Computational image analysis of nuclear morphology associated with various nuclear-specific aging disorders

Abstract

Computational image analysis is used in many areas of biological and medical research, but advanced techniques including machine learning remain underutilized. Here, we used automated segmentation and shape analyses, with pre-defined features and with computer generated components, to compare nuclei from various premature aging disorders caused by alterations in nuclear proteins. We considered cells from patients with Hutchinson-Gilford progeria syndrome (HGPS) with an altered nucleoskeletal protein; a mouse model of XFE progeroid syndrome caused by a deficiency of ERCC1-XPF DNA repair nuclease; and patients with Werner syndrome (WS) lacking a functional WRN exonuclease and helicase protein. Using feature space analysis, including circularity, eccentricity, and solidity, we found that XFE nuclei were larger and significantly more elongated than control nuclei. HGPS nuclei were smaller and rounder than the control nuclei with features suggesting small bumps. WS nuclei did not show any significant shape changes from control. We also performed principle component analysis (PCA) and a geometric, contour based metric. PCA allowed direct visualization of morphological changes in diseased nuclei, whereas standard, feature-based approaches required pre-defined parameters and indirect interpretation of multiple parameters. Both methods yielded similar results, but PCA proves to be a powerful pre-analysis methodology for unknown systems.

Figure 1. Automated pre-processing of nuclear images. (A) Raw images were collected with multiple fluorescence channels (see methods): red and blue channels for Lamin A/C and DNA, respectively. (B) Matlab code segmented the Lamin A/C channel using a level set active contour algorithm to delineate individual borders; here, after 320 iterations. (C) The code then showed the raw and computed nuclear image and allowed input from user to adjust the contour manually by dilating and eroding. Multiple views of the segmented nucleus (left to right: binary segmentation, segmentation with an outline and the result after segmentation) allowed rapid visualization and the possibility for manual adjustment after the auto-segmentation. Pop-up boxes allowed user to confirm segmentations. Only satisfactory results were used for computation.

Figure 2. Feature space analysis of nuclei in aging disorders. Segmented nuclei were analyzed for shape factors (Table 2), and perimeter was normalized to the average perimeter of the corresponding control group. (C and D) indicate control and disease groups, respectively. Bars are color-coded by the parameters. Solid bars indicate that the control and disease groups are statistically different, and outlined bars indicate that they are statistically similar based on a confidence of p < 0.01. The error bars represent the standard error of the mean. Representative images of each group are shown on the left of the graph, all to scale. (A) Nuclei in cells from the Ercc1^−/− mice, a model of XPE, showed altered circularity, perimeter and eccentricity. (B) HGPS patient cell nuclei at passage 22 showed differences in all features. (C) WS cell nuclei showed no differences to control nuclei.

Figure 3. Schematic of principal component analysis (PCA) in geometric space. Two dimensional shapes are assigned polar coordinates (x,y) so that many, disparate shapes can be statistically compared on one graph. From this graph, the principal components of variation can be determined by which lines best identify the largest variance of the data (red lines).

Figure 4. PCA of nuclei in aging disorders. Principal component analysis was performed on control and disease groups of each disease. The panels on the left show the average nuclear shape of both the disease and control groups (red box) and the first 8 modes of PCA for each group. The x-axis shows image variations from the average shape of the nuclei. The graphs on the right show the variance of the control and disease group from the average shape (x-axis) per mode (y-axis). (A) In XFE nuclei, the most significant deformation modes were size and elongation, and the disease group had greater variation; (B) HGPS nuclei showed higher variation and deformation modes of size and elongation, and modes with bumps; (C) WS control and disease groups has similar variations in all modes.

Figure 5. Comparison of control and disease group using PCA. The distribution of control and disease groups in the first three modes of PCA is shown. The x-axis represents the variation from the average, and the corresponding nuclear shape is shown below the x-axis. The y-axis represents the frequency of occurrence. (A) The XFE disease group was statistically larger (Mode 1) and more elongated (Mode 2) and had greater variation (width of the distributions); (B) the HGPS disease group was rounder and slightly more variant; (C) the WS disease and control groups were very similar.

Figure 6. Changes in nuclear shape in cells from HGPS patients with increasing passage. Shapes of HGPS nuclei were compared for multiple passages. For each passage, the distribution of nuclei in the first mode, variances of the first 8 modes, and average shapes of control and disease groups are shown, respectively. The nuclear images are to scale. (A) At early passage (p13), there was some variation between the control (C) and disease (D) group. (B) At passage 22, the difference became larger. (C) At late passage (p30) there were still differences between control and disease, but the control cell nuclei began to show greater cell-to-cell variability. (D) The difference in variance between the control and disease groups for the three passages (ΔVariance = Variance _disease – Variance _control) show a similarity for passage 22 and 30, which are significantly different than passage 13.