Pseudogenes are inactivated copies of known genes, which present an erosion or disruption either of their reading frames or their regulatory regions. They are usually detected by comparative analysis with their functional orthologs in close relative species. The alignment of such orthologous genes allows the detection of mutations, insertions/deletions (indels), frameshifts, premature stop codons and insertion of transposable elements, all of which cause the appearing of a truncated protein, or the inactivation or loss of essential functional domains. However, it is not always easy to assess the functional status of each annotated coding region within a genome for several reasons. First of all, orthology cannot be detected above a threshold of divergence between sequences. Second, not all frameshifts, indels, stop mutations and rearrangements will inactivate a gene and there is a wide length variation of homologous proteins within families. An additional problem is that there is no clear definition of what has to be considered a pseudogene and there are inconsistencies in the methods used in different studies, so that it is not always possible to compare the pseudogene content from different genomes. In some studies, any shortened open reading frame (ORF) is annotated as a pseudogene [23], while in others considerably shortened homologs are annotated as genes if they keep complete domains specifying some function [24]. Lerat and Ochman (2005) [25] considered a pseudogene only when more of 20% of the length of the primary sequence of the encoded protein was lost, while some other studies excluded from this category impairments in homology matches within a “cutoff” region at either end, considering that slightly shorter alignments can reflect functional protein [26]. The most widely used method to identify non-functional genes involves the comparison of nucleotide substitution rates at synonymous (K_s) and non-synonymous sites (K_a), using the K_a/K_s test [27]. Regions without functional constraints, such as pseudogenes, are expected to have K_a/K_s ratios not significantly different from one. However, comparative analyses can only detect pseudogenes in genes that have orthologs in other available genomes, and only if the inactivation is due to truncations or disruptions of the original ORF, which represents only a fraction of the potentially inactivated genes. Genes that have been inactivated by missense mutations or changes in regulatory regions will also remain undetected by these methods.

At the beginning of the Genomics era, pseudogenes were thought to be unusual in bacteria. This idea has completely changed now that it is possible to perform comparative genomics on different strains of the same species, on closely related species with different lifestyles or on those living in different environments. In fact, when the E. coli MG1665 genome was first reported, only one pseudogene was annotated among its 4288 coding regions, but recent studies revealed that this genome contains about 100 genes that retain less than 20% of the annotated sequence in other E. coli strains, some of which (at least) are likely to be non-functional [28]. Nevertheless, pseudogenes were early noticed in significant amounts in some bacterial pathogens such as R. prowazekii [13] or M. leprae [8], the latter being one of the most extreme cases known to date, with 1614 protein-coding genes and 1133 annotated pseudogenes, 40% of its 3.2-Mb genome. Since then, several comprehensive studies have been performed trying to extract general rules that explain the dynamics of pseudogenes within bacterial genomes.

Liu et al. (2004) [26] analyzed 64 prokaryote genomes and defined a conservative set of conditions to detect their proportion of pseudogenes. In this analysis, they do not include hypothetical or putative proteins, because a large proportion of them could be over-annotated. However, in order to maximize the efficiency of pseudogene finding, they searched for remnants of ancient genes in the regions that had been considered as IGRs in the original annotations. Using this approach, they found that pseudogenes are pervasive in prokaryotes, accounting for 1 to 5% of most analyzed genomes. They didn’t find a clear correlation between percentage of pseudogenes and lifestyle: taking aside the extreme case of M. leprae, the pseudogene fraction of the analyzed archaea, non-pathogenic and pathogenic bacteria were fairly similar (3.6, 3.9 and 3.3% respectively). However, and as mentioned in the introduction, it has been shown later on that some bacteria that have recently adopted an obligate host-dependent lifestyle (either mutualistic or parasitic) present higher amounts of pseudogenes, probably related with a rapid change to the new environments in which some previously encoded functions are no longer needed. The rapid adaptation to new environments also explains the situation of M. leprae and its differences with its close relative M. tuberculosis [29] in that, with 3959 protein-coding genes, has only six identified pseudogenes. A comparative analysis of both genomes suggested that the pseudogenes in M. leprae have degenerated by gene-by-gene inactivation mostly after the divergence of these two clades [30]. Additionally, genomes from host-dependent symbionts that are suffering a genome reductive syndrome are known to contain many pseudogenes in regions that were previously annotated as IGRs. Thus, when the genome of R. prowazekii, the a-proteobacteria that causes typhus was sequenced [13], it was found that a considerable portion of its 1.1-Mb genome (22.9%) is covered by ncDNA and pseudogenes. A comparative analysis of this genome with the genome of three representative species of the genus Rickettsia showed that most of the IGRs were, in fact, remnants of ancestral genes with a high degree of degradation [24,31]. Even in smaller genomes, such as those from four different strains of B. aphidicola (the bacterial endosymbiont of aphids), longer IGRs in one strain contain the remnants of lost genes in other lineages [32].

