Some reports have appeared in the recent literature advocating constructive neutral evolution (CNE) to explain the evolution of apparently complex cellular systems (Doolittle et al. 2011; Gray et al. 2010; Lukeš et al. 2011) based on the previous hypotheses of Covello and Gray (1993) and Stoltzfus (1999). Constructive neutral model suggests that a neutral mutation can allow the retention of a subsequent evolutionary change which would be otherwise lethal or that even slightly deleterious mutation can be compensated by another subsequent evolutionary change (Gray et al. 2010; Lukeš et al. 2011; Stoltzfus 1999). The repeated cycles of such stochastic evolutionary changes can lead to the addition of more and more macromolecules or their fragments to the system which becomes completely dependent on these components and thus the complexity of a cellular process can increase (Gray et al. 2010; Lukeš et al. 2011; Stoltzfus 1999).

Doolittle et al. (2011), Gray et al. (2010), Lukeš et al. (2011) and Stoltzfus (1999) suggest that the origin and evolution of huge molecular complexes such as ribosomes, spliceosomes, and editosomes can be mainly explained by CNE. Nevertheless, although CNE may have broad evolutionary implications, it is questionable if CNE can be equally applied for explaining the origins of these three complex structures in which RNAs exhibit their crucial function. While the origin of ribosomes dates certainly back to the last universal common ancestor (LUCA) (Woese 1998, 2000, 2001, 2002) and the spliceosomes were certainly present in the last eukaryotic common ancestor (LECA) (Collins and Penny 2005), the editosomes were invented only recently, probably in the ancestor of kinetoplastid lineage (Flegontov et al. 2011). Moreover, while catalytic ribosomal rRNAs are probably the most ancient ribozymes which still exist—they are clearly relics of the RNA world (Cech 2011), and the key catalytic function of spliceosomes is achieved by five cross-interacting snRNAs (Valadkhan and Jaladat 2010), gRNAs involved in RNA editing in kinetoplastid mitochondria just hybridize to pre-mRNAs (they serve only as a template for the correction of dismissed information) and they have no catalytic function—gRNAs are not ribozymes (Panigrahi et al. 2006). Finally, the spliceosomal RNP complex is almost as large as eukaryotic large ribosomal subunit (up to 60S) (Valadkhan and Jaladat 2010), while the editosome is a unique 20S complex composed solely of proteins (Panigrahi et al. 2006). In addition, the genomes of intracellular endosymbionts (such as mitochondria) are affected by higher mutation frequency, load, and drift and thus ratchet-like processes in comparison to nuclear genomes (Andersson et al. 2002; Wernegreen 2005; Woolfit and Bromham 2003). All common characteristics of spliceosomes and ribosomes and their differences from characteristics of editosomes suggest that the mechanisms for the origin and evolution of editosomes were partly different from those shaping the origin and evolution of ribosomes and spliceosomes. On the other hand, it is sure that ribosomes are obligatory for all organisms, spliceosomes are obligatory for all eukaryotes and editosomes are probably obligatory for the most of kinetoplastids. Hence, there is currently strong positive selective pressure to retain (or strong negative selective pressure to loose) these huge molecular complexes, because the survival of their carriers completely depends on them. It is also likely that a strong positive (and/or negative) selection largely shaped the origin and evolution of these molecular complexes in any transient stage, even if the function of a transient stage was different from their contemporary function.

Doolittle et al. (2011), Gray et al. (2010), Lukeš et al. (2011) and Stoltzfus (1999) suggest that spliceosomal introns and spliceosomes evolved by CNE from bacterial self-splicing group II introns which are often assumed to have been introduced to a prokaryotic ancestor of eukaryotes by the α-proteobacterial ancestors of mitochondria. This scenario is based on the proposal that prokaryotes (i.e., archaea) were ancestral to eukaryotes. However, no direct evidence supporting such notion exists. Instead, the current knowledge of biochemical properties of membranes, molecular phylogenies, distribution of genes among organisms, cell biology, and population genetics is rather consistent with the view that all organisms share an undefined common ancestor (LUCA) and that the root of the universal phylogenetic tree is between the Bacterial domain and the rest of organisms—the domains Archaea and Eukarya which share a more recent common ancestor (for review see Vesteg and Krajčovič 2011). With this respect, it is likely that bacterial group II introns share a common undefined ancestor with spliceosomal introns a spliceosomes—a proto-spliceosome.

