Background & Summary

Three-dimensional (3D) cell cultures are of increasing interest in research as they provide an intermediate model system between 2D cultures and animal models1. They overcome the limitations of 2D cell cultures, such as (i) the lack of natural structures or cell-to-cell and cell-matrix interactions of native tissues, (ii) the unrestricted access to oxygen, nutrients, metabolites, and signal molecules, and (iii) altered cell morphology, polarity and functions, e.g., differentiation, proliferation, vitality, gene expression patterns and responsiveness to stimuli2. As 3D cultures closely resemble native tissues, they have become an essential model system in a wide range of research areas, including regenerative medicine3 and cancer research4.

In particular, the so-called multiculture 3D tumour spheroids mimic solid tumours in many aspects, such as the heterogeneous architecture, growth kinetics, physical interactions or complex communication5. Unlike 2D mono-cultures, they also represent the tumour microenvironment (TME), which is an active participant of all stages of cancer development6,7. However, today, even the classic 3D single-culture spheroids seem too simple for reliable high-content screening (HCS) analysis for predicting in vivo effects of chemotherapy treatments8. Current technology allows co-culturing tumour cells with healthy stromal cells, e.g., fibroblasts, mesenchymal stem cells, endothelial cells or immune cells, for obtaining advanced but reproducible preclinical tumour models for a wide variety of uses, ranging from basic research for the development of therapeutic approaches to advanced HCSs of chemotherapy drugs9,10. In this scenario, 3D multiculture spheroids seem to offer the current best trade-off for drug HCS, providing advantages over both classical 3D single-culture spheroids6,7 and the more complex yet less reproducible patient-derived tumour organoids11. While organoids possess self-renewal and self-organization capacities recapitulating the original tumour characteristics, their clinical application is hampered by technical difficulties11,12.

Once the model has been defined, comprehending the complex nature of the tumour microenvironment is essential in order to gain an in-depth knowledge of cancer biology at a single-cell level13. Current high-resolution imaging systems can be used to understand cell-cell interactions within spheroids. For instance, light-sheet fluorescence microscopes (LSFMs) provide better imaging depth, faster imaging, less phototoxicity, and less photobleaching compared to wide-field or confocal microscopes. Furthermore, the optical sectioning results in less light-scattering and provides high-resolution plane-by-plane imaging, allowing the visualisation of single cells even in a dense 3D microenvironment14.

In this work, we created 3D tumour multiculture spheroids composed of three different fluorescently labelled cell lines analysed at different time points. In particular, we created three different tumour models composed of one cancer cell line (i.e., melanoma, breast cancer, or osteosarcoma), labelled with CellTracker (CT) Orange CMTMR dye, and two stromal cell lines, i.e., fibroblasts and endothelial cells, stained with CT DeepRed and CT Green CMFDA, respectively. To mimic chemotherapeutic stress, we applied low-dose doxorubicin (dox), which is a widely used antineoplastic drug in a myriad of malignancies15. Then, using the Leica True Confocal Scanning (TCS) SP8 Digital LightSheet (DLS) microscope, we acquired single-cell light-sheet fluorescence 3D images of the generated spheroids to analyse the cellular distribution after 24, 48, and 96 hours. While tracking structural changes in spheroids over an extended period can be informative, the high level of cell-to-cell communication led to the exchange of cytoplasmic dyes between different cell types, making it challenging to distinguish them. Finally, an expert microscope operator performed the segmentation of the entire spheroids and extracted morphological features using publicly available tools, i.e. AnaSP16,17 and ReViSP18.

The acquired LSFM 3D images of multiculture spheroids, the obtained segmentation masks and the extracted morphological features are freely available at figshare19. They can be used for further biological investigations on the cell distribution in 3D environments, for instance to investigate the position of the cancer cells over time when co-cultured with healthy stromal cells20. In addition, the 3D data can be exploited by the research community to compare various computational metrics quantitatively to assess image quality21, or to generate training sets for deep learning and machine learning techniques22, as well as to validate other segmentation approaches23.

Methods

Cell lines description

In this work, we used five different commercial human cell lines obtained from the American Type Culture Collection (ATCC, Manassas, Virginia, USA); three cancer cell lines, i.e. T-47D ductal carcinoma (ATCC, HTB-133, Lot: 63542749), A375 melanoma (ATCC, CRL-1619, Lot: 63905420), MG-63 osteosarcoma (ATCC, CRL1427, Lot: 70054704), and two stromal cell lines, i.e. MRC-5 fibroblasts (ATCC, CCL-171, Lot: 63405646), and EA.hy926 endothelial cells (ATCC, CRL-2922, Lot: 70030244). They were maintained following the corresponding ATCC guidelines.

Spheroid generation and staining

Multiculture tumour spheroids were generated by co-culturing three cell lines, specifically MRC-5, EA.hy926, and one type of tumour cell, either T-47D, A375, or MG-63. They were co-cultured in 384-well ultra-low attachment U-bottom plates (Greiner Bio-One, Kremsmünster, Austria), in DMEM supplemented with 10% Opticlone FBS (EuroClone, Milan, Italy), 2 mM L-glutamine and 1% Penicillin-Streptomycin-Amphotericin B mixture (all from Lonza, Basel, Switzerland). Before seeding, cells were stained with CellTracker dyes based on the manufacturer’s instructions (Invitrogen, Thermo Fisher Scientific, Waltham, Massachusetts, USA). Precisely, MRC-5, EA.hy926, and the tumour cells were stained with 1 µM Deep Red, 25 µM Green CMFDA, and 25 µM Orange CMTMR dyes, respectively. Then, the cells were mixed to generate tri-culture spheroids; we used 40% MRC-5, 40% EA.hy926, and 20% tumour cells. Spheroids were incubated for 24, 48, or 96 hours at 37 °C and 5% CO2 in the control medium or 0.6 µM dox-containing medium. Before live-cell imaging nuclei were stained with 1.5 µM Hoechst 33342 in DMEM for 60 min.

