A. Simulation and construction of concentration gradients in the chip

One of the advantages of CCEC is that it contains five collagen gel regions formed by the trapezoidal micropillar array barrier and five independent medium channels around the gel regions [shown in Fig. 1(b)]. The channels are filled with medium containing different drugs, diffusing through the gaps in the micropillar array into the gel region. Thus, different drug concentration gradients could simultaneously form in each gel region. To examine the space-time relationship of drug concentration gradient formation, we selected Rhodamine-dextran (4 kDa, red), Cascade Blue-dextran (3 kDa, blue), and FITC-dextran (4 kDa, green) with the same type as drug molecules, they are both organic and water-soluble molecules and diffuse in the form of molecules in collagen gel. These dye molecules have been widely used in drug gradient tests.27–2927. J. Yao, G. Li, Y. Jiao, Y. Zheng, Y. Liu, G. Wang, L. Zhou, H. Zhang, X. Zhang, J. Shuai, Q. Fan, F. Ye, S. Lou, G. Chen, K. Song, Y. Liao, and L. Liu, Lab Chip 21(15), 3004–3018 (2021). https://doi.org/10.1039/D0LC00951B28. M. R. Carvalho, D. Barata, L. M. Teixeira, S. Giselbrecht, R. L. Reis, J. M. Oliveira, R. Truckenmüller, and P. Habibovic, Sci. Adv. 5(5), eaaw1317 (2019). https://doi.org/10.1126/sciadv.aaw131729. Y. Pei, X. Wang, W. Huang, P. Liu, and L. Zhang, Cellulose 20(4), 1897–1909 (2013). https://doi.org/10.1007/s10570-013-9930-6 First, COMSOL (Multiphysics 5.6, Comsol, Sweden) software was used to simulate the diffusion of the three dyes in CCEC, confirming that a stable gradient was formed in the gel region of the chip. During the simulation, we used the “Transport of Diluted Species” physics field in the COMSOL software to simulate the diffusion of dyes in collagen. The initial concentration of the dyes and the porosity of collagen gel were set as 10.0 μg/ml and 36.8%. As shown in Fig. 2(a), the three dyes diffused to the adjacent gel regions, forming a composite concentration gradient and three single concentration gradients. Apparently, the closer the channels, the higher the dye concentration [shown in Figs. 2(a) and 2(b)]. Hence, the results indicate that our designed CCEC should be able to form stable single and composite biochemical concentration gradients through the diffusion of biochemical factors or small molecular compounds in collagen gels.Next, to verify the COMSOL simulation results, CCEC was constructed according to the fabrication process shown in Fig. 1(c). During the experiment, the experimental parameters were consistent with the simulation parameters. The three-dye solutions were added to the prepared CCEC, refreshed every 24.0h, and imaged by an inverted fluorescence microscope (Ti-E, Nikon, Japan) to visualize the concentration gradients in the gel region. As shown in Fig. 2(c), when the dye solution entered the channel, it gradually diffused to the adjacent gel regions, forming three single dye concentration gradients (I–III) and a composite dye concentration gradient (IV) in the chip. The fluorescence intensities of the three dyes over time and space were computed using a MATLAB script. It shows that the fluorescence intensity of each dye at a specific position is inversely proportional to the distance from the position to the dye channel, and the dye gradients in each region remain almost unchanged at 24, 48, 72, 96, and 120h, indicating that a continuous and stable concentration gradient can be formed in collagen region after 24h of dye diffusion. The result, that collagen with porous structures can maintain a stable biochemical concentration gradient, is consistent with previous research as Ref. 3030. Y. Jiao and S. Torquato, Phys. Biol. 9(3), 036009 (2012). https://doi.org/10.1088/1478-3975/9/3/036009. Therefore, the CCEC system can combine optical imaging technology to provide an excellent experimental platform for high-throughput drug combination screening.

B. Spatial dynamics of breast cancer cell numbers and migration in anti-cancer drug gradients

Another advantage of CCEC is that the chip contains 192 enclosed microchambers composed of collagen, each of which can serve as an independent research unit, providing cancer cells in the microchamber with a 3D culture microenvironment like that in vivo. By adding the medium containing different drugs to the microfluidic channels of the chip, stable drug concentration gradients are formed in various regions of the chip. In that case, each microchamber in each region is in a different local biochemical microenvironment. Therefore, combined with optical imaging and quantitative analysis, it is promising to assess the effect of single and composite drug gradients on cells in the microchamber at high-throughput levels.

