In the human genome, there are over 550 genes coding for proteases with another 150 being protease inhibitors . While the majority of protease genes belong to three different classes including serine, metallo- and cysteine proteases, a large portion (~25%) of human protease inhibitor genes belong solely to just one family: the serpin family.
The serpin family comprises hundreds of structurally homologous proteins that are present in both eukaryotic and prokaryotic organisms [2–4]. Its members include inhibitors of serine proteases as well as non-inhibitory members with other biological functions . In humans, a total of 35 serpin-coding genes were identified and grouped into 9 subfamilies designated SERPINA, SERPINB, SERPINC and so on till SERPINI subfamily [2, 3]. Each subfamily has its own specific chromosomal localization and well-conserved exon-intron arrangement. For example, nearly all members of the SERPINA subfamily are located at chromosome 14q32 and genes coding for the SERPINB subfamily are clustered at two distinct chromosomal loci: 6p25 and 18q21 [6–9]. Even for the last SERPINI subfamily, although consists of only two genes with one coding for a brain-associated protease inhibitor SERPINI1  and another for a pancreas-specific protease inhibitor SERPINI2 , its two members are closely adjacent at chromosome 3q26 and share a perfectly conserved gene organization pattern .
SERPINI1, also known as neuroserpin or protease inhibitor 12, is an axonally secreted serine protease inhibitor found in the nervous systems . It functions as a selective inhibitor of tissue-type plasminogen activator and is involved in a wide range of neuronal processes including synaptic plasticity [10, 13], neuronal migration [14, 15] and axogenesis [15, 16]. Genetic variations of SERPINI1 have been reported to render the enzyme conformation locked as an inactive folding intermediate [17–20]. These mutant proteins then undergo spontaneous polymerization within neurons and cause a severe progressive neurodegenerative encephalopathy named FENIB (f amilial e ncephalopathy with n euroserpin i nclusion b odies) [21, 22]. In addition to the mutations at the DNA level, the mRNA expression of SERPINI1 was found to be down-regulated or lost in brain tumors . These abnormalities, along with our latest findings that its adjacent gene SERPINI2 is also down-regulated in cancer , prompted us to study the regulatory mechanism of SERPINI1 gene in both normal and cancerous brain tissues. Interestingly, when the human genome sequence became available, it came to our attention that a gene termed PDCD10 was mapped to lie at less than 1 kb upstream of SERPINI1 in a divergent, head-to-head orientation.
PDCD10 was originally identified as TFAR15 (TF-1 cell a poptosis r elated gene-15) for being up-regulated after the induction of apoptosis by serum withdrawal in TF-1 human premyeloid cells . It was later grouped into a collection of genes thought to be involved in many apoptotic responses and thus renamed as PDCD10 (p rogrammed c ell d eath 10 gene). PDCD10 is highly conserved among different species, with over 99% amino-acid identity among human, dog, mouse and chicken. Although its physiological role remains unclear, a recent study showed that inhibition of the nematode PDCD10 ortholog led to embryonic lethality in 40% of the embryos and a dumpy phenotype in postembryonic viable embryos . Furthermore, PDCD10 mutations were identified in families with cerebral cavernous malformation, indicating that this gene may play a role in the development of cerebral vascular morphogenesis .
It is known that in prokaryotic and lower eukaryotic genomes, genes are often organized in a proximal pattern to facilitate gene co-expression and functional coupling. Increasing evidence suggests that many genes in mammalian genomes also tend to be in close physical vicinity. Recently, by conducting a genome-wide analysis of the human genome, Trinklein and his colleagues identified a class of head-to-head gene pairs whose transcriptional start sites are separated by less than 1 kb . Since mammalian genomes are far more spacious than prokaryotic genomes, it seems unnecessary for two neighboring genes to be arranged in an adjacent head-to-head pattern unless they share a common regulatory system such as an intergenic bidirectional promoter. It is conceivable that bidirectional promoters of gene pairs may be able to co-regulate the genes in a more efficient manner. For example, for homologous gene pairs such as the pair of collagen type IV α1 and α2 genes  and the pair of ATP-binding cassette molecule ABCG5 and ABCG8 genes [28, 29], their bidirectional promoters were shown to co-regulate the two genes thus ensuring the co-expressivity of the gene products. For non-homologous genes like SERPINI1 and PDCD10, however, the need for co-regulation is not as obvious. In addition, compared to all the other genes that are adjacently located to any of known SERPIN or PDCD genes, only the pair of SERPINI1 and PDCD10 genes were arranged in a head-to-head orientation and separated by an exceptionally short intergenic region. This close chromosomal proximity of SERPINI1 and PDCD10 and their divergent configuration are widely preserved in many species, implying that through the evolution there appears to have a need for SERPINI1 and PDCD10 to be tightly linked together to maintain gene co-regulation or functional association.
In this study, we present evidence showing that while SERPINI1 is predominantly expressed in brain and down-regulated in brain tumors, PDCD10 is ubiquitously expressed in all normal tissues but its gene transcription becomes aberrant in different types of cancers. We have also identified a 175-bp regulatory fragment within the intergenic region of SERPINI1 and PDCD10 that functions as a bidirectional promoter. Moreover, a critical fragment, from nt 176-473 outside the minimal promoter in this intergenic region, was found to possess a strong repressive activity for SERPINI1 and enhancing activity for PDCD10. These cis-acting elements may exist to coordinate the expression and regulation of the two flanking genes. To the best of our knowledge, of all non-homologous genes that have been reported to be closely adjacent in the human genome, the intergenic region of the head-to-head PDCD10-SERPINI1 pair at 3q26 exhibits by far the most complex regulatory function that governs the expression of both genes not only through an asymmetric bidirectional promoter, but also through regulations with some other cis-acting elements.