VCS MN neural net module identifies micronucleated cells

a) Diagrams of the two constructs transduced into hTERT-RPE1 cells to make the RFP703/Dendra cell line and how the proteins localize in micronucleated (MN+) cells before and after MN rupture. b) Diagram of VCS MN image analysis pipeline on RFP703/Dendra cells incubated in Mps1i. Raw image is of H2B-emiRFP703 taken on a 20x widefield objective. Nuclei are first segmented using the Deep Retina segmenter and these masks are used to generate image crops centered on each nucleus. Image crops are then resized to fit the ResNet18 UNet model (see Fig. S1) and analyzed by VCS MN to generate MN label predictions, which are mapped back to the raw image. After image reconstruction, MN are assigned to the nearest nucleus based on proximity. Nucleus masks from the Deep Retina segmenter are then classified as MN+ or MN- to generate the nuclei labels. c) Recall and positive predictive value (PPV) of MN identification within the image crops. MN were manually labeled as intact or ruptured using an image of NLS-3xDendra2 acquired at the same time. N = 5, n = 264, 158, 365, 283, 249. d) Recall and PPV of MN- and MN+ labeled nuclei. N = 2, n = 328, 186. e) Same as for d) except images were of U2OS cells expressing H2B-emiRFP703 and 3xDendra2-NLS. N = 1, n = 85, 95. f) RFP703/Dendra cells with ruptured (arrows) and intact (arrowhead) MN showing a loss of NLS- 3xDendra2 signal in ruptured MN. Scale bar = 10 µm. g) Recall and PPV for MN rupture-based nucleus classification. N = 3, n = 120, 91, 82.

MNFinder robustly segments MN across cell types and imaging conditions

a) Representative image showing undersegmentation and low recall by the VCS MN neural net on DAPI- labeled RPE1 cells imaged at 40x magnification after Mps1i incubation. Images were scaled down prior to VCS MN analysis to match 20x pixel size. Mean intersection-over-union (mIoU) and recall quantified from N =2, n = 33, 31 images. b) Overview of MNFinder module for classifying and segmenting MN and nuclei. Raw images of chromatin are first tiled by a sliding window and then processed by 2 groups of neural nets: one classifies pixels as either nuclei or MN (Nuc/MN) and one is an instance classifier for cells, with cells being defined as the smallest object that encloses a nucleus and its associated MN. Nuc/MN classifier results are post- processed to reassign MN that were misclassified as small nuclei and to expand MN masks. The cell instance classifier outputs gradient maps, which are used to define cells through watershed-based post-processing (see Fig. S4). Nuc/MN results are integrated with cell results to produce final labels of individual MN, nuclei, and their associations. Image crops are reassembled onto the final image by linear blending. c-d) Example images and MN pixel predictions using MNFinder on multiple cell types (RPE1 H2B-emiRFP703, U2OS, U2OS H2B- emiRFP703, HeLa H2B-GFP, and HFF), chromatin labels (DAPI, H2B-FP), and magnifications (20x, 40x). MN recall, PPV, and mIOU were quantified and averaged per image for each condition. Dotted line = performance of the VCS neural net on RPE1 H2B-emiRFP703 20x images. MNFinder performance is similar across conditions except images of H2B-GFP in fixed HeLa cells. N = 1. n (on graph) = cells.

Visual cell sorting can isolate RPE1 RFP703/Dendra micronucleated cells.

a) Overview of visual cell sorting protocol. Cells are plated in multi-well plates, cellular phenotypes are quantified on-demand during imaging by VCS MN, and labeled nuclei (e.g. MN+ and MN-) are photoconverted for either 200 or 800 ms, yielding two different ratios of red:green fluorescence (red and yellow nuclei). These differences are quantified by FACs and gated on the red:green ratios for cell sorting. Graphic created with BioRender.com. b) Histogram of nuclear red:green ratio quantification measured on repeated imaging of the same fields 0, 4, and 8 h after photoconversion. Representative images are pseudo-colored by log10 Dendra red:green ratio (below). N = 1, n = 82, 353, 285, 313. c) Design of MN cell isolation validation experiment. VCS MN classifier PPV calculated on images acquired during activation and frequency of MN- or MN+ cells manually quantified in cells plated and fixed post-FACs sorting. Pre-FACS: N = 2, n = 251, 263. post-FACS N=1, n = 338, 353.

