Journal: Nature Communications

Article Title: High-content morphological profiling by Cell Painting in 3D spheroids

doi: 10.1038/s41467-026-78596-5

Figure Lengend Snippet: a Composite image of an HCT116 spheroid at 30 µm depth, showing nuclei (Hoechst, blue), RNA and cytoplasm (SYTO 14, green), and cytoskeleton, plasma membrane, and Golgi (phalloidin and wheat germ agglutinin, WGA; red). Scale bar is 50 µm, and spheroids were imaged at 30 × (0.227 µm/pixel). Inset: magnified view showing individual cells (scale bar, 10 µm). b Individual channels from the same acquisition: Hoechst (nuclei, blue), concanavalin A (endoplasmic reticulum, cyan), SYTO 14 (cytoplasm and RNA, green), phalloidin and WGA (cytoskeleton, plasma membrane, and Golgi, red), and MitoTracker (mitochondria, gray). c An example of nuclei (left) and cell segmentation (right) by Cellpose. Objects are outlined in white. Scale bar is 50 µm. a – d Number of 2D segmentation object instances per spheroid at each z -section for HCT116 (gray) and HT29 (black). Right: total instances summed across the full z -stack per spheroid. e Detections per cell across the z -stack (duplicate estimate), reflecting how many times the same cell is counted across optical sections. Each datapoint is one well at one section; n = 58 HCT116 and 56 HT29 DMSO wells. Box plots: center line, median; box bounds, 25th and 75th percentiles; whiskers, most extreme value within 1.5 × the interquartile range; outliers not shown. f Representative 3D distribution of detected cell positions in an HCT116 spheroid exposed to 0.1% DMSO. g Principal Component Analysis (PCA) of single-cell morphological profiles normalized per-plate (left) versus per-section (right). Both panels use identical PCA loadings (computed on the per-plate-normalized DMSO data; PC1 and PC2 explain 23.64% and 13.19% of total variance, respectively) to facilitate direct comparison. Section dependence was assessed by two-sided Pearson correlation of PC2 against optical section, with η² from a one-way ANOVA as a descriptive effect size ( n = 348 well–section observations, 29 wells, no correction for multiple comparisons). Section-dependent effects are present before normalization (PC2: r = 0.79, η² = 68%, p = 1.1 × 10⁻ 76 ), but much less after (PC2: r = −0.00, η² = 1%, p = 0.96). Spheroid optical section (0–11) is color-coded in ( f , g ). Images in a – c are 16-bit, with display ranges of 190-1328 (Hoechst), 299-4332 (SYTO14), and 347-5876 (Phalloidin and WGA) out of 65535. Rolling-ball background subtraction (radius 50 pixels) was applied in an and b but not in c . Source data are provided as a Source Data file.

Article Snippet: Two human colorectal cancer cell lines were used: HCT116 (RRID:CVCL_0291; colorectal carcinoma, adult male donor; originally ATCC CCL-247) and HT29 (RRID:CVCL_0320; colorectal adenocarcinoma, female donor; originally ATCC HTB-38).

Techniques: Clinical Proteomics, Membrane, Spheroids, Analysis, Single Cell, Comparison, Reference Subtraction

Journal: Nature Communications

Article Title: High-content morphological profiling by Cell Painting in 3D spheroids

doi: 10.1038/s41467-026-78596-5

Figure Lengend Snippet: a Representative examples of HCT116 (top) and HT29 (bottom) max intensity projections (MIPs) for the compounds SN-38 (0.03 µM), binimetinib (10 µM), abemaciclib (3.0 µM), and DMSO (0.1%), as well as for ( b ) a single section at 30 µm depth. SYTO 14 (RNA) is shown in green, Hoechst (nuclei) is shown in blue, and phalloidin/WGA (cytoskeleton and golgi) are shown in red. Scale bars are 50 µm; images were acquired at 30 × (0.227 µm/pixel). Images are 16-bit, with display ranges of 190-1328 (Hoechst), 299-4332 (SYTO14), and 347-5876 (Phalloidin and WGA) out of 65535. No rolling-ball background subtraction was applied. c , d Grit scores measuring phenotypic effect are shown for MIP-based ( c ) and aggregated ( d ) profiles. Each bar is one compound–dose perturbation, 4 wells; error bars, mean ± 1 s.d. The dashed horizontal line marks the grit > 1.96 threshold. e , f Null and median pairwise replicate correlations distributions across dose steps for both HCT116 and HT29 cell lines. Null distributions are in grays and replicate in blues. Box plots (both series): center line, median; box bounds, 25th and 75th percentiles; whiskers, most extreme value within 1.5 × the interquartile range; outliers not shown. Each point is one compound–dose perturbation (4 wells); n = 41–52 per box, 208 per null. c – h pre-filter, treatment wells only. g , h Comparison of median pairwise replicate correlations of MIP-based and scAgg profiles in both HCT116 and HT29 cell lines. Data points are compound–dose perturbations ( n = 383; 196 HCT116, 187 HT29), color-coded for ( g ) dose, or ( h ) pathway. Source data are provided as a Source Data file.

