Spheroid arrays for high-throughput single-cell analysis of spatial patterns and biomarker expression in 3D

Abstract

We describe and share a device, methodology and image analysis algorithms, which allow up to 66 spheroids to be arranged into a gel-based array directly from a culture plate for downstream processing and analysis. Compared to processing individual samples, the technique uses 11-fold less reagents, saves time and enables automated imaging. To illustrate the power of the technology, we showcase applications of the methodology for investigating 3D spheroid morphology and marker expression and for in vitro safety and efficacy screens. First, spheroid arrays of 11 cell-lines were rapidly assessed for differences in spheroid morphology. Second, highly-positive (SOX-2), moderately-positive (Ki-67) and weakly-positive (βIII-tubulin) protein targets were detected and quantified. Third, the arrays enabled screening of ten media compositions for inducing differentiation in human neurospheres. Last, the application of spheroid microarrays for spheroid-based drug screens was demonstrated by quantifying the dose-dependent drop in proliferation and increase in differentiation in etoposide-treated neurospheres.

Figure 1. Spheroid microarray technology overview and mold making procedure.

(a) The current workflow to analyze spheroid histology requires separate processing of spheroids representing different conditions and results in many samples which need to be embedded (I), processed (II), sectioned (III), stained(IV) and imaged (V) separately. The random distribution of spheroids in different planes requires manual imaging and further takes up researcher and equipment time. Embedding multiple conditions on the same array (top) reduces the number of samples 11 times resulting in economies in reagents and hands-on time as well as the possibility for automated imaging of all spheroids located in the same plane. (b) Spheroid microarrays are made by pouring liquid agarose solution in histology molds and floating the Mold-maker on top of the solution. Once the agarose cools down and gels, the Mold-maker is removed and the spheroids are loaded into the wells of the resulting agarose mold. The mold is sealed with low-gelling agarose and is processed for histology.

Figure 2. The Mold-maker device, the agarose molds it produces and the final spheroid arrays.

(a) Three-dimensional plot of the Mold-maker device. It is made up of 66 paraboloid-shaped pegs on a single plane and features a handle for easy removal from the gel. (b) Agarose gel after Mold-maker has been removed. (c) Side view of the agarose gel. (d) Spheroids from 11 cell-lines embedded in a single agarose array. (e) Histogram of the average distance from the median lower edge for r = 500 μm beads from three independent experiments. Separate histograms and images are available in the Supplementary information. (f) Paraffin section of the spheroid microarray. (g) Scan of a representative spheroid array, conditions of interest are located in columns 1–10, column 11 contains multiple marking spheroids.

Figure 3. Hematoxylin and eosin staining of multi-spheroid arrays made up of different cell-lines.

Each column represents a different cell line, while rows are made up of replicate spheroids. OVCAR-3, HCC1806, MDA-MB-231, MCF-7, N87 were seeded at 16,000 cells/well, while RKO, BxPC3s, HCT116, U87 and 791T at 8,000 cells/well. All spheroids were cultured for 4 days as specified in the Methods section. Scale bar 500 μm.

Figure 4. Signal uniformity assessment for antibody-stained protein targets in spheroid microarrays of human fetal brain neurospheres (day 3 of culture).

(a) Percentage of nuclei positive for SOX-2 (high signal). (b) Percentage of nuclei positive for Ki-67(moderate signal), (c) Percentage of area positive for βIII-tubulin (low signal.) Signals for the stained samples (circles) are compared to negative isotype controls (squares). (d) Scanned images of a selection of the spheroid replicates used in the analysis. First two rows stained with SOX-2, third and fourth (Ki-67), fifth and sixth (βIII-tubulin). Blue-hematoxylin nuclear stain, brown-DAB positive stain. Scale bar 200 μm. In graphs a-c, the line represents the median, the error bars the interquartile range and the dots are individual values for the scoring. Note the difference in scale for the y-axis in panel c.

Figure 5. Neural stem cell differentiation in different media compositions.

(a) Scanned images of neurospheres stained for markers of neuronal (βIII-tubulin) and glial (GFAP) differentiation, progenitors (SOX-2) and morphology (H&E). Different media conditions are arranged in columns, while the markers are arranged in rows. Scale bar 500 μm. (b) Tukey boxplot of the area positive for GFAP (red, left box) and βIII-tubulin (blue, right box) compared to the total area of the sphere. (c) Tukey boxplot of the number of nuclei positive for SOX-2 compared to the total number of nuclei in the spheroid. For (b) and (c) the line represents the median percentage, the box is formed by the first and third quartile, whiskers are 1.5 times the quartiles and points represent outliers. (d) to (f)- Graphs of the magnitude of the difference between the various differentiation media and NSC medium for βIII-tubulin (d), GFAP (e) and SOX-2(f). Dots (d) and squares (e,f) represent the mean difference, while the error bars are 95%CIs from the ANOVA analysis with Dunnet’s multiple comparison test follow-up.

Figure 6. Effects of etoposide on the proliferation and neuronal differentiation of fetal brain neurospheres.

(a) Ratio of Ki-67 positive nuclei (in percent) as a function of increasing etoposide concentration. (b) Relative area positive for βIII tubulin (%). Dots are mean values derived from 3 sections from 3 experimental repeats, error bars are SDs. Four parameter-dose response curves were fitted to the data. (c) Representative images of untreated spheroids (left) and spheroids at increasing concentrations of etoposide stained for Ki-67(top) and βIII-tubulin (bottom). Scale bar 500 μm.