As it would be expected, the pseudogenes identified by Liu et al. (2004) [26] cluster mainly in specific families related with environmental responses (such as specific nutrient transporters or processing, and related with antigenic variation). Other disabled genes correspond to hypothetical and unknown proteins, belong to transposable elements and bacteriophages, or are remnants of horizontally-transferred genes. Later on, Lerat and Ochman (2005) [25] determined the pseudogene content of 11 available complete genomes from four bacterial genera (Staphylococcus, Streptococcus, Yersinia and Vibrio), each of which include at least one human pathogen, which were supposed to accumulate pseudogenes compared with their free-living relatives. In addition to the gene families identified in the previous study [26], in obligate host-associated bacteria several broadly distributed genes involved in nucleotide processing, repair or replication appear as pseudogenes. The loss of genes needed for DNA repair and recombination are also among the first losses detected in bacteria that have recently acquired an obligate endosymbiotic lifestyle [33], thus contributing to the accumulation of pseudogenes in these species [34].

Mobile genetic elements, such as phages, plasmids and transposable elements are widespread in prokaryotes, where they can represent a significant proportion of their DNA and play important roles in shaping their genomes. The simplest and most abundant transposable elements in bacteria are ISs. Although several classification schemes have been proposed, the one defined by Mahillon and Chandler (1998) [35] has become the most widely used. It is the format used by ISfinder (www-is.biotoul.fr) [36], a repository of ISs isolated from bacteria and archaea that also provides extensive background information, as well as an updated and comprehensive classification in IS families and subfamilies, and a proposal for a coherent nomenclature. These small elements are very compact and consist on a short DNA sequence, usually between 0.6 and 2.5 kb in length, able to translocate within and among replicons. A typical IS element only codes for proteins involved in its transposition, flanked by short inverted repeats (IR) of around 10 to 40 bp. Usually, their transposition generates a small duplication (2 to 14 bp) of the target DNA flanking the insertion point. Based on similarities in their transposases and IRs, and the length of their target site sequence, ISs can be grouped into 20 major families. IS elements are important factors involved in genetic variability, since they can promote genomic rearrangements, and generate mutations by their insertion within genes or regulatory sequences. This capacity to generate genome damage is probably one of the reasons why transposition is usually strongly regulated and maintained at a low level in free-living bacteria [37]. In fact, most bacterial genomes contain only a few copies of a limited number of IS types [38]. However, in bacteria that have recently adopted an intracellular lifestyle, there is a reduction in the selective pressure to keep the ISs regulated, and a massive expansion of those elements can take place (see below).

As it happens with the pseudogenes, the quality of IS annotation in sequenced genomes is highly heterogeneous due to the different annotation methods and nomenclature used by different researchers, and the diverse interests that drive the functional analysis of a given genome. In fact, it is not uncommon that annotation focuses only on the potential protein-coding sequences included in the elements, but ignores their IR boundaries and the directed repeats generated by their insertions, as well as the vestiges of ancestral ISs [36]. Nevertheless, several global surveys of available bacterial genomes have been performed in recent years, allowing conclusions to be obtained regarding the distribution, dynamics and evolution of these elements.