To test this function of the chip, we selected several breast cancer chemotherapy drugs. First, the three anti-cancer drugs were injected on opposing sides of a central region so that the concentration of these drugs spread across the central region in the CCEC system. Doing so allowed the inhibition effect of the various composite drugs on the numbers and migration of breast cancer cell to be observed. There were also three regions in our chip design containing only one drug. A control region that contained no drugs was also placed on the chip [Figs. 3(a) and 4(a)].Among the drugs considered, paclitaxel (PTX) is a commonly used first-line drug.3131. T. M. Abu Samaan, M. Samec, A. Liskova, P. Kubatka, and D. Busselberg, Biomolecules 9(12), 789 (2019). https://doi.org/10.3390/biom9120789,3232. S. Murray, E. Briasoulis, H. Linardou, D. Bafaloukos, and C. Papadimitriou, Cancer Treat. Rev. 38(7), 890–903 (2012). https://doi.org/10.1016/j.ctrv.2012.02.011 Gemcitabine (GEM) is an anti-cancer drug widely used in the clinical treatment of breast cancer.3333. C. Delfino, G. Caccia, L. R. Gonzales, E. Mickiewicz, J. Rodger, L. Balbiani, D. F. Morales, A. Z. Comba, and C. Brosio, Oncology 66(1), 18–23 (2004). https://doi.org/10.1159/000076330 Paclitaxel combined with gemcitabine has been used clinically.3434. B. Wirk and E. Perez, Semin. Oncol. 33(1), S6–S14 (2006). https://doi.org/10.1053/j.seminoncol.2005.07.028 7rh is a discoidin domain receptor 1 (DDR1) inhibitor that reduces the interaction between cancer cells and collagen by inhibiting the expression of DDR1 and thus inhibits the migration of cancer cells, 7rh is not currently in use in clinical practice.3535. M. Gao, L. Duan, J. Luo, L. Zhang, X. Lu, Y. Zhang, Z. Zhang, Z. Tu, Y. Xu, X. Ren, and K. Ding, J. Med. Chem. 56(8), 3281–3295 (2013). https://doi.org/10.1021/jm301824k,3636. Q. P. Lu, W. D. Chen, J. R. Peng, Y. D. Xu, Q. Cai, G. K. Feng, K. Ding, X. F. Zhu, and Z. Guan, Oncol. Lett. 12(5), 3598–3608 (2016). https://doi.org/10.3892/ol.2016.5088 Previous studies demonstrated that long-term use of a single drug leads to resistance to the drug, resulting in unsatisfactory efficacy, while synergistic combination drugs can reduce drug resistance and improve drug efficacy.37–3937. B. Xiao, L. Ma, and D. Merlin, Expert Opin. Drug Deliv. 14(1), 65–73 (2017). https://doi.org/10.1080/17425247.2016.120558338. F. Cappuzzo, F. Mazzoni, A. Gennari, S. Donati, B. Salvadori, C. Orlandini, G. L. Cetto, A. Molino, E. Galligioni, M. Mansutti, S. Tumolo, A. Lucentini, F. Valduga, S. Bartolini, L. Crino, and P. F. Conte, Br. J. Cancer 90(1), 31–35 (2004). https://doi.org/10.1038/sj.bjc.660151839. K. Y. Aguilera, H. Huang, W. Du, M. M. Hagopian, Z. Wang, S. Hinz, T. H. Hwang, H. Wang, J. B. Fleming, D. H. Castrillon, X. Ren, K. Ding, and R. A. Brekken, Mol. Cancer Ther. 16(11), 2473–2485 (2017). https://doi.org/10.1158/1535-7163.MCT-16-0834 Therefore, we tested the effects of single and composite drugs on cells under the anti-cancer drug group PTX+GEM+7rh in the CCEC system [shown in Fig. 3(a)]. Figure 3(b) shows the enlarged image of MDA-MB-231-RFP cells (red) and MDA-MB-231-RFP cells off-chamber migration under the anti-cancer drug group PTX+GEM+7rh for 120.0h. Specifically, under the single drug gradients (a1–c1), MDA-MB-231-RFP cell numbers and migration were changed along the direction of the diffusion of the drugs, but there was no significant difference in the drug-free region (d1). Under the composite drug gradient (e1), we find significant differences in different local regions. Such as, the local regions with higher concentrations of paclitaxel and gemcitabine have more substantial inhibitory effects on cells. This distribution indicates that the CCEC system can be used for high-throughput initial evaluation of single and composite drug effects through cell numbers and migration.Subsequently, to quantitatively analyze the numbers and migration of MDA-MB-231-RFP cell over time under the drug concentration gradient in each region. According to the studies, the fluorescence intensity of RFP-labeled cells is a reasonable parameter to indicate the numbers of cells.4040. A. Krtolica, C. O. D. Solorzano, S. Lockett, and J. Campisi, Cytometry 49(2), 73–82 (2002). https://doi.org/10.1002/cyto.10149 