Posted on 06/08/2005 2:57:23 PM PDT by <1/1,000,000th%
Explaining the evolution of complexity has been a challenge to darwinian theory since its conception. At the molecular level, biological complexity involves networks of ligand-protein, protein-protein and protein-nucleic acid interactions in metabolism, signal transduction, gene regulation, protein synthesis and so on. As organismal complexity increases, it has been observed that more control is required for the positive and negative regulation of genes such that complexity correlates with an increase in both the ratio and absolute number of transcription factors.
Both theoretical studies and genome analyses (Wagner, 1994, 2003; Mendoza & Alvarez-Buylla, 1998) have been used to examine the evolution of complex genetic networks. The duplication of genes is the predominant mechanism for the generation of new members of a protein family and so is central to the evolution of complexity. A duplicated gene can result in redundancy if multiple proteins have the same or overlapping function. Alternatively, due to the reduced selective constraint on protein evolution, one of the copies of the duplicated gene can become nonfunctional or, more significantly, can acquire a new function (Ohno, 1970; Wagner, 2001). The duplication that increases the size of a network may occur either via single-gene duplication events or by duplication of genes on a large scale, including the entire genome (Wagner, 1994). The need for networks to remain stable and functional in the cellular environment after the duplication event(s) is thought to favour whole-genome duplication (Papp et al, 2003). In addition, recent investigations on domain combinations have suggested that domain rearrangements (intrusion, loss and differential spacing between domains) are quite frequent and more important than previously assumed (Apic et al, 2001), especially for regulatory proteins. This means that pathways and networks not only evolve by the basic principles of gene evolution (gene duplication/loss and point mutation) but also they adapt by rearrangement of selectively advantageous 'building blocks'.
The basic helix-loop-helix (bHLH) protein family comprises an ancient class of eukaryotic transcription factors that are found in fungi, plants and animals (Moore et al, 2000). Due to their homo- and heterodimerization abilities, they form a complex protein-protein interaction network. The bHLH family is believed to have expanded together with the appearance of multicellularity (Ledent et al, 2002). In unicellular eukaryotes, such as yeast, bHLH proteins are involved in the regulation of several metabolic pathways (Massari & Murre, 2000). In contrast, the bHLH proteins of metazoa are involved in regulating the cell cycle (Luscher, 2001), sensing environmental signals (Gu et al, 2000), and also in developmental processes (Massari & Murre, 2000).
The bHLH transcription factors are named after their highly conserved 60-amino-acid-long domain that consists of a basic region followed by the helix-loop-helix motif, which comprises two amphipathic -helices separated by a variable-length loop (Littlewood & Evan, 1995). Two bHLH proteins form a functional homo- or heterodimer (that is, a four-helix bundle) through their HLH domains. Additionally, the two basic regions are responsible for recognizing and binding a core hexanucleotide DNA sequence, such as the E-box (Brownlie et al, 1997). It is usual for bHLH proteins to include additional domains that are responsible for the activation or repression of target gene activity. In some phylogenetic groups, additional dimerization domains (for example, leucine zipper (LZ), PAS or 'orange' domains) are found at the C-terminal side of the HLH domain (Ledent & Vervoort, 2001).
In this paper, we investigate the evolution of a network of bHLH transcription factors and, while recent studies (Atchley & Fitch, 1997; Ledent & Vervoort, 2001; Ledent et al, 2002) have concentrated on sequence analysis, we have combined genomic data, domain architecture and protein-protein interactions. We have used the bHLH system to examine the question of whether single-gene or large-scale duplications had an important role in the evolution of this network. The conceptual differences and implications for phylogenetic analysis and the development of protein-protein interactions are illustrated in Fig 1. Recent work on genetic networks (Conant & Wagner, 2003), protein-protein interactions and protein complexes (Papp et al, 2003) has suggested that large-scale duplications are the major driving force for network evolution. The results in this report do not exclude a role for large-scale duplications. However, the main point that emerges from our study is the demonstration that recurrence of single-gene duplications and domain rearrangements repeatedly give rise to complex networks.
