Quantification Killifish LIver image analysis

RNAscope quantification<br>To quantify the rate of hepatocyte efferocytosis, single Z-plane images were analyzed using the image analysis software QuPath v0.4.351. Cell boundaries were predicted using the “Cell Detection” function with parameters as modified from default settings as follows: Nucleus background radius = 5 µm, Nucleus minimum area = 5 µm2, Intensity Threshold = 5. Cells were then classified into three cell types: hepatocytes, erythrocytes, and efferocytosing hepatocytes using the “Train object classifier” function, which used the “Artificial neural network (ANN_MLP)” classifier option. Hepatocytes were cells with nearly exclusive fgg signal, erythrocytes were called as cells containing nearly exclusive cahz signal, and efferocytosing hepatocytes were called as cells containing cahz signal on a background consisting of primarily fgg signal. Any cells below the detection threshold likely represent other cell types and were ignored for this analysis. The ANN_MLP classifier was trained on manually annotated data from three biological replicates and checked manually for accuracy before being applied to all images in the dataset (40 images total). The percentage of each cell type was calculated using Microsoft Excel v16.75. To quantify ncFem1 expression in female and male liver cells, DAPI, ncFem1, and ppIB signals were quantified by using ImageJ2 v2.9.0/1.53t. For each animal, 2 images from 2 liver sections were processed (4 images per animal, 40 images total). Images were split into their different channels (DAPI, ncFem1, and ppIB) and a uniform threshold range (Image &gt; Adjust &gt; Threshold) of 30-255 for each channel was applied to all images. The “Analyze Particles” function on FIJI was used to obtain the areas of all detected particles, which were summed in Microsoft Excel v16.75 to generate the total area of the signal of interest per image. The mean of the total area was then calculated for each animal and the ratio of ncFem1:ppIB area was calculated. Those ratios were then divided by the median male ncFem1:ppIB area to compare ncFem1:ppIB ratio to a male baseline. To test for ncFem1 nuclear localization, images were analyzed in QuPath v0.4.3 and cell boundaries (including nuclear boundaries based on DAPI staining) were predicted as described earlier for efferocytosis classification. To infer nuclear and cytoplasmic signal strength, the mean signal intensity of ncFem1 and ppIB channels over the nucleus and cytoplasm of each cell were exported from QuPath. For each channel, the nuclear:cytoplasmic ratio was reported for each cell in each image. The mean nuclear:cytoplasmic ratio was calculated per image and the median value per animal was taken.

RNAscope 定量分析
为定量肝细胞胞葬作用（hepatocyte efferocytosis）的速率，本研究使用图像分析软件QuPath v0.4.351对单Z层图像开展分析。采用经默认参数修改后的"细胞检测"功能预测细胞边界，具体参数设置如下：细胞核背景半径=5 µm，细胞核最小面积=5 µm²，强度阈值=5。随后通过"训练对象分类器"功能将细胞分为三类：肝细胞、红细胞及胞葬性肝细胞，该分类器选用了"人工神经网络（ANN_MLP）"分类选项。肝细胞为几乎仅携带fgg信号的细胞，红细胞为几乎仅携带cahz信号的细胞，而胞葬性肝细胞则为在以fgg信号为主要背景的区域内携带cahz信号的细胞。所有低于检测阈值的细胞大概率属于其他细胞类型，本分析中不予纳入统计。ANN_MLP分类器基于3次生物学重复的人工注释数据进行训练，并在应用于数据集全部40张图像前，经人工核验分类准确性。各类细胞的占比通过微软Excel v16.75计算得到。
为定量雌雄肝细胞中ncFem1的表达水平，本研究采用ImageJ2 v2.9.0/1.53t对DAPI、ncFem1及ppIB信号进行定量分析。每只动物取2个肝切片的2张图像进行分析，即每只动物共4张图像，数据集总计40张图像。将图像拆分为DAPI、ncFem1及ppIB三个独立通道，并为所有图像统一应用各通道30~255的阈值范围（操作路径：图像>调整>阈值）。使用FIJI的"分析粒子"功能获取所有检测到的粒子面积，并通过微软Excel v16.75对面积求和，得到单张图像中目标信号的总面积。随后计算每只动物的总信号面积均值，并算出ncFem1与ppIB的面积比值。将该比值除以雄性样本ncFem1:ppIB面积的中位数，以将ncFem1:ppIB比值与雄性基线进行比对。
为检测ncFem1的核定位情况，本研究采用QuPath v0.4.3对图像进行分析，并按照前述肝细胞胞葬作用分类的流程预测细胞边界（包括基于DAPI染色的细胞核边界）。为推断核与胞质的信号强度，本研究从QuPath中导出每个细胞的细胞核及胞质内ncFem1与ppIB通道的平均信号强度。针对每个通道，报告每张图像中每个细胞的核质比。计算每张图像的平均核质比，并取每只动物的核质比中位数作为最终结果。