Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2019 Jun 1.
Published in final edited form as: SLAS Technol. 2018 Feb 7;23(3):231–242. doi: 10.1177/2472630318756058

Quantitative Live-cell Confocal Imaging of 3D Spheroids in a High Throughput Format

Elizabeth Leary 1,2, Claire Rhee 1, Benjamin T Wilks 1,2, Jeffrey R Morgan 1,2,*
PMCID: PMC5962438  NIHMSID: NIHMS966978  PMID: 29412762

Abstract

Accurately predicting the human response to new compounds is critical to a wide variety of industries. Standard screening pipelines (including both in vitro and in vivo models) often lack predictive power. Three-dimensional culture systems of human cells, a more physiologically relevant platform, could provide a high-throughput, automated means to test the efficacy and/or toxicity of novel substances. However, the challenge of obtaining high magnification, confocal z-stacks of 3D spheroids and understanding its respective quantitative limitations must be overcome first. To address this challenge, we developed a method to form spheroids of reproducible size at precise spatial locations across a 96-well plate. Spheroids of variable radii were labeled with four different fluorescent dyes and imaged with a high-throughput confocal microscope. 3D renderings of the spheroid had a complex bowl-like appearance. We systematically analyzed these confocal z-stacks to determine the depth of imaging, the effect of spheroid size and dyes on quantitation. Furthermore, we have shown that this loss of fluorescence can be addressed through the use of ratio imaging. Overall, understanding both the limitations of confocal imaging as well as the tools to correct for these limits are critical for developing accurate quantitative assays using 3D spheroids.

Keywords: HTS High Throughput Screening, Data Analysis, Fabrication

INTRODUCTION

The ability to accurately model the complex responses of human tissues and organs in a laboratory setting is crucial for many fields including drug discovery and toxicity testing1–5. However, the current methods and models, which include both in vitro assays and in vivo animal testing, are often inadequate in their ability to predict how a compound will interact within the human body1,3. In vitro assays, which predominantly utilize two-dimensional (2D) cell monolayers, are relatively inexpensive, easy to perform, and have the potential to utilize human cells from various sources. However, 2D cell monolayers are limited due to significant differences with the in vivo environment including low cell densities, a predominance of cell-plastic interactions, a lack of diffusion gradients and, for certain cell types, dedifferentiation and a loss of organ-specific functions6,7,8. Alternatively, while animal models are able to recapitulate the complexity of the in vivo environment, they are expensive and time-consuming, thus limiting their usefulness to the evaluation of only very small numbers of compounds. Furthermore, animal models, due to species differences, often fail to predict the human response9,10. Three-dimensional (3D) multi-cellular spheroids have been proposed as an intermediate test bed1,2,5,10. Compared to animal testing, spheroids are inexpensive, easy-to-use, and have the potential to use human cells1,2,3. Furthermore, spheroids better approximate the cell density found in vivo, have increased levels of cell-cell interactions and intercellular communication, as well as maintenance of the differentiated state, and recapitulation of organ-specific processes6,7,8,11–14.

Despite the biological advantages of spheroids, there are numerous other issues that must be overcome to utilize spheroids as a high-throughput screening tool. For example, imaging of thin 2D monolayers is straightforward, whereas spheroids require more complex techniques to visualize the additional biology occurring within these relatively thick microtissues2,15,16. Histological techniques, such as cryo- and plastic sectioning, work well, but are time-consuming, low throughput, not appropriate for screening large number of spheroids, and not amenable to imaging live cells15. Alternatively, confocal and two photon microscopy optically section spheroids and can visualize living cells17–18. Furthermore, confocal microscopy has been configured as a high throughput instrument for purposes of high content imaging of large numbers of samples. However, due to light scattering, confocal microscopy has inherent limitations with respect to imaging thick tissues15,18. To overcome this limitation, tissue clearing methods that remove scattering substances and/or provide refractive index matching have been developed18–21. Although the depth of imaging is significantly improved, clearing can only be used on fixed non-living cells as part of an endpoint assay18–21. Endpoint assays are important, but not all assays can be performed as single point assays. Live-cell imaging enables users to examine dynamic changes over time, and gather more information.

To evaluate the usefulness of spheroids as a high-throughput screening tool, we developed a method to form spheroids directly within a 96-well plate, as well as examine the quantitative limitations associated with live-cell confocal imaging of spheroids. Through molding agarose hydrogels directly within a 96-well plate, we formed four spheroids of reproducible size that are located at precise x, y, z-locations across the entire plate. By varying the number of cells seeded per well, the size of the spheroids could be varied ranging 50 – 200 μm in diameter. To determine the depth of confocal imaging, spheroids were formed from cells that were uniformly labeled with four fluorescent dyes each with different excitation/emission wavelengths. Confocal z stacks were acquired and analyzed to evaluate the total and positional fluorescent signal for spheroids of variable sizes. As spheroid radii increased, the overall proportion of fluorescent signal retention was diminished. Furthermore, the fluorescent signal decreased as a function of z depth into spheroids of all sizes. Interestingly, loss of fluorescence across each x, y confocal slice of the spheroid was not uniform. Due to the curved nature of the spheroid, loss of fluorescence was least around the outer edges of the spheroid and greatest in the center of the spheroid, resulting in a bowl-like appearance in a 3D rendering of the spheroid. Unlike solid tissue blocks, the irregular spheroidal shape and subsequent bowl-like appearance makes it difficult to accurately model, quantitate, and correct the fluorescence throughout a 3D microtissue. However, we have shown that loss of the fluorescent signal can be addressed by ratio imaging. Overall, understanding and accounting for the fluorescent changes throughout confocal z-stacks of variable sized spheroids is critical for designing quantitative biological assays using spheroids.

