ICIs development has deeply changed NSCLC therapeutic care of patients over the last decade, by demonstrating survival improvement. Importantly, platinum-based chemotherapy remained the backbone of treatment, in combination with ICIs, regardless tumor stage and histology. However, most patients will finally relapse, and rebiopsies are often difficult to obtain. Chemotherapies impact tumor cells and their microenvironment. However, molecular modifications remain largely unknown due to the challenging access to tumor tissues before and after systemic therapies. In order to overcome this limitation, the objective of our study is to develop and characterize 3D models of lung cancer, including or not TME cells. Indeed, the number of models with TME cells is limited particularly regarding the presence of macrophages. Moreover, to evaluate the impact of chemotherapies, we used a protocol based on chemotherapy cycles administrated in the clinic and not only a treatment during 72 h as it was usually the case in in vitro studies. Thus, we have developed Multicellular Tumor Spheroids (MCTS) models containing NSCLC cells, with or without fibroblasts and monocytes. First, we treated MCTS constituted of tumor cells alone with a combination of chemotherapeutic agents, carboplatin-paclitaxel (CaPa) or carboplatin-gemcitabine (CaGe), according to the treatment schedule used in patients (Fig. 1D). These treatments decreased cell viability and induced strong transcriptomic modifications. In particular, several genes involved in tumor cell immunogenicity [11] were deregulated, such as CD274 coding for PD-L1, which was induced. CD274 induction was confirmed at the protein level using flow cytometry and immunohistology. These experiments showed that PD-L1 induction was restricted to a fraction of cells localized at MCTS periphery, those in direct contact with chemotherapies. Then, chemotherapies impact was studied using complex MCTS constituted of NSCLC tumor cells, fibroblasts and monocytes. These treatments reduced cell viability depending on the cancer cell line used to form MCTS. Some MCTS being resistant to treatments, reduced the number of macrophages but had no impact on fibroblasts. A cell immunogenicity modification was also observed. Furthermore, CD274 gene expression increased, as previously observed, in NSCLC cells with low basal expression, but remained the same in NSCLC cells with a high basal CD274 expression. Herein, this PD-L1 induction was not either related to the treatment used nor to the presence or not of fibroblasts and macrophages. However, other genes regulation involved in anti-tumor immune response and resistance to immunotherapies, NECTIN4, HMGB1, TNFSF18, by CaPa or CaGe seem to be dependent on the cell line used and thus to be more heterogeneous.

Paclitaxel and gemcitabine belong respectively to taxane, a family of drugs interfering with microtubule during cell division, and nucleoside, a family of metabolic inhibitors. In regards of their different action mode, numbers of genes are differentially regulated when using CaPa or CaGe combinations. CaPa induced specifically genes expression involved in inflammatory pathways such as TNFα signaling, inflammatory response and interferon-γ response, and specifically reduced pathways involved in cell division such as E2F targets, G2M checkpoint and mitotic spindle (Figure S1B and Table S4). CaGe specifically induced genes expression involved in pathways such as DNA damages with P53 and UV response, and reduced, therefore, pathways involved in inflammation such as interferon-γ and interferon-α responses (Figure S1C and Table S5). Nevertheless, most of deregulated genes are common to the two combinations and suggests a strong impact of carboplatin, a platin salt. Mainly, genes induced are involved in pathways regulating inflammation and DNA damages like TNFα signaling and inflammatory response, as well as P53 and apoptosis respectively (Figure S1D and Table S6). At the level of immunogenicity, CaGe and CaPa had a similar impact on the hallmark genes of resistance/sensitivity to immune checkpoint inhibitors in ADACA 117. However, in H1437 and H1975, the effect seemed to depend on cells sensitivity to treatments. Indeed, CaGe induced a stronger decrease in cell viability compared to CaPa as well as a higher regulation of CD274 and TNFSF18 expression. Therefore, depending on cells, CaGe and CaPa effect might be different on immunogenicity. This illustrates the complexity and heterogeneity of the response to treatments observed in clinical practice [12, 13].

PD-L1, expressed on various solid tumor and immune cells, plays a crucial role in developing cancer immunoresistance by binding PD-1 expressed at the surface of T-lymphocytes, resulting in the inhibition of T-lymphocytes migration, proliferation and secretion of cytotoxic mediators [14]. Thus, PD-L1 is a major target in immunotherapy strategy. We observed that CaPa and CaGe combinations increased PD-L1 expression suggesting a possible conversion of negative/low PD-L1 cells to positive/high PD-L1 cells after chemotherapy. This PD-L1 induction had already been observed in a heterogeneous way in NSCLC tumors using immunohistochemistry and this seemed to depend on the type of chemotherapy used [15,16,17,18]. Interestingly, this PD-L1 induction following chemotherapies could explain, at least partially, the superiority of chemotherapy/immunotherapy compared to chemotherapy alone even in patients with a PD-L1 score < 1% in tumor [3, 5, 12, 19, 20]. Likewise, in patients with resectable NSCLC, a neoadjuvant combination of chemotherapy (carboplatin/cisplatin with paclitaxel) plus ICI showed a better efficacy than chemotherapy alone regardless of PDL-1 status suggesting a possible sensitization of tumors to ICIs by inducing PD-L1 expression. In addition, several studies showed a trend to an increase in PD-L1 level at relapse, after a cytotoxic chemotherapy in adjuvant or advanced stage [12, 13]. The reproduction of this chemotherapy effect on PD-L1 expression in MCTS suggests that this 3D model could be an interesting alternative to tumor samples to study treatments impact on tumor cells.

TME cells, including fibroblasts and macrophages, play a major role in the tumors development. Thus, we added these cells in our MCTS to take into account the interactions with tumor-associated macrophages (TAMs) and cancer-associated fibroblasts (CAFs) as well as physical properties of the three-dimensional cell culture [21, 22]. Monocytes differentiated into TAMs expressing CD14 and CD163 markers, usually tend to associate with immunosuppressive M2-like phenotype [23, 24]. We also observed that macrophages were highly sensitive to treatments whereas fibroblasts were not. Several studies have shown a depleting effect of gemcitabine on myeloid derived suppressive cells in NSCLC [25]. This effect was also observed in other types of cancers with chemotherapeutic agents including platinum derivatives, anthracyclines and taxane [25]. This suggests that after chemotherapy, residual tumours could be infiltrated by new monocytes in a new environment, with different phenotypes. It is difficult to generate data in order to confirm TME modification in tumours due to the difficult access to samples before and after treatments. Furthermore, CAFs are a major TME component that contribute to tumor growth, supporting resistance to treatments, migration and invasion [26]. As for monocytes, the presence of fibroblasts did not modify the induction of CD274 expression following treatments. As opposed to monocytes/macrophages, fibroblasts seemed less sensitive to chemotherapies suggesting a stability of fibroblast populations during tumors development and after treatments [27,28,29].