Application of a microfluidic-based perivascular tumor model for testing drug sensitivity in head and neck cancers and toxicity in endothelium

Abstract

A drug sensitivity test prior to clinical treatment is necessary for individualized cancer therapy. A useful in vitro model that mimics the in vivo tumor microenvironment is required to select optimal anti-cancer agents. To promote the microfluidic technology moving from the laboratory to the clinic, we developed a microfluidic-based perivascular tumor model that makes it possible to assess drug sensitivity in 3D cultured tumor spheroids and toxicity in endothelium in parallel. The sensitivity and toxicity of PTX, 5-FU, and CDDP were assessed using the microfluidic model. It was found that high concentrations of single drugs destroyed the human umbilical vein endothelial cell line (HUVEC) layer, although they can effectively induce apoptosis of head and neck cancer cells. A combination of low concentrations of these drugs presented a good curative effect on tumor cells and low toxicity to the HUVEC layer. Then, we applied the model to test anti-cancer drug sensitivity in cancer patients. Individual differences to candidate drug combinations were found among these patients. This suggests that different chemotherapy plans should be made for individual patients. This study demonstrated that the microfluidic model might be a useful platform for individual drug tests prior to clinical treatment.

---

Introduction

Chemotherapy is an important type of cancer treatment. Effective anti-cancer agents are expected to be as selective as possible against malignant cells, but have minimal toxicity to healthy tissues.¹ Nowadays, a number of candidate anti-cancer agents are available to patients.² However, patients usually show different responses to these agents due to individual variation and tumor heterogeneity.³ An in vitro drug sensitivity test prior to clinical treatment would be a significant step toward identifying the agent or combination of agents that is most suitable for a particular patient.

Cells are physically organized in three-dimensional (3D) patterns surrounded by extracellular matrix (ECM) in vivo. However, drug-testing models have often depended on two-dimensional (2D) cell culture studies, which might give misleading and non-predictive data for in vivo response.⁴ For anti-cancer drug testing, it has been concluded that the cancer cells growing in a 3D culture are more resistant to cytotoxic agents than in a 2D culture.⁵ Furthermore, drugs need to cross the vascular endothelium before reaching their targets inside the body. Vascular endothelium might be damaged due to cell toxicity during the delivery process, resulting in a wider range of inflammatory changes and thrombosis.⁶ The aim of modern anti-cancer therapy is to develop a more effective but less toxic treatment. Mimicking the features of the in vivo microenvironment in cell culture models is pivotal for designing a practical drug testing system.^7–10

With the recent advances in microfluidic technology, the utilization of microfluidic-based cell culture has been proven to be a promising alternative to the conventional cell culture methods.^11–16 A large number of different microfluidic-based 3D cell culture models have been developed.^17–21 3D culture in microfluidic devices can be achieved by embedding cells in various matrices or forming matrix-free spheroids. Anti-cancer drug tests with the 3D culture array on microfluidic devices have been performed.^22–26 In addition, most microfluidic cell culture systems for drug research exploit a perfusion cell culture format, in which medium flows are not only used to continuously feed the cultured cells but also to provide additional functionalities such as generating gradients of drug concentrations.^27–29 With this gradient-generation function, various controllable and definable drug conditions can be effectively created on a single device allowing the stimulation of a field of cells in a high-throughput manner. This holds great promise as a better substitute for current cell-based high-throughput drug testing schemes.

Squamous cell carcinoma (SCC) and salivary gland adenoid cystic carcinoma (ACC) are relatively common malignancies in the head and neck region.^30,31 The role of chemotherapy is slowly moving towards a more prominent position within the different treatment paradigms in patients with SCC and ACC.^1,32,33 In the present study, we developed a microfluidic-based model to reconstruct perivascular tumors integrated with a concentration gradient generator (CGG). 3D cultured SCC or ACC cells formed spheroids near the endothelium-lined channels. Drugs at multiple concentrations generated by the CGG passed through the endothelial layer before reaching the tumor spheroids, resembling drug delivery from blood vessel to target tissues. Drug sensitivity of the tumor and toxicity to endothelium can be assessed simultaneously. Furthermore, we applied the device to test the sensitivity of individual SCC and ACC clinical cases to combinations of drugs to facilitate individual cancer therapy.

