Pseudogenes can be classified into two main types, processed and duplicated [6]. Processed pseudogenes are generated via retrotransposition of the mRNA of their parent genes. After mRNAs of the parent genes are transcribed in the usual fashion by RNA polymerases, they are reverse transcribed and integrated into genomic DNA by reverse transcriptases and endonucleases encoded by long interspersed nuclear elements (LINEs) in primates and humans [14,5,17]. Because these pseudogenes are generated through mRNA intermediates, they are notable for their lack of introns, spliced out during mRNA maturation. On the other hand, duplicated pseudogenes are generated via direct DNA-to-DNA duplication followed by integration into genomic DNA and eventual disablement [18]. They retain most of the exon-intron arrangements with possible duplication of upstream and downstream regions.

We have developed computational methods for cataloguing processed and duplicated pseudogenes [19,3,4,20,2]. First we identify pseudogene candidates by aligning the genome in all six frames of the translated amino acid sequences to the known proteins in the organism [21]. Then we distinguish pseudogenes from their parent genes by identifying disablements such as insertions, deletions, and nonsense mutations, as these would interfere with the potential transcription and translation of the pseudogenes into a fully functional protein.

Because pseudogenes are released from the pressures of natural selection, they capture the sequences of genes at points in time and are subsequently subject to mutations at a neutral rate [22]. Understanding the subtleties of pseudogenes that effect their inactivation would aid in predicting genes de novo from genome sequences [23-25]. In addition to their passive role as genetic fossils, the functional roles of pseudogenes are still being characterized. Pseudogenes have been found to interact with the mRNA of their parent gene [26-28]. Some pseudogenes have also been implicated in chromosomal recombination and gene conversion events leading to diseases because of high sequence homology to their parent genes [7,29]. Others have been reactivated and become fully expressed variants of their parent genes [30].

In order to characterize the factors influencing the generation of pseudogenes, it is useful to study a selected set of genes that are common to multiple species and have many associated pseudogenes [22]. We identified such a set that encodes the enzymes in glycolysis, a fundamental metabolic pathway conserved since ancient anaerobic prokaryotes. Using our pseudogene pipeline, we assembled the first detailed catalog of the processed and duplicated pseudogenes of glycolytic enzymes in the well-annotated eukaryotic genomes: human, chimpanzee, mouse, rat, chicken, zebrafish, pufferfish, fruitfly, and worm genomes [20,31-39]. By comparing pseudogenes of orthologous genes in multiple genomes, we are able to identify general characteristics as well as species-specific characteristics. The dates of species divergence can be used as landmarks in the temporal evolution of the glycolytic pseudogenes.

Despite the high degree of conservation in the glycolytic enzymes, there is much more variation in their pseudogene abundances. Some genomes, like chicken, zebrafish, pufferfish, fruitfly, and worm, have very few or none, while others, like mouse and rat, have hundreds. The differences in pseudogene abundances alone suggests significant differences in the processes of gene expression, duplication, and retrotransposition in the different genomes. Previous studies have suggested that the difference lies in the prolonged lampbrush stage of oogenesis in mammalians as compared to non-mammalian organisms [48,49].

Most glycolytic pseudogenes are processed and can be assumed to be retrotransposed from an mRNA intermediate. It is possible that certain sequences intrinsic to the GAPDH and LDH genes may predispose them to be preferentially retrotranscribed, inserted, and preserved in the genome. These pseudogenes are classified as processed and not duplicated indicating their formation was the result of a retrotransposition event of the parent gene, rather than a duplication event. However, we must consider the possibility of formation of a processed pseudogene through a retrotransposition event and its subsequent duplication giving rise to so called "duplicated-processed" pseudogenes. Thus, while duplicated pseudogenes result from the duplication of parent gene, duplicated-processed pseudogenes result from the duplication of a processed pseudogene [50,51]. One way to differentiate processed pseudogenes from duplicated-processed pseudogenes is to check if the segments of the genome surrounding a pair of processed pseudogenes are also similar. Hence, we checked for the presence of 60 processed pseudogenes of human GAPDH in duplicated regions of the genome called segmental duplications [52]. A pair of processed pseudogenes located in segmental duplication pairs indicates that one of the pseudogenes was likely formed by the duplication of the other one and hence is a duplicated-processed pseudogene (Figure ​(Figure6).6). We identified eight duplicated-processed pseudogenes by this analysis, listed in Additional File 1. However, six of those eight pseudogenes occupy > 77% of the segments that are duplicated and could be the result of independent retrotransposition events. In this scenario perhaps the high sequence similarity of these segments led to their annotation as segmental duplications.

As a coincident finding, GAPDH has many more biological roles outside glycolysis as compared to the other glycolytic enzymes. For example, GAPDH functions in DNA repair, telomeric DNA binding, transcriptional regulation, nuclear RNA export, apoptosis, membrane fusion, phosphorylation, tubulin bundling, and sperm motility [53-59]. Because the molecular processes of retrotransposition are separate from the enzymatic functionalities, we can only speculate that the preponderance of non-glycolytic roles may be correlated to the enrichment of GAPDH pseudogenes.

In an intergenomic analysis, GAPDH pseudogenes have about five- to six-fold greater abundance in the rodent genomes as in the primate genomes even though overall the mouse genome was found to have about half as many pseudogenes as the human genome [3]. The mouse genome has higher rates of nucleotide substitution, insertion, and deletion [33] than the human genome, leading to a higher rate of pseudogene decay. However, the higher rate of pseudogene decay seems to have preferentially spared the GAPDH pseudogenes.

To further characterize the molecular history of pseudogenes in the human, chimpanzee, mouse, and rat genomes, it was necessary to identify the pseudogenes that were most likely present prior to the primate-rodent ancestral divergence. We used orthologous genes to identify regions of synteny between primate-rodent genome pairs. This approach is based on the assumption that gene-coding regions are much less variable than intergenic regions because of functional constraints and are therefore more reliably matched between genome pairs.