MATERIALS AND METHODS

Micro-mold fabrication and hydrogel formation

Molds, designed using computer-assisted design (CAD) (Solidworks, Concord, MA), consisted of two main components: a base platform upon which lay a series of 4 rows by 8 columns of pegs (Figure 1A). The design of the peg consisted of three main components. Directly upon the base platform lay a cylinder shape designed to fit tightly within the inner diameter of a well, thus preventing shifting in the x, y direction. This cylinder was then funneled into a flat-top conical structure ending with a 2 row by 2 column array of conical-shaped micro-posts each positioned at the center of an imaging grid of the Opera Phenix. Molds were 3D printed (Phenomyx, LLC, Cambridge, MA).

Figure 1. Molding system to form micro-wells for four spheroids per well in a 96-well plate.

Figure 1

Open in a new tab

Molds designed in CAD, consisting of an array of 4 rows by 8 columns of pegs, with each peg containing 4 micro-posts, were 3D printed (A). To create hydrogels, molten agarose was pipetted into each well of a 96-well plate, and the mold placed atop. After the agarose gelled, the mold was removed, leaving a hydrogel containing a square loading dock atop of 4 micro-recesses for spheroid formation (B).

To form hydrogels, 90-μ L of sterile molten UltraPure Agarose (Fisher Scientific, Waltham, MA) (2% weight/volume in phosphate buffered saline) was pipetted into each well and the mold was placed on top of the plate with the micro-posts submerged in agarose. After 10-minutes, the molten agarose solution solidified and the mold was removed. The resulting hydrogel that formed within each well contained a round loading dock above 4 micro-recesses (Figure 1B). To equilibrate the hydrogels, 150-μ L of serum-free DMEM (Life Technologies, Grand Island, NY) supplemented with 1% penicillin/streptomycin was added to each well, and incubated for 24 hours at 37°C with 10% CO2.

Cell culture, fluorescent dye labeling, and spheroid formation

Human ovarian granulosa (KGN) cells were grown in DMEM with 10% fetal bovine serum (FBS) (Fisher Scientific, Waltham, MA) and 1% penicillin/streptomycin at 37°C with 10% CO2. Once confluent, cell monolayers were labeled by first removing serum-containing medium from the culture flasks. Fluorescent dyes were reconstituted in serum-free DMEM, and incubated with the cell monolayer for 30-minutes at 37°C with 10% CO2. Four dyes were used to stain monolayers: 2μM CellTracker™ Red CMPTX (CTR) (Life Technologies, Grand Island, NY), 2μM CellTracker™ Green CMFDA (CTG) (Life Technologies, Grand Island, NY), 2μM CellTracker™ Violet (CTV) (Life Technologies, Grand Island, NY), and 2μM CellTracker™ Deep Red (CTDR) (Life Technologies, Grand Island, NY). After labeling, medium was exchanged with fresh serum-free DMEM, and incubated for 15-minutes at 37°C with 10% CO2. The labeled cell monolayers were harvested using 0.05% trypsin, concentrated by centrifugation at 120 x g for 6-minutes, and counted. Cells were washed once with serum-free DMEM and spun down at 120 x g for 6-minutes. Cells were re-suspended in serum-free DMEM at one of the following concentrations: 50,000-, 100,000-, 150,000-, 200,000-, 250,000-, 300,000-, 400,000-, 500,000-cells/mL A 20-μL aliquot of cell suspension was pipetted into the loading dock of each hydrogel to form four spheroids each composed of 250-, 500-, 750-, 1,000-, 1,250-, 1,500-, 2,000-, or 2.500-cells. After allowing the cell suspension to settle to the bottoms of the micro-recesses for 30-minutes, 150-μL of serum-free DMEM was added per well. Cells were allowed to self-assemble into spheroids for twenty-four hours prior to imaging.

Microscopy and image analysis measurements

To image spheroids, the Opera Phenix™ High Content Screening System (Perkin Elmer, Waltham, MA, USA), an inverted confocal microscope equipped with proprietary Synchrony™ Optics consisting of a Nipkow spinning microlens disk in conjunction with a pinhole disk and 2 sCMOS cameras, was used. Fluorescent images for each dye were acquired using the 20x water objective in conjunction with 4 excitation lasers: 405-nm for CTV, 488-nm for CTG, 561-nm for CTR, and 640-nm for CTDR. Confocal z-slices of spheroids were acquired every 5-μm for a total of 500-μm.

Confocal z-slice images were analyzed via two different image analysis software programs: Imaris (Bitplane, Belfast, UK) and ImageJ (National Institutes of Health, Bethseda, MD, USA). With Imaris, confocal slices were rendered as 3D objects, and a mask of the outer spheroid surface was created. Total volume was measured for each spheroid, and its average radii was computed by the following equation: r=34Vol.π3. ImageJ was used to measure the following parameters at every confocal slice: cross-sectional area, total fluorescence of cross sectional area, average fluorescent intensity across the radii.

Data analysis and derivation of ideal curves

To evaluate fluorescence, and its subsequent loss due to imaging limitations both the measured and maximum hypothetical fluorescence were computed per slice. The measured fluorescence for each slice was plotted as a function of its z-depth, and the area under the curve (AUC) was calculated (Supplemental Figure 1B, D). To calculate the hypothetical fluorescence, the following assumption was made: since all cells of the spheroid were evenly stained as a cell monolayer prior to spheroid formation, the signal was assumed to be homogenous throughout. Therefore, cross-sectional area should increase linearly with respect to fluorescence for each slice up until the equator. To determine the hypothetical signal, fluorescence was plotted as a function of cross-sectional area, and the best-fit line for the linear region of the curve was determined to be “y = mx + b” (Supplemental Figure 1A). The hypothetical values for each slice were computed by the following equation “Flhyp = m (CSA) + b” (Supplemental Figure 1B). The hypothetical fluorescence was plotted as a function of its z-depth, and the AUC was calculated (Supplemental Figure 1C). The total loss of fluorescence for each spheroid was computed as the difference between the hypothetical and measured AUC normalized by the hypothetical AUC (Supplemental Figure 1E). Spheroids were binned based upon size into 10-μm bins, and the average and standard deviation for each dye was computed.