Phylogenetic analyses based on the bHLH domain of the fungal and metazoan members of the bHLH protein family have classified them into four (A-D; Atchley & Fitch, 1997) or six (A-F; Ledent & Vervoort, 2001) major groups. The latter study split the B group of Atchley & Fitch (1997) into groups B and E, and also introduced the F group. The structural irregularities in the members of the F group raise considerable doubts about their classification as bHLH transcription factors, and so we excluded this group from our analysis. The short length of the bHLH domain, which consists of only 60 amino acids, compromises the reliability of phylogenetic analysis. However, the overall tree topology is congruent with previous studies (Atchley & Fitch, 1997; Ledent & Vervoort, 2001) and strongly supported by domain architecture as each group has a distinct domain arrangement (Fig 2A,B), apart from group B, which is paraphyletic (Fig 2A; Ledent & Vervoort, 2001). As yeast and plant sequences are only found in the B group, the paraphyly is consistent with group B being ancestral to the other four groups.
When we combined the phylogenetic analysis with heterodimerization/protein-protein interaction data, two distinct hub-based networks became apparent (Fig 2C). Moreover, distinct subnetworks can be identified. Each of these subnetworks contains at least one hub, a bHLH protein that interacts with a large number of other poorly connected 'peripheral' proteins. Although all hubs also homodimerize, only a few of the peripheral members do. The criterion for splitting a network into smaller subnetworks is the presence of hubs and peripheral members of the same phylogenetic group and with similar domain architecture (Fig 2C). Here, we designate each of the hub-based networks and subnetworks by using the name of the protein or the family that acts as the hub: (1) the 'Max' network in the B group, (2) the 'Arnt' subnetwork in the C group and (3) the 'E2A' subnetwork in the A group. Each hub-based network or subnetwork also has a distinct domain architecture (Fig 2B). The 'Arnt' and 'E2A' subnetworks are connected by the E group (HES family), together forming the 'E2A-Arnt' network. Note that the four members of the Mitf family in the B group are all connected and form an independent network, but as it is not hub based, they are not considered further.
Mathematical analysis confirms that the structure of these networks is nonrandom and hub based following the scale-free principle (see supplementary information online). This means that the distribution of connectivity decays as a power law P(k)k-, where k is the number of connections of a node and P(k) is the frequency of nodes with k interactions (Barabasi, 2002). Interestingly, the relative connectivity of the hubs in bHLH networks is higher than that found in the other biological networks that have been analysed. In terms of network theory, this is expressed as a lower value in the power-law equation for the bHLH network (where 1) than in most other networks (where ranges from 2 to 3; Goh et al, 2002). This means that there are fewer connections between the peripheral members of the bHLH network than in other networks. As it is conceivable that the higher connectivity of the bHLH hubs results from a bias in the data, we examined the literature carefully to check this result. Although we found a significant number of reports in which peripheral members were tested explicitly, they either exhibited a very restricted range of interactions or no interactions at all with other peripheral members (Hogenesch et al, 1997; Firulli et al, 2000; Luscher, 2001). We also note that most bHLH interactions have been confirmed by more than one independent method (see supplementary information online). The high connectivity of hubs in the bHLH networks appears to be a direct consequence of the fact that gene duplication events (single or large scale) have generated new peripheral proteins that then interact preferentially with the hub.
The hub proteins in the bHLH protein-interaction networks are usually widely expressed in different tissues and organs. They need to heterodimerize with peripheral members of the network, which have a more limited expression pattern, to exert their different effects. The peripheral members of the networks are either activators or inhibitors of transcription, and their formation of a heterodimer with the hub protein usually allows them to form a functional complex that binds specific promoter elements (for example, E-boxes). In several cases, dimerization may occur among peripheral members of the network alone, but these dimers are usually non-DNA-binding and therefore repress transcription.
The 'Max' network is involved in cell cycle control and it has no known bHLH protein interaction with the 'E2A-Arnt' network; this could be due to the nature of the additional dimerization domain (LZ) that seems to impose specificity on the range of bHLH interactions (Bornberg-Bauer et al, 1998). The members of the 'Arnt' subnetwork function mainly as environmental sensors controlling molecular clocks, the hypoxic response and the metabolism of toxic substances such as arylhydrocarbons. Finally, the 'E2A' subnetwork is involved in developmental processes.