Visual cell sorting pipeline identifies Mps1i transcriptional response.

a) Outline of experiment. Nuclei identified by the Deep Retina segmenter were randomly activated for 800 or 200 ms. b) PCA plot showing clustering of Mps1i-treated and DMSO-treated cells by treatment (major) and by replicate (minor). Each experimental replicate represents 2 technical replicates. c) MA plot of genes identified by RNASeq. Differentially expressed genes (FDR adjusted p-value < 0.05) are in green. Gray lines represent 1.5 fold-change in expression. d) Heatmap of Hallmark pathway enrichment between visual cell sorting data and data from (Santaguida et al., 2017) and (He et al., 2019) analyses of RPE1 cells after chromosome missegregation. Hallmark pathways (bottom) were clustered based on manually annotated categories (top).

Micronucleation and rupture result in few transcriptional changes.

a) Outline of experiment for MN+ and MN- cell isolation from Mps1i treated RFP703/Dendra cells. b) PCA plot showing clustering of MN+ and MN- cells by replicate (major) and condition (minor). c) MA plot of genes identified in MN+/- RNASeq. Of the identified DEGs, only 2 have fold-changes larger than 1.5. Both, TNFAIP3 and EGR1, are also significantly upregulated in Mps1i treated cells. d) Outline of experiment for MN rupture+ and rupture- cell isolation. e) PCA plot showing clustering of intact MN and ruptured MN cells by condition and replicate. f) MA plot of genes identified in MN rupture+/- RNASeq. The three DEGs with a fold-change higher than 1.5 and unique to this dataset are labeled on the plot. g) Heatmap of Hallmark pathway enrichment in datasets of DMSO vs Mps1i, Mps1i-treated cells with and without MN, and synchronized, Mps1i-treated, MN+ cells with and without MN rupture. All DEGs, defined as an FDR less than 0.05, are included. Pathways are grouped based on manual annotation (right). 6/43 pathways are unique to the MN+/- profile compared to Mps1i+/-, and 2/43 pathways are unique to MN rupture+/- compared to Mps1i+/-.

Micronucleation and rupture do not significantly alter the aneuploidy transcription response.

a) Heatmap of Mps1i+ DEGs (cutoff: absolute fold-change ≥ 1.5) compared to MN+ and rupture+ replicates. Euclidean distances calculated for features and samples and clustered by complete-linkage. Genes lacking values for one more experiment types were excluded. Line = gene cluster upregulated in MN rupture+ cells. b) Representative images and quantification of ATF3 and EGR1 labeling in RPE1 2xRFP-NLS cells after Mps1i incubation. Arrows = ruptured MN, arrowheads = intact MN. Scale bar = 20 µm. c) Quantification of normalized ATF3 and EGR1 mean nuclear intensity in manually classified cells. N = 2 (graph colors), n = on graph, p: ns > 0.05, * ≤ 0.05, *** ≤ 0.001 by GEE. d) Representative images of DNA FISH for chromosomes 1, 11, and 18 and H3K27Ac identification of intact MN. Arrows = ruptured MN, arrowheads = intact MN. e) Quantification of aneuploidy frequency (foci do not equal 2) per chromosome. Cells manually classified as MN- or MN+, and MN rupture- or rupture+. MN: Chr 1: N=2, n=429, 158; Chr 11: N=3, n=406, 313, 160; Chr 18: N=3, n=425, 202, 230. Rupture: Chr 1: N=2, n=187, 74; Chr 11: N=3, n=190, 108, 71; Chr 18: N=3, n=186, 102, 101.