To evaluate alterations of fluorescent signal within a spheroid, the center of the spheroid was identified, then on each confocal slice a series of concentric rings was propagated outwards from the center point (Supplemental Figure 5C). The fluorescent values around each concentric circle were then averaged to yield the average fluorescent signal across the cross-sectional radii (x, y) (Supplemental Figure 5C). The average fluorescent signal was then plotted as a function of its cross-sectional radii for each confocal slice throughout the spheroid.

To evaluate the fluorescent loss throughout the z-dimension, the center of the spheroid was identified, and the average fluorescent intensity was computed along that central z-axis for each slice (Figure 5B). The average intensity of each point was normalized by the maximum fluorescent intensity along the central z-axis for each spheroid. This normalized average intensity was plotted as a function of z-depth. The best-fit exponential decay function was calculated from 50-μ m past the peak signal for every spheroid, according to the following equation: “y = Cemx”, where “y” is normalized average fluorescence, “x” is z-depth, “m” represents the rate of loss. Spheroids were binned based upon 10-μm sized bins, and the average and standard deviation of the slope of the exponential decay function each dye was computed and plotted as a function of spheroid radii.

Figure 5. Fluorescent signal loss throughout the z-depth of spheroids exhibits a reproducible, exponential decay function.

Figure 5

Open in a new tab

KGN cell monolayers were labeled with CTG, trypsinized, counted, and seeded at varying cell numbers to form spheroids with diameters ranging from 50-μm to 200-μm. After 24-hours, confocal images were acquired and cross-sectional images (x, z) were reconstructed (A,B). Fluorescent signal loss results in a reproducible “bowl”-like pattern throughout spheroids, where brightest signal occurred along outer spheroid edge up until the equator, where the signal was dimmer in central portions (B). To assess the distance into live spheroids that could be imaged, the average fluorescent intensity was measured along that central z-axis for each confocal slice (B, yellow line). For each spheroid, the average intensity was normalized by peak fluorescent signal, plotted as a function of its z-depth, and the best-fit exponential decay curve was calculated from 50-μ m past the peak signal (B, red line). Representative graphs for spheroids with a 50-μm (C), 100-μm (D), and 150-μm (E) diameters are shown.

To evaluate the ability of ratio imaging to mitigate fluorescent loss throughout the z-depth, the average fluorescence along the central z-axis was measured for spheroids pre-labeled with both CTG and CTDR as 2D monolayers. To perform ratio imaging, the CTG fluorescence was divided by the CTDR fluorescence at each z-slice and vice versa. Both the average and normalized fluorescence were plotted as a function of spheroid z-depth. The average and standard deviation of the fluorescence throughout the z-depth was computed for both pre-and post-normalization. To compare the effect of ratio imaging, a coefficient of variation (CV) analysis was performed for both pre- and post-normalization of the fluorescence throughout the z-depth. Lower CV values indicate less variation throughout the z-depth.

RESULTS

Tri-axis positional control of spheroid formation in a 96-well plate for confocal imaging

To increase the throughput of spheroid formation, we designed a mold insert compatible with a 96-well plate. The micromold consisted of a series of pegs, where each peg was designed to fit within the dimensions of one well (Figure 1A). Furthermore, atop of each peg, were four conical shaped micro-posts, thus enabling the formation of four technical replicates per well. To minimize the imaging area and the number of confocal slices, the micro-pegs and were designed to fall within the same spatial location in every well. Therefore, after adding agarose into 96-well plate, the resulting hydrogel consisted of four micro-wells with the same x, y, z locations in every well (Figure 1B).

To evaluate the accuracy of the mold, spheroid formation was measured with respect to two different assessments: (1) the x, y, z location in each well, and (2) spheroid radii. Monolayers of KGN cells were labeled with CTR, trypsinized, counted, and seeded into hydrogels to form spheroids (~750 cells). After 24-hours, confocal slices were acquired using the 20X water objective and analyzed to measure x, y, z- position as well as spheroid radii. Spheroids formed in specific locations within the x, y- bounds of the 646- by 646-μ m imaging grid system of the microscope (Figure 2A). Furthermore, spheroids formed within a 200 μ m range of z-locations, approximately centered 650 μ m from the bottom of plate (Figure 2B). To assess reproducibility of size, we seeded plates with either the same or varying concentrations of monodispersed cells. When seeding ~750 cells per spheroid across the entire plate, the average radii was 54.2-μm +/− 5.2μm, thus deviation was approximately 10% of the mean (Figure 2C). Furthermore, when seeding varying concentrations, (~250- to ~2,000- cells/spheroid), the higher seeding densities yielded larger spheroids (Figure 2D). Regardless of the seeding density, the standard deviation was approximately 10% of its mean (Figure 2D, E).

Figure 2. Spheroids of consistent size are formed within specific x, y, z- locations.

Figure 2

Open in a new tab

Molds were used to produce agarose hydrogels directly into the wells of a 96-well plate. Monodispersed cells were seeded into the hydrogels at either i.) a single seeding density for the entire plate (A, B, C, E), or ii.) multiple seeding densities (D, E). After 24 hours of self-assembly, confocal images of the spheroids were obtained using the 20X water objective and analyzed. All spheroids were located within the bounds of one of the 646- × 646-μm (x, y)-imaging grids used by the 20X objective (A). Spheroids were located within approximately a 200-μm range of z-height, centered approximately 650-μm from the bottom of the plate (B). When using a single cell seeding density to form spheroids, the average spheroid radii was 54.2 μm +/− 5.2 μm standard deviation (C, E). Alternatively when using multiple seeding densities, spheroid size increased in a dose-dependent manner with higher seeding densities yielding larger spheroids (D, E). Regardless of seeding density used, the standard deviation accounted for approximately 10% of the average (C, D, E).

