Figure 5

Computational demonstration of quantitative advantage of organ-wide 3D histology. (a) CLARITY enables automated analysis of large populations that provide greater statistical confidence. In a simulated experiment measuring islet radius, N islets were randomly selected, with replacement, from 3 P6 pancreata and 3 P15 pancreata CLARITY datasets, and islet radius means were compared. The proportion of experimental comparisons that reached statistical significance (p < 0.05, unpaired t-test) when the experiment was performed 100 times is plotted against the number of islets measured per experiment. When sampling the size of approximately 250 islets in each pancreas, there is only a 50% chance the data will be significant even though the true trend exists. (b) Large populations greatly reduce experiment-to-experiment variability. Bars show the range of the mean of N islets when a simulation measuring islet size and neural crest interaction was repeated 100 times. The width of the experimental ranges, and hence experiment-to-experiment variability, is reduced with greater sample size per experiment. (c,d) CLARITY provided more accurate measures of islet size compared to 2D analysis. A whole pancreas was analyzed using CLARITY or eight non-adjacent 40 μm optical sections, and raw values for islet radius plotted. Features less than 10 μm were excluded from analysis due to being smaller than an islet (red). False negative values, where 2D estimation of radius would have incorrectly excluded the islet when the 3D size was large enough for analysis, represented 22% of the remaining islets (purple). False positive values, where the islet would have been excluded in 3D but were measured in the 2D analysis, represented 2% of the islets (green), and only 32% of the islets had sizes estimated in 2D that were within 10% of their 3D size. The remainder (68%) of measurements were underestimations of greater than 10%. (e–g) CLARITY provides more accurate measurements of islet number. A 2D and 3D analysis of the same sample demonstrated a significant excess of islets < 50 μm in sectioning, but undercounting of islets > 50 μm. (n = 3) (h) CLARITY can eliminate measurement errors due to sectioning. For each islet, 2D and 3D size was calculated. 2D sectioning significantly underestimated average islet size due to sectioning away from the islet equatorial maximum compared to volumetric analysis. (i) To assess the effect of 2D section thickness on islet radius measures, simulated optical sections of a range of slice thicknesses were analyzed. All 2D sections significantly underestimated islet radius compared to the measure in 3D. (j) Underestimation of islet size in 2D was calculated compared to 3D by taking the difference between values and dividing by size in 3D. 2D slice-based methodologies underestimate islet size by nearly 40%. (k) To explore whether error could be overcome by examining only large structures, the largest 20 islets in each CLARITY and 2D section (or total islets, if <20 were found) were averaged and compared for size. In addition, 20 islets were manually independently and blindly selected and measured in each section. 2D histology systematically underestimated islet radii in even the largest islets by >20% (p < 0.0001, Mann-Whitney test). 4 individual 2D sections shown to illustrate size variability by slice.