Cancer has been a severe public disease with millions of deaths every year over the world [1,2]. Practical approaches for cancer treatment are gradually focusing on cancer immunotherapy, especially T cell-based adoptive cell therapy (ACT) [3–5]. T cells that naturally infiltrate solid tumors possess the capacity to identify tumor antigens and are promising candidates as effector cells [6–11]. Tumor-reactive T cells, a subset of T cells possessing antitumor capabilities, can be specifically expanded from tumor biopsies and have shown impressive responses in several types of cancers in clinical practice [5,12–14]. Despite many successes, challenges to obtain tumor biopsies and the long-expanded periods limit their further application in clinical trials. Therefore, effective enrichment and expansion to obtain vast numbers of tumor-reactive T cells is a critical step for preparing therapeutic T cell productions [13,15,16]. However, the current methods of in vitro expansion are simply performed by mixing tumor cells with immune effectors, which lack the reconstruction of the tumor microenvironment [17–20]. Furthermore, the insufficient killing and the unremovable tumor cells might lead to a residual risk when T cells are reinfused. Thus, in order to overcome these shortcomings, effective strategies to enrich and expand tumor-reactive T cells are still being anticipated.

In this paper, a novel dynamic tumor immunology-on-a-chip that mimics tumor-immune crosstalk to realize the enrichment and expansion of tumor-reactive T cells was presented, as schemed in Fig. 1. Organ-on-a-chip is a microfluidic-based culture device that simulates organ structures, physiological environments for fluid and gas exchange, and intercellular interactions [21–28]. As one typical organ-on-a-chip, tumor-on-a-chip systems are often constructed by seeding tumor cells on predesigned 2-dimensional (2D) or 3D patterns and pose advantages over high throughput, low cost, and reproducibility [29–35]. Although they have shown promising applications in cancer metastasis and preclinical drug screening, the fixed structures of these tumor-on-a-chip systems made it challenging for efficient tumor spheroid growth, dynamic tumor environment recapitulation, and in vivo blood flow dynamic simulation. In addition, to our knowledge, recent tumor-on-a-chip is seldom utilized for tumor-reactive T cell screening, and their values in enhancing T cell immunotherapy have not been fully demonstrated [36–44].

Herein, we constructed the dynamic tumor immunology-on-a-chip with enhanced tumor spheroid-immune crosstalk and efficiently expanded tumor-reactive T cells under biomimetic flow conditions (Fig. 1). The microfluidic chip was assembled by a micropillar array at the bottom layer to trap tumor cells and herringbone-patterned microgrooves at the upper layer to generate helical flow and chaotic stirring for nutrient exchange. Tumor spheroids with spherical shape and uniform size distribution were formed in alginate-based porous hydrogel arrays. After the removal of hydrogels by enzymatic digestion, the pattern changes of fluid shear stress in tumor immunology-on-chip enabled direct crosstalk between tumor spheroids and T cells for taming tumor-reactive T cells. Notably, the tumor-reactive T cells could fully be activated and rapidly proliferate, exhibiting efficient and specific tumor-killing ability in vitro. Thus, our dynamic tumor immunology-on-a-chip holds great promise in achieving efficient 3D tumor growth and high-throughput screening of tumor-reactive T cells, offering a feasible approach for autologous T cell-based personalized cancer therapy.

In a typical experiment, microfluidic chips were fabricated by replicating the silicon mold and assembled after surface treatment with the oxygen plasma, as shown in Fig. 2A and Fig. S1. The microfluidic chip consisted of a bottom layer with micropillar arrays and an upper layer featuring herringbone-shaped microgrooves for medium mixing. The herringbone microgrooves were composed of symmetrical microchannels with an orientation angle of 60°, which were designed to generate turbulent flow, facilitate fluid transport from the herringbone to downstream edges of the microchamber, and increase the contact time between the culture medium and tumor spheres docked in the microcolumn (Fig. 2B). The micropillars were arranged in a hexagonal pattern, whose diameter, height, and separation distance were optimized to match the tumor spheroid formation needs (Fig. 2C).