Cumulative fluorescent loss increases as function of spheroid radii

To assess the loss of fluorescence associated with live-cell confocal imaging of spheroids, we imaged spheroids over a range of sizes and tested four different fluorescent dyes. To form uniformly labeled spheroids, KGN cell monolayers were first labeled with four different fluorescent dyes: CellTracker™ Violet (CTV), CellTracker™ Green CMFDA (CTG), CellTracker™ Red (CTR), CellTracker™ Deep Red (CTDR). After labeling, monolayers were trypsinized, counted, seeded at various cell densities to form spheroids with diameters ranging from 50μ m – 200μ m. After self-assemlbing for 24-hours, confocal slices for each spheroid were obtained (20X water objective). Representative images for a 50-μ m, 100-μ m, and 150-μ m diameter spheroids were reconstructed as orthogonal contrasts for the four dyes tested (Figure 3, Supplemental Figures 2, 3, 4). Since the cells were labeled as a monolayer prior to spheroid formation, hypothetically every cell in the spheroid should be evenly stained. Assuming there was no loss in fluorescence due to imaging limitations, the resulting spheroids should possess a homogenous signal throughout the entire spheroid. Qualitatively, for the four dyes tested, the signal was relatively homogenous throughout small spheroids (50-μ m diameter). However, as radii increased, the signal decreased for confocal slices located deeper into the spheroid (Figure 3, Supplemental Figures 2, 3, 4).

Figure 3. Loss of fluorescent signal occurs deeper within the z-depth for spheroids.

Figure 3

Open in a new tab

KGN cell monolayers were labeled with CTG, trypsinized, counted, and seeded at varying cell numbers to form spheroids with diameters ranging from 50-μm to 200-μm. Confocal images were acquired, and orthogonal contrast images of representative spheroids of 50-μm (A), 100-μm (B), and 150-μm (C) diameters were rendered (left column). For each spheroid, the measured fluorescence at each confocal slice was plotted as a function of its z-depth (white squares). The data are plotted versus the hypothetical fluorescence for each confocal slice, which was calculated assuming no loss of fluorescent signal (black squares). Representative graphs for spheroids with a 50-μm (A), 100-μm (B), and 150-μm (C) diameters are shown (right column). For small spheroids (50-μm), the plot of measured and hypothetical fluorescence were mostly overlapping throughout the z-depth (A, right column) As spheroids increased in radii, the plots of measured fluorescence fell short of hypothetical fluorescence as z-depth into spheroid increased (B, C right column). (scale bar is 40-μm).

To understand this loss of fluorescence with increasing depth into larger spheroids, we quantified the total fluorescent signal loss for each spheroid and compared populations of spheroids of varying sizes. The actual and hypothetical fluorescence at every confocal slice was measured and plotted as a function of its z-depth (Figure 3, Supplemental Figures 2, 3, 4). Assuming no loss, the plot of actual fluorescence should match the plot of hypothetical fluorescence. For small spheroids (diameter of 50-μ m), the plots for measured and hypothetical fluorescence were overlapping, indicating minimal loss for each of the following dyes: CTR, CTG, CTV, CTDR (Figure 3A, Supplemental Figures 2A, 3A, 4A). For medium and larger sized spheroids (diameters of 100-μ m and 150-μ m), the plots for measured and hypothetical fluorescence overlapped only through the lower portion of the z-depth. Beyond which, the plot of measured fluorescence falls short of the plot of hypothetical fluorescence (Figure 3B, C, Supplemental Figures 2B, C, 3B, C, 4B, C). Since all cells were homogenously labeled, the extent to which the plot of the measured fluorescence falls short of the plot of the hypothetical fluorescence is the loss due to imaging thick tissues.

To compare the effect of spheroid size on imaging, we quantified the difference between actual and hypothetical fluorescence for all dyes. The area under the curve (AUC) was calculated for actual and hypothetical fluorescence as a function of z-depth, and the percent loss was computed (Supplemental Figure 1C, D, E). The percent loss was binned as a function of radii, then plotted for each of the dyes tested (Figure 4). The percent loss of fluorescence increased as spheroid radii increased (Figure 4).

Figure 4. Cumulative loss of total spheroid fluorescence increases as a function of spheroid radii.

Figure 4

Open in a new tab

KGN cell monolayers were labeled with four fluorescent dyes: CTR, CTV, CTG, CTDR. After staining, monolayers were trypsinized, counted, and seeded at varying cell numbers to form spheroids of range of sizes. Confocal images were analyzed to determine the actual fluorescence per confocal slice and hypothetical fluorescence per slice, and both were plotted as a function of z-depth. By calculating the area under the curves for both actual and hypothetical fluorescence, the cumulative fluorescent loss of each spheroid was determined. Spheroids were binned into 10-μm bins based on radii. The average percent loss of fluorescent signal was calculated for each range of radii and plotted as a function of radii for each of following fluorescent dyes: CTG (A), CTV (B), CTR (C), and CTDR (D). For the four dyes tested, as radii increased, cumulative fluorescent loss also increased.