Materials and methods

Fabrication of the microfluidic model

The microfluidic device is composed of a glass substrate and two layers of a polydimethylsiloxane (PDMS) membrane. The PDMS layers were fabricated by repeated molding of the masters, which were prepared by spin coating a 100 μm-thick layer of SU8-3035 negative photoresist (Microchem Corp., Newton, CA, USA) onto a glass wafer respectively, and patterned by photolithography. The Sylgard 184 PDMS base and curing agent (Dow Corning, Midland, MI, USA) were mixed (10 : 1 by mass) thoroughly, degassed under vacuum, and poured onto the masters. The polymer was cured in an oven for 1 h at 80 °C. After cooling, the PDMS layers were gently peeled from the masters and trimmed to size. Inlet and outlet holes were punched out of the PDMS to generate reservoirs for the introduction of liquid (1.2 mm in diameter). The two layers of PDMS sandwiching a porous polycarbonate membrane (Whatman, Maidstone, UK) were sealed together with daubing mixture Sylgard 184 PDMS and cured in an oven for 1 h at 80 °C. The combined PDMS layers were bonded irreversibly to a glass substrate after 60 s of oxygen plasma treatment. The microfluidic device was UV sterilized for 2 h before use.

Validation of the CGG

Rhodamine 123 (Rh-123, Sigma-Aldrich, St. Louis, MO, USA) was used to validate the performance of the CGG. Rh-123 (2 μg ml⁻¹) and phosphate-buffered saline (PBS) were introduced into the device simultaneously from the drug inlets at a flow rate of 0.15 μl min⁻¹ using a double syringe pump (Baoding Longer Precision Pump Corp., Baoding, China), respectively. The fluorescent intensities of Rh-123 were quantified with an inverted fluorescent microscope (Olympus IX 71, Olympus, Tokyo, Japan). Images of data were analyzed with Image-pro plus 6.0. Background fluorescence was subtracted from the experimental data.

General cell culture

The human umbilical vein endothelial cell line (HUVEC) was purchased from American Type Culture Collection. ACC-M is a cell line established from a case of human salivary adenoid cystic carcinoma.⁴⁹ Both cell lines were maintained in DMEM/F12 medium (Hyclone, Logan, UT, USA), supplemented with 10% fetal bovine serum (Hyclone, Logan, UT, USA), 100 U ml⁻¹ penicillin, and 100 U ml⁻¹ streptomycin. The human tongue squamous cell carcinoma cell line (UM-SCC-6), kindly provided by the University of Michigan, USA, was maintained in DMEM/high glucose medium (Hyclone, Logan, UT, USA), supplemented with 10% fetal bovine serum, 100 U ml⁻¹ penicillin, and 100 U ml⁻¹ streptomycin. All cell lines were cultured at 37 °C with 5% CO₂ and 95% relative humidity.

Primary tumor cells were isolated from fresh SCC and ACC samples. The studies involving human materials were approved by the Research Ethics Committee, Dalian Medical University, China. The patients did not receive any chemotherapy or radiotherapy before surgery. The tissues were minced and digested with collagenase IV (Invitrogen, Carlsbad, CA, USA) for 1 h at 37 °C with shaking. The dissociated tissues were then incubated without shaking for 5 min at room temperature, followed by the separation of supernatant to a new tube and centrifugation (1000 rpm, 5 min). The cell pellet was resuspended in fresh culture medium and incubated at 37 °C with 5% CO₂ and 95% relative humidity.

Cell seeding on the microfluidic model

Tumor cells were cultured in ECM substitute (Cultrex Basement Membrane Extract, BME; R&D Systems, McKinley Place, MN, USA) on the microfluidic device. BME was thawed at 4 °C overnight. Tumor cells were trypsinized, centrifuged, and fully resuspended at 2.5 × 10⁷ cells per ml in ice cold BME. The cell–BME mixture was loaded into the 3D cell culture chambers on the bottom PDMS layer. BME gelled in less than 30 min at 37 °C and supported 3D distribution of tumor cells. Cell culture medium was introduced into the channels on the upper PDMS layer. HUVEC cells were resuspended in fresh medium and loaded into the channels on the upper PDMS layer. The device was incubated at 37 °C with 5% CO₂ to allow attachment of HUVEC. The device was then incubated at 37 °C with 5% CO₂ and 95% relative humidity for 3 days to allow the tumor cells to grow into spheroids.

Drug sensitivity assays

A traditional 96-well plate was used to investigate the effects of paclitaxel (PTX), cisplatin (CDDP), and 5-fluorouracil (5-FU) (Sigma-Aldrich, St. Louis, MO, USA) on 2D cultured tumor cells with 1 × 10⁴ cells per well. After 24 h incubation with different concentrations of drugs, cellular viability was evaluated.