To meet differentiated fluid demands in tumor spheroid culture and tumor-reactive T cell recovery during coculture, a transformable microfluidic chip was constructed. Thus far, transformable microfluidic chips were rarely reported in literatures due to the lack of environmentally responsive materials for chip fabrication. In this system, we utilize an enzymatic-degradable methacrylated alginate (AlgMA) hydrogel as a scaffold to support the formation of tumor spheroids. AlgMA was obtained by chemical modification of alginate, a natural polymer widely found in brown seaweeds. The excellent biocompatibility, rapid gelling, and hydrophilic properties of alginate would consequently benefit the application of AlgMA in tissue engineering and regenerative medicine. The results of nuclear magnetic resonance (NMR) spectroscopy showed that the methacrylate group was successfully grafted onto sodium alginate (Fig. S1A). Due to the existence of vinyl groups, AlgMA could be photopolymerized into hydrogels with the presence of initiators when exposed to ultraviolet light irradiation. In addition, AlgMA hydrogels could be further dual-crosslinked by divalent ions, including calcium and iron ions, to improve the mechanical strength. As shown in Fig. S1B, dual-crosslinked AlgMA hydrogel (AlgMA-Ca and AlgMA-Fe) possessed a higher mechanical strength compared with the single photopolymerized AlgMA hydrogel, which was attributed to the newly formed ionic bonds between guluronate residues and divalent ions. Apart from the enhanced mechanical strength, the implementation of dual-crosslinking network also strengthened the structural stability of hydrogel and mitigated its swelling. To prove this hypothesis, AlgMA hydrogels were immersed in different solutions to investigate their swelling behaviors. The results showed that the ions in phosphate-buffered saline (PBS) and Dulbecco's modified Eagle's medium (DMEM) culture medium markedly reduced the swelling of photopolymerized AlgMA hydrogels and prolonged their stabilization times (Fig. S2A and B), which facilitated the maintenance of designed structure.

In order to achieve the transformation of microfluidic structure, it is necessary to remove the prefilled AlgMA hydrogels without disassembling the chip device. Alginate lyase is a specific enzyme from bacteria that could oligomerize alginate. As shown in Fig. S2C, 100 U/ml of alginate lyase was able to degrade photopolymerized AlgMA hydrogels within 4 h. In contrast, calcium ion or iron ion cross-linking hindered the degradation process of alginate lyase (Fig. S2D). Besides, biocompatibility testing of alginate lyase showed that it was virtually noncytotoxic to Hepa 1-6 cells (Fig. S3A and B). Therefore, the photopolymerized AlgMA hydrogel was used to construct dynamic microfluidic chips. As shown in Fig. 2D and Fig. S4, a hydrogel-embedded microfluidic chip was successfully constructed. The hydrogel-embedded microchambers were first prepared by turning the micropillar array through specific polydimethylsiloxane molds and filled with photopolymerized AlgMA hydrogels. Apart from the spaces created by herringbone-shaped microgrooves and micropillars shown in Fig. 2E, the AlgMA hydrogel could not only completely fill the voids among micropillars but also create new microwells for tumor cell cultivation (Fig. 2I).

These delicate structures could furthermore generate turbulent flows to increase the exchange rate of culture medium in the chip. To confirm this, the distribution of fluids within the microfluidic chip was studied via numerical simulation and was analyzed through a streamline plot. It could be seen that rotational fluid distributions occurred in both the vertical and horizontal cross-section of the microfluidic channel when the flow rate of the inlet was kept at 3 ml/h (Fig. 2F to H). In detail, the highest velocity was observed in the gaps of diagonal micropillars, whereas the lowest appeared in the gaps of horizontal micropillars due to the blockage of micropillar (Fig. 2F). Contributed by the upper layer of herringbone-shaped microgrooves, turbulent flow was successfully generated in the rectangular grooves, as shown in Fig. 2G and H. Similarly, the differences in flow rates between the microwells as well as turbulent flow generated by herringbones still occurred in hydrogel-embedded microfluidic chips (Fig. 2J to L). Taken together, the designed herringbone-shaped microgrooves could generate the chaotic fluids in a microfluidic chip and improve the mixing efficiency of the culture medium, thus making it suitable for tumor spheroid culture.