Within spheroids, loss of the fluorescent signal is dependent upon the positional (x, y, z) location

To assess where signal loss was occurring throughout the spheroid, we imaged spheroids over a range of radii and tested different fluorescent dyes. KGN cell monolayers were homogenously labeled with four different dyes (CTG, CTV, CTR, CTDR), trypsinzed, and seeded at varying cell densities to form spheroids with diameters ranging from 50μ m – 200μ m. After 24-hours, z-stacks were acquired, and the center point of each spheroid was identified. Representative confocal slices (x, y) were assessed at 30°north of, 30°south of, 60°south of, and the equator of a spheroid (~100μ m diameter) stained with CTG (Supplemental Figure 5B). For confocal slices south of the spheroid equator (30° and 60°south), the fluorescent signal was relatively homogenous and bright across the entire slice (Supplemental Figure 5B). However, deeper into the z-depth (equator and 30°north), the fluorescent signal began to decrease preferentially in the center of each slice, whereas the outer edge of the spheroid retained a brighter signal (Supplemental Figure 5B). To quantify this pattern, the average fluorescence across the cross-sectional (x, y) radii was computed by averaging individual fluorescent values from a series of concentric rings (Supplemental Figure 5C). The average fluorescent signal was plotted as a function of its cross-sectional radii for the confocal slices 30° and 60°north of, 30° and 60°south of, and the equator for spheroids of variable radii (Supplemental Figure 5C, D, E). For smaller spheroids (~50μ m diameter), the average fluorescent intensity across the cross-sectional radii was relatively constant across the confocal slice up until the equator, above which there was a decline in signal (Supplemental Figure 5C). For larger spheroids (~150μ m diameter), fluorescent signal was relatively homogenous throughout the confocal slice corresponding to 30°south, however, fluorescence of the confocal slices deeper into the spheroid had a measurable decrease in the center as compared to the spheroid edge (Supplemental Figure 5E). Similar trends were observed for all dyes tested (data not shown). This pattern of fluorescent signal yields a “bowl-like” appearance in 3D renderings, with the brightest signal occurring along the bottom hemispheric edge of the spheroid (Figure 5B).

To compare the effect of spheroid size on this fluorescent loss throughout a spheroid, we first quantified the distance into live spheroids that could be imaged along the central z-axis (Figure 5B). For each spheroid and dye, the average intensity was first normalized by the maximum fluorescent signal along the central axis. The normalized signal was plotted as a function of its z-depth and the best-fit exponential decay curve was calculated from 50-μ m past the peak signal (Figure 5C, D, E). Fluorescent dyes exhibited different staining patterns. For example, CTG staining was a homogenous uniform signal throughout the spheroid, whereas CTR staining was a more randomized punctate signal (Figure 3 left column, Supplemental Figure 3 left column). Dyes with a homogenous signal, also exhibited a reproducible exponential decline in fluorescent signal throughout the spheroid (Figure 5). Alternatively, dyes with a punctate signal exhibited a more variable sawtooth-like decrease in signal throughout (Supplemental Figure 6B). For all dyes, best-fit lines containing correlation values lower than 0.9 were excluded from further analysis.

To compare the effect of size on depth of imaging, we compared the slope of the exponential decay function from spheroids of variable radii. Slope was binned as a function of radii, then plotted for each dye (Figure 6). For the four dyes tested, the slope of the exponential decay function increased as radii increased, indicating that smaller spheroids possessed a steeper decline in signal throughout the z-depth when compared to larger spheroids (Figure 6).

Figure 6. Cumulative loss of total spheroid fluorescence increases as a function of spheroid size.

Figure 6

Open in a new tab

KGN cell monolayers were labeled with four fluorescent dyes: CTR, CTV, CTG, CTDR. After staining, monolayers were trypsinized, counted, and seeded at varying cell numbers to form spheroids of range of sizes. Confocal images were analyzed to determine the average fluorescent signal along the central z-axis, which was then normalized to peak fluorescence, and plotted as a function of z-depth for every spheroid. The best-fit exponential decay was calculated from 50-μ m past the peak signal. To evaluate the effect of spheroid size on the rate of loss, the slope of the exponential decay function was compared. Spheroids were binned into 10-μm bins based on radii. The average slope was calculated for each range of radii and plotted as a function of radii for each of following fluorescent dyes: CTG (A), CTV (B), CTR (C), and CTDR (D). For all dyes tested, the slope of the exponential decay function increased as a function of spheroid radii, which implied that the rate of signal loss was greater for small spheroids versus larger spheroids despite being the same cells and labeled with the same dye.

Ratio imaging corrects for loss of fluorescence throughout the z-depth

To determine if ratio imaging could be used to correct for loss throughout the z-depth, KGN cell monolayers were uniformly labeled with both CTR and CTDR, trypsinized, and seeded at varying cell densities to form spheroids of variable radii. After 24-hours, confocal z-stacks were acquired and fluorescence along the central z-axis was measured (Supplemental Figure 5B). Fluorescence of both CTG and CTDR were plotted as a function of z-depth (Figure 7A, B). Since the spheroid was uniformly with both dyes, fluorescence should hypothetically be the same throughout the z-depth, however, the signal of both dyes declined as a function of z-depth (Figure 7A, B). To normalize the attenuation, CTG fluorescence was divided by CTDR fluorescence at each confocal slice, and vice versa. This normalized signal was then plotted as a function of z-depth (Figure 7C, D). Once normalized, both dyes exhibit a more uniform signal throughout the z-depth (Figure 7C, D).

Figure 7. Ratio imaging reduces variation throughout the z-depth.

Figure 7

Open in a new tab

KGN cell monolayers were labeled with both CTG and CTDR, trypsinized, counted, and seeded at varying numbers to form spheroids. After 24-hours, confocal images were acquired and analyzed to measure the fluorescent signal along the central z-axis. Data from two representative spheroids are shown (A, C − 50μ m radii; B, D − 75μ m radii). The signal for both CTG (green) and CTDR (red) was plotted as a function of z-depth (A, B). To perform ratio imaging, the signal of each dye was divided by the other dye at each slice, and plotted as a function of its z-depth (C, D). To evaluate the effect of ratio imaging, the average and standard deviation of fluorescent signal along the central z-axis was computed. A CV analysis was performed for each spheroid for each dye both pre- and post-normalization (E). Lower CV values indicate less variation throughout the z-depth (E).

