On the basis that the vertebrate phylotypic stage should express the highest ratio of genes that conceivably existed in ancestral vertebrates, or Vertebrate genes (see Methods and Figure 1 for the definition of Vertebrate genes, also see Additional file 1: Taxonomic classification of homologues of mouse protein-coding genes according to taxonomic range.), we established the following index:

Vertebrate ancestor index at developmental stage k = V_k/N_k

Where V_k: number of non-redundant Vertebrate genes expressed at stage k; N_k: number of non-redundant total genes expressed at stage k.

Similarly, analyses with Bilaterian, Chordate, Tetrapod, and Amniote genes would also reveal the nature of each embryonic stage in a taxonomic context (see Methods for conserved gene definition).

To analyze the ancestor indexes of mouse embryonic stages, we collected mouse gene sets from the ENSEMBL database 17 and gene expression profiles from the Expressed Sequence Tags Database (dbEST) 18, and considered a gene to be "expressed" if the corresponding UniGene 19 ID was present in any of the categorized EST libraries. EST libraries categorized to each stage are listed in Table 1. Additionally, to make this analysis tolerant to the fluctuation in timing of embryonic stages, we took advantage of moving group analysis (similar to the well known "moving average analysis" technique) to calculate the ancestor index of two sequential stages:

Vertebrate ancestor index at grouped stage k = V_k, _k+1/N_k, _k+1

Where V_{k, k+1}: number of non-redundant Vertebrate genes expressed at stage k or k+1; N_k, _k+1: number of non-redundant total genes expressed at stage k or k+1.

With these methods, we calculated the bilaterian, chordate, vertebrate, tetrapod, and amniote ancestor indexes (Figure 2A–E).

What surprised us initially was the appearance of an early peak around grouped stages 4 to 10 from the analyses of Bilaterian and Chordate genes (Figure 2A,B). Broadly evaluated analysis also supported this tendency; both the early (stages 2–7) and middle embryonic periods (stages 11–18) have significantly higher ancestor indexes than the late period (stages 25–31). The prominently higher ancestor indexes shown in Figure 2A,B indicate the existence of novel, well-conserved stages conceivably carrying characteristics of the bilaterians at around the cleavage to blastula stages in mouse embryogenesis.

We next calculated ancestor indexes to search for a putative phylotypic stage in mouse embryogenesis. In contrast to the bilaterian and chordate ancestor indexes, vertebrate, tetrapod, and amniote ancestor indexes did not show prominent peaks (Figure 2C–E. Instead, we obtained weakly curving graphs peaking around grouped stages 8 to 18 (Figure 2C–E), covering the presumptive periods of the vertebrate phylotypic stage proposed by comparative morphological studies (such as the pharyngula stage 10, the early somite segmentation stage 11, and the tailbud stage 12). Importantly, the vertebrate ancestor index of the middle embryonic period was significantly higher than those of both the early and late embryonic periods (Figure 2C), as also seen in the tetrapod and amniote ancestor indexes (Figure 2D,E). All of these curves (Figure 2C–E) were significantly different from those obtained and shown in Figure 2A,B (P < 0.001; see Methods for the evaluation of ancestor index distribution).

If we were to see the body plan and morphological patterning as readouts of genetic programs for development, then the ancestor indexes calculated for developmental genes would refine our understanding of the conservation of the stages between 8 and 18. On this assumption, we focused on mouse developmental genes defined in the Gene Ontology database (see Methods for the definition of developmental genes) to derive ancestor indexes of individual stages. The highest ancestor index occurred at grouped stage 14 (stages 14–15), and was significantly higher than values of both the early and late embryonic periods (Figure 2H–J). Similar peaks in vertebrate, tetrapod, and amniote ancestor indexes indicate that E8.0–8.5 has been the stage most conserved throughout the evolution of vertebrates. Additionally, the peak of the amniote ancestor index at grouped stage 14 (Figure 2J) indicates that these stages express the smallest ratio of newest developmental genes. In light of the recent report that newer genes tend to evolve faster than older genes 20, the E8.0–8.5 period is the most genetically unmodifiable or contains the strongest developmental constraints in mouse embryogenesis. Thus, our analyses with developmental genes reveal that mouse embryogenesis indeed has a highly conserved stage characteristic of ancestral vertebrates, conceivably at around E8.0 to E8.5 (the period of onset of pharyngeal arch formation and somitogenesis), which is in accord with the concepts of a vertebrate phylotypic stage 6. Notably, many of the mice with mutations in vertebrate developmental genes expressed at the putative vertebrate phylotypic stage have fatal or systemic phenotypes: embryonic or perinatal lethal phenotypes (52%), and abnormal phenotypes of the nervous (33%), craniofacial (14%), cardiovascular (24%), respiratory (12%), or skeletal (18%) system (see Additional file 2: Mutant phenotypes of vertebrate developmental genes expressed at the phylotypic stage). Thus, many of these gene products could be critical factors for defining fundamental morphological features of embryos. This information in turn implies that mutations of developmental genes expressed at the putative phylotypic stage are not easily accommodated, which also supports the idea that this stage tends to be evolutionarily conserved.

