Repeated exposure of anticancer agents to tumorspheres in open-surface microwell arrays for modeling chemotherapy-induced dormancy in colorectal cancer

Abstract

Dormant cancer cells (DCCs) serve as crucial contributors to tumor drug resistance and recurrence; however, the mechanisms underlying their formation and biological characteristics remain inadequately understood. The establishment of reliable DCC models is essential for elucidating resistance mechanisms and formulating intervention strategies. Microfluidic chips represent valuable tools for conducting efficient cell-based drug testing assays; however, they encounter challenges related to cell recovery, which restricts their applicability in offline analytical contexts. This study reports the development of tumorspheres within open-surface microwell arrays, designed to establish a chemotherapy-induced colorectal cancer dormancy model. Uniform tumorspheres were generated through suspension culture in the microwells, resulting in a sphere diameter of 98.2 ± 9.8 μm and cell viability of 94.6 ± 3.0%. Following repeated exposure of an anticancer agent combination (5-fluorouracil/oxaliplatin/SN-38, FOLFIRINOX), tumorspheres were retrieved and subjected to various off-line assays. Cells within the tumorspheres demonstrated dormancy phenotypes, including diminished drug sensitivity, impaired migration, suppressed metabolism, and inhibited proliferation. Transcriptomic analysis reveals significant upregulation of drug resistance genes and cell cycle regulators, suggesting molecular mechanisms underlying dormancy. The agent verteporfin, which targets the signaling pathway associated with dormancy, exhibited improved efficacy (25.0–27.7%) in the elimination of dormant cells when administered in conjunction with FOLFIRINOX. This developed dormancy cancer model offers an efficient tool for dissecting the mechanisms of tumor dormancy and advancing the discovery of anticancer agents.

---

Introduction

Chemotherapy has the potential to eliminate the majority of cells within a solid tumor; however, the residual tumor cells regain their proliferation capacity after the withdrawal of the drug.¹ These cells often enter a dormant state, characterized by cell cycle arrest in the G0/G1 phase.² These dormant cancer cells (DCCs) exhibit a high resistance to chemotherapy but re-enter the proliferation cycle under favorable conditions.³ Studies suggest that the high plasticity of DCCs plays a pivotal role in tumor drug resistance and recurrence. Consequently, targeting DCCs has been proposed as a potential strategy to combat tumor drug resistance. However, the mechanisms underlying the formation and the biological characteristics of DCCs remain incompletely understood. Therefore, the construction of dormant tumor models is essential for elucidating tumor drug resistance mechanisms and developing targeted interventions.

Classical dormant tumor models predominantly rely on in vivo tumor cultivation, a process that is both time-consuming and costly.⁴ Recently, in vitro three-dimensional (3D) cultures have emerged as a valuable alternative because they not only mimic the in vivo tumor microenvironment but also offer advantages such as ease of use, clear testing results, and efficient drug screening capabilities.⁵ Chemotherapeutic agents, such as 5-fluorouracil, can induce G0/G1 phase arrest to establish dormant models.⁶ However, current 3D cultures typically depend on hydrogel or agarose scaffolds, which often exhibit uneven porosity, leading to significant heterogeneity in tumorsphere size. As reported, cell viability decreases with increasing tumorsphere size.⁷ This variability affects drug diffusion and elicits inconsistent responses, thereby hindering the standardization of these models.

Microfluidic technology presents a novel approach to address the aforementioned limitations, primarily through two strategies: first, the incorporation of microstructures on microfluidic chips, such as micro-sieves⁸ and micro-wells,^9,10 facilitates cell aggregation to form tumorspheres, thereby obviating the necessity for hydrogels; second, microfluidic chips can be readily designed with high-density microstructures that promote the formation of uniformly-sized tumorspheres. The uniformity in tumorsphere sizes mitigates variations in cell–drug responses, rendering the tumorsphere arrays extensively applicable in anticancer drug testing. However, with the widespread use of closed format microfluidic chips that facilitate perfusion culture, current microfluidic technologies continue to encounter complexities associated with tumorsphere retrieval, limiting the application of offline analytical tools necessary to elucidate cellular responses to drugs.

To address this issue, the current work developed a method for forming tumorsphere arrays on an open-surface microwell array format microfluidic device. We used a hydrophilic polymer-coated microwell array, which facilitates the formation of tumorspheres through cell self-assembly. We demonstrated that the microwell array is capable of forming uniformly sized tumorspheres while ensuring high cell viability. The microfluidic device presented herein achieves a microwell density an order of magnitude greater than those of currently available microwell plates, while also offering enhanced operational convenience. In comparison to closed-format microfluidic chips,^11,12 the open-surface microwell array facilitates easy retrieval of tumorspheres, thereby enabling the correlation of cell–drug response phenotypic outcomes with a range of offline analyses.

Building on this, the tumorsphere arrays were repeatedly exposed to a drug combination (5-fluorouracil/oxaliplatin/SN-38) to investigate the biological responses and dormancy characteristics of colorectal cancer cells within. After two rounds of drug treatment, the tumorspheres were retrieved for off-line analysis and cells within were verified to have entered a dormant state. This dormancy model was also used to screen drugs and it was found that the combination of the selected agent with conventional chemotherapy increased the efficacy of killing dormant cells.

Experimental

Microfluidic device

The microfluidic device comprises four open-surface modules, each containing 10 000 hexagonal micro-wells (with an edge length of 86.66 μm and inter-row/column spacing of 80 μm). A triangular buffer zone (10 mm × 10 mm) is incorporated at the near-center site to mitigate fluid shear disturbances. The upper layer structure features four reservoirs with an edge length of 30 mm and a circular boundary (with an inner diameter of 85 mm and an outer diameter of 88 mm), ensuring the stability and operational compatibility of the device. The microwell array device was fabricated using the conventional SU-8 photolithography and polydimethylsiloxane (PDMS) casting method. Detailed information is described in the SI.

