This article demonstrates a standardized method for constructing three-dimensional tumor spheroids. A strategy for spheroid observation and image-based deep-learning analysis using an automated imaging system is also described.
In recent decades, in addition to monolayer-cultured cells, three-dimensional tumor spheroids have been developed as a potentially powerful tool for the evaluation of anticancer drugs. However, the conventional culture methods lack the ability to manipulate the tumor spheroids in a homogeneous manner at the three-dimensional level. To address this limitation, in this paper, we present a convenient and effective method of constructing average-sized tumor spheroids. Additionally, we describe a method of image-based analysis using artificial intelligence-based analysis software that can scan the whole plate and obtain data on three-dimensional spheroids. Several parameters were studied. By using a standard method of tumor spheroid construction and a high-throughput imaging and analysis system, the effectiveness and accuracy of drug tests performed on three-dimensional spheroids can be dramatically increased.
Cancer is one of the diseases most feared by human beings, not least because of its high mortality rate1. In recent years, the possibility of treating cancer has increased as new therapies have been introduced2,3,4,5. Two-dimensional (2D) and three-dimensional (3D) in vitro models are used to study cancer in a laboratory setting. However, 2D models cannot immediately and accurately assess all of the important parameters that indicate antitumor sensitivity; therefore, they fail to fully represent in vivo interactions in drug therapy testing6.
Since 2020, the global three-dimensional (3D) culture market has been greatly boosted. According to one report from NASDAQ OMX, the global value of the 3D cell culture market will exceed USD 2.7 billion by the end of 2025. Compared with 2D culture methods, 3D cell culturing exhibits advantageous properties, which can be optimized not only for proliferation and differentiation but also for long-term survival7,8. By such means, in vivo cellular microenvironments can be simulated to obtain more accurate tumor characterization, as well as metabolic profiling, so that genomic and protein alterations can be better understood. Due to this, 3D test systems should now be included in mainstream drug development operations, especially those with a focus on screening and evaluating novel antitumor drugs. Three-dimensional growths of immortalized established cell lines or primary cell cultures in spheroid structures possess in vivo features of tumors such as hypoxia and drug penetration, as well as cell interaction, response, and resistance, and can be regarded as a stringent and representative model for performing in vitro drug screening9,10,11.
However, these 3D culture models also suffer from several problems that may take some time to solve. Cell spheroids can be formed using these protocols, but they differ in certain details, such as culture time or embedding gels12, so these constructed cell spheroids cannot be well controlled under a restricted size range. The size of the spheroids may influence the consistency of the viability test and imaging analysis. The growth microenvironments and growth factors also vary, which may lead to different morphologies due to differences in the differentiation among cells13. There is now an obvious need for a standard, simple, and cost-effective method for constructing all types of tumors with controlled sizes.
From another perspective, although homogeneous assays and high-content imaging approaches have been developed to evaluate morphology, viability, and growth rate, the high-throughput screening of 3D models remains a challenge for various reasons reported in the literature, such as the lack of uniformity in the position, size, and morphology of tumor spheroids14,15,16.
In the protocol presented here, we list each step in the construction of 3D tumor spheroids and describe a method for spheroid observation and analysis using a high-throughput, high-content imaging system that involves auto-focus, auto-imaging, and analysis, among other advantageous characteristics. We show how this method can produce 3D tumor spheroids of uniform size that are suitable for high-throughput imaging. These spheroids also demonstrate a high sensitivity to cancer drug treatment, and morphological changes in the spheroids can be monitored using high-content imaging. In summary, we demonstrate the robustness of this methodology as a means to generate 3D tumor constructs for drug evaluation purposes.
1. Spheroid construction
2. Drug treatment
3. Spheroid viability
4. Spheroid observation and deep learning analysis through images in the drug test
Figure 1A,B shows the process used for constructing tumor spheroids in this study. We first seeded the cells in a 48-well U-bottom plate. This step is almost the same as that used in 2D cell culture. We kept the plate in a common incubator with water surrounding the wells so that the deposited cells started to form spheroids in a self-assembly process. Under normal operational conditions, most types of tumor spheroids were completely formed after 5 days when a targeted medium was used (Supplementary Table 2). We checked the status of spheroid construction within 5 days with a digital microscope (Figure 1D). After the growth process was completed, gel was added to wrap up the individual spheroids in each well, producing an in vivo-like extracellular matrix (ECM) for each spheroid. Drug testing was then carried out. As we used a high-content imaging system with deep-learning analysis software (Figure 1C), we could clearly observe individual spheroid growth (Figure 1E) and obtain values for suitable parameters to determine characteristics during drug therapy, including viability, diameter, and roughness.