To confirm quantitatively that ratio imaging yielded a more uniform signal throughout, we performed a coefficient of variation (CV) analysis, according to the following equation: CV=st.dev.average. For every spheroid and each dye, the average and standard deviation of fluorescence throughout the entire z-depth was computed both pre- and post-normalization. Hypothetically, assuming there was no loss, fluorescence should be homogenously labeled throughout the z-depth, thus possessing a low standard deviation relative to the average. However, without normalization, fluorescence decreases throughout the z-depth for both CTG and CTDR, resulting in a CV of approximately 0.5 (Figure 7E). However, after normalization of the CTG signal to the CTDR signal and vice versa, the CV value was reduced to approximately 0.12 showing that variation in fluorescence throughout the z-depth can be reduced by ratio imaging.

DISCUSSION

Needed for drug discovery and toxicity testing are new, more predictive high throughput in vitro assays that more accurately mimic the biological complexity of human organs and tissues1,2,5,22. Multi-cellular spheroids are emerging as a possible solution, but challenges still remain with respect to work flow, spheroid formation, long term stability of spheroids and the high magnification imaging of spheroids necessary to acquire the quantitative high content information that spheroids can provide1,2,5,10,23. In addition to improved in vivo-like differentiation and function, much of this high content information is due to the fact that, like native tissues, spheroids have multiple cell layers that create gradients (e.g., oxygen, nutrients, metabolites) and barriers (e.g., drug transport, signaling)4,6–8,24. Biological activities in these microenvironments can be probed using a wide variety of fluorescent dyes along with confocal microscopy. But, to assess these activities as a function of radial position within a spheroid, it is critical that the images are high magnification to get accurate quantitative depictions.

In this paper, we aimed to increase assay throughput by designing and evaluating a mold system that formed spheroids directly within a 96-well plate, and then imaging the plate using a 20X water objective of a high-throughput confocal microscope. We acquired z stacks of living spheroids that were uniformly labeled with four different fluorescent dyes and quantified the loss of fluorescent signal as we imaged deeper into each spheroid. Unlike planar specimens where signal loss is a relatively simple function of z-depth, the curved nature of spheroids creates a more complex pattern. In 3D renderings, spheroids appear as bowl-like with a bright signal around the edges and a dampened signal in the interior. To better understand fluorescence signal distribution prior to developing biologically-driven assays, we quantified fluorescence of four dyes as a function of overall signal from entire spheroid, the rate of decrease throughout the z-depth, as well as the influence of spheroid size on these stacks of images. Furthermore, we used ratio imaging to demonstrate that loss throughout the z-depth can be compensated.

Prior to assessing fluorescence signal, we first designed and evaluated the mold’s ability to form spheroids of reproducible size at precise locations. We molded four micro-wells into agarose in each well of a 96 well plate, thus forming four replicate spheroids per well that were placed at precise x, y and z locations within each well. The x, y locations corresponded to the centers of the four boxes in the imaging grid used by the 20X water objective. These x, y locations improve work flow by obviating the need to pre-scan a plate to find spheroid locations and/or having to stitch together adjacent images of spheroids that span grid lines. Placement of spheroids in a precise z location also improves work-flow since spheroids can be found at a predictable distance from the plane of initial focus. This fixed position reduces the number of confocal z slices needed to reliably image the entire spheroid. Moreover, the autofocus function can be used with our design because, unlike round bottom low attachment plates typically used to form spheroids in a high-throughput format, our mold is compatible with flat bottom well plates. Although agarose micro-wells are curved to aid in the self-assembly of spheroids, agarose has a refractive index nearly equal to that of water (1.334 vs. 1.3329 of water), thus should not scatter light more significantly than a water interface would25. Lastly, the agarose micro-wells have been shown to provide a stable cell culture platform for the long term, up to 4-week culture of spheroids of certain cell types14,15,26.

We controlled spheroid size by altering the number of cells seeded into each well. When the same number of cells were seeded across an entire plate, the variation of spheroid radii was approximately 10% of its mean. Spheroid size could be controlled by altering the input number if cells, with higher seeding densities yielding larger spheroids. Over the range of sizes tested, variation of spheroid radii was also approximately 10% of the mean for all sizes. A number of factors probably contribute to variation in spheroid radii including pipetting errors and variation of the radius of the input cells (KGN cells 12 +/− 1μm). Control over spheroid size and uniformity of size are important because the biology of spheroids is often linked to their size. For example, spheroids with diameters significantly greater than 200μm often have necrotic cores with dead and dying cells due to the limitations of diffusion. For some cancer models, large spheroids with necrotic cores are thought to mimic the complexity of tumors in vivo12,22,27. But for other assays, such as those that use primary cells or that wish to replicate normal physiology, smaller spheroids are preferable with their high viability and lack of confounding influences associated with apoptotic/necrotic core.

Furthermore, the ability to perform confocal microscopy of an entire spheroid is dependent on their size. For small spheroids with diameters in the range of 50μm, we found there was minimal loss of fluorescence throughout the z stack. The entire spheroid could be imaged and for each slice the increase in cross-sectional area of the spheroid correlated with an increase in the total fluorescence. This was true for all four dyes tested. Depending on the assay, small spheroids may be sufficient. Our spheroids with 50μm diameters contain approximately 2–3 cell layers from the surface of spheroid to its edge. This number would vary depending on a number of factors including the oblate shape of the spheroid, the size of the cell type used and if the spheroid undergoes morphological changes such as the formation of luminal structures15. For assays using small spheroids that want to quantitate activity as a function of radius, an entire z stack could be obtained or a single confocal slice at about the equator may be sufficient for analysis.

