Automated Detection of siRNA Cytosolic Release

HeLa cells stably expressing YFP-galectin-9 were imaged using live-cell microscopy for 8 h during treatment with lipoplex-formulated siGFP-1 at 2:4 pmol:µl ratio of siRNA to LF2000. Endosomal siRNA release was then automatically and manually detected in each cell. For manual event detection, each cell was observed frame by frame until a release event was visible or until the end of the acquisition. For event detection, both automated and automated combined with manual quality control, the procedure was performed as described above. De novo recruitment and colocalization of galectin-9 with siRNA-lipoplexes was then manually evaluated in all cells and the first galectin-9 positive event was recorded. Cells located partially outside the image border at the time of siRNA release or galectin-9 recruitment were excluded. For the manual and automated detection of cytosolic siRNA release, cells were evaluated up until the first detected event. To determine the sensitivity and specificity of the cytosolic release detection to correctly identify the first siRNA release event in an evaluated cell, observations were classified as follows: cytosolic release events detected within five frames before or after galectin-9 recruitment to the releasing lipoplex were considered true positive observations. If no cytosolic siRNA was detected even though galectin-9 was recruited to a visible lipoplex, the observation was considered false negative. Observations were classified as true negative if no cytosolic siRNA release was detected and no recruitment of galectin-9 to lipoplexes could be observed, and false positive if the cytosolic release was detected in the absence of observable galectin-9 recruitment to the releasing lipoplex within five frames before or after the detection. Manual quality control was then performed of all siRNA release events identified by the automated detection algorithm, providing the opportunity to correct false positive events, that were then reclassified as true negative observations. This approach is analogous to the manual quality control of release events identified in experiments evaluating d1-eGFP knockdown (without galectin-9 reference). Release detection sensitivity was calculated as all true positive events divided by the sum of true positive and false negative observations. Specificity was calculated as true negative observations divided by the sum of true negative and false positive events.