Therefore, we evaluated the effect of the drug concentration gradients by dividing and analyzing the fluorescence images of MDA-MB-231-RFP cells at any time point relative to 0h. Specifically, each of the three single drug regions and one drug-free region was divided into 32 sub-regions, and the composite drug region was divided into 64 sub-regions. Then, the fluorescence intensity of each sub-region was analyzed by MATLAB. The fluorescence intensity of cells in a specific sub-region n and time t defines as I (n, t), where n = 0, 1, …, 32 or n = 0, 1, …, 64, and t = 0, 24, …, 120h. To analyze cell behavior within each sub-region, the cell fluorescence intensity rate R (n, t) defines aswhere I (n, 0) is the fluorescence intensity of cells in sub-region n at 0h, and I (n, t) is the fluorescence intensity of cells in sub-region n at t h. In this study, the fluorescence intensity rate of MDA-MB-231-RFP cells and MDA-MB-231-RFP cells off-chamber represent cell numbers and migration, respectively, as shown in Fig. S2 and S3 in the supplementary material. Moreover, the drugs had little effect on cells within 72.0h, and the analysis of the fluorescence intensity rate within this time may not provide more valuable information, so we mainly analyzed the fluorescence intensity rate of the cells at 120.0h.Figures 3(c) and 3(d) show the spatial distribution diagram of the fluorescence intensity rate of MDA-MB-231-RFP cell numbers and migration at 120.0h, respectively. Under the single drug gradient of paclitaxel (a1), gemcitabine (b1), and 7rh (c1), the fluorescence intensity rate of cells is low and changes from low to high from the side containing drugs to the side without drugs. In contrast, the fluorescence intensity rate of cells in the drug-free region (d1) is higher and exhibits non-spatial differences. These results illustrate that paclitaxel, gemcitabine, and 7rh have formed stable drug gradients in their respective regions and affected the numbers and migration behavior of MDA-MB-231-RFP cell. Moreover, under the composite gradient of the three drugs (e1), the fluorescence intensity rate of cells in the local area with higher concentrations of paclitaxel combined with gemcitabine (paclitaxel+gemcitabine, bottom left), as well as paclitaxel combined with 7rh (paclitaxel+7rh, upper left), are lower than those of gemcitabine and 7rh single drug gradients, respectively. Meanwhile, the local fluorescence intensity rate of paclitaxel+gemcitabine is lower than that of paclitaxel+7rh. Therefore, these results show that the concentration gradients formed by the three drugs in the chip caused the spatial difference in the MDA-MB-231-RFP cell numbers and migration. The inhibitory effect of combination drugs on cancer cells is more substantial than that of the single drug alone, and this effect of paclitaxel+gemcitabine is more effective than that of paclitaxel+7rh.Besides, to further illustrate that the CCEC system could evaluate different schemes of combined chemotherapy settings, we chose another anti-cancer drug group, PTX+5-Fu + PP2 [Fig. 4(a)], for testing. Among them, PTX was selected and regarded as the key drug in this drug group. Fluorouracil (5-Fu), like paclitaxel, is also one of the most used anti-cancer drugs in breast cancer.4141. W. Zhang, M. Feng, G. Zheng, Y. Chen, X. Wang, B. Pen, J. Yin, Y. Yu, and Z. He, Biochem. Biophys. Res. Commun. 417(2), 679–685 (2012). https://doi.org/10.1016/j.bbrc.2011.11.142 Paclitaxel combined with fluorouracil also has been used clinically.4242. N. Yamamoto, P. Jiang, M. Yang, M. Xu, K. Yamauchi, H. Tsuchiya, K. Tomita, G. M. Wahl, A. R. Moossa, and R. M. Hoffman, Cancer Res. 64(12), 4251–4256 (2004). https://doi.org/10.1158/0008-5472.CAN-04-0643 PP2 is an Src kinase inhibitor, which can enhance the expression of E-cadherin (E-cad), activates the function of the E-cad-mediated cell adhesion system, and inhibits the proliferation and migration of cancer cells, PP2 is not currently used in clinical practice.4343. M. P. Sanchez-Bailon, A. Calcabrini, D. Gomez-Dominguez, B. Morte, E. Martin-Forero, G. Gomez-Lopez, A. Molinari, K. U. Wagner, and J. Martin-Perez, Cell. Signal. 