Bilaterian and chordate ancestor indexes calculated with developmental genes again showed patterns significantly different from those of vertebrate genes (Figure 2F,G vs 2H–J; P < 0.001), with their highest values at grouped stages 3 and 10. Similar to the results obtained with genome-level analysis, both of these peaks were earlier than the putative vertebrate phylotypic stage, spanning from the period of cleavage to the onset of gastrulation (Table 1). These results suggest that this early embryonic period carries the basic body plan of bilaterians (i.e. bilaterally symmetric), making it the "bilateriotypic" stage. Gene expression information of Ciona intestinalis embryogenesis further allowed us to verify the existence of the putative bilaterian-related stage in this organism at the early embryonic stage (see Additional file 3: Ancestor indexes obtained for genes expressed during Ciona intestinalis embryogenesis). The highest value of bilaterian ancestor indexes of genes expressed during C. intestinalis embryogenesis occurred at the cleavage stage, and was significantly higher than the ancestor index of the later embryonic stage (combined expression data of larval and juvenile stages). However, we could not obtain evidence at the level of "developmental genes" owing to the scarcity of gene ontologies associated with C. intestinalis genes.

Genome and homologue information for each gene were taken from ENSEMBL database 17 (v37, February 2006) using the data mining system BioMart 23. Only "protein coding" gene types were used in our analysis (total of 25,613 ENSMUSG IDs in the Mus musculus genome; 14,278 ENSCING IDs in the C. intestinalis genome). Homologues represent reciprocal BLAST hits (including UBRH: unique best reciprocal hits; MBRH: multiple best reciprocal hits; and RHS: reciprocal hits based on synteny information) with an expectation (E)-value < 1e-10 (WU BLASTP, BLOSUM62). The genome data sets were versions NCBI m34 for M. musculus and JGI 2 for C. intestinalis. Taking gene loss during evolution into consideration 24, we defined the following evolutional classifications, which depend on the presence or absence of at least one homologue in a taxon of interest (see also Figure 1). Bilaterian genes: at least one homologue present in protostomes. Chordate genes: in addition to genes classified in Bilaterian genes, genes that have at least one homologue in urochordates. Vertebrate genes: in addition to genes classified in Chordate genes, genes that have at least one homologue in teleosts. Tetrapod genes: in addition to genes classified in Vertebrate genes, genes that have at least one homologue in amphibians. Amniote genes: in addition to genes classified in Tetrapod genes, genes that have at least one homologue in aves.

All the stage-traceable EST libraries of M. musculus and C. intestinalis were manually downloaded from the NCBI UniGene database 19 and categorized by their developmental stage descriptions (Table 1). We extracted all UniGene IDs with at least one entry in each categorized EST library, and further linked them to the ENSEMBL gene IDs. ENSEMBL genes with at least one of these UniGene IDs were considered as "expressed".

For comparing ancestor indexes between stages, we constructed a 2 × 2 contingency table and determined significant differences between the ratios (e.g. non-redundant Vertebrate gene numbers expressed at stages k or k+1 and non-redundant total genes expressed at stages k or k+1) by Fisher's exact test (two-tailed). To test the distribution pattern of ancestor indexes, we converted the number of expressed Bilaterian, Chordate, Vertebrate, Tetrapod, and Amniote genes in each grouped stage to cumulative frequency distribution, and compared these survival curves (Kaplan-Meier method) by the Generalized Wilcoxon test. For all the statistical tests, P-values less than 0.05 were considered significant.