Many different hypothesis have been proposed to explain the abundance and variability of IS elements among prokaryote genomes. Wagner et al. (2007) [38] used IScan, a free open source package, to examine the more than 2000 IS elements present in 438 completely sequenced bacterial genomes in the most comprehensive analysis to date. They found a high homogeneity pattern of IS families across vast taxonomic scales, consistent with previous and more limited works [39,40,41]. Such high-sequence homogeneity could be explained by the rapid spreading of ISs within a genome, in addition to other genetic mechanisms such as gene conversion. In the evolutionary scenario proposed by Wagner et al., after an IS enters a genome, its copy number expands rapidly through transposition. Consequently, there is a low degree of diversity among the different copies of an IS present in a given genome. Eventually, due to the deleterious effect of their accumulation, IS elements tend to disappear and may become extinct from the lineage. However, later on, it may be reintroduced through HGT. The impact of HGT in the spreading of these sequences is revealed by the presence of closely related IS elements of the same family in non-closely-related genomes. However, HGT appears to be necessary but not sufficient for the presence of ISs, since their abundance within a genome does not depend on the level at which genomes are invaded by the elements [42].

Touchon and Rocha (2007) [42] performed a comprehensive reannotation and analysis of the putative functional ISs present in 262 prokaryote genomes in order to test for IS family specificity, the influence of host genome size, pathogenicity, or human association in the IS abundance or density. They found that IS distribution in prokaryotic genomes strongly correlates only with genome size, probably due to a decrease on the density of highly deleterious insertion sites with genome size. They hypothesized that IS abundance is mostly determined by selection and, therefore, the effective population sizes of the microorganism would strongly influence IS abundance. Thus, the high increase in the amount of ISs that has been detected in some bacterial pathogens, even though they have smaller genomes, would be the consequence of a recent reduction of effective population sizes, not of their parasitic lifestyle. However, it must be taken into account that it is their host-dependent lifestyle that determines that many genes become redundant or superfluous, which also diminishes the effectiveness of natural selection, because there are more sites that can be substrate of IS transposition without affecting the fitness of the microorganism.

On a more recent study, Newton and Bordenstein (2011) [43] analyzed 384 bacterial genomes on the light of their phylogenetic relationships, genome sizes and ecology, in order to test whether there is a correlation between any of these factors and the mobile elements density (which in this study included phages, plasmids and transposable elements). They found that the density of mobile DNA elements only correlates with bacterial ecology. ISs account for the majority of mobile DNA elements in nearly half of the analyzed genomes, and there is a significant variation on the amount of them depending on the bacterial lifestyles (free-living, facultative intracellular and obligate intracellular bacteria), and also between horizontally and vertically transmitted obligate intracellular bacteria. The increase in the amount of IS elements in bacteria that have recently evolved as specialized pathogens or acquired an intracellular lifestyle had already been noticed in previous studies [44]. Since these bacteria sequestered inside eukaryotic cells can not exchange material with other bacteria through HGT, the massive presence of ISs must be due to an increase in the replicative transposition of elements that were resident at the onset of the obligate symbiosis [45], when many genes that had become non-essential can accumulate ISs without a detrimental effect. In addition, the high abundance of IS elements is also an important source of chromosomal rearrangements suffered by these genomes at this point [46]. Some studies emphasize that the persistence of IS elements in bacterial genomes can not be only explained by the balance of their propagation as selfish elements and the control of the host genome to avoid deleterious effects, but also due to their ability to promote adaptive evolution of the host genomes by generating beneficial mutations that increase the fitness of the host [47]. However, such benefits do not seem to be acting on extremely reduced genomes with a large evolutionary history with their hosts, since IS elements have been completely lost in their genomes. Their absence in the latter stages of the endosymbiosis can mostly explain the extreme chromosomal stasis observed in the genomes of the different B. aphidicola strains that have been sequenced [48].

