Image processing and analysis for quantifying gene expression from early Drosophila embryos

Abstract

Correlation of quantities of transcriptional activators and repressors with the mRNA output of target genes is a central issue for modeling gene regulation. In multicellular organisms, both spatial and temporal differences in gene expression must be taken into account; this can be achieved by use of in situ hybridization followed by confocal laser scanning microscopy (CLSM). Here we present a method to correlate the protein levels of the short-range repressor Giant with lacZ mRNA produced by reporter genes using images of Drosophila blastoderm embryos taken by CLSM. The image stacks from CLSM are processed using a semiautomatic algorithm to produce correlations between the repressor levels and lacZ mRNA reporter genes. We show that signals derived from CLSM are proportional to actual mRNA levels. Our analysis reveals that a suggested parabolic form of the background fluorescence in confocal images of early Drosophila embryos is evident most prominently in flattened specimens, with intact embryos exhibiting a more linear background. The data extraction described in this paper is primarily conceived for analysis of synthetic reporter genes that are designed to decipher cis-regulatory grammar, but the techniques are generalizable for quantitative analysis of other engineered or endogenous genes in embryos.

FIG. 1.

Parabolic and nonparabolic background forms. Representative parabolic (A) and flat (B) backgrounds. Data were taken from the middle 45–55% egg length of embryos. In general, we observed parabolic background for flatter embryos. Embryo cross-sectional radii were measured and classified as flattened (14–23 μm) or round (24–33 μm). Considerable variability in curvature was noted, but flatter embryos tended to exhibit higher curvature overall. Data were obtained for Giant imaging ventral–dorsal (C) and anterior–posterior (D), and lacZ imaging ventral–dorsal (E) and anterior–posterior (F). Parabolicity of background is measured by curvature of the parabola fit to curve, with apex generally at center of embryo cross section. For anterior–posterior Giant background measurements, young gt^X11 mutants expressing virtually no detectable Giant protein were used. For lacZ background imaging embryos without the reporter gene were utilized.

FIG. 2.

Proportionality of signal to mRNA levels. The average relative amounts of lacZ mRNA in heterozygous and homozygous transgenic embryos were measured by RTqPCR analysis (A) and by immunofluorescent in situ hybridization and CLSM (B). Quantitation of mRNA levels by RTqPCR is representative of two biological assays; error bars indicate standard deviation from five technical replicates. Fluorescent imaging of 39 embryos were quantitated in (B).

FIG. 3.

Proportionality of signal to Giant protein levels. Embryos with apparently normal intensity of Giant (A) and low intensity of Giant (B) were observed for age-matched embryos. Normalized levels of 22 background-subtracted Giant images were plotted, and a roughly 10-fold range of signal intensities were observed (C). Overlapping data points were separated into two columns for clarity for the anterior and posterior stripe.

FIG. 4.

Sampling of integrated mRNA and Giant data. lacZ reporter constructs were introduced into Drosophila by germline transformation and mRNA detected by fluorescent imaging. Dorsal and Twist activators drive lacZ gene expression in ventral regions (A). When regulated by Giant, gaps in this pattern are evident (B). A sampling mesh is imposed upon the background-subtracted image, and values for Giant and lacZ are collected (C, D). The mesh size here is exaggerated for clarity. Sampling proceeds from anterior region of anterior stripe to posterior region of posterior stripe.

FIG. 5.

Gene regulatory representations from Giant-regulated lacZ reporter gene. To generate gene regulatory plots, data derived from slices (lacZ mRNA channel [A] and Giant protein channel [B]) are plotted in a 2D space. Only two slices are shown for clarity. The plots numbered 1–6 show how robust lacZ levels present in ventral portions of the embryo are repressed in regions containing peak Giant levels such as in plot 1–4. In more dorsally located regions (plot 5–6), limiting levels of activators reduce lacZ expression regardless of Giant levels. The overlaid lines, obtained by fitting the data with a rational function, show the trend of the relation between levels of Giant and lacZ.