diffused out of the hydrogel and into the water was measured by collecting samples every 15 min over 4 h. Equal volumes of water were added back to each sample during each time point. Diffused dextran intensity was measured with a plate reader (Biotek Synergy, excitation: 490 nm, emission: 525 nm). Intensity values were normalized to total intensity released throughout the entire experiment. Using fluorescence intensities, the cumulative mass of released FITC-dextran was calculated and used to determine the hydrogel diffusion coefficient and theoretical mesh size using the hindered solute diffusion in a solvent-filled-pore model as previously described.^21-23

To calculate swelling ratio, PEG-PQ hydrogels with either 0 or 9.4 mM NVP (n = 4) were polymerized and placed in PBS overnight, allowing gels to swell to equilibrium before being weighed. Weighed gels were dried in ambient air for 6 h and reweighed. The swelling ratio was calculated for each sample as previously described.^21-23

2.5. Characterization of Hydrogel-Induced Phenotypic Changes.

Induced phenotypic changes were determined by quantifying morphology, viability, proliferation, metabolic activity, and early apoptosis as previously described.^21-23 To measure morphological changes, cells were cultured for 15 days before being fixed with 4% paraformaldehyde (30 min). Fixed cells were permeabilized using PBS-T (PBS + 0.2% (w/v) bovine serum albumin with 0.1% (v/v) Triton-X) and blocked using bovine serum albumin blocking buffer for 30 min. The cytoskeletal protein, F-actin, was labeled with Alex Fluor 568 phalloidin (25 μL/mL) (Invitrogen) for 1 h and counterstained with Hoechst 33342 (10 μL/mL) for 30 min. Hydrogels were rinsed and imaged using confocal microscopy (Zeiss LSM 710). Solitary cells and cell clusters were traced using FIJI software (NIH) to quantify the percentage of cells residing as single or clustered cells, rounded or invasive, non-invasive or invasive clusters and to determine the single cell and cluster cell density. Invasive cells/clusters were defined by a roundness value less than 0.80. n = 5 z-stacks from five individual hydrogels for each formulation were analyzed.

Cell viability was measured at days 0 (6 h post encapsulation) and 15. Hydrogels were labeled using a Live/Dead cell viability kit (Invitrogen) following the manufacturer’s instructions. After rinsing, labeled hydrogels were placed on coverslips and imaged via structured illumination using a Zeiss AxioObserver Z1 inverted fluorescence microscope equipped with a Zeiss Apotome imaging system and a Hamamatsu ORCA-Flash 4.0LT camera. Using FIJI software (NIH), the number of live and dead cells were counted to determine the cell viability and viable cell density (n = 6 z-stacks from three hydrogels for each formulation).

Proliferation was measured via 5-ethynyl-2-deoxyuridine (EdU) incorporation using the Click-It EdU Imaging kit (Invitrogen) according to the manufacturer’s instructions at days 0 and 15. Cell-laden hydrogels were incubated in 10 μM EdU in the medium for 24 h. Hydrogels were rinsed, fixed using 4% paraformaldehyde at room temperature for 30 min, permeabilized with PBS-T for 15 min, and blocked using PBS with 3% fetal bovine serum for 30 min. Cells were labeled with Alexa Fluor 647 azide (proliferating nuclei) and Hoechst 33342 (all nuclei), imaged via structured illumination, and quantified (n = 6 z-stacks from three hydrogels for each formulation).

Metabolic activity was measured using an Alamar Blue assay (Thermo Fisher) at days 0 and 15. Encapsulated cells were cultured using a phenol red-free medium. At both time points, cell-laden hydrogels were rinsed and incubated in a working solution of Alamar Blue (10 μL of 10× stock + 100 μL phenol red-free media) for 4 h at 37 °C. Hydrogels were transferred into fresh wells containing a phenol red-free medium. A plate reader (Biotek Synergy, excitation: 550 nm, emission: 600 nm) was used to measure the fluorescence intensity and to quantify the relative metabolic activity. Measured day 15 values were normalized to day 0 (n = 6 z-stacks from three hydrogels for each formulation). Wells with only the medium were used as background controls.

To measure cells undergoing early apoptosis at days 0 and 15, hydrogels were rinsed with binding buffer (HEPES-buffered saline (HBS) with 2.5 mM CaCl₂) for 15 min and incubated at 37 °C with a solution with CF568 Annexin V (Biotium, 1 μg/mL), an early apoptosis marker, for 30 min. Cells were counterstained with Hoechst 33342 (10 μg/mL), rinsed, and imaged. Cells were counted using FIJI software (NIH) to quantify those undergoing early apoptosis (n = 6 z-stacks from three hydrogels for each formulation).