24(6), 1276–1286 (2012). https://doi.org/10.1016/j.cellsig.2012.02.011,4444. J. S. Nam, Y. Ino, M. Sakamoto, and S. Hirohashi, Clin. Cancer Res. 8(7), 2430–2436 (2002). https://doi.org/10.1111/j.1349-7006.2002.tb02479.x After experiments, we found that the concentration gradients formed by paclitaxel, fluorouracil, and PP2 in different regions had spatially differentiated effects on the numbers and migration of MDA-MB-231 cell [shown in Fig. 4(b)]. At the same time, the inhibitory effect of composite drugs on the numbers and migration of MDA-MB-231 cell in the corresponding regions was greater than that of single drugs, which was consistent with the trend observed in the PTX+GEM+7rh group [shown in Figs. 4(c) and 4(d)]. Shapiro–Wilks test and Kolmogorov–Smirnov test were conducted on the cell fluorescence intensity rates under drug groups PTX+5-Fu+PP2 and PTX+GEM+7rh (Fig. S4 in the supplementary material), which were in accordance with the normal distribution. A few of the outliers were derived from experimental errors. By loading two groups of anti-cancer drugs into the chip and analyzing their effects on the numbers and migration of MDA-MB-231 cell, it is shown that our CCEC system is very suitable for high-throughput analysis of cell responses to different drugs. In particular, the 192 collagen microchambers constructed by this chip system can study both the effects of drugs on cell numbers and the effects of drugs on cell migration, which makes the CCEC system unique from other 2D or 3D drug screening systems.

C. Statistical analysis of cell dynamics

In general, the concentration of drugs at a location is related to the distance of drug diffusion. The farther the drug was from the source channel, the lower its concentration. Here, to analyze the effects of individual drugs on cells, as well as the distribution of these in our design, we divided the single drug regions (a1–c1, a2–c2) and the drug-free regions (d1, d2) into four quadrants (Q1, Q2, Q3, Q4), and characterized the numbers and migration behavior of cell by calculating the fluorescence intensity rate of cells in each quadrant at 120.0h. As shown in the left of Figs. 5(a) and 5(b), for the two groups of anti-cancer drugs PTX+GEM+7rh and PTX+5-FU+PP2, the fluorescence intensity rate of cell numbers and migration in single drug regions gradually increased from Q1 to Q4. At the same time, it did not have a noticeable gradient variation in the drug-free regions, and the fluorescence intensity rate in drug-free regions is higher than that in the single-drug regions.Moreover, in the case of composite drugs, due to the boundary of adjacent drugs represents the region with the most effective drug combination, which has the most effect on cells. Therefore, we selected the four corners (QI, QII, QIII, and QIV) in the composite drug region (e1, e2) under two groups of anti-cancer drugs to further quantify the fluorescence intensity rate of cells and analyze the inhibitory effects between single and composite drugs of two anti-cancer drug groups. The results were showed in Figs. 5(a) and 5(b). For the PTX+GEM+7rh group, the fluorescence intensity rate of QI and QII in Region (e1) was lower than that of Q1 in regions (a1–c1). Similarly, for the PTX+5-FU+PP2 group, the fluorescence intensity rate of QI and QII in Region (e2) was also lower than that of Q1 in regions (a2–c2). Meanwhile, the fluorescence intensity rate of partial quadrants (QIII and QIV for PTX+GEM+7rh and PTX+5-FU+PP2) in composite drugs was higher than that of the corresponding quadrant (Q1) of the single drug, which may be due to the dilution of the medium in the adjacent channel, leading to its weakened inhibitory effect on cells. These results indicate that the combination of PTX+GEM+7rh and PTX+5-FU+PP2 has a more substantial inhibitory effect on MDA-MB-231 cells than for single drugs. In addition, by comparing the composite drug corners (QI, QII) of the two composite drug regions (e1, e2), we found that the fluorescence intensity rates of QI in region (e1) and QI, QII in region (e2) are significantly lower than that of QII in region (e1), indicating that the paclitaxel+gemcitabine, paclitaxel+fluorouracil, and paclitaxel+PP2 have better cell numbers and migration inhibition effect than paclitaxel+7rh. This also indicates that our CCEC system has the potential to evaluate both combination and single drugs simultaneously.