Figure 1: Tumor spheroid formation and automated imaging and analysis. (A) Schematic showing the standard tumor spheroid culture procedure with the timeline below. (B) Schematic of drug treatment testing using tumor spheroids showing different levels of invasiveness in 3D matrices. (C) Image of a deep-learning-based algorithm-related system for automated imaging and analysis showing details as follows: plate platform; condenser with light source; motorized x, y stage; motorized z-axis module; objective wheel; filter wheel; CCD; and computer with the developed system software. (D) The formation of NCI-H23 tumor spheroids from monolayer cells observed from the eyepiece of a microscope. The scale bar represents 500 µm. (E) Variation in the invasiveness and size of the NCI-H23 spheroids without drug treatment over 10 days. The scale bars represent 200 µm. Please click here to view a larger version of this figure.
In this study, we used the antitumor drug AMG510 to test its effect on treating non-small-cell lung cancer. The analysis system provided image data without complex image processing. The brightfield images in Figure 2 show that the growth of NCI-H23 spheroids could be inhibited by AMG510. This was indicated by the decreased size along with increased concentration. In contrast, size increased in the control group for 10 days. During this period, the roughness of the edge side, indicating invasiveness, also changed. The spheroids exhibited flat edges on day 1 but rough edges on day 10. However, it should be noted that comparing roughness through brightfield images alone is a challenging task. It should also be mentioned that the standard method involves spheroids of appropriately average size. Finally, we note that a clean background was achieved using this method, as few cells adhered to the bottom.
Figure 2: Brightfield images of NCI-H23 cell spheroids treated with different concentrations of AMG510 automatically captured by a high-content microscope. The columns represent different days, and the rows represent different drug concentrations. The results include three spheroids for each condition. All the images were automatically focused at 2x magnification and then captured at 10x magnification by the high-content imaging system using artificial intelligence-based software and a microscope-associated microenvironment controller. The scale bars represent 200 µm. Please click here to view a larger version of this figure.
The cell viability results, shown in Figure 3A, demonstrated reduced viability in the 10 day cultures of drug-treated samples at all dosage levels when compared to the control. Samples with concentrations over 0.01 µM exhibited a rapid decrease in tumor cell viability, indicating the sensitivity of AMG510 therapy to non-small-cell lung cancer. On day 10, these samples exhibited similar levels of final viability, with values below 70%, as shown in Figure 3B.
Figure 3: Tumor spheroid viability of the AMG510-treated sample groups. Drugs were added on day 1. (A) Tumor spheroid viability was measured on day 1, day 4, day 7, and day 10. (B) Terminal cell viability of the samples with concentration gradients. The drug concentrations used to treat each tumor spheroid were set from 0.001 µM to 5 µM. All data are presented as mean ± SEM (n = 3). Please click here to view a larger version of this figure.
Figure 4A reveals a similar trend with respect to the spheroid diameter analysis. Samples with concentrations above 0.01 µM exhibited an obvious reduction in average diameter. Uniformity in the spheroid size could also be seen on the first day of drug treatment, with values of about 800 µm. The diameter ratios of these spheroids (Figure 4B) indicated contraction during the AMG510 co-culture in a visibly more obvious way than the viability test. In addition, we calculated the spheroid tumor growth inhibition values in terms of their diameters, as shown in Figure 4C. The TGI value is a relative evaluation parameter, and the values could indicate decreased growth as a result of original seeding differences among the spheroids; however, we again obtained the result that concentrations over 0.01 µM had an obvious effect on the success of the AMG510 treatment. In an earlier study, we experimented with the IC50 value, which is a significant index for drug testing; however, we found that NCI-H23 spheroids under AMG510 treatment exhibited no changes in this parameter.