2.6. Measuring Drug Responses: 2D Controls.

231s were trypsinized, counted, and seeded in a fibronectin coated (10 μg/mL), 96-well plate at a density of 5000 cells/well in a serum-containing medium. To measure the drug response of cells independent of encapsulation, we replicated pre-encapsulation conditions by allowing the cells to adhere overnight followed by culture in a serum-free medium for 48 h. DOX, PAC, and 5-FU were purchased (Cayman Chemical) and reconstituted in DMSO to a stock concentration of 172.4, 58.5, and 384.6 mM, respectively. Stock solutions were diluted in the medium to a range of concentrations: DOX (0.001, 0.01, 0.1, 1, 10, 100, or 1000 μM), PAC (0.001, 0.01, 0.1, 1, 10, or 100 μM), and 5-FU (0.001, 0.01, 0.1, 1, 10, 100, or 1000 μM) and added to the 231s for 48 h (n = 6 wells per concentration).³⁵ At the highest drug concentration tested, the DMSO concentration remained at ≤0.1% (v/v); thus, 0.1% (v/v) DMSO was used as a vehicle control in 2D studies. Cells were rinsed in PBS twice to remove the excess medium followed by labeling with a Live/Dead cell viability kit (Invitrogen). Only live cells were labeled as dead cells detached from the wells and were washed away. Labeled cells were imaged using the proper filter cubes and analyzed using FIJI software (NIH).

2.7. Measuring Drug Responses: Cells Cultured in Hydrogels.

DOX, PAC, and 5-FU stocks were diluted in the medium to desired concentrations: DOX (0.001, 0.01, 0.05, 0.1, 0.55, 1, or 2 mM), PAC (0.05, 0.1, 0.55, or 1 mM), and 5-FU (1.0, 4.0, 7.0, or 10 mM). At the highest concentrations tested, the DMSO concentration ranged between 1 and 2% (v/v) for all drugs used. A vehicle study was conducted and showed no significant differences in cell viability between 1 or 2% (v/v) DMSO in serum-containing medium (Figure S1). Thus, all vehicle controls used in this study contained the medium with 1% (v/v) DMSO. We also tested higher concentrations (18 and 26% (v/v)) of DMSO required to use higher doses of PAC and 5-FU; however, these concentrations were cytotoxic to 231s (Figure S1) and therefore not used. Fifteen days post encapsulation, cell-laden hydrogels were rinsed with PBS and transferred to a new well plate for drug exposure for 48 h.³⁵ The hydrogels were rinsed twice with PBS to remove the excess medium followed by labeling with a Live/Dead cell viability kit (Invitrogen) in PBS for 30 min, according to the manufacturer’s instructions. The hydrogels were rinsed with PBS for 10 min to remove the excess stain. Labeled hydrogels were placed on coverslips and imaged using structured illumination on a Zeiss AxioObserver Z1 inverted fluorescence microscope as previously described.^23-25 Fluorescence z-stacks (z-height: 150 μm) were acquired approximately at the center of the hydrogels, between the bottom and top surfaces of each sample, and analyzed using FIJI software (NIH). Due to the red autofluorescence of DOX, only the live cells were imaged for viability studies. To quantify the cell viability, the number of live cells was normalized to the number of live cells measured in the vehicle control. A minimum of six hydrogels were analyzed for each condition.

2.8. Doxorubicin Accumulation Studies.

Cells were encapsulated in hydrogel formulations that induced growth, balanced dormancy, or cellular dormancy for 15 days as described above. The medium was replaced with the DOX-containing medium at a concentration of 0.05 mM. Four hydrogels for each formulation were imaged at varying time points (1, 2, 3, 5, 12, 20, 24, and 48 h).^36-38 The mean cellular fluorescence intensity was measured and plotted against time to quantify intracellular DOX accumulation.^39-41 To calculate the nuclear to cytoplasmic ratio, the fluorescence intensity measurements of the cytoplasmic and nuclear regions were acquired for each cell, with at least 70 cells measured per image.^42-44 Cells were co-stained with Hoechst 33342 to locate cell nuclei.

2.9. Statistical Analysis.

For statistical comparison between viability values, distribution normality was assessed via quantification of skewness and kurtosis, where values within ±2 indicated a normal distribution. Equal variance among groups was additionally evaluated. A one-way analysis of variance (ANOVA) with a post-hoc Tukey–Kramer test was implemented to determine the statistical significance between multiple groups. Unless otherwise indicated, p < 0.05 was considered statistically significant. OriginLab software was used to fit DOX viability data. A nonlinear, growth/sigmoidal curve fit using a dose–response function with a Levenberg–Marquardt iteration algorithm was performed and allowed to converge for all conditions with a chi-square tolerance of 1 × 10⁻⁹.

3. RESULTS

3.1. Response of MDA-MB-231s to Doxorubicin, Paclitaxel, and 5-Fluorouracil in 2D Culture.

The response of 231s cultured on standard tissue culture plastic (2D) to DOX, PAC, and 5-FU was quantified as a control. 231s were cultured on 96-well plates and serum starved for 48 h, following the same protocol used for encapsulation for ease of comparison. The 231s were exposed to a drug-containing medium at the following concentrations (DOX: 0.001, 0.01, 0.1, 1, 10, 100, or 1000 μM; PAC: 0.001, 0.01, 0.1, 1, 10, or 100 μM; and 5-FU: 0.001, 0.01, 0.1, 1, 10, 100, or 1000 μM) for 48 h. The live 231s were fluorescently labeled, imaged, and quantified (Figure 1).

Figure 1.

Response of MDA-MB-231s cultured on tissue culture plastic (2D) to 5-fluoruracil, paclitaxel, and doxorubicin. (A) Quantification of MDA-MB-231 cell viability cultured on tissue culture plastic (2D) after 48 h drug exposure. n = 6 images from six individual wells. Values represent mean ± standard deviation. (B) Dose–response curves for doxorubicin, paclitaxel, and 5-fluorouracil with EC₅₀ and IC₅₀ values. Data points represent mean ± standard deviation.