D. Effects of drug gradients on the expression of DDR1 and E-Cad in MDA-MB-231-RFP cells

In biology, combining immunofluorescence and microscopy imaging technologies allows the evaluation of intracellular proteins' expression and studying the influence of external factors on cells.4545. Y. H. Chan, G. J. Chang, Y. J. Lai, W. J. Chen, S. H. Chang, L. M. Hung, C. T. Kuo, and Y. H. Yeh, Cardiovasc. Diabetol. 18(1), 125 (2019). https://doi.org/10.1186/s12933-019-0928-8 In our study, the effects of drug concentration gradients on cell numbers and migration were assessed by characterizing drug-related proteins. Due to the 7rh and PP2 could inhibit the proliferation and migration of cancer cells by inhibiting the expression of DDR14646. L. Castro-Sanchez, A. Soto-Guzman, M. Guaderrama-Diaz, P. Cortes-Reynosa, and E. P. Salazar, Clin. Exp. Metastasis 28(5), 463–477 (2011). https://doi.org/10.1007/s10585-011-9385-9 and enhancing the expression of E-Cad to activate the E-Cad-mediated cells adhesion,4747. E. G. Graham, E. M. Wailes, and N. H. Levi-Polyachenko, J. Biomed. Nanotechnol. 12(2), 308–319 (2016). https://doi.org/10.1166/jbn.2016.2175 respectively. Here, we selected DDR1 and E-cad as the biomarker of PTX+GEM+7rh and PTX+5-Fu+PP2, and labeled them with yellow fluorescence dye and blue fluorescence dye, respectively. Therefore, under the confocal microscope, we were able to obtain the overlay fluorescence image of DDR1 (yellow) and E-cad (blue) in MDA-MB-231-RFP (red) cells in each microchamber region and adjacent matrix region of the biochip [Figs. 6(a) and 6(c)]. As expected, DDR1 is less expressed on the side close to 7rh, while E-CAD is more expressed on the side close to PP2, indicating that the concentration gradients formed by the two drugs in the chip affect the protein expression of the cells.Moreover, to analyze the distribution of relative expression levels of DDR1 protein and E-cad protein further quantitatively, we divided (c1) and (c2) regions into 32 sub-regions and divided the (e1) and (e2) regions into 64 sub-regions, respectively. Then, the relative protein expression was quantified by MATLAB software to calculate the fluorescence intensity of each sub-region. As shown in Fig. 6(b), the relative expression of DDR1 in the (c1) and (e1) region is lower on the 7rh side (about 0.30), while it is higher on the medium or gemcitabine side (about 0.7), which suggest that 7rh can inhibit the numbers and migration of cancer cell by inhibiting the expression of DDR1. Similarly, for the PTX+5-Fu+PP2 group [Fig. 6(d)], the relative expression of E-cad on the PP2 side is higher in both (c2) and (e2) regions (about 0.7), while it is lower on the medium or fluorouracil side (about 0.35), suggesting that PP2 can enhance the expression of E-cad and inhibit the numbers and migration of cancer cell. Besides, comparing the relative expression of DDR1 and E-Cad proteins under the single drugs with composite drug conditions, the protein expression of DDR1 in the single drug region was slightly lower than that in the corresponding region of composite drugs. In contrast, the expression of E-Cad exhibited the opposite situation. These results of protein expression in MDA-MB-231-RFP cells under the drug concentration gradients were consistent with the above statistical analysis results, indicating that the effects of drugs indirectly reflect the inhibition of 7rh and PP2 drugs on cell numbers and migration on cell proteins. Therefore, the CCEC system can assess high-throughput drug-related proteins, evaluate the mechanism of drugs, and provide an efficient platform for preclinical drug screening.