Quantitative image analysis

To quantify vRNA copies from R-ISH stained tissues, photographic images using epifluorescence were taken with a digital camera, and the TIFF images were used to analyze the area occupied by silver grains using Photoshop (Adobe Systems, San Jose, California) with Fovea Pro 4 (Reindeer Graphics, Asheville, North Carolina), corrected by the background silver grain density of the slide. Section weights were estimated from their volume (5-μm thickness × their area). The number of copies of vRNA was calculated as follows: silver grains observed/cell × 2 disintegrations/grain × 1 Ci ÷ 2.2x10¹² disintegrations/min ÷ exposure time (min) ÷ specific activity of probe (Ci/mmole probe RNA) × 6.02 × 10²⁰ copies/mmole = vRNA copies/cell) [5, 6, 10]. For virions associated with the follicular dendritic cell network (FDCn), silver grains were enumerated over follicles, excluding grains over vRNA+ cells, and the number of vRNA copies in the follicle estimated as just described, dividing by 2 based on 2 vRNA copies/virion to estimate the number of virions.

To quantify RNAscope stained tissues, whole slides were scanned at 40× magnification with an AT2 slide scanner (Aperio Technologies, Vista, California), and regions of interest (ROIs), including B cell follicles (BCF) were saved as TIFF images for analysis. ROIs were analyzed with Photo-shop CS6 (Adobe Systems, San Jose, California) using Noiseware 5 (Imagenomic, Vienna, Virginia) for noise reduction and Fovea Pro 4 (Reindeer Graphics, Asheville, North Carolina) for the analysis. To estimate the number of virions in BCF, ROIs were segmented by reducing noise with Noiseware, duplicating the ROI image, converting the duplicate from RGB to CMYK color mode, thresholding the magenta channel, and then copying and pasting the revised image as a new layer on the original image; the threshold value was determined empirically for each set of slides. To refine the segmentation, Fovea's fill holes command was used to consolidated positive areas, and then the reject features command was applied to remove any objects < 2 pixels or touching an edge of the image. Stained areas of each ROI were recorded with Fovea's area fraction command. Image quantitation data were recorded in text files, and analyzed images were saved in Photoshop format. The fraction of BCF ROIs that stained for vRNA by RNAscope was divided by the average virion size calculated from discretely resolved virions from both SupT1 CCR5+ cells and LTs stained with RNAscope and Fast Red (average virion size = 0.9695 μm²).

To quantify the number of productively infected vRNA+ cells from RNAscope stained tissues, Fast Red staining in an ROI was segmented by reducing noise with Noiseware, duplicating the ROI image, converting the duplicate from RGB to CMYK color mode, thresholding the magenta channel, and then copying and pasting the revised image as a new layer on the original image. The segmentation was refined by morphological opening and closing using Fovea's Euclidean distance map-based morphology operations; when counting positive cells, a morphological opening removed small objects, and then a closing consolidated positive areas within cells. Watershed segmentation was also performed to segment individual stained objects and the Fast Red-positive objects were counted with Fovea. The data were recorded in a text file, and the analyzed image was saved in Photoshop format. The number of vRNA+ cells was calculated based on area (mm²) or per 10⁵ cells. Nuclei in an ROI were segmented by thresholding the red channel with the Pun algorithm in Fovea, morphological opening, watershed segmentation, and then counting the segmented objects. The number of vRNA+ cells was divided by the estimated number of cells determined by the nuclei segmentation in that image and multiplied by 10⁵.

The number of vDNA+ cells was determined by either manually counting by at least two individuals who were blinded to the proportion of 3D8 or ACH2 cells on each slide or quantified using a similar approach to virion quantification with noise reduction and thresholding the CMYK magenta channel as above, and rejecting objects < 9 pixels or touching an edge of the image.