Surface coating

The surface of the microwell arrays was coated with an epoxy-modified hydrophilic polymer.^13,14 In brief, a solution of 5.0% (w/v) dimethyl acrylamide (DMA) and 0.1% (v/v) glycidyl methacrylate (GMA) was prepared in distilled water and thoroughly degassed for 10 min. To initiate polymerization, 0.1% (v/v) tetramethylethylenediamine (TEMED) and 0.05% (w/v) potassium persulfate were added to the mixture, and the reaction was allowed to proceed for 30 min at room temperature. Subsequently, the polymer solution was extensively dialyzed using 3500 molecular weight cut-off dialysis membranes for three days, with distilled water being changed every 12 h. The resulting polymer solution was diluted 20-fold with pure water prior to being introduced into the bonded microdevice. After an incubation period of 15 min at room temperature, the solution was completely pumped out, and the microwell array device was directly heated at 110 °C for 10 min.

Cell culture

Colorectal cancer cell lines, HCT116 and HT29 (Wuhan Procell Life Science & Technology Co., Ltd.), were cultured in DMEM basic medium supplemented with 10% FBS in T25 culture flasks and incubated at 37 °C in a 5% CO₂ cell incubator. The cells were digested using a 0.25% trypsin–EDTA solution for 3 min to form a suspension. Following centrifugation and washing with phosphate-buffered saline (PBS) to eliminate residual trypsin, the cells were re-suspended in culture medium based on the cell count. After cell seeding, the excess fluid was aspirated. High-glucose DMEM medium containing 10% FBS was then added. The microwell array device was maintained in a 100 mm culture dish and placed in a 37 °C cell incubator.

For tumorsphere culture in ultra-low attachment (ULA) 96-well U-bottom plates, HCT116 cell suspensions were plated at densities ranging from approximately 50 to 50 000 cells per well in DMEM containing 10% FBS. The cells were grown in plus ULA 96-well U-bottom plates (BeyoGold™) at 37 °C with 5% CO₂ for 1 to 5 days. Tumorsphere morphology was examined using bright-field microscopy, and cell viability was evaluated through calcein-AM (Dojindo, CS26) and propidium iodide (PI, Beyotime, ST511) staining.

Exposure of anticancer agents to tumorspheres

The FOLFIRINOX drug combination contained the following concentrations of each agent: 5-fluorouracil (5-Fu, 4 μM), oxaliplatin (Ox, 0.5 μM), and SN-38 (12.5 nM). Each round of drug treatment lasted 72 hours. Then, the drug-containing medium was replaced with fresh medium.

Cell viability assay

Cell viability was evaluated utilizing a combination of calcein-AM and PI in PBS. Following the removal of the medium, the mixed solution was introduced and incubated at room temperature for 15 min. Subsequently, the mixture was replaced with PBS, and fluorescence images were captured employing a fluorescence microscope (Leica, Dmi8 M). The images were processed using LAS X software, and the areas of viable and dead cells were analyzed with ImageJ software.

Transwell assay

Tumorspheres were digested using a 0.25% trypsin–EDTA solution, subsequently suspended in serum-free high-glucose DMEM medium, and counted. Cells were then seeded into the upper chambers of a Transwell device, while 20% FBS-containing high-glucose DMEM was introduced into the lower chamber. After 48 h, the medium was removed, and the cells were fixed with a 4% paraformaldehyde solution, followed by staining with crystal violet solution. Following washing, the migrated cells were observed under a microscope.

Glucose consumption assay

Cells resuspended in high-glucose DMEM medium supplemented with 10% FBS were seeded into a 96-well plate. After a 24 h incubation period, the supernatant was aspirated and replaced with distilled water. Glucose consumption was assessed using a glucose detection kit in accordance with the manufacturer's instructions (Boxbio, AKSU001C). Following the mixing of the working solution with the sample, incubation was conducted at 37 °C for 15 min. Absorbance was measured using a microplate reader, and glucose concentration was determined based on the standard curve and the dilution ratio.

Immunofluorescence analysis

Tumorspheres were subjected to fixation, dehydration, permeation, embedding, and sectioning. Endogenous peroxidase activity was inhibited, followed by permeabilization and blocking. Primary antibodies were incubated overnight, after which incubation with secondary antibodies was performed. The antibodies used included Ki67 (Abcam, AB15580), MUC2 (Servicebio, GB14110), chromogranin A (Servicebio, GB111316-100), P27 (Proteintech, 25614-1-AP), SOX9 (IMMblot, IM41268), DEC2 (IMMblot, IM53147) and NR2F1 (Proteintech, 24573-1-AP). Following nuclei staining, the sections were subsequently mounted on a fluorescence microscope for observation. Images were processed using LAS X and ImageJ softwares.

Cell cycle analysis

Cells were harvested and washed with PBS, and cell proliferation was assessed using a BeyoClick™ EdU-488 Cell Proliferation Kit (Beyotime, C0071S). The cells were incubated with the EdU working solution, followed by centrifugation and subsequent incubation with propidium iodide solution. After washing with PBS, the cells were fixed with paraformaldehyde, permeabilized with Triton X-100 in PBS, and washed again with PBS. The click reaction solution was prepared and mixed with the sample cells. Following the removal of the click solution, the cells were washed with PBS, and flow cytometry was conducted. Cell populations were analyzed using a flow cytometer (Guava, Millipore).