The microfluidic model was used to test the drug sensitivities of 3D cultured tumor cells. To investigate the responses of tumor cells to a single drug, PTX, CDDP, or 5-FU at an initial concentration of 0.5 μg ml⁻¹, 5 μg ml⁻¹, or 400 μg ml⁻¹ was introduced into one drug inlet and drug-free medium was introduced into another drug inlet by a double syringe pump simultaneously. To assess the effects of combined drugs, PTX or 5-FU with CDDP were pumped into the CGG continuously via two drug inlets. After 24 h, the viability of tumor cells was assessed. A tumor cell viability assay was performed using Hoechst 33342 (Molecular Probes, Eugene, OR, USA) and propidium iodide (PI; Sigma-Aldrich, St. Louis, MO, USA). Cells were incubated with 4 μg ml⁻¹ Hoechst 33342 at 37 °C for 60 min, rinsed with PBS, labeled with 2 μg ml⁻¹ PI for 5 min, rinsed with PBS, added to fresh cell culture medium, and imaged immediately. Hoechst 33342 labeled living cells, while PI labeled dead cells. The cell survival rate was calculated as the ratio of the number of living cells to the number of living plus dead cells. The half maximal inhibitory concentration (IC₅₀) of each drug was confirmed as the concentration of a drug that obtained a 50% death rate of cancer cells.

Immunofluorescent staining

HUVEC cells were washed with PBS, fixed in 4% paraformaldehyde for 10 min and blocked with 10% normal goat serum at room temperature for 30 min. Cells were then incubated with rabbit anti-ZO-1 (Proteintech, Chicago, Illinois, USA) at dilution of 1 : 50 overnight at 4 °C. The next day, cells were incubated with DyLight 488 conjugated goat anti-rabbit IgG (H + L) (Abbkine, Redlands, California, USA) with dilution 1 : 100 for 1 hour. The cell nucleus stained with DAPI (1 : 2000; Life Technology) and wash with PBS buffer. Images were record with an inverted fluorescent microscope.

Statistical analysis

Statistical analyses were performed using SPSS version 13.0 for Windows. Student's t test was used to confirm comparisons of binary variables. All experiments were repeated at least three times. Significance was defined at p ≤ 0.05.

Results

Construction of the perivascular tumor model based on microfluidic technique

The microfluidic model is composed of a glass substrate and two layers of PDMS membrane (Fig. 1A). The top PDMS layer has a CGG with two drug inlets. The CGG is integrated with six downstream 2D cell culture channels terminating at HUVEC inlets. The bottom PDMS layer has six 3D cell culture units. Each unit has three cell culture chambers connecting with a respective inlet and outlet. 3D cell culture units are located just under 2D cell culture channels, respectively. There is a porous polycarbonate membrane between them. As illustrated in Fig. 1B, tumor cells were mixed with BME matrix and seeded in the 3D cell culture units firstly; then HUVEC cells were seeded in the 2D cell culture channels. HUVEC cells attached to the porous membrane served as the endothelial barrier, while tumor cells growing in 3D matrix mimicked perivascular tumors.

Fig. 1 The microfluidic model for testing drug sensitivity and toxicity. (A) The microfluidic device is composed of a glass substrate and two layers of PDMS membrane sandwiching a porous membrane. (B) Schematic illustration of cell seeding on the model. (C) Photograph of the established model. Blue stain was filled in the CGG and 2D cell culture channels a to f, while red stain was filled in the bottom 3D cell culture units. Scale bar = 5 mm. (D) Validation of the CGG. Fluorescent intensities of Rh-123 (green) in channels a to f are shown. Scale bar = 200 μm.

The established model is shown in Fig. 1C. When two streams of different solutions were driven into the CGG via drug inlets and flowed downstream, they were repeatedly split, mixed, and recombined. The characteristics of the concentration gradients produced through the CGG were validated using Rh-123. As shown in Fig. 1D, the intensity of Rh-123 gradually increased from channel a to f. A relatively stable concentration gradient was generated by the CGG.