Article Snippet: Two human colorectal cancer cell lines were used: HCT116 (RRID:CVCL_0291; colorectal carcinoma, adult male donor; originally ATCC CCL-247) and HT29 (RRID:CVCL_0320; colorectal adenocarcinoma, female donor; originally ATCC HTB-38).

Techniques: Reference Subtraction, Comparison

Journal: Nature Communications

Article Title: High-content morphological profiling by Cell Painting in 3D spheroids

doi: 10.1038/s41467-026-78596-5

Figure Lengend Snippet: a A quantification of the number of drugs with grit > 1.96 for cell lines HCT116 and HT29. Color-coded are the drugs unique to each culture format, as well as the non-unique drugs. The total amount of drugs is 52. b Semi-supervised UMAPs of 2D morphological profiles. c Heatmap displaying hierarchical clusters of cosine similarities for 2D profiles. d Comparison of cosine scores between pairs of compounds in 2D and 3D. The x-axis shows the difference in similarity scores (3D – 2D), while the y-axis shows the highest similarity score across both modalities for each pair. Pairs whose absolute between-format difference (3D – 2D) fell in the top 5% (95th percentile) across all matched pairs were flagged as divergent; among these, pairs with a cosine score of at least 0.5 in either 2D or 3D were labeled. e Representative images of cells in 2D (top) and 3D (bottom) exposed to olaparib (3 µM), 5-fluorouracil (5-FU) (10 µM), and the negative control DMSO (0.1%). SYTO 14 is green, phalloidin/WGA is red, and Hoechst is blue. 2D images were acquired at 20 × (0.34 µm/pixel) and 3D images at 30 × (0.227 µm/pixel) from a 30 µm depth. Images are 16-bit, with display ranges of 190-1328 (Hoechst), 299-4332 (SYTO14), and 347-5876 (Phalloidin and WGA) out of 65535. No rolling-ball background subtraction was applied. f Preprocessed morphological profiles are shown for 2D (top) and 3D (bottom) exposed to 5-FU (10 µM) and olaparib (3 µM). Scale bars are 50 µm. Source data are provided as a Source Data file.

Article Snippet: Two human colorectal cancer cell lines were used: HCT116 (RRID:CVCL_0291; colorectal carcinoma, adult male donor; originally ATCC CCL-247) and HT29 (RRID:CVCL_0320; colorectal adenocarcinoma, female donor; originally ATCC HTB-38).

Techniques: Comparison, Labeling, Negative Control, Reference Subtraction

Journal: Nature Communications

Article Title: High-content morphological profiling by Cell Painting in 3D spheroids

doi: 10.1038/s41467-026-78596-5

Figure Lengend Snippet: a Microscopy images of spheroids exposed to 5-FU (10 µM; n = 13; green), olaparib (3 µM; n = 13; purple), and etoposide (2.5 µM; n = 8; orange), plus a DMSO control (0.1%; n = 13; gray). Stained for EdU (red), Hoechst (blue), and γH2AX (green); 12-bit; channels background-subtracted in Fiji across the whole image (rolling ball = 50, 50, 5); display ranges Hoechst 0–2500, γH2AX 11–160, EdU unscaled (0–4095). Scale bars, 50 µm. Images were acquired at 30 × (0.227 µm/pixel) from a 30 µm depth. b Quantification of DNA damage (γH2AX foci/nucleus), S-phase entry (fraction of EdU-positive nuclei), and spheroid size (nuclei/spheroid) across the four conditions (n as in ( a ); unit of study, individual spheroid in a separate well); error bars, mean ± SEM. Significance by Kruskal–Wallis followed, where significant, by two-sided Mann–Whitney U; comparisons against DMSO were Bonferroni-corrected, the 5-FU versus olaparib comparison was not. Exact P values versus DMSO for 5-FU, olaparib and etoposide respectively, and for 5-FU versus olaparib: γH2AX foci, 0.0122, 6.2 × 10⁻⁵, 2.9 × 10⁻⁵ and 2.7 × 10⁻⁴; EdU-positive fraction, 4.9 × 10⁻⁵, 6.2 × 10⁻⁵, 4.8 × 10⁻⁴ and 5.9 × 10⁻⁴; nuclei per spheroid, 0.0088, 0.0359, 2.9 × 10⁻⁵ and 0.412 (not significant). c Hallmark gene-set enrichment analysis (GSEA): NES for 5-FU (x-axis) versus olaparib (y-axis), colored by agent as in ( a ) or black where significant for both (FDR < 0.1). d Per-gene log2FC for the 15 strongest effectors per signature (top to bottom: p53 Hallmark, E2F Hallmark, p53-apoptosis Reactome); colored as in ( a ), opacity indicating significance (FDR < 0.1). e Pairwise cosine-similarity maps of the 10 drugs most similar to 5-FU (10 µM) in HCT116 (left; subset of Fig. ) and HT29 (right); 5-FU boxed. Source data are provided as a Source Data file.

Article Snippet: Two human colorectal cancer cell lines were used: HCT116 (RRID:CVCL_0291; colorectal carcinoma, adult male donor; originally ATCC CCL-247) and HT29 (RRID:CVCL_0320; colorectal adenocarcinoma, female donor; originally ATCC HTB-38).

Techniques: Microscopy, Spheroids, Control, Staining, MANN-WHITNEY, Comparison, Analysis