Interestingly, the topology of the 'Max' network parallels that of the 'E2A-Arnt' network (see supplementary information online). In 'Max', there are two hubs (Max and Mlx) with peripheral partners that are either activators or repressors. The two hubs are connected through the Mad family, which act as repressor proteins. In the 'E2A-Arnt' network, there are also two hub families (Arnt and E2A families), and their peripheral partners are either activators or repressors. As with 'Max', repressor proteins (this time, of the HES family) connect these two hub families within the 'E2A-Arnt' network.
The most striking finding is that the parallelism between the 'Max' and the 'E2A-Arnt' networks extends to their phylogenetic relationships (Fig 2A). In the 'Max' network, the two hubs are not clustered together phylogenetically, which argues against their evolution by genome duplication or other large-scale events. The 'bridge', that is, the molecules that link the two hubs, is a family of repressors that are phylogenetically quite distant from the hubs. The same pattern appears in the 'E2A-Arnt' network. The two hub families (Arnt and E2A) are not clustered together in the sequence-based phylogeny, but (instead) cluster with their respective activators or repressors. Again, the 'bridge' between them is a family of repressors (the HES family) that are phylogenetically quite distant from the hubs. All of this suggests that there are similar restraints on the evolution of the different networks in the bHLH protein family. Newly emergent networks and subnetworks have followed this same evolutionary process at least twice. Thus, these bHLH networks show evolutionary convergence, giving rise to a symmetrical topology. The hub proteins have a key role in this convergence, as they act as central regulators that enable appropriate choices to be made between alternate patterns of gene expression.
Further evidence for a common mechanism of network evolution comes from another interesting parallelism between the 'Arnt' and 'E2A' subnetworks. In each, there exists a family of proteins (Period, which is not included in our analysis, and Id, respectively) that can bind with the hub or some peripheral members. As Period and Id lack either the bHLH or the basic region domain, they form nonfunctional heterodimers that cannot bind E-boxes. Thus, in both the 'E2A' and 'Arnt' subnetworks, there exists a mechanism that functionally sequesters the hubs or the activators by forming nonfunctional heterodimers. In other words, the same inhibitory mechanism is used on both 'sides' of the network.
Many of the binding specificities and their evolution can be understood from a structural perspective. For example, the inclusion of the PAS domain in group C restricts the range of interactions in which the bHLH domain may participate (Pongratz et al, 1998). Moreover, in the case of the AP-4 protein, the presence of two LZ domains inhibits its dimerization with the E2A hub (Hu et al, 1990). Furthermore, the differential spacing of the LZ from the HLH domain in the B group has been regarded as a mechanism of restricting interactions in the TFE3 factor (Beckmann & Kadesch, 1991). It is also well known that a small number of point mutations may cause subtle structural alterations in the surface of dimerization domains that result in significant changes in dimerization specificity (LZs of the 'Max' network) (Nair & Burley, 2003). These findings increase our confidence that many peripheral members do, in fact, have only a limited number of interactions and that this result is not due to experimental bias.
There are several conclusions from our analysis. First, our results suggest that, for the evolution of networks based on one kind of binding domain (such as the bHLH), a model of single-gene duplication followed by domain rearrangements, point mutations and ongoing gene duplication is sufficient to generate quite complex interaction patterns, which mediate activation and repression. This finding does not preclude a role for large-scale gene duplication. For example, the postulated two rounds of whole-genome duplication in early vertebrates (Ohno, 1970) increased the complexity of networks by increasing the number of paralogous genes in each family. Nevertheless, our results strongly suggest that it is via single-gene duplications and domain rearrangements that new networks first arose in early metazoan evolution, before the divergence of arthropods and chordates. The mechanism of frequent gene duplication and domain rearrangement has also been demonstrated for signal transduction proteins involved in metazoan development (Miyata & Suga, 2001). To our knowledge, the emergence of a hub-based network with scale-free properties has not been explicitly explained before using phylogenies based on real data. This mechanism of evolution could be extrapolated to other families of transcription factors or signal transduction proteins that form complex dimerization networks and expanded during early metazoan evolution.