As expected, 231s demonstrated significant decreases in cell viability with increased drug concentration for all three drugs tested. DOX was the most effective followed by PAC and 5-FU (Figure 1). The results demonstrate that viability decreased to 37 ± 9% at 0.01 μM before gradually dropping to 1 ± 0% at 100 μM for DOX (Figure 1A). Fitting the dose–response curve for DOX resulted in an EC₅₀ of 0.006 μM and IC₅₀ of 0.007 μM (Figure 1B). PAC was less effective than DOX. At 0.01 μM, viability remained high at 94 ± 13% but decreased to 6 ± 2% at the highest concentration tested (100 μM) (Figure 1A). Calculated EC₅₀ (0.716 μM) and IC₅₀ (0.651 μM) values were higher for PAC compared to DOX (Figure 1B). 5-FU was the least effective with an EC₅₀ of 1.81 μM and with 65 ± 3% of the cell population viable at the highest dosage of 1 mM (Figure 1). An IC₅₀ could not be calculated for 5-FU as viability did not drop below 65% (Figure 1B).

3.2. Hydrogel Characterization and 231 Phenotypic Definitions.

We previously developed and implemented 16 poly(ethylene glycol) (PEG)-based hydrogel formulations constructed from three simple components: (i) proteolytically degradable PEG-PQ, (ii) integrin ligating PEG-RGDS, and (iii) the non-degradable cross-linker NVP to induce phenotypic changes in 231s.^30-32 PEG-PQ was held constant at 5% (w/v) for all formulations, and the PEG-RGDS and NVP concentrations were varied (Figure 2A). In this study, we implemented three hydrogel formulations that allowed for induction of three different 231 phenotypes: (i) a growth formulation containing high ligand density (10 mM PEG-RGDS) and high degradability (0.01 min⁻¹), (ii) a balanced dormancy formulation containing lower adhesivity (1 mM PEG-RGDS) and lower degradability (0.005 min⁻¹) achieved by the addition of the non-degradable NVP (9.4 mM), and (iii) a cellular dormancy formulation with no adhesivity (0 mM PEG-RGDS) and high degradability (0.01 min⁻¹) (Figure 2A,C). NVP incorporation increases the cross-linking density through addition of non-degradable monomers resulting in an increased compressive modulus and lower degradability (Figure 2). Limiting PEG-RGDS concentration inhibits integrin-mediated adhesion, while NVP incorporation decreases matrix degradability leading to increased confinement. Tuning these properties mediates cell–matrix interactions necessary for controlling cellular phenotype.

Figure 2.

Hydrogel characterization and MDA-MB-231 phenotypic metrics. (A) Schematics of three hydrogel formulations used to induce growth, balanced dormancy, or cellular dormancy in MDA-MB-231s. (B) Representative maximum intensity z-projections from 3D image stacks of MDA-MB-231s fluorescently labeled with phalloidin (red: F-actin) and Hoechst (cyan: nuclei) after 15 days in culture. SB = 100 μm. (C) Table displaying hydrogel properties (compressive moduli at days 0 and 15, degradation rate, PEG-RGDS concentration, pore size, and swelling ratio) for the three formulations. (D) Table displaying the influence of hydrogel composition on MDA-MD-231 proliferation (EdU⁺), metabolic activity, live to dead cell density ratio post day 0, and morphology. Asterisk indicates statistical significance (p < 0.05). Hydrogel schematics in (A) reproduced with permission from ref 21. Copyright 2019 Biomaterials.

The response of 231s to the three hydrogel formulations was measured to verify hydrogel-induced changes in phenotype. The growth state was defined when statistically significant increases in the ratio of actively proliferating cells, metabolic activity, and density of new live cells compared to new dead cells at day 15, relative to day 0, were observed (Figure 2D). The majority of cells residing in this state existed as invasive cell clusters (Figure 2B). Balanced dormancy was defined when a balance in the density of new live and dead cells post day 0 was achieved (indicated by a ~1:1 live:dead cell ratio), indicating that the increase in density of live cells was nearly perfectly balanced by the increase in density of dead cells over 15 days preventing growth of larger micrometastases as observed in the growth formulation.^21-23 Cells in this state presented a rounded morphology (Figure 2B). Cellular dormancy was defined by low proliferation, no statistically significant increase in metabolic activity at day 15, relative to day 0, and no new live cells at day 15. Cells in this state existed as solitary, rounded cells (Figure 2B). Hydrogel formulations that induced each of these phenotypic states were used for all drug response studies presented (Figure 2A).

3.3. Response of MDA-MB-231s to Doxorubicin.

Measured cellular metrics used to define phenotypic states demonstrate the ability to induce breast cancer dormancy using engineered hydrogels. It is well established that dormant cells often display increased chemoresistance.^27-29 To quantify the degree of chemoresistance imparted on 231s in dormant states compared to the growth state, the response of 231s cultured in the three hydrogel formulations to DOX was quantified. Post encapsulation, the cells were cultured for 15 days in the desired hydrogel formulation to provide enough time for the cells to respond to the hydrogel properties; culturing for longer time periods did not significantly alter phenotype.^21-23 Cells were exposed to DOX (0.001–2 mM) for 48 h, and cell viability was quantified (Figure 3A,B).³⁵ To assess the drug response of 231s in the three phenotypic states, the cells were labeled with Calcein AM to fluorescently label live cells, imaged (Figure 3A), and the number of live cells normalized to the vehicle control (media containing 1% DMSO) were measured to quantify cell viability (Figure 3B).