Single drug sensitivity of 3D cultured UM-SCC6 and ACC-M

We investigated the sensitivities of 3D cultured UM-SCC6 to PTX, CDDP, or 5-FU, respectively. After being cultured for 3 days in BME on the microfluidic model, UM-SCC6 formed numerous multicellular spheroids in culture chambers. We pumped PTX (0.5 μg ml⁻¹), CDDP (5 μg ml⁻¹), and 5-FU (400 μg ml⁻¹) into the CGG via one drug inlet, respectively, while culture medium without drug was pumped into the CGG via another drug inlet simultaneously. Six drug concentrations were generated in channels a to f. The concentration range of each drug from channel a to f was calculated. PTX concentrations were 0.01 ± 0.007 μg ml⁻¹, 0.035 ± 0.016 μg ml⁻¹, 0.16 ± 0.009 μg ml⁻¹, 0.35 ± 0.020 μg ml⁻¹, 0.47 ± 0.010 μg ml⁻¹, 0.49 ± 0.010 μg ml⁻¹, CDDP concentrations were 0.09 ± 0.08 μg ml⁻¹, 0.37 ± 0.10 μg ml⁻¹, 1.68 ± 0.06 μg ml⁻¹, 3.66 ± 0.14 μg ml⁻¹, 4.82 ± 0.10 μg ml⁻¹, 4.95 ± 0.05 μg ml⁻¹, and 5-FU concentrations were 2.3 ± 2.3 μg ml⁻¹, 28.3 ± 11.7 μg ml⁻¹, 129.5 ± 4.8 μg ml⁻¹, 282.3 ± 12.5 μg ml⁻¹, 372.9 ± 12.3 μg ml⁻¹, 390 ± 10.0 μg ml⁻¹, respectively. The responses of UM-SCC6 cells to different concentrations of each drug are shown in Fig. 2A. It was found that the cell survival rate reduced from channel a to f along with an increase in the drug concentration. The IC₅₀ values of PTX, CDDP, and 5-FU for 3D cultured UM-SCC6 were 0.54 ± 0.02, 5.5 ± 0.3, and 454 ± 40 μg ml⁻¹, respectively.

Fig. 2 Responses of 3D cultured UM-SCC6 and ACC-M to PTX, CDDP, and 5-FU. (A) Images and cell survival analysis of 3D cultured UM-SCC6 stimulated by different concentrations of PTX, CDDP, and 5-FU on the microfluidic device. (B) Images and cell survival analysis of 3D cultured ACC-M stimulated by different concentrations of PTX, CDDP, and 5-FU on the microfluidic device. Blue (Hoechst 33342), red (PI). Scale bar = 200 μm.

We subsequently investigated the sensitivities of 3D cultured ACC-M to PTX, CDDP, or 5-FU, respectively. The responses of ACC-M cells to different concentrations of a single drug are shown in Fig. 2B. It was found that the cell survival reduced from channel a to f with an increase in the drug concentration. The IC₅₀ values of PTX, CDDP, and 5-FU for 3D cultured ACC-M were 0.45 ± 0.03, 5.2 ± 0.35, and 400 ± 40 μg ml⁻¹, respectively.

We compared the IC₅₀ values of PTX, CDDP, and 5-FU between 2D and 3D cultured tumor cells. It was found that the IC₅₀ values of each drug for 3D cultured UM-SCC6 and ACC-M cells were significantly higher than those for 2D cultured UM-SCC6 and ACC-M cells (p < 0.05) (data not shown). This suggested that tumor spheroids had significantly increased resistance to PTX, CDDP, and 5-FU compared to 2D cultured tumor cells.

Single drug toxicity to HUVEC

On the microfluidic model, drugs in the top channels diffused through the HUVEC layer before reaching tumor spheroids, resembling the drug diffusion process in vivo. Therefore, the toxicities of PTX, CDDP, and 5-FU to HUVEC could be assessed along with testing drug sensitivity of tumor cells. HUVEC cells attached on the porous membrane and contacted each other tightly without drug stimulation. The expression of ZO-1, a biomarker indicating tight junctions between endothelial cells, was used to determine drug toxicity to HUEVC. With administration of PTX, CDDP, and 5-FU for 24 h, respectively, toxicities to the HUVEC layer could be found in the channels with high drug concentrations (Fig. 3A–C). ZO-1 expression decreased significantly in channels e and f with the administration of PTX, in channels c to f with the administration of CDDP and 5-FU.

Fig. 3 HUVEC damage from different concentrations of PTX, CDDP, and 5-FU. The concentration of each drug was increased from channel a to f. (A) The expression of ZO-1 decreased significantly in channels e and f by the administration of PTX. (B) The expression of ZO-1 decreased significantly in channels c to f by the administration of CDDP. (C) The expression of ZO-1 decreased significantly in channels c to f by the administration of 5-FU. Scale bar = 50 μm. *, p < 0.05; **, p < 0.01.