Proposing that group II introns were ancestral to spliceosomal ones is not neutral (nor slightly costly) but seriously a costly scenario. It has been known for more than two decades that spliceosome-mediated splicing is slower than splicing of group II introns (Baurén and Wieslander 1994; Beyer and Osheim 1988; Lang and Spritz 1987; Schmidt et al. 1996). Although group II introns are among the largest and slowest ribozymes whose folding to the native structure takes approximately 30 s in vitro, they are spliced much faster in vivo in the presence of protein co-factors (Fedorova et al. 2010; Pyle et al. 2007; Roitzsch and Pyle 2009). In contrast, in vitro and in vivo experiments suggest that spliceosomal introns are excised in approximately 1–20 min (Audibert et al. 2002; Kessler et al. 1993; Singh and Padgett 2009). Lukeš et al. (2011) argue that five spliceosomal RNAs are unlikely to get together to make a group II intron. However, the spliceosomal RNAs are the components of the same RNP complex processing genetic information. Hence, the possibility that these RNAs would be reversely transcribed and inserted into the host DNA genome tightly linked together does not seem to be so unlikely. Their further reduction to group II introns does not appear to be an unlikely proposal as well, because group II introns are just faster. As we have recently argued (Vesteg and Krajčovič 2011), the evolution of slower and more complex RNA-based catalytic system from faster and simpler one could be hardly considered as being selectively neutral (nor nearly neutral) in every transient stage. In addition, much more efficient purely protein-based catalytic system would be predicted to evolve from simple RNA-based system rather than a less efficient RNA-based system requiring interaction and catalysis of various RNA molecules. For example, the protein complexity of mitochondrial ribosomes has increased in comparison to the ribosomes of their bacterial ancestors, while the rRNA complexity has decreased (Lukeš et al. 2011). Therefore, the RNA complexity of the spliceosome would be also expected to decrease in evolution, while the spliceosomes-late scenario proposes exactly the opposite. Moreover, the expansion of group II introns from the α-proteobacterial ancestors of mitochondria would be either impossible in an asexual hypothetical prokaryotic ancestor of eukaryotes or it would be dramatically fitness-reducing, because it would slow down both DNA replication and gene expression of a hypothetical prokaryotic host (for review see Vesteg and Krajčovič 2011). Since the spliceosomes-late scenario avoids any adaptive reason to explain the origin of the spliceosome, it can be considered as a virtual creation of less efficient and more complex RNA-based system from more efficient and simpler RNA-based system in the DNA–protein world in which enzymes are approximately million times more efficient than ribozymes (Jeffares et al. 1998; for review see Penny et al. 2009). This model is definitely not neutral and it would need to suggest clear selective advantages compensating for the serious costs to become plausible.

In our opinion, a proto-spliceosome could have likely been more complex than group II introns and less complex than contemporary spliceosome, while it could have been less efficient and slower than both these RNA-based machines. The selection for higher efficiency of gene expression, and higher speed and fidelity of splicing could have been the main evolutionary forces shaping the origin of both spliceosomes and group II introns. However, the diversification of these two RNA-based structures from their common ancestor could have probably occurred in different populations of cells living in different conditions under different selective pressures.

Although slower spliceosomes need not to be necessarily disadvantageous in comparison to faster group II introns, if gene regulation and expression control become more critical than speed during the advent of complexity, the key problem with the proposal for the origin of spliceosomal introns from group II introns is that this scenario largely ignores the RNA continuity theory—the proposal that the ancestry of most (if not all) catalytic RNAs dates back to the RNA world. The ancestors of these catalytic RNAs could have been initially likely selected for different function from that of their descendants. Nevertheless, the ancient ribozymes were likely pre-adapted and later selected (adopted) for similar function like they achieve presently (Penny et al. 2009; Poole et al. 1999). Although the regulatory or other non-catalytic short RNAs are expected to arise also after the invention of translation and genetic code, the ribozymes (whose catalysis is 103–109 times less efficient than that of enzymes) are not expected to evolve and to substitute for the catalysis achieved already by the enzymes (Penny et al. 2009; Poole et al. 1999). Hence, the origin of spliceosomal snRNAs most likely dates back to the RNA world as well as the origin of rRNAs (Cech 2011).