Figure 3.

Response of MDA-MB-231s to doxorubicin. (A) Representative maximum intensity z-projections of 3D image stacks of MDA-MB-231s fluorescently labeled with Calcein AM (live cells: green) after 48 h exposure to a media + 1% v/v DMSO vehicle control (DMSO) or varying concentrations of doxorubicin. Top row: growth; middle row: balanced dormancy; bottom row: cellular dormancy. SB = 100 μm. (B) Quantification of cell viability. Asterisk indicates p < 0.05. n = 5 z-stacks from five individual hydrogels for each formulation. Values represent mean ± standard deviation. (C) Dose–response curves with EC₅₀ and IC₅₀ values. Data points represent mean ± standard deviation.

231s cultured in the hydrogel formulation that induced cellular dormancy displayed statistically significant higher viability than 231s cultured in the growth formulation for DOX concentrations ranging from 0.001 to 1 mM (Figure 3B). For instance, at 0.01 and 0.05 mM, viability remained at 91 ± 2% and 62 ± 5% for cells undergoing cellular dormancy and 82 ± 1% and 44 ± 3% for cells in the growth state, respectively (Figure 3B). Increased differences in viability were observed at higher concentrations (0.55 and 1.00 mM) where viability was four to six times higher for cells undergoing cellular dormancy compared to those in the growth state (Figure 3B). For example, at 0.55 mM, dormant 231s remained 24 ± 3% viable compared to 5 ± 1% in the growth state (Figure 3B). 231s undergoing balanced dormancy also displayed higher viability compared to 231s in the growth state over a range of concentrations, with significant differences observed at 0.05 mM (balanced dormancy:growth, 57 ± 2%:44 ± 3%) and 0.55 mM (balanced dormancy:growth, 16 ± 1%:5 ± 1%) (Figure 3B).

Dose–response curves were fit to the viability data for the growth and two dormant states to quantify EC₅₀ and IC₅₀ values (Figure 3C). EC₅₀ and IC₅₀ values from 231s in 2D culture were much lower compared to cells cultured in hydrogels as expected (Figures 1B and 3C). In 2D culture, DOX had EC₅₀ and IC₅₀ values of 0.006 and 0.007 μM, respectively, while the EC₅₀ values for cells in the three hydrogel formulations ranged between 47.400 and 83.300 μM and IC₅₀ values ranged from 43.600 to 79.900 μM (Figures 1B and 3C). Specifically, 231s in the growth state had an EC₅₀ of 43.6 μM, while those undergoing balanced dormancy or cellular dormancy had higher EC₅₀ values of 67.6 and 83.3 μM, respectively (Figure 3C). Similarly, IC₅₀ values followed the same trend (growth: 43.6, balanced dormancy: 58.3, and cellular dormancy: 79.9 μM) (Figure 3C). Together, this data indicates that 231s in either dormant state exhibited significantly higher chemoresistance to DOX compared to cells in the growth state. In particular, 231s undergoing cellular dormancy maintained the overall higher viability than those undergoing balanced dormancy, as indicated by the higher EC₅₀ and IC₅₀ values (Figure 3C).

3.4. Response of MDA-MB-231s to Paclitaxel.

To assess the chemoresistance of dormant 231s to other common breast cancer drugs, encapsulated cells were cultured for 15 days and exposed to varying concentrations of PAC for 48 h. As described above, the live cells were fluorescently labeled with Calcein AM, imaged, counted, and normalized to the vehicle control for each hydrogel formulation (Figure 4A,B). 2D studies demonstrated that compared to DOX, PAC was less effective (Figure 1A,B) and therefore required a higher concentration to achieve a comparable cellular response as DOX. 231s undergoing cellular dormancy had statistically significant higher viability for most PAC concentrations compared to 231s in the growth state. For instance, at 0.10 and 0.55 mM, 231s undergoing cellular dormancy remained 91 ± 6% and 84 ± 5% viable compared to 55 ± 4% and 51 ± 4% for cells in the growth state, respectively (Figure 4B). Furthermore, at the highest PAC concentration tested (1 mM), 75 ± 3% of cells undergoing cellular dormancy remained viable while only 39 ± 4% of cells in the growth state were viable (Figure 4B).

Figure 4.

Response of MDA-MB-231s to paclitaxel. (A) Representative maximum intensity z-projections of 3D image stacks of MDA-MB-231s fluorescently labeled with Calcein AM (live cells: green) after 48 h exposure to a media + 1% v/v DMSO vehicle control (DMSO) or varying concentrations of paclitaxel. Top row: growth; middle row: balanced dormancy; bottom row: cellular dormancy. SB = 100 μm. (B) Quantification of cell viability. Asterisk indicates p < 0.05. n = 5 z-stacks from five individual hydrogels for each formulation. Values represent mean ± standard deviation.