RNA-Seq

Total RNA was extracted from the collected cells and mRNA was enriched using oligo(dT) magnetic beads. The enriched mRNA, or total RNA following rRNA removal, was fragmented, and complementary DNA (cDNA) was synthesized. A sequencing library was constructed, followed by PCR amplification, as well as assessments of library size distribution and concentration. The library was then loaded onto a sequencing platform. The resulting data were analyzed utilizing bioinformatics tools. The RNA-seq data was uploaded to the NCBI database, with a BioProject ID of PRJNA1259694.

Statistical analysis

Data were analyzed using GraphPad Prism 9.5 software. For comparisons between two groups, two-tailed t-tests were used, with results expressed as mean ± standard deviation. p-Values and significance markers were as follows: *(p < 0.05), **(p < 0.01), ***(p < 0.001), ****(p < 0.0001). IC₅₀ analyses were performed by converting the X-axis to logarithmic scale (base 10) and fitting the dose–response curve using nonlinear regression.

Results and discussion

Suspension cell culture in microwell arrays

This study employed open-surface microwell arrays for developing tumorspheres (Fig. 1A). The surface of the microwell arrays was coated with a hydrophilic copolymer, specifically poly(dimethylacrylamide-co-glycidyl methacrylate) (PDMA-co-GMA). As illustrated in Fig. 1B, the hydrophilic coating effectively reduced the contact angle of the PDMS substrates from 101.31° ± 0.92° to 21.56° ± 1.58°. As reported by Wu et al.,¹⁴ the stable hydrophilic coating formed by poly(PDMA-co-GMA) is attributable to the covalent coupling between epoxy groups and silicon hydroxyl groups, as well as the hydrophilic interaction between the polymer and the surface. This epoxy-functionalized polymer coating demonstrated a significant inhibition of protein adsorption, effectively preventing cell attachment.

Fig. 1 Microwell arrays for forming tumorsphere arrays. A) Schematic of microwell arrays' structure and operation. B) Hydrophilic polymer coating to eliminate cell adhesion. Surface contact angle of the PDMS substrates with or without poly(PDMA-co-GMA) coating (left); cell cultures in uncoated and coated microwells (scale bar: 150 μm).

The hydrophilic coating facilitated the rapid filling of the micro-wells with cell suspensions upon loading, resulting in stochastic capture of cells. To minimize disturbance to the cells within the wells, a buffer zone was established downstream of the liquid loading points. The fluid simulation and experiment (Fig. S1) indicate that with a buffer zone of 8 mm or longer, the disturbance to cells in micro-wells (with a depth of 250 μm) was negligible. Continuous observation revealed that cells within the micro-wells did not adhere to the surfaces; rather, they spontaneously aggregated to form tumorspheres. By day 2 post-seeding, tumor cells began to form loosely aggregated spheres spontaneously, which subsequently grew and became more compact. These findings suggest that the hydrophilic coating provides a reliable non-adherent environment, thereby promoting the spontaneous aggregation of tumor cells into tumorspheres. Furthermore, this hydrophilic-coated microwell arrays proved to be highly convenient for the retrieval of tumorspheres. Experimental results indicated that after the microwell array was inverted and gently rinsed with PBS, all tumorspheres detached from the micro-well array without any residual material (Fig. S2). This stable hydrophilic coating can be reused at least three times without any cell attachment.

Formation of uniformly-sized tumorspheres in microwell arrays

It has been reported that tumorsphere size significantly influences the kinetics of cellular responses to pharmacological agents.¹⁵ Larger tumorsphere sizes not only affect the uptake of oxygen, nutrients and drugs and the removal of metabolic waste within the spheres, but also influence signaling pathways, DNA damage and repair mechanisms and cell cycle processes.¹⁶ When the size of tumorspheres exceeds 200 μm, it has a pronounced effect on cell viability. Therefore, the use of uniformly sized tumorspheres is essential for studies investigating cellular response to drugs.

Our results demonstrated that tumorspheres were successfully formed in microwells with varying input cell numbers. The size of the tumorspheres was found to depend on both the input cell number and the culture duration (Fig. 2A and S3). By modulating the input cell density and culture time, tumorspheres of specific sizes could be precisely achieved.

Fig. 2 Tumorspheres formed via the microfluidic suspension culture. A) Light microscope images of HT29 tumorspheres and statistical analysis of tumorsphere size (scale bar: 100 μm). B) Tumorspheres generated following 48 h of culture with approximately 60 cells seeded in each microwell (left); size distribution of tumorspheres in different regions of the microwell array (scale bar: 100 μm). C) Variability in tumorsphere size and cellular viability in the ULA 96-well plate cultures (green fluorescence indicates living cells, red fluorescence indicates dead cells, scale bar: 100 μm). D) Histological and immunofluorescence images of HT29 and HCT116 tumorspheres (scale bar: 50 μm).

Using the microwell arrays, tumorspheres ranging from 50 to 150 μm in size were generated by regulating the input cell density. In the subsequent experiments, we selected tumorspheres with a size of approximately 100 μm, achieved by seeding approximately 60 cells per well and culturing for 2 days. The results revealed that the tumorsphere sizes in the micro-well array exhibited excellent uniformity, with a coefficient of variation (CV) of only 4.22% across different regions (Fig. 2B). This highly uniform tumorsphere array in our device ensures the reliability and reproducibility of cellular responses to drugs.

The performance of the open-surface microwell array was compared with the ULA 96-well plate (Fig. 2C). As shown in Table 1 and Fig. S4 and S5, cell viability within tumorspheres in the open-surface microwell array was significantly improved (94.6 ± 3.0% vs. 69.6 ± 11.1%) in comparison with ULA 96-well plate. The open-surface microwell array is also superior to the multi-well plate with respect to plating efficiency, reagent consumption, size uniformity of tumorsphere, and operation time.