Image acquisition

The imaging setup and all imaging parameters were previously described in Diosdi et al.24,25. Briefly, for each spheroid, fluorescent 16-bit images were produced using a Leica TCS SP8 DLS microscope with a sCMOS DFC9000 camera (Leica Microsystems, Wetzlar, Germany). The images were acquired at three different time points (i.e., 24, 48, and 96 hours after seeding) using a 200 ms exposure time and a 25x/0.95 detection objective with a 2.5 mm mirror device on the objective. The laser intensity was adjusted for each channel at 405, 488, 552, and 638 nm (maximum laser intensity 350 mW). For imaging, DPBS mounting medium was used for every spheroid. The resolution of the images is 2048 × 2048 pixels, with a pixel size of 0.14370117 µm and a distance of 3.7 µm between each image in z-stack. The Leica Application Suite X (LAS X) software produced Maximum Intensity Projection (MIP) 8-bit images with a pixel resolution of 1683 × 1683 and a pixel size of 0.17486631 µm.

Image analysis

Morphological features have been extracted by analysing the MIP of each spheroid reported in the collection using AnaSP version 3.017. Briefly, MIP images were obtained by merging all the fluorescent channels and these were manually segmented by an expert microscope operator for obtaining binary masks, with the foreground in white over a black background. Several morphological features (e.g., Diameter, Perimeter, Area, Circularity, and Sphericity) have been extracted, including the Volume, which was reconstructed using ReViSP. The mathematical equations describing the features are available in the AnaSP’s user manual and Supplementary File 1.

Data Records

The dataset is available at figshare19, with this section being the primary source of information on the availability and content of the data being described. The collection includes images of spheroids acquired at three time points from the generation of the multiculture 3D tumour models. The images are single-cell light-sheet fluorescence 3D data saved in a multi-tiff file format. Each spheroid was stained with the same four fluorescent dyes. For each staining, a 3D stack is provided for each spheroid. Figure 1 reports the MIP for the four channels (Nuclei - ch001, EA.hy926 - ch002, tumour - ch003, and MRC-5 - ch004) available in the collection for a random spheroid shown as an example. All spheroids are acquired from top to down and each layer of the tiff file corresponds to a different z-section. The number of layers ranges from 38 to 90 sections, and the z-step between subsequent layers is the same for all the spheroids. In a separate folder MIP images are also provided for each spheroid and all of the four channels and a merged one.

Fig. 1
figure 1

Representative Maximum Intensity Projection (MIP) images of a 3D tumour multiculture spheroid of the A375 cell line imaged 24 hours after seeding. Four channels were acquired for visualisation, precisely Blue - Nucleus (Hoechst 33342), White - MRC-5 (Deep Red), Green - EA.hy926 (Green CMFDA), and Red – A375 (Orange CMTMR). The scale bar represents 50 µm for all the images. Images were taken and visualised using Leica LAS X software.

Full size image

For each tumour cell line and time point, images of 5 spheroids have been acquired for both the control and the drug-treated condition. The three different tumour types were labelled as A - T-47D, B - A375, and C - MG-63, and dox-treated samples were differentiated with “+” symbol. In total, 90 spheroids are included in the collection according to the scheme reported in Table 1. All the data are freely available at figshare19. The uploaded files follow the naming convention: “TumourType_ + _TimePoint_SpheroidID_channelID.tif” where (+) labels doxorubicin treated samples. Figure 2 shows a random spheroid for each condition obtained using the different fluorescence channels, excluding the nuclei signal for a better understanding of the cell distribution.

Table 1 Collection statistics: number (n.) of spheroids and layers for each condition considered.
Full size table
Fig. 2
figure 2

Representative Maximum Intensity Projection (MIP) images of the 3D tumour multiculture spheroid dataset. The three different tumour types were labelled as A - T-47D, B - A375, and C - MG-63, and dox- treated samples were differentiated with “+” symbol. For visualisation, three channels were merged, precisely White - MRC-5 (Deep Red), Green - EA.hy926 (Green CMFDA), and Red - Tumour (Orange CMTMR dyes). The scale bar represents 50 µm. The images were taken and visualised using Leica LAS X software.

Full size image

For each spheroid included in the collection, several morphological features have been extracted using merged MIP images. The MIP images and related binary masks are freely available at figshare19. For each spheroid analysed, Supplementary Table 1 reports values for selected features, Also, a table reporting all the values for the different extracted features is available in a .csv and .xls format at figshare19. The average values are reported in Table 2. The single values are reported in µm in the case of Diameter and Perimeter, in µm2 for Area, in µm3 for Volume, and dimensionless (i.e. [−]) for Circularity and Sphericity.

Table 2 Average values of selected morphological features for each group of five spheroids considered for each condition.
Full size table

Technical Validation

All images were collected as part of routine wet-lab analyses using a commercial microscope. Accordingly, quality assurance was performed just with daily calibrations done on the system. The raw images were uploaded without editing to figshare. Two expert microscopists carefully checked the uploaded data in blind.