Similar to the trend observed with DOX, 231s undergoing balanced dormancy exhibited significant chemoresistance to PAC. For example, at 0.55 and 1.00 mM, viability was 86 ± 5% and 72 ± 3% for cells undergoing balanced dormancy and 51 ± 4% and 39 ± 4% for cells in the growth state, respectively. There were no statistically significant differences in viability between the two dormant states for any concentration of PAC tested (Figure 4B).

PAC was dissolved in minimal DMSO according to its solubility and manufacturer’s instructions. Due to the upper limits of PAC efficacy, we were unable to test higher drug doses as higher doses would exceed the concentration of DMSO that 231s could tolerate and thus would no longer represent cells targeted by the mechanism of the drug but instead by DMSO itself (Figure S1). For this reason, we were unable to fit a dose–response curve to this data. However, viability data demonstrates that cells in both dormant states showed significant chemoresistance to PAC similar to DOX.

3.5. Response of MDA-MB-231s to 5-Fluorouracil.

We investigated the effect of a third, commonly used breast cancer drug, 5-FU, on 231s in the three phenotypic states. The same protocol was followed to encapsulate, label, image, and quantify live cells in the growth, cellular dormancy, and balanced dormancy states. Compared to both DOX and PAC, the 2D studies indicated that 5-FU was the least effective and therefore required a higher concentration (Figure 1A,B), which was limited by DMSO toxicity as discussed. Nonetheless, a significant increase in chemoresistance was observed for 231s undergoing cellular or balanced dormancy compared to those in the growth state (Figure 5A,B).

Figure 5.

Response of MDA-MB-231s to 5-fluorouracil. (A) Representative maximum intensity z-projections of 3D image stacks of MDA-MB-231s fluorescently labeled with Calcein AM (live cells: green) after 48 h exposure to a media + 1% v/v DMSO vehicle control (DMSO) or varying concentrations of 5-fluorouracil. Top row: growth; middle row: balanced dormancy; bottom row: cellular dormancy. SB = 100 μm. (B) Quantification of cell viability. Asterisk indicates p < 0.05. n = 5 z-stacks from five individual hydrogels for each formulation. Values represent mean ± standard deviation.

At 1 mM, 231s remained highly viable in all three hydrogel formulations. However, at 4 mM, 231s in the growth state decreased to 86 ± 7% viability while those undergoing cellular dormancy remained 100 ± 6% viable and 231s undergoing balanced dormancy displayed 92 ± 4% viability (Figure 5B). A similar trend was observed at 7 mM, with 231s in the growth state further decreasing to 80 ± 2% viability while cells in both dormant states remained >90% viable (Figure 5B). At the highest concentration tested (10 mM), 231s undergoing cellular dormancy were 99 ± 4% viable compared to only 75 ± 5% for those in the growth state (Figure 5B). No statistically significant differences were measured between 231s in either dormant state (Figure 5B). These results indicate that dormant 231s remain >90% viable at all concentrations of 5-FU tested and display greater chemoresistance to 5-FU than 231s in the growth state (Figure 5B).

3.6. Doxorubicin Accumulation and Localization.

To determine the cause of increased chemoresistance to DOX displayed by dormant 231s compared to those in the growth state, we took advantage of DOX autofluoresence^36,37 to quantify intracellular drug accumulation^39-41 (Figure 6A,B). Cell-laden hydrogels were exposed to DOX (0.05 mM) and imaged at 1, 2, 3, 5, 12, 20, 24, and 48 h post exposure (Figure 6A). Image analysis was used to measure the mean intracellular fluorescence intensity of DOX in 231s cultured in the three hydrogel formulations over time (Figure 6B). Note that, while cells in the growth formulation appear invasive after 15 days in culture, they revert to a rounded morphology after 48 h DOX exposure. The results demonstrate that DOX steadily increased in 231s with no statistically significant differences between 231s in the three phenotypic states (Figure 6B). This data suggests that drug uptake and efflux was similar in 231s in all three phenotypic states and therefore did not play a role in the observed chemoresistance in cells undergoing dormancy.

Figure 6.

Doxorubicin accumulation and localization in MDA-MB-231s cultured in three hydrogel formulations. (A) Representative maximum intensity z-projections of 3D image stacks of doxorubicin (red) in MDA-MB-231s cultured in growth, balanced dormancy, or cellular dormancy states after 48 h exposure to 0.05 mM doxorubicin. Blue indicates Hoechst counterstain (nuclei). Green boxes in the top row indicate zoomed-in regions shown in the two bottom rows. SB = 100 μm (top row). SB = 50 μm (middle and bottom rows). (B) Quantification of doxorubicin accumulation in MDA-MB-231s over 48 h measured by mean fluoresce intensity. (C) Ratio of nuclear to cytoplasmic localization of doxorubicin at 48 h from fluorescence intensity measurements of the cell cytoplasm and nucleus. Asterisk indicates p < 0.05. n = 4 z-stacks from four individual hydrogels for each formulation. Values represent mean ± standard deviation.