Table 1 Comparison of the performance between the open-surface microwell array and the ULA 96-well plate

	Open-surface microwell array	ULA 96-well plate
Plating efficiency	40 000 tumorspheres formed per operation	96 tumorspheres formed per operation
Reagent consumption	∼4 mL per operation	∼20 mL per operation
Size of tumorsphere	Highly uniform	Heterogeneously sized
Time to tumorsphere formation	2 days	3 days
Viability	94.6 ± 3.0%	69.6 ± 11.1%
Time to seed the cells	<1 min	∼15 min
Anti-adhesion performance	Complete elimination of cell adhesion	Incomplete elimination of cell adhesion
Medium exchange disturbance	Minimal	Significant

---

We conducted a morphological analysis of the retrieved tumorspheres. Hematoxylin and eosin (HE) staining revealed that cells with in the microwells formed 3D structures without any necrosis (Fig. 2D). The tissue architecture of tumorspheres derived from various colorectal cancer cell lines exhibited notable differences. HT29 tumorspheres displayed tightly arranged cells in a solid clump like configuration, whereas HCT116 tumorspheres presented more vacuole-like regions. AB-PAS staining confirmed that both HT29 and HCT116 tumorspheres contained blue staining areas, indicating the presence of mucus (Fig. 2D).

Induction of dormancy in colorectal cancer cells through repeated chemotherapy treatment

Building upon the tumorsphere array, we employed the FOLFIRINOX drug combination to repeatedly treat the tumorsphere array (Fig. 3A). We evaluated the response of colorectal cancer cells within the tumorspheres following two rounds of drug treatment (Fig. 3B). After the initial round of drug treatment, the sizes of the tumorspheres (HT29-C1 and HCT116-C1) decreased by 46.04% and 25.12%, respectively, while cell viability diminished by 5.68% and 21.58% (Fig. 3C and D). In contrast, the response of tumorspheres to the second round of drug treatment was markedly diminished. In comparison to the C1 group, HCT116-C2 tumorspheres exhibited only a 5.38% reduction in size and a slight decrease of 0.76% in cell viability, whereas HT29-C2 tumorspheres demonstrated a 2.94% increase in size alongside a 0.64% decrease in cell viability. However, the IC₅₀ values of HT29-C2 and HCT116-C2 tumorsphere cells increased by 214% and 445%, respectively, when compared to the C0 group (no drug treatment) (Fig. 3E). These findings suggest that chemotherapy fails to eradicate all the tumor cells, and the surviving tumor cells exhibit enhanced resistance upon subsequent exposure to chemotherapeutic agents.

Fig. 3 Repeat administration of anticancer agents to the tumorspheres to induce drug-resistance in colorectal cancer cells. A) Scheme of FOLFIRINOX drug treatment. C1 and C2 indicate tumorspheres that have undergone one or two rounds of drug treatment, respectively. Numbers in black squares indicate the days of cell culture, while those in red squares indicate the days for drug treatment. B) Live/dead staining of cells (scale bar: 100 μm). C) Sizes of C1 and C2 tumorspheres. D) Viabilities of cells in C1 and C2 tumorspheres. E) Increased IC₅₀ values of C2 cells in comparison to C0 cells (no drug treatment).

Confirmation of dormancy in colorectal cancer cells: inhibition of cell migration and metabolism

Transwell migration assays revealed that the proportion of migrating cells in HCT116-C1 tumorspheres decreased by 82.2% compared to HCT116-C0, while that in HCT116-C2 tumorspheres was nearly undetectable (Fig. 4A and B). This finding suggests that the migration capacity of colorectal cancer cells significantly diminished following the initial round of drug treatment and was almost entirely lost after the second round of treatment. Glucose consumption assays indicated a reduction in the metabolic activity of cells within tumorspheres subsequent to drug treatment (Fig. 4C). The glucose consumption of HCT116-C1 cells was measured at 2.20 ± 0.01 mg mL⁻¹, which was lower than that of HCT116-C0 cells (2.89 ± 0.04 mg mL⁻¹). Similarly, HT29-C1 cells exhibited a glucose consumption of 1.94 ± 0.01 mg mL⁻¹, which was lower than the 2.63 ± 0.06 mg mL⁻¹ observed in the C0 group. In contrast to C1, no significant decrease in glucose consumption was noted in the C2 group. These results indicate that the metabolic activity of cells significantly declined following the first round of drug treatment. The above findings show that cells within the tumorspheres enter a state of diminished activity following the repeated drug treatment.

Fig. 4 Functional changes of colorectal cancer cells within the tumorspheres after repeat drug treatment. (A and B) Transwell assay shows reduced migration potential of C1 and C2 cells (scale bar: 100 μm). C) Glucose consumption assay shows reduced glucose consumption of C1 and C2 cells. D) C1 and C2 cells represented decreased sphere forming rates (scale bar: 50 μm).

Decrease in cell proliferation capacity

The proliferative capacity of colorectal cancer cells was assessed through a single-cell tumorsphere formation assay. The cell suspension obtained from tumorspheres of each group was subsequently distributed in the microwell array to form a single cell culture array and cultured for the duration of 7 days (Fig. 4D). In the HT29-C0 group, 13.41% of the cells formed tumorspheres with a diameter of ≥30 μm, whereas this proportion diminished to 4.59% and 0% in the HT29-C1 and HT29-C2 groups, respectively.

We investigated the expression of proliferation markers in the drug-treated cells (Fig. 5). Ki67, a well-established marker of cell proliferation, exhibited significantly lower expression levels in dormant cells.¹⁷ P27, a cyclin-dependent kinase (CDK) inhibitor that plays a crucial role in regulating cell cycle arrest during the G0 phase, demonstrated increased expression in dormant cells.¹⁸ As demonstrated in Fig. 5A, the Ki67(+) rate exhibited a reduction from 48.4 ± 9.2% in HCT116 cells and 56.3 ± 9.6% in HT29 cells within the C0 group to 29.2 ± 8.6% in HCT116 cells and 32.3 ± 8.8% in HT29 cells in the C2 group. Concurrently, the P27(+) rate increased from 14.4 ± 6.3% (HCT116) and 2.1 ± 2.7% (HT29) in the C0 group to 22.7 ± 7.9% (HCT116) and 6.7 ± 4.8% (HT29) in the C2 group.