Based on the universal presence of ribosomes, Poole et al. (1998, 1999) suggested that proto-ribosomes of the first riboorganisms might have been ancestrally involved in RNA replication which might have been mediated by tRNAs—the donors of complementary trinucleotides. The selective advantage of such a replication mechanism could have been the proofreading (or elimination of unnecessary and foreign RNA)—only those RNAs recognized by tRNAs were replicated by a step-wise addition of three nucleotides at once (Poole et al. 1999). The amino acids might have initially served only for stabilization of tRNA-SSU rRNA interaction and the LSU rRNA might have been only later integrated into the complex (Poole et al. 1999). The proto-ribosomes were thereafter, probably after the invention of the first peptides with the RNA-dependent RNA-polymerasing activity, fully adopted for translation. The evolution of such a proto-ribosome could have been mediated by the series of Darwin–Eigen hypercycles (Poole et al. 1999). A step-wise addition of proteins to the RNA backbone of a proto-ribosome could have led to the origin of a primitive ribosome specialized for translation. The primitive translation system of LUCA is proposed to have been ancestrally much more error-prone and a step-wise selection for higher efficiency and fidelity of translation is expected to have formed the structure of ribosomes (Woese 1998, 2000, 2001, 2002).

Similarly like proto-ribosome, the proto-spliceosome of the first riboorganisms might have ancestrally achieved completely different function from that of contemporary spliceosomes. The introns-first theory suggests that the genome of the first RNA-based organisms was composed of RNA modules encoding functional ribozymes ancestral to contemporary introns and a small portion of waste information ancestral to contemporary exons (Penny et al. 2009; Poole et al. 1999). The proto-spliceosomal RNAs could have been selected for their capability of the excision of functional ribozymes and joining together unnecessary information (the ancestor of mRNAs), eventually recognized, and degraded by other ribozymes. Following this scenario, the proto-spliceosomal RNAs could be viewed as selfish elements, since they excised themselves from host RNAs to yield their functional catalytic forms, but they were also altruistic and advantageous for their carriers, because they could have promoted maturation of other host ribozymes. It is also likely that the invention of primitive proteins contributed to the complexity of the proto-spliceosomes and that translation and splicing somehow co-evolved in a step-wise manner mediated by Darwin–Eigen hypercycles.

Concomitant with the evolution of genetic code and translation, the proto-spliceosomes could have been adopted to reshuffle small-peptide encoding regions and thus to increase the variability of arising proteins—the hypothesis that splicing allowed recombination of RNA modules encoding protein domains (exon-shuffling) and thus it was adaptive (Doolittle 1978; Gilbert 1978; Kurland et al. 2007). If this scenario is considered purely as occurring exclusively at the DNA level via recombination of introns, it can be criticized as a creationistic scenario including a foresight (Blake 1978). However, the exon-shuffling at the RNA level (either via recombination or via trans-splicing) could have been rather an obligatory part of gene expression in the RNP-based LUCA. The spliceosome-dependent trans-splicing could have helped to bring together dispersed information and to increase the efficiency of gene expression. The exons could have ancestrally encoded independent protein domains which could have functioned in domain complexes—the hypothetical ancestors of modern proteins composed of various protein domains. Bringing the exons together by trans-splicing, translating them as a single unit, and producing a single protein containing many coupled domains, could have increased the efficiency of both gene expression and protein domain catalysis. The translation of many domains at once would at least minimize the time to assemble separate domains into catalytic complexes. Following this scenario, the increased variability of proteins amenable to selection would be only a side effect of the RNA exon-shuffling or spliceosome-dependent trans-splicing process which was ancestrally likely selected as an innovative and more parsimonious way of gene expression saving the time and energy in the RNP world.