Combined drug sensitivity of 3D cultured UM-SCC6 and ACC-M

The sensitivities of UM-SCC6 and ACC-M to combined drugs were evaluated using the microfluidic model. PTX (0.5 μg ml⁻¹) or 5-FU (400 μg ml⁻¹) with CDDP (5 μg ml⁻¹) were pumped continuously into the CGG via two separate drug inlets. Channels a to f had gradually decreased concentrations of PTX or 5-FU and increased concentrations of CDDP. The combined drugs induced apoptosis of UM-SCC6 and ACC-M effectively in all channels (Fig. 4A and B). The cell survival from channel a to f is similar, near 50%. These results demonstrated that a combination of low concentrations of PTX with CDDP, or 5-FU with CDDP have effects similar to those of high concentrations of a single drug on tumor cells.

Fig. 4 Responses of 3D cultured tumor cells to combined PTX–CDDP and 5-FU–CDDP on the microfluidic device. (A) Images and cell survival analysis of 3D cultured UM-SCC6. The cell survival rates are similar in channels a to f. (B) Images and cell survival analysis of 3D cultured ACC-M. The cell survival rates are similar in channels a to f. Scale bar = 200 μm.

Combined drug toxicity to HUVEC

Toxicity to HUVEC by combination of PTX or 5-FU with CDDP was assessed. With PTX–CDDP administration, high expression of ZO-1 was detected in channels b and c, while its expression decreased significantly in channels a, d, e, and f (p < 0.05) (Fig. 5A). With 5-FU–CDDP administration, high expression of ZO-1 was detected in channels c and d, while its expression decreased significantly in channels a, d, e, and f (p < 0.05) (Fig. 5B). It suggested that high concentrations of PTX, 5-FU, and CDDP were toxic to HUVEC cells although they can effectively induce apoptosis in over 50% of tumor cells. Combination of low concentration of PTX or 5-FU with CDDP exhibited low cytotoxicity to HUVEC, and this concentration could induce apoptosis in approximately 50% of tumor cells. Therefore, channels b and c were recommended with PTX–CDDP administration, and channels c and d were recommended with 5-FU–CDDP administration.

Fig. 5 HUVEC damage from PTX–CDDP and 5-FU–CDDP. (A) The expression of ZO-1 showed the highest level in channels b and c, and decreased significantly in channels a, d, e, and f with PTX–CDDP administration. (B) The expression of ZO-1 showed the highest level in channels c and d, and decreased significantly in channels a, b, e, and f with 5-FU–CDDP administration. Scale bar = 100 μm. *, p < 0.05.

Application of the microfluidic model on SCC and ACC cases

We applied the microfluidic device on clinical SCC and ACC cases to evaluate the drug sensitivity in different patients and toxicity in HUVEC cells. Primary cancer cells isolated from three SCC cases (SCC1 to SCC3) and three ACC cases (ACC1 to ACC3) to a combination of PTX or 5-FU with CDDP was assessed using the microfluidic model. As shown in Fig. 6A–F, individual drug sensitivity of these patients was found. SCC1, SCC2, ACC1, and ACC2 showed relatively high sensitivities to both combinations of PTX–CDDP and 5-FU–CDDP with the cell survival rates near 60%. By contrast, SCC3 and ACC3 showed low sensitivity to both combinations of PTX–CDDP and 5-FU–CDDP with the cell survival rates near 80%. Toxicities to HUVEC cells by combination of PTX or 5-FU with CDDP were similar to those of cell line culture assay shown in Fig. 5A and B (data not shown). Therefore, channels b and c were recommended with PTX–CDDP administration, and channels c and d were recommended with 5-FU–CDDP administration. Considering both drug sensitivity and toxicity, combination of PTX–CDDP was recommended to SCC1 and ACC1 patients, while combination of 5-FU–CDDP was recommended to SCC2 and ACC2 patients.

Fig. 6 Responses of clinical SCC and ACC cases to PTX–CDDP and 5-FU–CDDP on the microfluidic device. (A) SCC1 showed high sensitivity to PTX–CDDP. (B) SCC 2 showed high sensitivity to 5-FU–CDDP. (C) SCC 3 showed low sensitivity to both PTX–CDDP and 5-FU–CDDP. (D) ACC1 showed high sensitivity to PTX–CDDP. (E) ACC2 showed high sensitivity to 5-FU–CDDP. (F) ACC 3 showed low sensitivity to both PTX–CDDP and 5-FU–CDDP.