Fig. 5 The expression of proliferation markers in the drug-treated cells and the cell cycle distribution of tumor cells following drug treatment. A) Immunofluorescence staining (left) and statistics of Ki67 and P27 expressions (right) (scale bar: 50 μm). B) Flow cytometry analysis of the cell cycle for cells derived from tumorspheres across various groups. C) Expressions of NR2F1 in C0 and C2 subgroups of tumorspheres. Similarly, in the HCT116-C0 group, 17.39% of the cells formed tumorspheres, while this proportion decreased to 7.27% and 0% in the HCT116-C1 and HCT116-C2 groups, respectively. These findings indicate that the proliferative capacity of tumor cells was significantly reduced following the drug treatment (scale bar: 50 μm).

We conducted an analysis of the expression levels of the dormancy regulator NR2F1 within the C2 subgroups in comparison to the C0 counterparts. As depicted in Fig. 5C, NR2F1 expression exhibited significant differences between the C0 and C2 subgroups. Specifically, in HT29 tumorspheres, NR2F1 expression was markedly elevated in the C2 subgroup at 22.2%, relative to 0.5% in the C0 subgroup (P < 0.0001). Similarly, in HCT116 tumorspheres, NR2F1 expression was 21.6% in C2 compared to 11.5% in C0 (P < 0.05). These results substantiate that cancer cells transitioned into a dormant state subsequent to repeated exposure to anticancer agents.

We also investigated the cell cycle distribution of tumor cells following drug treatment. The results indicated that the proportion of G0/G1 phase cells within the tumorspheres increased significantly after the drug treatment. As illustrated in Fig. 5B, for HT29-derived tumorspheres, the proportions of G0/G1 phase cells were 30.6%, 56.1%, and 84.1% in the C0, C1, and C2 groups, respectively. For HCT116-derived tumorspheres, the proportions were 50.2%, 60.9%, and 96.0% in the C0, C1, and C2 groups, respectively. These findings suggest that a majority of the cells in the tumorspheres transitioned into the G0/G1 phase following drug treatment, indicating their entry into a dormant state.

To assess the capacity of C2 cells to reinitiate proliferation, C2 tumorspheres were cultured in a drug-free medium. After a 40 day culture period, the C2 cells successfully reestablished three-dimensional structures. Nevertheless, live/dead staining revealed reduced cell viability, with values of (50.6 ± 2.1%) for HT29 tumorspheres and (40.3 ± 8.1%) for HCT116 tumorspheres. These findings confirm that C2 cells can sustain long-term viability (Fig. S6); however, based on the current experimental data, these cells did not fully restore their proliferative potential. We propose that the complete reactivation of these dormant cells may require more complex microenvironmental conditions.

Gene expression alterations in cells Induced by repeated drug treatment

We conducted a transcriptomic analysis to identify gene expression alterations in colorectal cancer cells induced by repeated drug treatment. Compared to the C0 group, 12 416 and 11 047 differentially expressed genes in HT29-C2 and HCT116-C2 cells were identified, respectively (Fig. S7). KEGG pathway enrichment analysis indicated that HT29-C2 cells predominantly exhibited up-regulated expression of genes associated with drug resistance signaling pathways, including those related to ABC transporters and cytochrome P450 drug metabolism (Fig. 6A), which is consistent with previous reports.^19–22

Fig. 6 Transcriptomic analysis reveals gene expression alterations in CRC cells subjected to repeat drug treatment. A) Comparative analysis of signaling pathways affected in HT29 and HCT116 cells after the drug treatment. B) Changes in expressions of drug resistance-related and cell cycle-related genes, with red squares indicating up-regulation and blue ones indicating down-regulation.

In contrast, HCT116-C2 cells demonstrated down-regulated expression of genes involved in various fundamental metabolic pathways, aligning with the enhanced drug resistance and diminished metabolic capacity observed in the C2 cells.^23–25 Furthermore, genes that promote cell cycle progression were down-regulated, while genes that induce cell cycle arrest were up-regulated in the C2 cells. Specifically, genes such as SPDYA, MYC, and CDK3 in HT29-C2, as well as CAPN3, CCND2, AURKA, and KIF14 in HCT116-C2, exhibited down-regulation (Fig. 6B). These genes typically facilitate cell cycle progression and cell proliferation.^26–32 Conversely, genes such as CDKN2C, ORC1, GPNMB, and CDC25C in HT29-C2, along with PLK3, CDKN2D, and PLK2 in HCT116-C2, were found to be upregulated (Fig. 6B). These up-regulated genes are implicated in inducing cell cycle arrest, which leads to suppressed cell proliferation and failed cytokinesis, resulting in cells being arrested in the G0/G1 phase.^33–39

Moreover, drug resistance-related genes such as UGT1A7, ABCA3, ALDH3A1, ABCB4, CA9, SLC19A3, and ABCG2 in HT29-C2, as well as E2F1, CYP2F1, ANKRD1, and AOX1 in HCT116-C2, were also up-regulated (Fig. 6B). These upregulated genes enhance drug metabolism and contribute to increased drug resistance.^{20,22,37,40–48}

In summary, the majority of cells within tumorspheres subjected to repeated drug treatment were observed to be arrested in the G0/G1 phase, exhibiting diminished migration, metabolic activity, and proliferation. The results of RNA sequencing revealed a gene expression profile consistent with increased drug resistance and reduced metabolic activity. Consequently, this study successfully established a model of colorectal cancer dormancy through the repeated drug treatment of the tumorsphere arrays.

