Acquisition of quantitative data on gene expression
in individual embryos
1. Confocal scans of embryo.
The panel of antibodies to Drosophila segmentation genes was used to fluorescently stain embryos as described (Kosman98).
Gene expression is monitored by confocal scanning microscopy of fixed embryos stained with fluorescence tagged antibodies. The scans were
obtained using the x16 oil immersion plan objective of Leica TCS4D confocal microscope. Three 8-bit
channels of the microscope were used to detect the proteins separately. The dyes were excited
with single wave-length at a time to ensure no leakage between channels, using a BP-FITC filter
with a 488-nm excitation line for FITC and BP-60030 filter with 568-nm excitation line for
Texas Red and the RG655 filter with the 647 nm excitation line for Cy5.
For each stain two 1024 x 1024 pixel images of the blastoderm at two-micron depth intervals were
obtained.
2. Preparation of whole embryo mask
Next, the three channels are compared at each pixel and the maximum value is selected for the pixel maximum image, which forms the basis of the whole embryo mask image.
This mask is needed in the subsequent processing steps which function is to put all embryos into a common orientation before data collection can occur, as described below.
Download description of this step in PDF format
Download Khoros 2.2 workspace:
import.wksp
3. Examination of whole embryo mask to decide necessary processing steps before image rotation and cropping
Thus, each embryo image presents a unique image processing problem to prepare it for data collection. In dealing with hundreds of such embryo images, it was observed that roughly five cases are encountered.
The main tool that guides the processing steps is the whole embryo mask. A mask is a binary image (i.e. composed only of full "on" or full "off" pixels), in which "on" corresponds to regions of high signal and "off" to regions of no or little signal. Thus, the mask provides an easy-to-interpret representation of the size and shape of the embryo-of-interest region.
During examination of different cases of the mask cleaning procedure, more difficult problems are reduced stepwise to simpler ones, until all can be treated in the same way.
Download description of this step in PDF format
Download Khoros 2.2 workspace:
show.wksp
4. Mask cleaning procedures for different cases
Re-masking the embryo image
Changing the threshold level from 5 to 10 accomplishes this. Further the mask should be processed as in cases 3 and 4.
Download description of this step in PDF format
Download Khoros 2.2 workspace:
remask.wksp
Manual extraction of the region, which connects the desired embryo with the nearby embryo
Download description of this step in PDF format
Download Khoros 2.2 workspaces:
Cleaning of the embryo by simple segmentation of the whole embryo mask
Download description of this step in PDF format
Download Khoros 2.2 workspaces:
Correct whole embryo mask
5. Rotating the embryo image to align A-P axis with the horizontal
It can be applied to all embryos, including embryos with modified whole embryo mask.
Download descriptions of this step in PDF format
Download Khoros 2.2 workspaces:
6. Flipping embryos to a common orientation
The resultant image is named as Embryo Image.
Download description of this step in PDF format
Download Khoros 2.2 workspaces:
7. Making a smooth whole embryo mask in preparation for data collection
This in necessary because we would like to make expression data relative to the physical dimensions of the embryo.
Download description of this step in PDF format
Download Khoros 2.2 workspaces:
8. Obtaining the numerical expression data from the embryo image
Overview
First step is the conversion of 3-channel pixel maximum into the binary nuclear mask.Here mask means exactly the same as before, but it contains many elements, composed of contiguous "on" pixels, separated by intervening "off" pixels. Each of these elements corresponds to a single nucleus in the embryo image. The idea is to convert the embryo into a form in which the computer can "see" individual nuclei. After that it can measure their location within the embryo, their average brightness, and tabulate the results.
Thus, the embryo image is boiled down to 6 columns of numbers which describes the location of ~2000 nuclei and a quantity value for each nucleus of three different gene products.
9. Embryo image segmentation to construct the binary nuclear mask
Next, this watershed image is used to define the nucleus scale brightness gradients which are a very good approximation of the edges of the nuclei. Once these nucleus borders have been established, the image has been effectively "segmented": each "segment" becomes a single element of the nuclear mask.
Download descriptions of this step in PDF format
Download Khoros 2.2 workspace:
data.wksp
Download description of this step in PDF format
Download all preccessing steps PDF's in ZIP file
Download all workspaces in ZIP file
Back to preccessing steps list
10. Reduction of embryo images to expression data
First labeling of the binary nuclear mask makes groups of pixels (nuclei)
and gives this group a numerical label.
Shape analysis calculates the average x and y location of all pixels
in the group (centroid) thus defining the position of the nucleus.
The blob extractor decomposes a grayscale expression pattern into a
series of subimages corresponding to individual nuclei. It operates separately
on each of the three grayscale channels, and the three intensity averages
are combined with the position measurements of each nucleus, which are
expressed as a percentage of the total length and height of the embryo.
Download Khoros 2.2 workspace:
data.wksp