Some assays may require larger spheroids with increased cell layers, however, we found that for spheroids with diameters above 50μm, there was an associated loss of signal. For all dyes tested, the signal loss increased as spheroid radii increased. Most importantly, this loss was not uniform throughout the spheroid. Unlike a planar specimen that will undergo uniform loss across each slice of a z-stack, in a spheroid, the loss of fluorescence was far greater in the interior versus its outer edges due to its curvature. Additionally, confocal slices beyond a certain z-depth into a spheroid are not accurate quantitative representations of fluorescence as a function of radial position within that slice, since the signal from interior portions would not differ from background noise. Furthermore, we hypothesized that regardless of size, the rate of loss throughout the central z-axis would exhibit an exponential decay function with similar rate of loss for first 50μm. However, even the rate of fluorescent decrease throughout the z-depth was dependent upon spheroid size, with smaller spheroids possessing a steeper rate than larger ones. This might imply that smaller spheroids scatter light more than large spheroids, despite possessing the same starting cell types and dyes. Therefore, to avoid differences between large and small spheroids, restricting spheroid size to a set range of radii would be beneficial when developing quantitative fluorescent assays.

This non-uniform overall fluorescent signal and its variable rate of loss complicates the ability to compensate for signal loss throughout a z-stack. To address the shortcomings associated with confocal imaging of large spheroids, there are various methodologies and approaches. For example, the depth of imaging in cleared samples is increased19–22. Although we have not tested it, we suspect that, like our live cell experiments, there is a reliable depth into cleared samples from which accurate radial information can be obtained from a given slice. However, utilizing this technique limits the type of assays that can be performed to fixed, end-point assays, since clearing is not compatible with live cell imaging. An alternative approach to acquiring live cell images deep within large spheroids could be the use of two photon microscopy, however we are not aware of a two photon microscope designed for high throughput automated imaging. Alternatively, one could limit spheroid size, 3D portions, or single confocal slices of spheroids to analyze only a set size and z-depth, where reliable quantitative information can be obtained. As shown by the data with small spheroids, this distance may be about 50μm. However, depending on the experimental assay, data from small spheroids or portions of spheroids may not be accurate quantitative depiction and require the use of larger spheroids. Additionally, to compensate for signal loss and reduce variability throughout the z-depth, one could utilize ratio imaging, however, ratio imaging does not overcome the physical limitations of confocal microscopy, and thus is still only effective for certain range of sizes.

Overall, the use spheroids for high throughput screening affords numerous advantages, including in vivo-like differentiation, function and architecture, as well as drug sensitivities comparable to tumors6,7,8,27,29. Moreover, spheroids have been formed from immortal cell lines as well as primary cells14,15,26. Additionally, since they are self-assembled from mono-dispersed cells, spheroids have been formed from mixtures of two or more cell types to make designer spheroids that can mimic even more complex tissue units30. Applications of spheroids, some of which have been adapted to high throughput formats, include toxicity testing, drug transport, and drug discovery3,5,23,24. Spheroids also create complex microenvironments that include gradients, barriers, polarity, cell-cell signaling as well as significant changes to cell physiology, morphology and differentiation2,4,5. These changes to the microenvironment will vary along the 3D radius of a spheroid. To accurately quantify these biological changes as a function of radii, the quantitative limitations of confocal imaging of spheroids must be understood. In this paper, we addressed some of the challenges of quantifying the bowl-like 3D image of spheroids. These studies set the stage for designing quantitative spheroid based assays for automated high throughput screening.

Supplementary Material

supp

Acknowledgments

We thank Microtissues, Inc. for the mold used to make the 96 well plates.

FUNDING

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: this work was funded in part by seed funds from Unilever and the generous support of Donna McGraw Weiss ‘89 and Jason Weiss. Also, this work was funded under the U.S. Department of Education, GAANN Award P200A150037, the National Science Foundation Award OIA1655221, the National Institute of Environmental Health Sciences Award 1U01ES028184.

Footnotes

DECLARATION OF CONFLICTING INTEREST

J.R.M has an equity interest in Microtissues, Inc. This relationship has been reviewed and managed by Brown University in accordance with its conflict of interest policies.

