Patterns of Gene Expression During Drosophila Mesoderm Development Eileen E. M. Furlong, Erik C. Andersen, Brian Null, Kevin P. White, Matthew P. Scott

Several levels of verification were used to assess the quality of the data obtained from these experiments.

1) The experiments were conducted with three independently collected and hybridized samples for each time point (a total of 42 hybridizations). These samples were stage-verified by immunostaining and only tightly stage-matched mutant and wild-type collections were used.

2) The raw data were filtered to eliminate any low-intensity spots, as indicated by not having >75% of the pixels >2 standard deviations above the background signal. Sample and array variability was determined by calculating the correlation coefficient for each gene on the array for all pair-wise combinations of repeated samples. The median correlation coefficient is 0.92 and the median standard deviation/mean is 0.246 (using log₂). To identify genes whose expression differed in the mutant condition, the ratio of mutant-mean to wild-type mean for each developmental stage was used. SAM analysis (Significance Analysis of Microarrays) was used to select a statistically significant group of genes for each of the seven conditions (three twist and four Toll). A two-class unpaired test was conducted using a twofold threshold and a false detection rate of ~2.5%. The total number of significant and false-positive genes is the sum of all the significant genes and false positive genes from each of the seven tests. This identified 1062 genes with significant transcription changes, of which 26 are predicted to be false positives. A number of genes were significant in more than one test (i.e., experimental condition); therefore the 1062 significant genes was reduced to 643 unique genes (with 26 predicted false positives), and this is the data set that was used in this study.

3) The expression profiles for all of the known mesodermal genes match the published data (Fig. 3).

4) The middle developmental time period gave a transcription profile intermediate between the early and late time periods, effectively providing further confirmation of these expression patterns. This is most clearly seen in the graph shown in the paper (Fig. 1B).

5) In situ hybridizations have been conducted on 50 of the genes identified in this analysis. Forty-five are expressed in twist regulated tissues, which include early mesoderm, somatic muscle, visceral mesoderm, heart, pharyngeal muscle, gonadal mesoderm, procephalic mesoderm, hemocytes, fat body, and anterior and posterior midgut. Four are expressed in the amnioserosa (these genes may be involved in immune response, which could be mesoderm dependent). One is expressed in the ventral nerve cord.

Supplemental Figure 1. twist homozygous mutant embryos do not develop any mesoderm. The upper panels show two stage 11 embryos immunostained with antibody to dMef2 (-dMef2). The panel on the upper left shows a wild-type embryo while the panel on the upper right shows a twist^ey53 homozygous mutant embryo. The only cells that immunostain with -dMef2 in twist homozygous embryos are endodermal cells that are misplaced near the pole cells. Toll gain-of-function embryos (Toll^10B) produce ectopic Twist. The two lower panels show stage 5 embryos immunostained with antibody to Twist. The panel on the left is a wild-type embryos while the panel on the right is a gain-of-function embryo for the maternal gene Toll. Note that Twist is produced throughout the embryo in all embryos produced by mothers of this genotype. This study compares the gene transcription profiles of wild-type embryos (Canton S) to twist^ey53 homozygous mutant embryos (mesoderm-deficient) and Toll^10B embryos (mesoderm-enriched) at different stages of development.

Medium version | Full size version

Supplemental Figure 2. The twist^ey53 allele was placed in trans to the Kr-GFP balancer chromosome. A transgenic embryo sorter was used to isolate the non-GFP twist homozygous mutant embryos from their GFP-balancer siblings [E. E. Furlong, D. Profitt, M. P. Scott, Nature Biotechnol. 19, 153 (2001)]. The twist embryo collections were sorted in solution (PBS/2% Tween-20). The wild-type embryos (Canton S) were placed in the same solution for the same amount of time that the mutant embryos were sorting in the machine (usually ~30 to 45 minutes). Sorting and placing Drosophila embryos in solution has little or no effect on their development or viability. Although most of these sorted embryo collections were used to make RNA, a small percentage of the embryos were formaldehyde-fixed and immunostained with -dMef2. These stained embryos were used to verify the presence of the twist phenotype in the sorted embryos and also to ensure that the wild-type embryos and the homozygous twist mutant embryos were tightly stage-matched. The sorting accuracy was >99% and only tightly stage-matched wild-type and twist mutant embryo collections were used for this analysis.

Medium version | Full size version

Supplemental Figure 3. Complete twist/Toll clustergram. The microarrays used for this study were composed of a total of 9200 elements including over 8500 cDNA elements corresponding to 5081 unique genes from the Drosophila RA2 expressed sequence tag (EST) collection [a set assembled by Berkeley Drosophila Genome Project (BDGP) prior to the DGC1 set]. The dataset that is described in this paper consists of all genes that had a twofold change in expression levels and were termed significant by SAM analysis (see Validation section for more details). The data from one representative EST were chosen in cases where duplicate and triplicate ESTs were in the data set. The hierarchical cluster below shows the 643 unique genes that passed these selection criteria. The five example clusters in Fig. 2 in the paper are indicated in red on the left-hand side of the cluster. The annotation for each gene is as follows:

• + or -. A plus sign means the EST has been sequence confirmed on the 5' end. A negative sign means there is no sequence confirmation. The sequence analysis will be described in detail in a manuscript in preparation describing the developmental profile of gene expression throughout the Drosophila lifecycle, by the authors listed below.
• Gene name, if the gene has been previously characterized.
• GO function as reported on Gadfly and in Adams et al., [Science, 287, 2185 (2000)] combined with sequence family groupings from BDGP.
• Blast hit.
• Protein sequence motifs, from Gadfly.
• CG name.
• EST name.
• Cytological location (from FlyBase).
• Our unique identifier for each gene.

Both the sequencing and the annotation of these ESTs was done as a collaborative effort by the following people: M. Arbeitman (B. Baker's lab), E. Furlong (M. Scott's lab), F. Imam (M. Krasnow's lab), E. Johnson (M. Krasnow's lab), B. Null (M. Scott's lab), and K. White.

This file is in Adobe Acrobat PDF format. If you have not installed and configured the Adobe Acrobat Reader on your system, please see Help with Printing for instructions.

Download supplement

Supplemental Table 1. 130 genes had reduced expression in twist mutant embryos compared to staged matched wild-type embryos, referred to as the twist-low genes (these genes had > 2 fold reduction and were statistically significant by SAM analysis (see validation section)).  A Self Organizing Map (SOM) clustering program was used to group these genes in classes with similar trends in expression.

This analysis identified three groups: genes that are reduced in all three timepoints, genes that are significantly reduced only in the early timepoint, genes that are significantly reduced only in the late timepoint.  The graph in figure 1B (in the paper) shows the average gene expression for all the genes in each class (the centroids).  The table below shows the identity and ratio (log₂) of the genes within each of the 3 classes.  The EST that was verified by in situ hybridization from each group (Fig 1C) is indicated in red.

• View the supplemental table for Furlong et al.