Development of resistance antagonistic strategies targeting dormancy-related signaling pathways

We further endeavored to utilize the established dormancy model to investigate resistance antagonistic strategies by targeting dormancy-related signaling pathways. According to the literature,⁴⁹ tumor dormancy is intricately associated with the YAP/TAZ-TEAD signaling pathway. Consequently, we selected five small-molecule drugs that target functional proteins linked to the YAP/TAZ-TEAD pathway (conteltinib, PF-562271, ifebemtinib, defactinib, and verteporfin) to evaluate their efficacy in reversing tumor drug resistance. These individual drugs were combined with FOLFIRINOX to administer to C2 tumorspheres (Fig. S8). The results indicated that following treatment with the combinations, cell viability in the C2 tumorspheres diminished to varying degrees. Notably, verteporfin demonstrated the most pronounced effect, with the combination treatment of FOLFIRINOX and verteporfin enhancing cell-killing efficacy by 25.0% in HT29-C2 tumorspheres and 27.7% in HCT116-C2 tumorspheres (Fig. 7).

Fig. 7 The drug resistance reversal effects of various YAP/TAZ-TEAD signaling pathway-related drugs when combined with the FOLFIRINOX regimen. A) The cell survival rate of C2 tumorspheres after treatment with FOLFIRINOX (1×) in combination with different concentrations of verteporfin. B) Schematic of the drug treatment process about the response of C2-tumorspheres to the FOLFIRINOX–verteporfin combination. C) The survival rate of C2 tumorsphere cells after the chemical treatment.

Although verteporfin has been demonstrated to enhance FOLFIRINOX-induced eradication of dormant cells, direct evidence supporting the involvement of the YAP/TEAD pathway in the dormancy model remains to be confirmed. Owing to the lack of direct evidence for YAP/TEAD pathway activation—e.g., YAP nuclear localization, alterations in target gene expression—this study cannot yet confirm the association between the YAP/TEAD pathway and the dormant phenotype. The underlying mechanisms require further verification via subsequent experiments.

Conclusions

This study establishes a colorectal cancer dormancy model through tumorsphere array engineering on an open-surface microfluidic platform. By dispensing cell suspensions onto hydrophilic polymer-coated microwell-arrays, uniformly sized tumorspheres were formed through the suspension culture. The hydrophilic polymer-coated microwell arrays circumvent two critical limitations in conventional microfluidic systems: (1) achieving precise size control through surface energy-guided self-assembly of cells, and (2) enabling non-destructive tumorsphere retrieval via open-access architecture. These enhancements render our model particularly suitable for simulating colorectal cancer dormancy and provide a robust experimental platform for investigating the mechanisms underlying dormant tumor drug resistance.

Repeated exposure to a drug combination induced characteristic dormancy phenotypes: (i) cell cycle arrest in the G0/G1 phase, (ii) up-regulation of quiescence markers, and (iii) acquired chemoresistance. Transcriptomic profiling revealed coordinated activation of dormancy-associated pathways, establishing molecular signatures of therapy-induced dormancy. Utilizing this dormancy model, we discovered that a dormancy inhibitor, when combined with chemotherapy, significantly improved the efficacy of tumor cell killing.

Beyond current applications, this system enables: (i) high-content drug screening through parallelized architecture, (ii) dormancy stage-specific biomarker discovery, and (iii) mechanistical dissection of tumor dormancy through integrated molecular profiling. By bridging the gap between conventional 2D models and in vivo systems, this technology provides a transformative approach for investigating tumor repopulation dynamics and developing dormancy-targeted therapies.

While the current study presents a technically advanced microfluidic platform for modeling cancer dormancy, the dormancy model does not yet fully recapitulate the in vivo tumor microenvironment. This limitation is attributable to the absence of stromal cells, immune cells, and endothelial cells, all of which play critical roles in regulating dormancy maintenance and drug resistance through cytokine signaling pathways.^50,51 Future investigations will aim to incorporate these cellular constituents into the dormancy model to enhance its physiological relevance.

Author contributions

Kang H. and Wang X.: methodology, investigation, formal analysis, data curation, writing – original draft, and visualization. Ye W., Luo Y. and Chen J.: methodology and investigation. Ma Z. and Li J.: software, formal analysis, and investigation. Wang L., Lin D. and Liu D.: conceptualization, methodology, writing – review & editing, supervision, project administration, and funding acquisition. Kang H. and Wang X. contributed equally and share the first authorship.

Conflicts of interest

There are no conflicts to declare.

Data availability

The raw sequencing data supporting the findings of this study have been deposited in the NCBI Sequence Read Archive (SRA) under the BioProject accession number PRJNA1259694. These data are publicly available and can be accessed through the NCBI SRA database. The data that support the findings of this study are available from the corresponding author upon reasonable request.

Supplementary information (SI): the SI contains the fluid simulation, the recovery of tumorspheres, the tumorsphere formation in microwell array, the tumorsphere formation in the ULA 96-well plate, the cell viability of C2 tumorspheres after a 40-day drug-free interval, the volcano plots of expressed genes, the survival rate of cells within C2 tumorspheres following drug treatments with FOLFIRINOX in combination with potential dormancy inhibitors, the methods for fluid simulation and contact angle measurement. See DOI: https://doi.org/10.1039/d5lc00453e.