Discussion

In this study, we developed a microfluidic-based model for testing drug sensitivity of 3D cultured cancer cells and toxicity in 2D cultured endothelial cells. On the microfluidic model, drugs at multiple concentrations passed through the endothelial barrier and subsequently induced apoptosis of tumor spheroids. This process mimics the intravenous administration of anti-cancer drugs in vivo. We further applied this model to evaluate the anti-cancer drug sensitivity in clinical cases.

Solid tumors are supported by the ECM framework in vivo. After extravasation into the interstitial space, drugs penetrate tumor tissue by diffusion through the ECM. Water-soluble drugs traverse the ECM more readily than hydrophobic agents, but large molecules have poor diffusion coefficients, and may be sequestered by binding to epitopes in the matrix before they reach the intended target.^34,35 Platforms for 3D culture, such as scaffolds, hydrogels, and hanging drops can provide enhanced models for testing of drug delivery and toxicity.^36,37 Cancer cells tend to form spheroids by self-assembly in 3D culture. Spheroids have a closely packed 3D architecture and have diffusional limits to mass transport of drugs, nutrients and other factors, similar to in vivo tissues.³⁸ Thus, spheroids are appropriate models for testing drug delivery systems in vitro, and testing in spheroids is increasingly being regarded as an essential step in drug development.^8,39 In this study, cancer cells were embedded in BME matrix on the microfluidic model. After 3 days, both UM-SCC6 and ACC-M cells formed spheroids in matrix resembling the nest-like tumor structures in vivo. These cancer spheroids present significantly higher resistances to PTX, CDDP, and 5-FU than 2D cultured cancer cells. As the model is not confined to SCC and ACC tumor, it also can be used to study other types of cell in 3D microenvironment. Therefore, this microfluidic model would be a better platform to provide predicted results than standard 2D culture models.

Great progress has been made in microfluidic technology to predict drug-related toxicities at the preclinical stage and avoid costly mistakes in the clinic.^40,41 The microfluidic model developed in this study enabled parallel assessment of drug sensitivity of tumor cells and toxicity to endothelium. It was found that a high concentration of CDDP or 5-FU destroyed the tight contact between HUVEC cells, although it could induce the apoptosis of UM-SCC6 and ACC-M effectively. By contrast, a combination of low concentrations of these drugs presented a good curative effect on tumor cells and low toxicity to the HUVEC layer. Therefore, combined drugs gain an advantage over single drug usage.

Microfluidic-based gradient generation devices have advantages over conventional methods.^42–44 These advantages include increased throughput, reduced experimental operating costs, stable gradient generation that can be maintained for a long period of time, and the ability to observe the cellular responses to these concentration gradients in realtime. In the microfluidic model developed in this study, a CGG was integrated with cell culture channels. It could generate stable concentration gradients. To assess the effect of monotherapy, six concentrations of the candidate drugs were generated and each concentration of drug acted on cancer cells in three chambers. In the 6 × 3 array, 18 experiments can be performed in parallel, each with a unique condition. This device can also be used to assess the effect of combined therapy if two candidate drugs were introduced into the CGG via different inlets. With the 6 × 3 × 2 array, 36 experiments can be performed in parallel. The high-throughput ability meets the requirements of modern drug sensitivity tests.

To promote clinical application of this microfluidic technique, we applied the model to test anti-cancer drug sensitivity in SCC and ACC cases. PTX, 5-FU, and CDDP are three major drugs used in SCC and ACC treatment. A CDDP-based combination of these drugs has been shown to result in increased overall survival. To identify the combination of agents that is most suitable for a particular patient, we tested two combinations, PTX–CDDP and 5-FU–CDDP, on primary cells isolated from SCC and ACC tissues. Individual differences were found among these cases. Some cases showed good response to the PTX–CDDP combination, while some cases showed sensitivity to the 5-FU–CDDP combination. This result might offer guidance on individualized cancer therapy. In addition, we found that primary cells showed higher drug resistance than SCC and ACC cell lines. It was reported that stromal cells are important mediators of tumor cell survival and treatment resistance in primary tumors.^45–47 We considered that a certain number of stromal cells might be included in the primary cells, resulting in increased drug resistance in SCC and ACC cases. A drug sensitivity test based on a mixture of tumor and stromal cells may provide a better prediction of clinical treatment outcome.⁴⁸

In summary, we developed a microfluidic-based perivascular tumor model that enables parallel assessment of drug sensitivity in 3D tumor spheroids and toxicity in endothelium. Studies based on clinical cases demonstrated that the microfluidic model might be a useful platform for individual drug sensitivity tests prior to clinical treatment. The microfluidic technology is moving from the laboratory to the clinic.