To quantify potential differences in the intracellular distribution of DOX, the nuclear to cytoplasmic (N/C) ratio was calculated by measuring the mean fluorescence intensity of DOX in the cytoplasm and nucleus of each cell (Figure 6C).^42-44 A nuclear co-stain, Hoechst 33342, was used to distinguish the cell nucleus from the cytoplasm for these measurements. The N/C ratio was 0.43 ± 0.3 for the growth state, 0.29 ± 0.02 for cells undergoing balanced dormancy, and 0.29 ± 0.04 for cells undergoing cellular dormancy (Figure 6C). The data demonstrates that dormant 231s had a ~1.5-fold decrease in DOX nuclear localization compared to those in the growth state (Figure 6C). Since DOX induces DNA damage and cell death via methods including intercalation into DNA,⁴⁵ this data suggests there is more DNA-bound DOX in 231s in the growth state compared to those undergoing dormancy. Therefore, while DOX accumulates equally in dormant cells, a less amount of drug localizes to the nucleus, leading to an increased chemoresistance in dormant 231s.

4. DISCUSSION

An important step toward preventing and treating cancer metastasis and recurrence includes forming a better understanding of the various factors that induce and mediate cancer latency and dormancy and mechanisms of chemoresistance imparted on dormant cells. A significant roadblock in preventing cancer metastasis is that many current chemotherapeutics fail to effectively treat dormant cells. In previous studies, we developed and implemented a suite of hydrogel formulations that induced distinct phenotypic states (growth, balanced dormancy, and cellular dormancy) in triple negative breast cancer cells, 231s, by altering cell–matrix interactions.^21-23

Metastatic cancer cells require a permissive premetastatic niche, which sustains the metabolic needs of migrating and actively proliferative cells; ECM components that permit invasion, migration, and proteolytic degradation; and protumorigenic immune signaling that suppress anti-tumor responses. Subsequent dynamic interactions between cancer cells and their microenvironment result in the metastatic niche. In the hydrogel platform utilized here, we focused on using ECM biochemical and mechanical properties to dictate cancer cell phenotype in the metastatic niche as it is well established that integrin-engagement between tumor cells and the altered ECM regulate downstream signaling pathways that control cell cycle progression and migration. The hydrogel containing 5% w/v PEG-PQ, 10 mM PEG-RGDS, and 0 mM NVP provided 231s with adhesive ligands and an enzymatically degradable matrix, which promoted cell–matrix interactions and growth in 231s. In contrast to this permissive niche, inhibiting cell–matrix interactions led to dormant phenotypes in 231s. The matrix containing 5% w/v PEG-PQ, 0 mM PEG-RGDS, and 0 mM NVP mimicked a niche in which DTCs fail to form integrin-mediated adhesions, leading to poor matrix engagement and survival. The hydrogel containing 5% w/v PEG-PQ, 1 mM PEG-RGDS, and 9.4 mM NVP mimicked a niche in which DTCs failed to locally degrade their matrix. By altering the hydrogel properties of the premetastatic niche, we were able to induce growth and dormant states in 231s. Here, we implemented three of these engineered hydrogels (one for each phenotypic state) to quantify the drug response of dormant cells to commonly used chemotherapeutics. While other methods can be used to induce dormancy in cancer cells, this study particularly focuses on hydrogel-induced dormancy.

Control studies were performed in 2D to assess the 231 drug response to DOX, PAC, and 5-FU (Figure 1). Following the same protocol used for encapsulation, 231s were cultured, serum starved for 48 h, and exposed to common breast cancer drugs, DOX, PAC, or 5-FU, for 48 h. Viability data demonstrates IC₅₀ values of 0.007 and 0.651 μM for DOX and PAC, respectively (Figure 1 and Table S1), and EC₅₀ values of 0.006, 0.761, and 1.810 μM for DOX, PAC, and 5-FU, respectively (Figure 1 and Table S1), which are comparable to previously reported values from 2D studies.^46-48 For 3D studies, 231s were cultured for 15 days in the three hydrogel formulations and cells undergoing growth, balanced dormancy, or cellular dormancy^21-23 were exposed to DOX, PAC, or 5-FU for 48 h.³⁵ Viability assays were conducted to quantify the response to each drug (Figures 3A, 4A, and 5A). As expected, 231s in 3D were less sensitive to drugs compared to 2D culture. Differences between 2D and 3D systems in the context of drug screening have previously been reviewed,⁴⁹ and it is accepted that 3D drug platforms better represent drug responses observed in vivo.^49,50 Cells may respond differently to drugs in 3D due to oxygen gradients similar to those in in vivo tumors, spatial organization of cell surface receptors, matrix diffusion, and physical constraints that influence gene expression.⁵⁰ Differences in our system may be do physical constraints imparted by the matrix and spatial organization of encapsulated cells. However, we do not expect there to be oxygen gradients or transport limitations since hydrogels are only 500 μm in thickness, which is below the distance range necessary to form these gradients. Additionally, we expect no limitations in drug transport as drugs used in these studies are small molecules with molecular weights ranging from 130 to 854 g/mol. For instance, doxorubicin was estimated to be <2 nm in diameter using a molecular model.⁵¹ Additionally, molecules of similar molecular weights do not exceed 2 nm in size.⁵² The average mesh size of the hydrogels used to induce growth and cellular dormancy states in these studies was 69 ± 5 nm and 55 ± 4 nm for the balanced dormancy hydrogel formulation. Since the drugs used here were an order of magnitude smaller than the hydrogel mesh size, we anticipate high levels of diffusion throughout the hydrogel.