Second, we find a compelling symmetry between the two networks ('Max' and 'E2A-Arnt') as well as similar structures within them. The 'E2A-Arnt' network resembles an expanded and duplicated version of the 'Max' network. However, the phylogenetic analysis does not support large-scale gene or genome duplication as a mechanism for the generation of the two networks. Rather, our results indicate that each of the networks evolved independently by single-gene duplication events towards a similar topology (in terms of hubs, repressors and activators), thus providing a striking example of convergent evolution at the level of protein networks. The reason for this reliance on single-gene duplications is that, if a new network was generated by large-scale gene duplication, then all members of that network would have to be isolated from the 'parental' network. This is highly unlikely, given that we know that network specificity is defined by domain architecture (Hu et al, 1990; Bornberg-Bauer et al, 1998); thus, after entire network duplication, it would be necessary for the new architecture to have arisen simultaneously in all members of the new network (Fig 1). Accordingly, new domain architecture is only significant if it is initially duplicated in a singular fashion such that a new and isolated network forms.
In conclusion, the initiation of the evolution of new subnetworks by gene duplication in pre-existing networks, coupled with domain rearrangements and point mutations, provides the possibility for a multitude of new protein-protein interaction pairs to develop. We infer that, in the case of bHLH, early duplication events of genes ancestral to a particular network were followed by domain intrusions and losses. Although the picture might be different with other gene families, it appears that single-gene duplications cannot be dismissed as a feasible mechanism to generate complex networks, and that such duplications have facilitated the evolution of complexity.
The tables are on the site.
It's ironic that biologists are on the verge of being able to explain how evolution works at the molecular level at the same time that there's a growing movement to block or dilute the teaching of science.
I wonder what the new Creationism will call itself?
Well, that's always been a problem - carrying over the principles of molecular evolution to macro organismal evolution. It is easy to envision evolution occurring after genetic duplication since the copy is able to mutate without damaging or interfering with the function of the "parent" or original gene (assuming the absence of dosage effects, of course). It is much more difficult to explain evolution of an arm into a flipper since there presumably is a large period of time when the evolvibg limb is neither an arm or a flipper. In that intermediate stage you've lost the selective advantage of a functional arm but not yet achieved the advantage of a flipper. Now, if the organism had duplicated the limb allowing the two to evolve independently, it would make more sense!
Not really. If you were to try to encode certain characteristics into a genome, how would you do it?
Wouldn't you, perhaps, initiate "new subnetworks by gene duplication in pre-existing networks, coupled with domain rearrangements and point mutations" -- by actively manipulating "hubs", "activators," and "repressors?"
Sounds like some interesting research on a very interesting mechanism. But it doesn't provide you any protection from the idea that the dreaded Intelligent Designer might be in the loop for any or all of the phenomena described.
Indeed, it seems to me that this research will prove to be quite useful for allowing human Intelligent Designers to achieve new heights in genetic manipulation.
But seriously, the genes that control arms and flippers have been identified and catalogued. Once the molecular networks are described and compared, the intermediate changes from flipper to arm and back to flipper can be described at the molecular level with the types of changes that caused them to occur.
ID is still a rhetorical argument that says that unknowns are examples of complexity. Once the unknowns are gone, what will be left of the argument?
You've nailed this one. All the more reason not to leave the field to the atheists.
Doesn't look all that predictive to me.
I think I'll wait until they can tell us where the "seat of consciousness" might be located.
In order for macroevolution to work you need a mechanism whereby external selection pressures work at the DNA level. To my knowledge (and I'm a cancer geneticist, other than in response to a small number of environmental mutagens)no such mechanisms exists. And in order for such selection to work the intermediate forms must provide for an advantage....if they don't, then they will be selected against and the original structure will be preserved. Such cannot occur in the long transition from arm to flipper.
The problem there, though, is the mechanism by which the changes came about. The basic idea in evolution is that each of the changes has to confer an advantage in order to be passed along.
We also have to recognize the tendency of people to focus on single attributes -- the flipper, say -- rather than on the entire biological system. It takes a heck of a more than one flipper to make a dolphin, for example, and thus it's got to be a very special series of mutations that maintains a "system-level" set of advantages over time.
We do see an lot of rhetoric of the "nature came up with an ingenious solution" type, trying to explain (or gloss over) that....
PING
|
Can you please explain how the recent advances in astrophysis have crushed evolution??
Never heard that before.
Beat me by 56 seconds. LOL
Well, that's rather wrong, actually. Successive forms may very well be a neutral adaptation, insofar as they may confer no more or less of an advantage than the original form.
Granted. Loss of a functional limb during the transition would be hard to explain as a neutral adaptation.
Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.