Fig 1.
Human lung adenocarcinoma is characterized by dysregulation of cell and nuclear size.
63X images shown. A., schematized alveolus, related to B., showing AT2 cells (marker: proSPC, green), presumptive wide and flat AT1 cells, non-CK7-staining macrophage, M, and stroma, S. Epithelial cells express marker CK7, but weakly in AT2 cell bodies. B. and C., 5 μm z-projections of representative normal distant control alveoli, B., and representative EGFR+ LA tumor, AT2 cells marked by proSPC (green), C. Cytokeratin 7, CK7 (red); overexpressed in tumor cells. DNA stained by TO-PRO-3 (cyan). White arrows: thin cellular processes (shown in 1E, below). Right, full-stack max-intensity projections from each axis. Scale bars, 20 μm. Row D., 3D segmentation approach for tumor cells. Left, a high-density confocal stack often referenced during early development of our machine-learning algorithm. Middle, Machine Learning from CK7 staining predicts cell boundaries. Right, Commercial Imaris software fits cell boundaries, with rainbow coloring showing individually segmented cell surfaces. High-coverage case shown. E. Left, 3D rendering of AT2 cells (proSPC) showing thin cellular processes; right, processes shown with AT2 cell body segmentation (gray). White box shows perspective scale, 2 μm per small tick mark. Yellow arrow: AT2 cell body; white arrow: cellular process.
Fig 2.
Quantitation of cell and nucleus dimensions in EGFR+ and KRAS+ lung tumors.
A., Schematic: potential outcomes for cell and nucleus scaling in cancerous cells, hypothesis 2 is that LA retains nuclear scaling function. B., histograms of cell volumes of near-euploid (1.6–4.4n) cells (N = 21, n = 2683) as a function of EGFR+ (N = 10, n = 1411) or KRAS+ (N = 11, n = 1260) genotype and stage. Green, cell volumes for tumor cells within the boundaries of the 2n normal AT2 cells; Red, cell volumes outside the Gaussian distribution of 2n population of Normal AT2 cell volumes. C., violin plots for near-euploid cells show very significantly abnormal scaling and significant changes in distributions between stages. 1–2 and 2–3. Colors alternating for visibility, normal in grey. Kruskal-Wallis ANOVA, p<1 x 10−4 (****). Select pairwise p-values for Mann-Whitney tests in Tables 1–3. D.-E., quantitation of differences between near-euploid tumor cells and 2n wild-type AT2 cells. Circles, cell volume. Diamonds, nuclear volume. Crosses, N:C ratio. D., Corresponding fold change (tumor-to-normal) for cell sizes (top); standard deviation (bottom). Error bars ≤ 2.7 x 10−3 omitted for clarity Magenta, EGFR+, Blue, KRAS+. E., percentages of cells which fall within the regression fits to histograms for tumor cells, but outside the Gaussian fit to 2n AT2 cells; showing cell volume, nucleus volume and N:C ratio; colored by stage. Error bars, SE. Lines connect points to show trends. F., variance of cell size parameters as a function of stage. Dotted lines and solid markers, ploidy most significantly affects the variance of cell and nuclear volume. Related to Table 4 and S1 Data gives detail.
Table 1.
P-value results of pairwise Mann-Whitney testing select conditions shown in Fig 2C, cell volume of near-euploids.
Table 2.
P-value results of pairwise Mann-Whitney testing select conditions for nuclear volume of near-euploids.
Table 3.
P-value results of pairwise Mann-Whitney testing select conditions for N:C ratios of near-euploids.
Table 4.
Percentages of abnormal cell volumes, nuclear volumes, and N:C ratios vs. stage, genotype, and DNA content.
Fig 3.
Ploidy alone cannot explain a majority of cell and nuclear size dysregulation.