Acknowledgements

This work was supported by the National Nature Science Foundation of China (Grants 22174048, 82373453, 82372345 and 82404576); the Postdoctoral Science Foundation of China (Grant 2020M682665); the Department of Science and Technology of Guangdong Province (Grant 2021B1515420004, 2022A1515110751, and 2023A1515010910); the Guangzhou Science Technology and Innovation Commission (Grants 202201020363, 202206010156, 2023A04J1276, and 2024A03J1032); the Science Foundation of Guangzhou First People's Hospital (Grant PT22174048, PT82372345).

References

1. J. Liu, N. Niu, X. Li, X. Zhang and A. K. Sood, Semin. Cancer Biol., 2022, 81, 132–144.
2. T. G. Phan and P. I. Croucher, Nat. Rev. Cancer, 2020, 20, 398–411.
3. M. P. F. Damen, J. van Rheenen and C. L. G. J. Scheele, FEBS J., 2021, 288, 6286–6303.
4. S. Sant and P. A. Johnston, Drug Discovery Today: Technol., 2017, 23, 27–36.
5. Y.-C. Chen and E. Yoon, in 3D Cell Culture, ed. Z. Koledova, Springer New York, New York, NY, 2017, vol. 1612, pp. 281–291.
6. K. Toshimitsu, A. Takano, M. Fujii, K. Togasaki, M. Matano, S. Takahashi, T. Kanai and T. Sato, Nat. Chem. Biol., 2022, 18, 605–614.
7. J. Friedrich, W. Eder, J. Castaneda, M. Doss, E. Huber, R. Ebner and L. A. Kunz-Schughart, J. Biomol. Screening, 2007, 12, 925–937.
8. Y.-C. Chen, P. N. Ingram, S. Fouladdel, S. P. McDermott, E. Azizi, M. S. Wicha and E. Yoon, Sci. Rep., 2016, 6, 27301.
9. X. Wang, T. He, Z. Chen, J. Chen, Y. Luo, D. Lin, X. Li and D. Liu, Lab Chip, 2024, 24(6), 1702–1714.
10. Y. Liu, X. Chen, J. Chen, Y. Luo, Z. Chen, D. Lin, J. Zhang and D. Liu, ACS Biomater. Sci. Eng., 2022, 8, 3623–3632.
11. H. Su, Y. Chen, Z. Xuan, H. Ren, P. Lu, M. Zhao and H. Wang, Smart Med., 2025, 4, e70013.
12. Y.-C. Chen, X. Lou, Z. Zhang, P. Ingram and E. Yoon, Sci. Rep., 2015, 5, 12175.
13. X. Liu and F. A. Gomez, Anal. Bioanal. Chem., 2009, 393, 615–621.
14. D. Wu, B. Zhao, Z. Dai, J. Qin and B. Lin, Lab Chip, 2006, 6, 942.
15. S. K. Singh, S. Abbas, A. K. Saxena, S. Tiwari, L. K. Sharma and M. Tiwari, BioTechniques, 2020, 69, 333–338.
16. F. Hirschhaeuser, H. Menne, C. Dittfeld, J. West, W. Mueller-Klieser and L. A. Kunz-Schughart, J. Biotechnol., 2010, 148, 3–15.
17. K. K. Payne, R. C. Keim, L. Graham, M. O. Idowu, W. Wan, X.-Y. Wang, A. A. Toor, H. D. Bear and M. H. Manjili, J. Leukocyte Biol., 2016, 100, 625–635.
18. M. Zhang, R. Peng, H. Wang, Z. Yang, H. Zhang, Y. Zhang, M. Wang, H. Wang, J. Lin, Q. Zhao and J. Liu, Cell Death Dis., 2022, 13, 159.
19. J. R. Reed and W. L. Backes, Pharmacol. Ther., 2012, 133, 299–310.
20. R. Zarbock, E. Kaltenborn, S. Frixel, T. Wittmann, G. Liebisch, G. Schmitz and M. Griese, Biochim. Biophys. Acta, 2015, 1851, 987–995.
21. N. Tournier, S. Goutal, S. Mairinger, I. Hernández-Lozano, T. Filip, M. Sauberer, F. Caillé, L. Breuil, J. Stanek, A. F. Freeman, G. Novarino, C. Truillet, T. Wanek and O. Langer, J. Cereb. Blood Flow Metab., 2021, 41, 1634–1646.
22. X. Xu, Y. Zheng, L. Luo, Z. You, H. Chen, J. Wang, F. Zhang, Y. Liu and Y. Ke, Cell Death Dis., 2024, 15, 318.
23. Y.-H. Cho, E. J. Ro, J.-S. Yoon, T. Mizutani, D.-W. Kang, J.-C. Park, T. Il Kim, H. Clevers and K.-Y. Choi, Nat. Commun., 2020, 11, 5321.
24. M. Huang, D. Zhang, J. Y. Wu, K. Xing, E. Yeo, C. Li, L. Zhang, E. Holland, L. Yao, L. Qin, Z. A. Binder, D. M. O'Rourke, S. Brem, C. Koumenis, Y. Gong and Y. Fan, Sci. Transl. Med., 2020, 12, eaay7522.
25. L. Lin, A. J. Sabnis, E. Chan, V. Olivas, L. Cade, E. Pazarentzos, S. Asthana, D. Neel, J. J. Yan, X. Lu, L. Pham, M. M. Wang, N. Karachaliou, M. G. Cao, J. L. Manzano, J. L. Ramirez, J. M. S. Torres, F. Buttitta, C. M. Rudin, E. A. Collisson, A. Algazi, E. Robinson, I. Osman, E. Muñoz-Couselo, J. Cortes, D. T. Frederick, Z. A. Cooper, M. McMahon, A. Marchetti, R. Rosell, K. T. Flaherty, J. A. Wargo and T. G. Bivona, Nat. Genet., 2015, 47, 250–256.