Changes in drug response were also observed between 231s undergoing cellular dormancy, balanced dormancy, or growth. For exposure to DOX, cells in either dormant state displayed significantly higher viability than cells in the active growth state for a majority of the concentrations tested (Figure 3B). Fits to the dose–response curves demonstrated that cells undergoing balanced or cellular dormancy had a 1.3- and 1.8-fold increase, respectively, in IC₅₀ values compared to actively proliferating cells in the growth state (Figure 3C). This indicates that a higher dosage of DOX is required to induce 50% death in dormant cells compared to actively proliferating cells. While our IC₅₀ and EC₅₀ values fall within the range of other 3D DOX studies, it is worth noting that current IC₅₀ and EC₅₀ values reported in 3D drug studies vary significantly in magnitude due to differences in key factors including matrix properties, experiment duration, and cell growth rates.³⁵ Similar results were seen in PAC and 5-FU (Figures 4 and 5). In both cases, dormant cells in either state had significantly higher viability than those in the growth state. PAC and 5-FU were less effective in reducing cell viability and therefore required higher concentrations (Figures 1, 4, and 5). Even though all drugs tested were dissolved in minimal DMSO according to solubility, we were unable to test higher drug doses as higher doses would lead to cytotoxic effects from DMSO alone and would confound interpretation of the results. Therefore, we were not able to generate a dose–response curve for these drugs; however, the viability data demonstrates that cells in either dormant state showed significant chemoresistance to PAC and 5-FU, similar to DOX (Figures 4 and 5). These results agree with current data that indicates dormant cells show increased chemoresistance in vivo and in vitro.^{18,24,25,27-29} This data also serves to further validate the ability to induce dormant states using engineered hydrogels composed of simple components as demonstrated here.

All drugs used in these studies target proliferating cells through various mechanisms. DOX interferes with cell cycle progression by preventing DNA replication via intercalation and inhibition of topoisomerase II. Additionally, DOX can be oxidized to a semiquinone radical, which can generate reactive oxygen species causing DNA damage.⁴⁵ PAC disrupts proliferating cells by preventing microtubule disassembly during mitosis. 5-FU inhibits RNA transcription and interferes with DNA synthesis.^35,53 Cells unable to proliferate due to DNA damage will activate apoptotic mechanisms during cell cycle check points.⁵⁴ Furthermore, cancer cells often have inhibited or deficient repair mechanisms that prevent them from repairing damaged DNA.⁵⁵ According to the fractional kill theory, a drug will only eliminate those cells that pass through the relevant cell cycle phase during drug exposure.^54,56 Positive correlations between clinical response of DNA-targeted chemotherapy and rate of proliferation support this theory.^54,57 Thus, DOX, and other cancer drugs, preferentially target proliferating cells and may even kill rapidly dividing cells that are not cancerous such as blood cells in the bone marrow. Therefore, dormant cells in a quiescent state (G₀-G₁ arrest) that are slow cycling, such as those in this study, would not be susceptible to drug-induced death for drugs that target cell division.^27,54 Dormancy-associated resistance is supported by multiple patient studies that indicate dormant DTCs are resistant to common forms of chemotherapy.^24-29

Studies that quantify DOX accumulation and localization have been used to understand chemoresistance mechanisms. To test if drugs accumulated differently in dormant 231s compared to those in the growth state, DOX autofluoresence^36-38 was implemented to quantify drug accumulation^39-41 by temporally measuring the intracellular fluorescence intensity (concentration) of DOX over 48 h. The results indicated no significant differences in intracellular concentration and, therefore, DOX accumulation between actively proliferating and dormant cells (Figure 6A,B). These results indicate that drug accumulation (difference in uptake and efflux)⁴³ was similar for 231s in all three phenotypic states. If dormant 231s used efflux pumps that actively expel DOX out of the cell, similar to known chemoresistant cells,^43,44 we would expect accumulation to be less in 231s undergoing dormancy. However, this data indicates drug accumulation does not play a role in dormant 231 chemoresistance, at least for DOX.

While drug accumulation can indicate if DOX made it into a cell, translocation into the nucleus better indicates its chemotherapeutic effect since DOX works by intercalating into DNA.⁴⁴ To measure this, we calculated the N/C ratio by measuring the DOX fluorescence intensity in the nucleus and cytoplasm of each cell using a nuclear co-stain (Figure 6C).^42-44 Data showed that dormant 231s had a ~1.5-fold decrease in DOX nuclear localization compared to those in the growth state (N/C ratio: growth: 0.43 ± 0.03; balanced dormancy: 0.29 ± 0.02; cellular dormancy: 0.29 ± 0.04) (Figure 6C). Since DOX can passively diffuse across the cell,⁵⁸ this data suggests that DOX equally accumulates in 231s undergoing dormancy or growth; however, drug distribution within the cell leads to a difference in chemotherapeutic response.

Since proliferation is low in dormant cells, likely leading to a lower DNA content,⁵⁹ it is possible that dormant cells are simply less susceptible to DOX cytotoxicity since there are fewer binding sites available. Furthermore, since dormant cells are not progressing through the cell cycle, DOX that does accumulate in dormant cells is likely to have lesser cytotoxic affects.⁵⁴ This may also contribute to why dormant cells in our platform show increased chemoresistance to DOX. Further experiments are required to confirm this theory. Some studies have suggested that dormant cells may additionally employ other survival pathways to resist treatment. For instance, endoplasmic reticulum (ER) stress can activate pERK signaling and downstream activation of transcription factors that relieve drug-induced stress.^60-62 Additionally, the unfolded protein response (UPR) can downregulate topoisomerase II expression, leading to DOX resistance.^60,63 Future experiments need to be conducted to determine if these responses are playing a role in maintaining dormant cell survival during treatment.