It is also possible that the origin of spliced-leader (SL-)RNA-mediated spliceosome-dependent trans-splicing occurring in a variety of unrelated eukaryotes (for review see Hastings 2005) may trace back to the RNP-based LUCA. This process is obligatory for these organisms, because mRNAs obtain cap from 5`-end of SL-RNA via this process and the cap is necessary for translation. This process does not allow effective regulation of gene expression at the level of transcription initiation, because SL-RNAs and pre-mRNAs are transcribed from different genomic regions (Martínez-Calvillo et al. 2010; Vesteg et al. 2009). Therefore, the process is energetically costly in comparison to the effective regulatory condition seen in prokaryotes and many model eukaryotes. It has been proposed that SL-RNA-mediated trans-splicing evolved many times independently in eukaryotes (Derelle et al. 2010; Douris et al. 2010; Lukeš et al. 2009). However, assuming a secondary de novo appearance of such a costly process in the DNA–protein world even once appears to be highly improbable from the selectionistic point of view. The recent bioinformatic comparison of SL-RNAs from various unrelated eukaryotes revealed that their secondary structures are strikingly similar suggesting that all SL-RNAs share a common ancestor which was present in LECA (Marz et al. 2010) and likely also in LUCA.

The role for selfish replicative elements (i.e., viruses) in the origin and evolution of spliceosomes as well as ribosomes can also not be ruled out. Such viruses were most likely abundant in the RNA and the RNP world (Gilbert and de Souza 1999) and they were probably attacking the first riboorganisms all the time. The theoretical simulations suggest that altruistic features can easily appear in hypercyclic interactions of self-replicators and the hypercycle can easily become resistant against other parasites (Boerlijst and Hogeweg 1991). The experimental evidence also exists that a conflict mediation system (e.g., two competitive viruses invading the same host) can result in the evolution of a highly cooperative system in quite a short time (Sachs and Bull 2005).

DNA could have been invented by viruses, because it was more stable molecule than RNA and it would not be recognized by “immunity system” (RNA inactivation mechanisms) of an RNP-based host (Forterre 2002, 2005, 2006). Owing to the enhanced stability and thus coding capacity of DNA, DNA was later adopted by the RNP-based host(s), although it is questionable if such “DNA-transition” happened once in LUCA (Vesteg and Krajčovič 2008; Woese 1998, 2000, 2002), twice—in the ancestor of Bacteria, and in the common ancestor of Archaea and Eukarya (Forterre 2005; Poole and Logan 2005), or three times—in the ancestors of each domain independently (Forterre 2005, 2006).

Various selective forces for the simplification of cell structure in the ancestors of prokaryotes and for the origin of parsimoniously organized prokaryotic genomes (genes without spliceosomal introns are organized in operons) have been suggested: thermoreduction (Forterre 1995), r-selection (Poole et al. 1998), efficient selection in large populations (Lynch 2007), chronic energy stress (Valentine 2007), or other reasons (for review see Penny et al. 2009). All these selective mechanisms can be also applied for explaining the reduction of proto-spliceosomes to group II introns and almost complete intron loss in prokaryotes. On the other hand, the persistence of spliceosomes in eukaryotes can be explained by complete dependency of eukaryotes and their ancestors (LECA) on this RNP complex. An intermediary evolutionary stage analogous to that leading to the origin prokaryotic operonal gene structure from eukaryotic-like one can be exemplified by trypanosomatids. In contrast to their relatives—free-living euglenids (Vesteg et al. 2010), trypanosomatid parasites are almost completely devoid of cis-spliceosomal introns. Nevertheless, trypanosomatid pre-mRNAs are polycistronic containing transcripts of genes separated by trans-introns and the individual mRNAs are resolved and capped by spliceosome-dependent SL-RNA-mediated trans-splicing (for review see Martínez-Calvillo et al. 2010). In this case, the origin of prokaryotic gene expression would require nothing else than replacement (or reduction) of trans-introns by (to) something like Shine–Dalgarno sequence (ribosome binding site) and its recognition by ribosome accompanied by the loss of nuclear membrane, the loss of spliceosomes and SL-RNAs, and the reduction of ribosomes.

Taken altogether, although CNE could have somehow contributed to the origin and evolution of spliceosomes and ribosomes in any transient stage described here, there is no reason to suggest it as the dominant mechanism. The origin and evolution of these huge molecular ribomachines can be largely explained by applying selectionistic criteria.