Figure 4: Tumor size represented by spheroid diameters in the control and AMG510-treated sample groups. (A) Tumor spheroid diameters were measured on day 1, day 4, day 7, and day 10. (B) The spheroid growth ratio was defined as the terminal volume relative to the original volume and calculated using the spheroid diameters. (C) The spheroid growth inhibition ratio was defined relative to the volume and calculated using the spheroid diameters. The AMG510 concentrations used to treat each tumor spheroid were set from 0.001 µM to 5 µM. All data are presented as mean ± SEM (n = 3). Please click here to view a larger version of this figure.
We then carried out a further evaluation of the invasiveness of the spheroids in terms of the roughness values produced as output by the software (Figure 5A) and also the excess perimeter index based on the spheroid boundaries, produced by the same means (Figure 5B). Higher roughness means more cells are migrating from the spheroid to the gel, which, thus, represents higher invasiveness. The results of controls were compared with different drug concentration levels. For NCI-H23, an invasive cell line, the spheroid roughness and EPI values both increased with culture time without any drug treatment18, and this could be proven by the brightfield images and diameter increases in the control samples. However, growth was attenuated by AMG510 treatment, wholly in proportion with variations in size.
Figure 5: Tumor boundary recognition in the untreated control and AMG510-treated sample groups. Drugs were added on day 1. (A) Tumor roughness was measured by the software on day 1, day 4, day 7, and day 10, indicating the invasiveness of the tumor spheroids. (B) The spheroid perimeter was located and drawn by the software using deep-learning algorithms. The spheroid areas at the focal plane were then measured by Image J, and an excess perimeter index was calculated based on these data. The AMG510 concentrations used to treat each tumor spheroid were set from 0.001 µM to 5 µM. All data are presented as mean ± SEM (n = 3). Please click here to view a larger version of this figure.
Supplemental Table 1: The composition of the plate. The anti-lung-tumor drug AMG510 was used in the experiment, and groups were set according to drug gradients. Two plates were used: one for the viability assay kit test and one for the imaging analysis.
Supplemental Table 2: Cell lines that have been used in 3D tumor construction and drug testing. Please click here to download these Tables.
The microenvironment plays an important role in tumor growth. It may affect the provision of extracellular matrices, oxygen gradients, nutrition, and mechanical interaction and, thus, affect gene expression, signal pathways, and many functions of tumor cells19,20,21. In many cases, 2D cells do not produce such effects or even produce opposite effects, thus affecting the evaluation of drug treatments. However, the emergence of 3D models has addressed this problem. For instance, our evaluation of the effect of AMG510 on a lung cancer treatment system was completed using 3D methods. We constructed a 3D tumor-spheroid-ECM (TSE) model by creating tumor spheroids embedded within a gel to monitor AMG510 treatment on a lung-related invasive cell line. AMG510 is a new inhibitor that targets G12C-mutant KRAS22, and its effectiveness was confirmed by our finding that the non-small-cell lung cancer (NSCLC) experienced by some patients responding to this mutant. From our series of experimental data and analysis, some very valuable data can be discerned. Sensitivity to AMG510 was significantly enhanced under a 3D spheroid condition, compared with the 2D condition, and improved still further within a hypoxia condition.
The models constructed through the present method offer several advantages. Firstly, as a 3D model, it can simulate the tumor and the extracellular matrix well. The simple but effective method of constructing 3D tumor spheroids allows the cells to survive for a 2 week period. After the 5 days of spheroid construction, a period of around 10 days is available, which is sufficient to carry out any drug testing. In addition, the method of 3D cell culture production is almost the same as the 2D method, and its cost is far lower than that of 3D bioprinting. Compared with other methods such as the hanging-drop method, the method described here helps to produce more uniform spheroids, and the cell size is fully controlled by the type and density of cells. Finally, this model has been shown to be applicable to various types of cancer cells, and this flexibility may prove to be of especially high value. In the future, researchers engaged in the formation of tumor spheroids might add not only fibroblasts and endothelial cells to their tumor cell suspensions but also immune cells such as T cells and NK cells to promote stromal cell migration and boost the immunotherapeutic effects of new drugs23,24.
