Screen image analysis

All image processing was done in the open source software package, ImageJ using custom plugins available at http://research.stowers.org/imagejplugins. First, DAPI images were thresholded with the default ImageJ autothreshold (IsoData) method. Nuclei adjacent to the image border as well as nuclei with area under 4 pixels or over 1000000 pixels were eliminated. The cytoplasmic region was estimated as a 4 iteration dilation of the nuclear region, where closely spaced nuclei were not allowed to merge. The minimum intensity of the GFP image (ignoring dead pixels) was subtracted from each image as an estimate of background intensity. Tests with more complex methods like rolling ball background subtraction did not change the resulting statistics significantly. Then the average intensity of each nucleus and associated cytoplasmic region were measured and the nuclear/cytoplasmic ratio was calculated.

All pilot screens with DMSO showed a biphasic response with nuclear enriched (nuclear/cytoplasmic ratio ~3) and unenriched (nuclear/cytoplasmic ratio ~1) populations. On a logarithmic scale, the midpoint between these populations is 1.2. Therefore, 1.2 or in other words 20% above the cytoplasmic value was set as a threshold for “nuclear” classified cells. For each screen image, the fraction of nuclear cells was calculated along with the standard error in the mean for the 3 images. 28 control DMSO wells were scattered on each plate. They did not show a spatial or plate dependence, so they were pooled for hit analysis. Wells were considered hits if the fraction nuclear was more than 3 standard deviations different from the DMSO distribution center. Given that the screen was run twice, hits were listed in S1 and S2 Tables if they were hits in either run.

P values for hits were calculated as follows. The average fraction nuclear and its standard error were calculated from all six images (three from each run). The t distribution was then used to calculate the probability that these values were greater than or equal to the DMSO average for positive hits (nuclear export) and less than or equal to the DMSO average for negative hits (nuclear enrichment).