63X images shown. A., schematic: subgroups of scaling relationships between DNA content and cell volume: supra: cell volumes larger than expected for ploidy; sub: DNA content higher than expected for cell volume. Colors, individual patient data. Grey box, near-WT, with AT2 volume μ ± σ and near-euploid (n = 1.6–4.4). B., distributions of cell ploidy and volume for individual cells; colored to distinguigh patients. Solid black line indicates simple multiplication of the estimated cell volume and ploidy of a normal AT2 cell, or the equivalent slope; showing +/- 1 S.D., dotted lines. For example, the solid passes through the point (2 x 582 fL, 2 x 2n), where 582 is the mode of the first gaussian fit to the AT2 cell volume distribution (See S6 Fig). The three plots each represent a binning of patients from all stages and both genotypes, but with either no preference, Proportional, or a departure from this line, Subproportional, or Supraproportional. C., top, the fraction of cells classified within each category of abnormal (EGFR+, N = 10, n = 2195, KRAS+, N = 11, n = 1939, S1 Data, tab E). More cells in stage 2 display the supraproportional phenotype, and subproportional cells are enriched in stage 3 KRAS+ cancer. C., bottom, Near-WT, cells with cell volumes close to WT (455–709 fL) that are also near-euploid (1.6–4.4n) are rare in LA, but are depleted in stage 2. Magenta, EGFR+, Blue, KRAS +. D., images of high-quality KRAS+ cell models for each representative category, right, with corresponding 2D images; N, typical AT2 controls. 1, subproportional, 2 proportional, 3 subproportional. White arrows, cells corresponding to models seen in 3D renderings. Scale bar, 20 μm.
Fig 4.
Tumor tissue organization correlates to the extent of cell size dysregulation.
A, Schematic: sheet, solid, and sparse patterns of tissue organization describe most lung tumor data. Top: representative IF signal for three types, immunostained for CK7 (red) and Lamin A+C (green), observed in both EGFR+ and KRAS+ genotypes, EGFR+ patients shown. middle: Renderings of 2D segmentations; nuclei (gray) and cell surfaces (rainbow), Scale bars, 20 μm. bottom: rendering of 3D segmentations, with stromal controls shown as bubble depictions. White box shows perspective scale, 2 μm per small tick mark. B., cell volume distributions taken from segmented image stacks similar in morphology to the canonical ones shown, exclusively from stage 2 LA and from both genotypes. Sample sizes N patients and n cells, parentheses show respective (EGFR, KRAS) samples: NSheet = 4 (1,3), nsheet = 155 (63,92), NSolid = 4 (2,2), nSolid = 438 (275,163), NSparse = 2 (1,1), nSparse = 73 (22,51).
Fig 5.
Near-euploid tumor cells can be purified and sorted by cell size using flow cytometry.
A. 9.43 x 105 total events shown for a dissociated tumor from patient 5296 of unspecified genotype and grade. Signal gates to purify epithelial cells that were alive when fixed, and with near-euploid DNA amounts are shown left-to-right overlaid on fluorescence histograms for the parent digest, top row, and CD45- MACS flow-through, bottom row. The DNA gate is 1.5x the mode shown for CD45-. B., The SSC-H vs SSC-W plot is gated as shown to separate Zombie Red-/EPCAM+/DNA- cells selected from A. C., Contour plot of the SSC-W as a function of DNA for Zombie Red-/EPCAM+ cells, green, as compared with DNA-size-sorted cells as shown. Size-sorted tumor cells have SSC-W that is essentially independent of DNA amount. D., Representative histograms of cell areas measured from brighfield images of cells on a hemacytometer as shown in E., for normal lung tissue, left, and a LA tumor, right. Top to bottom, each row shows the parent, the MACS CD45- flow-through, and the results from the three size gates. Percentages, relative cell yields for gates G1-G4. G4 always contained low yield aggregates. E., brightfield images of cells sorted by gates G1-G3, and with mean cell areas close to those shown in 5D. Scale bar, 20 μm.
Fig 6.
AT2-like cells display cellular processes whose organization is altered in lung adenocarcinoma.