In recent years, new information has been obtained by the study of bacterial endosymbionts in different stages of their relationship with their respective hosts. Several cases are especially useful to illustrate the dynamics of junk DNA in the genomes of bacteria that acquire a specialized lifestyle and will be analyzed in detail in this section and summarized in Table 1.

Comparative studies have revealed that bacterial genomes are under selective pressures that have a deep impact on their shape and gene content, including a previously unexpected degree of diversity in gene repertoires within and between species. In addition to a set of genes that are shared by all members of a monophyletic group, there are a considerable number of genes and other functional features that can be highly specific depending on the environmental context. The evolution of these gene repertoires can be explained by processes of gene gains through HGT and duplications, and gene loses, through pseudogenization and gene excision [28]. Since it is well known that HGT has a great impact on prokaryotes [64], reductive evolutionary processes due to gene degradation and elimination must also occur very frequently in order to maintain compact genomes [31,65]. Pseudogenes must be quickly removed from the genomes, as it can be deduced from the small proportion of them those that are present simultaneously in different strains. When mutations occur in genes that are no longer needed as an adaptation to particular living conditions, the inactivated genes can remain in the genome for some time and they gradually erode in a random manner until they are completely removed [57,66], because within bacteria (as well as within several eukaryotes) there is a mutational bias toward deletions over insertions [31,67,68,69].

The dynamics of degradation and elimination of junk sequences has been studied in intracellular specialists by comparative genomics with their free-living relatives (Figure 1). Free-living bacteria have larger genomes with a moderate content of repeated sequences and self-propagating DNA, such as transposons, bacteriophage, a moderate amount of pseudogenes (usually between 1% and 5%) and relatively short IGRs. In the transition from free-living to host-dependent lifestyles, junk DNA increases their content, leading some times to an expansion of genome size parallel to a decrease in gene-coding capacity. The accumulation of pseudogenes in different degrees of decay (sometimes leading to long IGRs in which no traces of former pseudogenes can be detected except based on genome synteny studies), as well as the accumulation and losses of IS elements in the genomes of these bacteria can be explained by genetic and population factors. At the beginning of the association, the new rich, protected and stable niche provided by the host makes unnecessary or redundant some gene functions. Due to a decreased efficiency of the purifying selection, they can rapidly accumulate slightly deleterious mutations on genes that have become non-essential. The inefficiency of the purifying selection also explains the increase in the amount of IS elements present in these genomes at the onset of symbiosis, which in some cases seems to be accompanied by an increase in the transposition rate and the concomitant production of pseudogenes created by IS insertions. At the same time, the random genetic drift increases due to a drastic reduction of the bacterial effective population size when they are transmitted from one host to another. The uncontrolled proliferation of IS elements can lead to the loss of large stretches of genomic DNA by unequal recombination when they appear in direct orientation. This was one of the predicted effects in the model for the reductive syndrome of endosymbiont genomes, which takes place in two stages: first, a massive reduction and a second stage of pseudogenization and progressive disappearing of non-functional sequences by gene-by-gene erosion and deletion, leading to the highly compact and streamlined genomes of symbionts with a long established obligatory intracellular relationship with their hosts. IS elements also suffer the accumulation of mutations that render them inactive for transposition, thus becoming real junk DNA. The process occurs in a random manner so that some genomes can accumulate dramatic amounts of ISs (as in the case of SOPE) or lose most of them at the end of the first stage. The possibility for rapid genome degradation by means of deletion events is unlikely in the latter genomes and therefore, as it is the case of S. symbiotica SCc, they enter the second step of gene-by-gene degeneration while still retaining great amounts of pseudogenes which, with evolutionary time, will appear as long IGRs.