These results indirectly show a potential relationship between hydrogel properties and dormancy signaling in 231s. By restricting adhesion-mediated interactions and matrix degradability and thus reducing proliferation, 231s entered a dormant state, which led to increased chemoresistance and decreased DOX nuclear localization. However, the mechanistic relationship underlying hydrogel-induced dormancy and drug resistance has not been investigated in this study but is of interest in future studies. Additionally, future experiments that quantify changes in phenotype (morphology, cluster size, etc.), and those that implement secondary drug exposure or combined drug dosage could provide detailed mechanistic insight into the observed chemotherapeutic response.

It is well established that metastases originating from many types of primary cancer (breast, lung, prostate, etc.) undergo periods or latency or dormancy and whose length depends on a multitude of factors including origin, receptor status, and microenvironmental properties of the secondary organ infiltrated.^4,7 It is also well established that dormant DTCs in secondary organs often display enhanced chemoresistance.^26-31 Since metastasis is one of the leading causes of cancer-associated morbidity and mortality, there is an urgent need for simple platforms that allow for thorough investigations of how dormancy is induced and maintained. Such platforms could be beneficial in developing and testing new therapeutic strategies targeted at eliminating dormant cancer cells or preventing dormant cells from escaping dormancy.^3,4,35 Although cancer dormancy can be induced in animal models and dormant cells isolated,^64,65 implementation of these models for drug discovery can be cumbersome due to the difficulty in monitoring their response over time due to resolution limitations of current in vivo imaging modalities. Accordingly, in vitro platforms could potentially accelerate therapeutic development assuming that the cellular responses induced by these platforms are biomimetic enough to recapitulate important aspects of what occurs in vivo. Toward this goal, we previously demonstrated the ability to induce growth, cellular dormancy, or balanced dormancy states in the triple negative MDA-MB-231 cell line via measurement of gross cell behavior and phenotypic classification.^21-23 Since it is well established that dormant cancer cells often display chemoresistance, the work presented here further validates that our simple hydrogel platform can be used to induce dormant states in metastatic breast cancer with increased chemoresistance. Such a platform potentially provides the ability to better understand how dormancy is initiated, the mechanisms of chemoresistance used by dormant cells, and the ability to screen new therapeutics. While a multitude of factors can induce dormancy, this platform focuses solely on the influence of matrix properties with respect to ligand density and degradability. While the influence of other dormancy inducing factors including immune responses, hypoxia, nutrient deprivation, soluble factors, and secondary cell signaling were excluded in this study they could potentially be included in future studies although additional complexity needs to be carried out in a well-controlled, well-characterized, and thorough manner to understand their synergistic roles in influencing the observed cancer cell response. Similarly, although we focused exclusively on triple negative breast cancer, which has a relatively short latency period, we speculate that other breast cancers with different receptor statuses (e.g., ER⁺/PR⁺, HER2⁺) or cancers originating from different primary tumor sites (lung, prostate, etc.) could be induced to undergo dormancy, which could be tested in future studies.

5. CONCLUSIONS

A significant roadblock in preventing metastatic relapse is the inability to target and effectively treat dormant cells. In vitro models that can provide a platform to screen new therapeutic strategies that target dormant cells and help prevent cancer recurrence. In this study, we implemented a hydrogel-based breast cancer dormancy model to determine dormant cell chemotherapeutic responses. We demonstrated that dormant cells, either in balanced or cellular dormancy, showed increased chemoresistance to DOX, PAC, and 5-FU. Specifically, when exposed to DOX, cells in either dormant state had increased EC₅₀ and IC₅₀ values. We further demonstrated that DOX accumulation is consistent in all phenotypic states and therefore did not play a role in observed chemoresistance. Measurements of drug distribution showed decreased DOX nuclear colocalization in dormant populations, potentially accounting for increased chemoresistance in dormant cells. Overall, this study serves to further validate the utility of this hydrogel platform for cancer dormancy studies and provides insight into dormancy-associated chemoresistance.

ABBREVIATIONS

DTC

disseminated tumor cell

ECM

extracellular matrix

PEG

poly(ethylene glycol)

231

MDA-MB-231

IC₅₀

half maximal inhibitory concentration

EC₅₀

half maximal effective concentration

DOX

doxorubicin

PAC

paclitaxel

5-FU

5-fluorouracil

N/C

nuclear:cytoplasmic

DMEM

Dulbecco’s modified Eagle’s medium

FBS

fetal bovine serum

PQ

GGGPQG↓IWGQGIK

M

molar

DMSO

dimethyl sulfoxide

DIPEA

N,N-diisopropylethylamine

PEG-PQ

PEG-PQ-PEG

DI

deionized

GPC

gel permeation chromatography

PBS

phosphate-buffered saline

NVP

N-vinylpyrrolidone

PDMS

poly(dimethyl siloxane)

ANOVA

analysis on variance

ER

endoplasmic reticulum

UPR

unfolded protein response