Based on the improved flow patterns resulting from the designed hydrogel architecture, the transformable hydrogel-embedded microfluidic chip was used to produce tumor spheroids. As shown in Fig. 3A and Fig. S5A, tumor cells clustered at the bottom at day 1 and became spheroid after 3 d. In order to reproducibly generate tumor spheroids, an examination was conducted to assess the relationship between the quantity of tumor cells and the resultant diameter of tumor spheroids. It was found that the diameter of the tumor spheroid increased with the seeding number of tumor cells (Fig. S5A and B). In addition, the proliferation of tumor spheroid was monitored and analyzed using cell counting kit-8 (CCK-8) assay. As shown in Fig. S5C, tumor spheroids exhibited continuous growth and revealed good cell viability. Furthermore, tumor spheroid arrays with similar size and a relatively round shape were yielded in hydrogel-embedded microfluidic chips (Fig. 3B). A tight compaction status among tumor cells in spheroid was presented in the images obtained by microscope and scanning electron microscope (SEM) (Fig. 3C and D). To investigate the ratio and location of dead cells, tumor spheroids were stained with calcein-AM and propidium iodide (PI) dyes. As shown in Fig. 3E, a slight apoptosis was found with several dead cells in the center region of the spheroid, proving the biocompatibility of AlgMA hydrogel, the appropriate structure for the constraint of tumor spheroid, and the interaction with medium flows. Further, to improve the contact efficiency between tumor cells and T cells, AlgMA hydrogel was removed after tumor spheroid formation. It could be seen that the treatment of alginate lyase could sufficiently degrade the hydrogel while keeping tumor spheroids docked between the micropillar arrays (Fig. 3F). Specifically, benefitting from the quick and mild enzymatic reaction, the structural integrity of the tumor spheroid could be maintained for further coculture (Fig. 3G and Fig. S6). Before using tumor spheroids to expand tumor-reactive T cells, tumor cells should be inactivated to avoid the immunosuppression on T cells. Therefore, we irradiated the tumor cells and tumor spheroids with 0-, 10-, 20-, 40-, and 80-Gy x-ray doses through an x-ray irradiator. To ensure the growth inability of tumor cells and select an appropriated x-ray dose, a proliferation evaluation of post-irradiative cells was carried out after 1, 3, 5, and 7 d using CCK-8 assays. It showed that the counts of irradiated tumor cells were markedly decreased compared to nonirradiated cells during the observation period (Fig. S7A and B). Furthermore, such decrease was demonstrated in a dosage-dependent manner. As a result, to achieve the most sufficient inactivation, an x-ray dose of 80 Gy was chosen to inactivate tumor spheroids (Fig. S7C). Apart from the inactivated tumor cells as feeder cells, the tumor immune microenvironment (TME) influenced tumor progression and determined the therapeutic outcome of immunotherapy. Therefore, the expression level of PD-L1 on Hepa 1-6 cells or spheroids was analyzed. We found that PD-L1 had a base-level expression on Hepa 1-6 cells or spheroids, whereas interferon-γ (IFN-γ) treatment dramatically enhanced the PD-L1 expression even on Hepa 1-6 spheroids (Fig. S8).

Research has reported the presence of lymphocytes with tumor-reactive ability in the peripheral blood of tumor-bearing patients [45–47]. These tumor-reactive lymphocytes could be prepared by coculturing peripheral blood lymphocytes with tumor cells or tumor-derived organoids. CD8⁺ cytotoxic T cells derived from OT-1 mice have been demonstrated to effectively eradicate tumor cells expressing the peptide from chicken ovalbumin-derived peptide (OVA), both in vitro and in vivo. Therefore, we isolated lymphocytes containing anti-OVA T cell receptor (TCR) from the peripheral blood of OT-1 mice (Fig. 4A). As shown in Fig. 4B, the analysis of T lymphocyte subpopulation was conducted utilizing flow cytometry, and the percentage of CD3⁺ T cells in OT-1 mice was similar to that in nontransgenic C57 mice, whereas the transgenic cells were strongly skewed toward the CD8⁺ subpopulation (Fig. S9), aligned with results in previous studies [48]. To ensure such responsiveness, T cells from peripheral blood were activated by OVA-overexpressing Hepa 1-6 spheroids, and the proliferation of T cells was assessed (Fig. 4B and C and Fig. S10). The results showed that OT-1 T cells proliferated well and responded strongly to the tumor spheroids. Moreover, the specific recognition between anti-OVA TCR and OVA peptide antigen was further confirmed by the H-2Kb antibody (Fig. 4D). After stimulation with tumor spheroids, degranulation markers of activated T cells, such as CD107a and IFN-γ, were dramatically increased, suggesting that T cells were fully activated for cytotoxic degranulation (Fig. 4E to G).

In order to assess the cytotoxicity of enriched T cells, OVA-reactive T cells were collected from microfluidic chips and subsequently cocultured with OVA-overexpressing tumor cells (Hepa 1-6) at an effector-to-target (E:T) ratio of 10:1 (Fig. 5A). T cells activated solely by interleukin-2 (IL-2) served as the control group (referred to as IL2-T cells). Following a 24-h incubation with T cells, survival of Hepa 1-6 cells was examined by calcein-AM staining. Fluorescence images as well as the analysis on fluorescence intensity demonstrated the substantial decrease in cell viability with PD-1 blocking antibody treatment (Fig. 5B and C). In addition, the expression of CD107a and IFN-γ was markedly higher in tumor spheroid-activated T cells than in the IL-2-activated T cell group (Fig. 5D and E). These results demonstrated that OVA-specific T cells could achieve complete activation through direct interaction with tumor spheroids, enhancing the tumor cell elimination efficiency.