References

  • 1.Mazzoleni G, Di Lorenzo D, Steimberg N. Modelling tissues in 3D: the next future of pharmaco-toxicology and food research? Genes Nutr. 2009;4(1):13–22. doi: 10.1007/s12263-008-0107-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.Pampaloni F, Reynaud EG, Stelzer EH. The third dimension bridges the gap between cell culture and live tissue. Nat Rev Mol Cell Biol. 2007;8(10):839–845. doi: 10.1038/nrm2236. [DOI] [PubMed] [Google Scholar]
  • 3.Hirschhaeuser F, Menne H, Dittfeld C, et al. Multicellular tumor spheroids: an underestimated tool is catching up again. J Biotechnol. 2010;148(1):3–15. doi: 10.1016/j.jbiotec.2010.01.012. [DOI] [PubMed] [Google Scholar]
  • 4.Atala A, Kasper FK, Mikos AG. Engineering complex tissues. Sci Trans Med. 2012;4(160):160rv12. doi: 10.1126/scitranslmed.3004890. [DOI] [PubMed] [Google Scholar]
  • 5.Elliot NT, Yuan F. A review of three-dimensional in vitro models for drug discovery and transport studies. J Pharm Sci. 2011;100:59–74. doi: 10.1002/jps.22257. [DOI] [PubMed] [Google Scholar]
  • 6.Chang TT, Hughes-Fulford M. Monolayer and Spheroid Culture of Human Liver Hepatocellular Carcinoma Cell Line Cells Demonstrate Distinct Global Gene Expression Patterns and Functional Phenotypes. Tissue Eng Part A. 2008;15(3):559–567. doi: 10.1089/ten.tea.2007.0434. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Chitcholtan K, Asselin E, Parent S, et al. Differences in growth properties of endometrial cancer in three dimensional (3D) culture and 2D cell monolayer. Exp Cell Res. 2012;319(1):75–87. doi: 10.1016/j.yexcr.2012.09.012. [DOI] [PubMed] [Google Scholar]
  • 8.Astashkina AI, Mann BK, Prestwich GD, et al. Comparing predictivie drug nephrotoxicity biomarkers in kidney 3D primary organoid culture and immortalized cell lines. Biomaterials. 2012;33(18):4712–4721. doi: 10.1016/j.biomaterials.2012.03.001. [DOI] [PubMed] [Google Scholar]
  • 9.Knight A. Animal experiments scrutinised: systematic reviews demonstrate poor human and clinical toxicological utility. ALTEX. 2007;24:320–325. doi: 10.14573/altex.2007.4.320. [DOI] [PubMed] [Google Scholar]
  • 10.Krewski D, Acosta D, Jr, Andersen M, et al. Toxicity testing in the 21st century: a vision and a strategy. J Toxicol Environ Health B Crit Rev. 2010;13:51–138. doi: 10.1080/10937404.2010.483176. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Griffith LG, Swartz MA. Capturing complex 3D tissue physiology in vitro. Nat Rev Mol Cell Biol. 2006;7:211–224. doi: 10.1038/nrm1858. [DOI] [PubMed] [Google Scholar]
  • 12.Yamada KM, Cukierman E. Modeling tissue morphogenesis and cancer in 3D. Cell. 2007;130:601–610. doi: 10.1016/j.cell.2007.08.006. [DOI] [PubMed] [Google Scholar]
  • 13.Bartosh TJ, Ylostalo JH, Mohammadipoor A, et al. Appregation of human mesenchymal stromal cells (MSCs) into 3D spheroids enhances their antiinflammatory properties. Proc Natl Acad Sci USA. 2010;107:13724–13729. doi: 10.1073/pnas.1008117107. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Vantangoli MM, Madnick SM, Huse SM, et al. MCF-7 human breast cancer cells form differentiated microtissues in scaffold-free hydrogels. PLoS One. 2015;10:e0135426. doi: 10.1371/journal.pone.0135426. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Kabadi PK, Vantangoli MM, Rodd AL, et al. Into the depths: Techniques for in vitro three-dimensional microtissue visualization. Biotechniques. 2015;59:279–286. doi: 10.2144/000114353. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Mohapatra S, Nandi S, Chowdhury R, et al. Spectral mapping of 3D multi-cellular tumor spheroids: time-resolved confocal microscopy. Phys Chem Chem Phys. 2016;18(27):1881–1890. doi: 10.1039/c6cp02748b. [DOI] [PubMed] [Google Scholar]
  • 17.Konig K, Uchugonova A, Gorjup E. Multiphoton fluorescence lifetime imaging of 3D-stem cell spheroids during differentiation. Microsc Res Tech. 2011;74(1):9–17. doi: 10.1002/jemt.20866. [DOI] [PubMed] [Google Scholar]
  • 18.Boutin ME, Hoffman-Kim D. Application and assessment of optical clearing methods for imaging of tissue-engineered neural stem cell spheres. Tissue Eng Part C Methods. 2015;21(3):292–302. doi: 10.1089/ten.tec.2014.0296. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Hama H, Kurokawa H, Kawano H, et al. Scale: a chemical approach for fluorescence imaging and reconstruction of transparent mouse brain. Nature Neurosci. 2011;14:1481–1488. doi: 10.1038/nn.2928. [DOI] [PubMed] [Google Scholar]
  • 20.Kuwajima T, Sitko AA, Bhansail P, et al. ClearT: a detergent- and solvent-free clearing method for neuronal and nonneuroanal tissue. Development. 2013;140(6):1364–1368. doi: 10.1242/dev.091844. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Ke MT, Fujimoto S, Imai T. SeeDB: a simple and morphology-preseriving optical clearing agent for neruonsal circuit reconstruction. Nat Neurosci. 2013;16(8):1154–1161. doi: 10.1038/nn.3447. [DOI] [PubMed] [Google Scholar]
  • 22.Tung YC, Hsiao AY, Allen SG, et al. High-throughput 3D spheroid culture and drug testing using a 384 hanging drop array. Analyst. 2011;136(3):473–478. doi: 10.1039/c0an00609b. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Mehta G, Hsiao A, Ingram M, et al. Opportunities and challenges for use of tumor spheroids as models to test drug delivery and efficacy. Journal of Controlled Release. 2012;164:192–204. doi: 10.1016/j.jconrel.2012.04.045. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Achilli TM, McCalla S, Meyer J, et al. Multilayer spheroids to quantify drug upake and diffusion in 3D. Mol Pharm. 2014;11:2071–2081. doi: 10.1021/mp500002y. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Jain A, Yang AH, Erickson D. Gel-based optical waveguides with live cell encapsulation and integrated microfluidics. Opt Lett. 2012;37(9):1472–1474. doi: 10.1364/OL.37.001472. [DOI] [PubMed] [Google Scholar]
  • 26.Dingle YT, Boutin ME, Chirilla AM, et al. Three-dimensional neural spheroid culture: An in vitro model for cortical studies. Tissue Eng Part C Methods. 2015;21(12):1274–1283. doi: 10.1089/ten.tec.2015.0135. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Vinci M, Gowan S, Boxall F, et al. Advances in establishment and analysis of three-dimensional tumor spheroid-based functional assays for target validation and drug evaluation. BMC Biol. 2012;10:29. doi: 10.1186/1741-7007-10-29. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Visser TD, Groen CA, Brakenhoff GJ. Absorption and scatter correction in fluorescence confocal microscopy. J Microscopy. 1990;163(2):189–200. [Google Scholar]
  • 29.Robertson FM, Ogasawara MA, Ye K, et al. Imaging and analysis of 3D tumor spheroids enriched for a cancer stem cell phenotype. J Biomol Screen. 2010;15:820–829. doi: 10.1177/1087057110376541. [DOI] [PubMed] [Google Scholar]
  • 30.Curran S, Vantangoli MM, Boekelheide K, et al. Architecture of chimeric spheroids controls drug transport. Cancer Microenviron. 2015;8(2):101–109. doi: 10.1007/s12307-015-0171-0. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

supp
NIHMS966978-supplement-supp.docx (2.1MB, docx)

RESOURCES