In recent decades, few reports have focused on describing the techniques and tools currently available to extract significant biological data from 3D models25. This information can be considered fundamental to drug testing and therapeutic discovery using 3D cell culture models26. The study of Zanoni et al. described novel open-source software that carries out automatic image analysis of 3D tumor colonies. The authors showed that morphology parameters influenced the response of large spheroids to drug treatment and that spheroid size and shape could both be sources of variability27. However, with tumor spheroids cultured in 3D gel, which mimic ECMs in vivo, cell behavior is much more complex, and invasiveness may be exhibited. While the data are only concerned with size and shape, such limitations will remain. Many reliable and diversified parameters can be obtained using a high-content imaging system. Such systems can not only determine the cell viability and spheroid shape but can also detect and verify the characteristics of tumors and the inhibition effect of drugs upon tumors. Invasiveness can also be determined by such means. For example, an invasive tumor may have a relatively higher roughness value, while a noninvasive tumor may have a higher EPI value. Changes in such values can indicate different drug effects. In the future, we hope to combine these techniques with microfluidic chips to obtain more convenient and effective strategies.
The standardized method described here can meet the requirements of most drug screening and evaluation tests. However, there are still some problems that may occur during future experiments. For example, cells may adhere to the inner walls of the wells after centrifuging, or the culture medium may not be suitable. In addition, the gel used in the protocol, as well as the widely used Matrigel, can easily gelate during the long process. However, most of these problems can be solved through standard operation and optimization techniques. There are other possible limitations in this method. Due to the spheroid structure, nutrients, oxygen, and waste diffusion through the spheroid are influenced by the size. The viability of the cells in the spheroid core may be compromised when the spheroid diameter is over 1,000 µm. Conversely, it becomes hard to observe the invasion when the diameter is below 400 µm.
The authors have nothing to disclose.
We thank all the members of our laboratories for their critical input and suggestions. This research was supported by the Key Project of Jiangsu Commission of Health (K2019030). Conceptualization was conducted by C.W. and Z.C., the methodology was performed by W.H. and M.L., the investigation was performed by W.H. and M.L., the data curation was performed by W.H., Z.Z., S.X., and M.L., the original draft preparation was performed by Z.Z., J.Z., S.X., W.H., and X.L., the review and editing was performed by Z.C., project administration was performed by C.W. and Z.C., and funding acquisition was conducted by C.W. All the authors have read and agreed to the published version of the manuscript.
0.5-10 μL Pipette tips | AXYGEN | T-300 | |
1.5 mL Boil proof microtubes | Axygen | MCT-150-C | |
100-1000μL Pipette tips | KIRGEN | KG1313 | |
15 mL Centrifuge Tube | Nest | 601052 | |
200 μL Pipette tips | AXYGEN | T-200-Y | |
3D gel | Avatarget | MA02 | |
48-well U bottom Plate | Avatarget | P02-48UWP | |
50 mL Centrifuge Tube | Nest | 602052 | |
Alamar Blue | Thermo | DAL1100 | |
Anti-Adherence Rinsing Solution | STEMCELL | #07010 | |
Certified FBS | BI | 04-001-1ACS | |
Deionized water | aladdin | W433884-500ml | |
DMEM (Dulbecco's Modified Eagle Medium) | Gibco | 11965-092 | |
DMSO | sigma | D2650-100ML | |
Excel sofware | Microsoft office | ||
Graphpad prism sofware | GraphPad software | ||
High Content Microscope and SMART system | Avatarget | 1-I01 | |
Image J software | National Institutes of Health | ||
Insulin-Transferrin-Selenium-A Supplement (100X) | Gibco | 51300-044 | |
Parafilm | Bemis | PM-996 | |
PBS | Solarbio | P1020 | |
Penicillin/streptomycin Sol | Gibco | 15140-122 | |
RPMI 1640 | Gibco | 11875-093 | |
Scientific Fluoroskan Ascent | Thermo | Fluoroskan Ascent | |
T25 Flask | JET Biofil | TCF012050 | |
Trypsin, 0.25% (1X) | Hyclone | SH30042.01 |