A., 25X images at 0.3779 μm resolution showing a single plane display RAGE, (green) is expressed highly in AT1 cells in normal tissue, top, but we could not detect expression in either genotype of tumor cells bottom. KRAS+ shown. Cells immunostained for CK7 (red) and DNA (cyan). B., 63X images shown. Left three columns, raw data shown in a maximum-intensity projection (MIP) 5 μm thick illustrates the behavior of the AT2 process network in normal distant controls, top, and tumor lung, bottom. Right column, a snap from the Imaris workstation of a MIP over an entire processed image stack shows the overall structure of the process network. Each row shows a different specimen (see Table 5). proSPC (green), CK7 (red). Bottom row, transition regions are observed in both genotypes of LA and contain presumptive diseased AT2-like cells that neighbor enlarged CK7-expressing cells. EGFR+ tumors shown (also S4 and S5 Videos).
Table 5.
Patient identifiers corresponding to data used to create each image shown and select figures.
Table 6.
Primary and secondary antibodies used in this study.
Fig 7.
Enlarged EGFR+ or KRAS+ lung adenocarcinoma cells can proliferate, possess similar protein density to stroma, and display organelle size scaling.
A. Nucleus sizes from proliferating EGFR+ and KRAS+ tumor cells identified by Ki-67 staining, relative to non-proliferative normal AT2 cells (NEGFR = 8, NKRAS = 8, n = 435). B., Zombie Red (red) is a broadly amine-reactive dye that in these FFPE samples reacts equally with small stromal cells, S, as it does with enlarged cancer cells, T, which is inconsistent with cells swelling via water uptake (N = 2). DNA stained with TO-PRO-3 (cyan). Right, Zombie Red intensity of KRAS+ tumors, a proxy for protein density in the cell, does not show the downward trend with cell area expected for water-diluted cells (N = 2, n = 20). C. Allometry of nuclear volume as a function of cell volume for LA (N = 21, n = 2082). Red line, double exponential fit to data. D., Fibrillarin (green) marks nucleoli comparing tumor cells of varying size in a KRAS+ stage 3 patient’s tumor (ID# 4855). CK7 (red), TO-PRO-3 (cyan). E., Total fibrillarin content in nuclei scales with segmented nuclear volume as expected for productive nucleoli (related to D., N = 1, n = 212). F., SON (green) marks nuclear speckles (also KRAS+ patient 4855). CK7 (red), TO-PRO-3 (cyan). G., Quantification of total SON content shows speckles also scale with segmented nuclear volume as they do in living cells (related to G., N = 1, n = 125).
Fig 8.
Diversity of nuclear organization and a revised 2D perspective of human lung adenocarcionoma.
A., typical patterns of DNA staining observed in both EGFR+ and KRAS+ LA. n, normal AT2 cells with proSPC (yellow), CK7 (red), DNA (cyan). gr, granular DNA. cy, cytoplasmic inclusions, with fibrillarin (green). lad, lamin associated DNA, with Lamin A+C (magenta). v, DNA voids. w, “wagon-wheel” DNA. Idealized representations of nuclear structures shown for clarity. See Table 5 for associated patient ID information. B., Visual depiction of the relative equatorial cross-sectional area in μm2 of typical LA cells, if they were spheres, relative to the nucleus (cyan). green, AT2, cyan, DNA, black, DNA voids, magenta, nucleoli/speckles.
Fig 9.
Epithelial cell-size dysregulation is also found in mouse tumors and may proceed with loss of tissue organization and upregulation of transcription.
A., KrasLSL-G12D/+; p53fl/fl; Rosa26LSL-eYFP/LSL-eYFP mouse lung tissue shows nuclear size dysregulation, despite volumes of ~50% that of humans. B., Model 1: Oncogenesis is characterized by lesions, loss of flattening to an AT1 form and thus shortening of processes. At stage I, a contact-dependent mechanism preserves some cell size regulation. Stage II involves more mixed stroma/tumor regions and disorganization, leading to sparse cells and increased cell size dysregulation. Late stage (III-IV) cancers are dominated by adaptation towards maximal cell division rate, evolution of immune resistance, and also loss of cytokeratin staining. Smaller cells, and those of the more subproportional volume:ploidy relationship in KRAS+ cancers (Fig 3) may become enriched in late stages.
Table 7.
Stage, genotype, and pathological classification of lung adenocarcinoma patients.