Disclosure statement

The authors have no conflict of interest.

Authorship contributions

Tingjiao Liu and Bingcheng Lin designed research. Dong Jin, Xiaochi Ma, Yong Luo, Shimeng Fang, Zhaorong Xie, Xiaojie Li, Lilu Liu, Jing Kong, and Jiao Li performed research and analyzed the data. Dongyuan Qi and Fuyin Zhang collected the clinical cases. Tingjiao Liu, Bingcheng Lin and Dong Jin wrote the paper.

List of abbreviations

2D	Two-dimensional
3D	Three-dimensional
5-FU	5-Fluorouracil
ACC	Salivary gland adenoid cystic carcinoma
CDDP	Cisplatin
CGG	Concentration gradient generator
ECM	Extracellular matrix
HUVEC	Human umbilical vein endothelial cell line
IC50	Half maximal inhibitory concentration
PBS	Phosphate-buffered saline
PDMS	Polydimethylsiloxane
PI	Propidium iodide
PTX	Paclitaxel
Rh-123	Rhodamine 123
SCC	Squamous cell carcinoma

Acknowledgements

This work was supported by National Natural Science Foundation of China (No. 81171425) and National Major Scientific and Technological Special Project for “Significant New Drugs Development” (2013ZX09507005).

References

1. P. M. Specenier and J. B. Vermorken, Oral Oncol., 2009, 45, 409–415.
2. Y. Molin and J. Fayette, Anti-Cancer Drugs, 2011, 22, 621–625.
3. T. K. Hoffmann, GMS Curr. Top. Otorhinolaryngol. Head Neck Surg., 2012, 11, Doc03.
4. L. David, V. Dulong, D. L. Cerf, L. Cazin, M. Lamacz and J.-P. Vannier, Acta Biomater., 2008, 4, 256–263.
5. J. L. Horning, S. K. Sahoo, S. Vijayaraghavalu, S. Dimitrijevic, J. K. Vasir, T. K. Jain, A. K. Panda and V. Labhasetwar, Mol. Pharm., 2008, 5, 849–862.
6. T. K. Stutchbury, K. L. Vine, J. M. Locke, J. S. Chrisp, J. B. Bremner, P. R. Clingan and M. Ranson, Anti-Cancer Drugs, 2011, 22, 24–34.
7. K. M. Yamada and E. Cukierman, Cell, 2007, 130, 601–610.
8. S. Breslin and L. O'Driscoll, Drug Discovery Today, 2013, 18, 240–249.
9. J. Friedrich, C. Seidel, R. Ebner and L. A. Kunz-Schughart, Nat. Protoc., 2009, 4, 309–324.
10. F. Pampaloni, E. G. Reynaud and E. H. K. Stelzer, Nat. Rev. Mol. Cell Biol., 2007, 8, 839–845.
11. J. El-Ali, P. K. Sorger and K. F. Jensen, Nature, 2006, 442, 403–411.
12. D. Wlodkowic and J. M. Cooper, Curr. Opin. Chem. Biol., 2010, 14, 556–567.
13. E. W. Young and D. Beebe, Chem. Soc. Rev., 2010, 39, 1036–1048.
14. M. Mehling and S. Tay, Curr. Opin. Biotechnol., 2014, 25, 95–102.
15. K. Ziółkowska, R. Kwapiszewski and Z. Brzózka, New J. Chem., 2011, 35, 979–990.
16. F. Tanweer, V. L. Green, N. D. Stafford and J. Greenman, Head Neck, 2013, 35, 756–763.
17. M. Marimuthu and S. Kim, Anal. Biochem., 2011, 413, 81–89.
18. D. Huh, G. A. Hamilton and D. E. Ingber, Trends Cell Biol., 2011, 21, 745–754.
19. S.-Y. C. Chen, P. J. Hung and P. J. Lee, Biomed. Microdevices, 2011, 13, 753–758.
20. X. J. Li, A. V. Valadez, P. Zuo and Z. Nie, Bioanalysis, 2012, 4, 1509–1525.
21. D. Sun, J. Lu, Z. Chen, Y. Yu and Y. Li, Microfluid. Nanofluid., 2014, 17, 831–842.