To generate autologous tumor-reactive T cells and explore their antitumor activity, C57 mice were injected with Hepa 1-6 tumor cell lysate, and polyclonal T lymphocytes were isolated from mouse peripheral blood (Fig. 6A). After 2 weeks of coculture, tumor spheroid-activated T cells formed numbers of clusters and a dramatically increased percentage of CD3⁺ T cells were observed (Fig. 6B). The purity of enriched T cells exhibited an increase from 10% to 68% (Fig. 6C). Characterization of T cell subsets revealed that almost all tumor spheroid-activated T cells were CD8⁺, while IL-2-activated T cells were CD4⁺ (Fig. 6D). Furthermore, the degranulation of T cells was assessed in both experiment groups. It was found that T cells activated by tumor spheroids exhibited higher CD107a and IFN-γ expression than those activated by IL-2 (Fig. 6E to G). Next, the antitumor efficacy of tumor spheroid-activated peripheral blood mononuclear cells (PBMCs) was explored. Cytotoxic T cells were quantified and subsequently cocultured with Hepa 1-6 cells at varying E:T ratios for 72 h. It was found that the apoptotic number of tumor cells was positively correlated with tumor-reactive T cell numbers and a more pronounced tumor-killing activity can be observed at an E:T ratio of 10:1 (Fig. 6H and I). These results demonstrated that tumor-reactive T cells can be enriched from peripheral lymphocytes and were effective in killing autologous tumors.

Tumor-on-a-chip platforms serve as effective alternatives to animal models for the preclinical assessment of the therapeutic efficacy. In order to recapitulate the progress of tumor growth, angiogenesis, and metastasis, strategies including photolithography and 3D bioprinting have been devised to reproduce 1 or 2 critical components of the tumor microenvironment [49,50]. Nevertheless, owing to the intricate nature of both physiological and tumor microenvironments, current tumor-on-a-chip platforms with rigid structures still face many challenges prior to their implementation in industrial practices and clinical applications. In this study, we developed a dynamic tumor immunology-on-a-chip for tumor-reactive T cell expansion. Benefiting from the anti-adhesion properties and good biocompatibility of alginate hydrogel arrays, tumor cells seeded in the chips could successfully form uniform tumor spheroids. After removing the alginate hydrogels, the microfluidic chips exhibited a completely different structure and fluidic pattern. It was demonstrated that dynamic tumor immunology-on-a-chip, integrated with the herringbone mixer, accelerated fluid and gas exchange and effectively enhanced tumor spheroid formation.

In clinical practice, the sources of T cells used for ACT include effector T cells obtained from patients, as well as T cells that have been genetically modified to express TCRs or chimeric antigen receptors (CARs) [13,51,52]. As a subset of effector T cells able to recognize tumor antigens and exhibit antitumor capability, tumor-reactive T cells have been found in several patient's own tissues, such as tumor-infiltrating lymphocytes (TILs) in tumor lesions, PD-1⁺ lymphocytes in peripheral blood, CD8⁺ T cells in regional lymph nodes, and CD8⁺ T cells in the spleen [16,53–55]. Despite the notable variations in clonotypes identified across different sources, tumor-reactive T cells obtained from peripheral blood remain a promising alternative source of TILs for the development of engineered cell therapies against cancer.

The success of TIL therapy trials in metastatic melanoma showed an abundance of tumor-reactive T cells existing in melanoma that can be harnessed for ACT [13]. In addition, tumor-reactive T cells are also easily found in lung cancer, ovarian cancer, pancreatic cancer, colorectal cancer, and bladder cancer [14,56–59]. Nevertheless, the proportion of tumor-reactive CD8⁺ T cells within the overall immune cell population is relatively low and varies according to the type of cancer. Consequently, there is an urgent need for effective strategies to enrich and expand tumor-reactive T cells while minimizing clonal depletion and cellular exhaustion in order to preserve their cytotoxic efficacy prior to reinfusion. In our study, tumor spheroids were used as a target for the screening of peripheral blood-derived tumor-reactive T cells. We found that tumor-reactive T cells can be fully activated in microfluidic chips with potent antitumor activity proved by tumor cytotoxicity experiments. These findings indicate that our dynamic tumor immunology-on-a-chip provides a valuable platform for culturing tumor spheroids and in vitro screening tumor-reactive T cells.