26. Q. Jin, G. Liu, L. Bao, Y. Ma, H. Qi, Z. Yun, Y. Dai and S. Zhang, Cancer Manage. Res., 2018, 10, 2757–2765.
27. G. Bretones, M. D. Delgado and J. León, Biochim. Biophys. Acta, Gene Regul. Mech., 2015, 1849, 506–516.
28. T. Teo, S. Kasirzadeh, H. Albrecht, M. J. Sykes, Y. Yang and S. Wang, Pharmacol. Res., 2022, 180, 106249.
29. S. Zhao, D. Huang and J. Peng, J. Genet. Genomics, 2021, 48, 955–960.
30. S.-Y. Park, C.-J. Lee, J.-H. Choi, J.-H. Kim, J.-W. Kim, J.-Y. Kim and J.-S. Nam, J. Exp. Clin. Cancer Res., 2019, 38, 399.
31. J. M. Mosquera, H. Beltran, K. Park, T. Y. MacDonald, B. D. Robinson, S. T. Tagawa, S. Perner, T. A. Bismar, A. Erbersdobler, R. Dhir, J. B. Nelson, D. M. Nanus and M. A. Rubin, Neoplasia, 2013, 15, 1–10.
32. P. Pejskova, M. L. Reilly, L. Bino, O. Bernatik, L. Dolanska, R. S. Ganji, Z. Zdrahal, A. Benmerah and L. Cajanek, J. Cell Biol., 2020, 219, e201904107.
33. G. Guan, L. Zheng, J. Xi, X. Yang, X. Chen and F. Lu, Virol. Sin., 2021, 36, 810–813.
34. Y. Tatsumi, S. Ohta, H. Kimura, T. Tsurimoto and C. Obuse, J. Biol. Chem., 2003, 278, 41528–41534.
35. M. Suda, I. Shimizu, G. Katsuumi, C. L. Hsiao, Y. Yoshida, N. Matsumoto, Y. Yoshida, A. Katayama, J. Wada, M. Seki, Y. Suzuki, S. Okuda, K. Ozaki, M. Nakanishi-Matsui and T. Minamino, Sci. Rep., 2022, 12, 6522.
36. K. Liu, M. Zheng, R. Lu, J. Du, Q. Zhao, Z. Li, Y. Li and S. Zhang, Cancer Cell Int., 2020, 20, 213.
37. S. Xie, H. Wu, Q. Wang, J. P. Cogswell, I. Husain, C. Conn, P. Stambrook, M. Jhanwar-Uniyal and W. Dai, J. Biol. Chem., 2001, 276, 43305–43312.
38. H.-A. Lee, K.-B. Chu, E.-K. Moon and F.-S. Quan, J. Cancer, 2021, 12, 5086–5098.
39. F. Li, M. Jo, T. E. Curry and J. Liu, PLoS One, 2012, 7, e41844.
40. S. Kukal, D. Guin, C. Rawat, S. Bora, M. K. Mishra, P. Sharma, P. R. Paul, N. Kanojia, G. K. Grewal, S. Kukreti, L. Saso and R. Kukreti, Cell. Mol. Life Sci., 2021, 78, 6887–6939.
41. T. K. L. Kiang, M. H. H. Ensom and T. K. H. Chang, Pharmacol. Ther., 2005, 106, 97–132.
42. G. Muzio, M. Maggiora, E. Paiuzzi, M. Oraldi and R. A. Canuto, Free Radical Biol. Med., 2012, 52, 735–746.
43. H.-C. Huang, B.-H. Shiu, Y. Nassef, C.-C. Huang, Y.-E. Chou, W.-C. Ting, L.-C. Chang, J.-C. Lin, L.-K. Hsiao, S.-F. Yang and S.-C. Su, J. Cancer, 2022, 13, 2775–2780.
44. Y. Dang, T. Zhang, S. Pidathala, G. Wang, Y. Wang, N. Chen, C. Song, C.-H. Lee and Z. Zhang, Cell Res., 2024, 34, 458–461.
45. J. I. Øvrebø, M.-R. Bradley-Gill, N. Zielke, M. Kim, M. Marchetti, J. Bohlen, M. Lewis, M. van Straaten, N.-S. Moon and B. A. Edgar, Proc. Natl. Acad. Sci. U. S. A., 2022, 119, e2113704119.
46. J. Wan, B. A. Carr, N. S. Cutler, D. L. Lanza, R. N. Hines and G. S. Yost, Drug Metab. Dispos., 2005, 33, 1244–1253.
47. X. Xu, D. Zhong, X. Wang, F. Luo, X. Zheng, T. Feng, R. Chen, Y. Cheng, Y. Wang and G. Ma, Sci. Rep., 2024, 14, 5268.
48. H. R. Nejabati, K. Schmeisser, V. Shahnazi, D. Samimifar, Y. Faridvand, Z. Bahrami-Asl, N. Fathi-Maroufi, S. Nikanfar and M. Nouri, Ageing Res. Rev., 2020, 62, 101131.
49. A. V. Pobbati and W. Hong, Theranostics, 2020, 10, 3622–3635.
50. P. Tallón de Lara, H. Castañón, M. Vermeer, N. Núñez, K. Silina, B. Sobottka, J. Urdinez, V. Cecconi, H. Yagita, F. Movahedian Attar, S. Hiltbrunner, I. Glarner, H. Moch, S. Tugues, B. Becher and M. van den Broek, Nat. Commun., 2021, 12, 769.
51. H. Park, I. Kang, S. Lee, M. Park, S. Kim, S. Y. Lim, H. Nam, D. Yun, S. Kim, Y. Kim, J. H. Jeong, K. Lee, H. K. Lee, Y. Lee and Y.-C. Kim, J. Controlled Release, 2025, 385, 113970.

---

Footnote

† Authors Kang H. and Wang X. contributed equally and share the first authorship.

---