22. A. Y. Hsiao, Y. Torisawa, Y. C. Tung, S. Sud, R. S. Taichman, K. J. Pienta and S. Takayama, Biomaterials, 2009, 30, 3020–3027.
23. S.-M. Ong, C. Zhang, Y.-C. Toh, S. H. Kim, H. L. Foo, C. H. Tan, D. Noort, S. Park and H. Yu, Biomaterials, 2008, 29, 3237–3244.
24. L. Y. Wu, D. D. Carlo and L. P. Lee, Biomed. Microdevices, 2008, 10, 197–202.
25. Y. Markovitz-Bishitz, Y. Tauber, E. Afrimzon, N. Zurgil, M. Sobolev, Y. Shafran, A. Deutsch, S. Howitz and M. Deutsch, Biomaterials, 2010, 31, 8436–8444.
26. L. Yu, M. C. W. Chen and K. C. Cheung, Lab Chip, 2010, 10, 2424–2432.
27. N. Ye, J. Qin, W. Shi, X. Liu and B. Lin, Lab Chip, 2007, 7, 1696–1704.
28. C. Yang, Y. Wu, Z. Xu and J. Wang, Lab Chip, 2011, 11, 3305–3312.
29. C. Liu, L. Wang, Z. Xu, J. Li, X. Ding, Q. Wang and C. Li, J. Micromech. Microeng., 2012, 22, 065008.
30. M. Sadick, S. O. Schoenberg, K. Hoermann and H. Sadick, GMS Curr. Top. Otorhinolaryngol. Head Neck Surg., 2012, 11, Doc08.
31. C. A. Moskaluk, Head Neck Pathol., 2013, 7, 17–22.
32. S. A. Bhide and C. M. Nutting, Oral Oncol., 2010, 46, 436–438.
33. P. P. Savvides, Semin. Plast. Surg., 2010, 24, 137–147.
34. R. A. Sheth, R. Hesketh, D. S. Kong, S. Wicky and R. Oklu, J. Vasc. Interv. Radiol., 2013, 24, 1201–1207.
35. G. Mehta, A. Y. Hsiao, M. Ingram, G. D. Luker and S. Takayama, J. Controlled Release, 2012, 164, 192–204.
36. A. Roth and T. Singer, Adv. Drug Delivery Rev., 2014, 69–70, 179–189.
37. N. T. Elliott and F. Yuan, J. Pharm. Sci., 2011, 100, 59–74.
38. S.-M. Ong, Z. Zhao, T. Arooz, D. Zhao, S. Zhang, T. Du, M. Wasser, D. Noort and H. Yu, Biomaterials, 2010, 31, 1180–1190.
39. K. Kwapiszewska, A. Michalczuk, M. Rybka, R. Kwapiszewski and Z. Brzózka, Lab Chip, 2014, 14, 2096–2104.
40. Y. Toh, T. C. Lim, D. Tai, G. Xiao, D. Noort and H. Yu, Lab Chip, 2009, 9, 2026–2035.
41. T. G. Fernandes, M. M. Diogo, D. S. Clark, J. S. Dordick and J. M. Cabral, Trends Biotechnol., 2009, 27, 342–349.
42. M. Wu, S. Huang and G. Lee, Lab Chip, 2010, 10, 939–956.
43. Y. Wen and S. Yang, Expert Opin. Drug Discovery, 2008, 3, 1237–1253.
44. F. Lin, W. Saadi, S. W. Rhee, S. Wang, S. Mittal and N. Jeon, Lab Chip, 2004, 4, 164–167.
45. S. Ingthorsson, V. Sigurdsson, A. J. Fridriksdottir, J. G. Jonasson, J. Kjartansson, M. K. Magnusson and T. Gudjonsson, BMC Res. Notes, 2010, 3, 184.
46. U. Joimel, C. Gest, J. Soria, L.-L. Pritchard, J. Alexandre, M. Laurent, E. Blot, L. Cazin, J. Vannier, R. Varin, H. Li and C. Soria, BMC Cancer, 2010, 10, 375.
47. K. D. Grugan, C. G. Miller, Y. Yao, C. Z. Michaylira, S. Ohashi, A. J. Klein-Szanto, J. A. Diehl, M. Herlyn, M. Han, H. Nakagawa and A. Rustgi, Proc. Natl. Acad. Sci. U. S. A., 2010, 107, 11026–11031.
48. Z. Xu, Y. Gao, Y. Hao, E. Li, Y. Wang, J. Zhang, W. Wang, Z. Gao and Q. Wang, Biomaterials, 2013, 34, 4109–4117.
49. H. Ma, T. Liu, J. Qin and B. Lin, Electrophoresis, 2010, 31, 1599–1605.

---

Footnote

† Co-first authors.

---