Research ArticleEVOLUTION

Evolution of the TSC1/TSC2-TOR Signaling Pathway

See allHide authors and affiliations

Science Signaling  29 Jun 2010:
Vol. 3, Issue 128, pp. ra49
DOI: 10.1126/scisignal.2000803

Abstract

The TSC1/TSC2-TOR signaling pathway [the signaling pathway that includes the heterodimeric TSC1 (tuberous sclerosis 1 protein)–TSC2 (tuberous sclerosis 2 protein) complex and TOR (target of rapamycin)] regulates various cellular processes, including protein synthesis, in response to growth factors and nutrient availability. Homologs of some pathway components have been reported from animals, fungi, plants, and protozoa. These observations led to the perception that the whole pathway is evolutionarily conserved throughout eukaryotes. Using complete genome sequences, we show that, contrary to this view, the pathway was built up from a simpler one, present in the ancestral eukaryote, coupling cell growth to energy supplies. Additional elements, such as TSC1 and TSC2, were “bolted on” in particular eukaryotic lineages. Our results also suggest that unikonts [Opisthokonta (including animals and fungi) and Amoebozoa] form a monophyletic group with the Excavata and Chromalveolata. A previous proposal, that the root of the eukaryotic “tree of life” lies between the unikonts and other organisms, should therefore be reevaluated.

Introduction

The central role of the TSC1/TSC2-TOR signaling pathway [the signaling pathway that includes the heterodimeric TSC1 (tuberous sclerosis 1 protein)–TSC2 (tuberous sclerosis 2 protein) complex and TOR (target of rapamycin)] in the integration of various physiological processes in animal cells is well understood at the molecular level (1, 2). Disruption of the TSC1/TSC2-TOR pathway is involved in a number of genetic disorders resulting in physical and neurocognitive manifestations, such as seen in tuberous sclerosis complex (TSC), neurofibromatosis (NF1), and fragile X syndrome (1, 35). Several signaling pathways converge on the TSC1-TSC2 complex, which can be thought of as a “switchboard” collecting input signals and using them to regulate fundamental cellular processes (6) (Fig. 1). The best-characterized TSC1-TSC2 output is through the GAP (guanosine triphosphatase–activating protein) function of TSC2, whereby it inhibits Rheb (Ras homolog enriched in brain) and thereby TOR. TOR exists in two pools: one associated with Raptor as part of TOR complex 1 (TORC1) and one associated with Rictor as part of TORC2. The TORC1 complex promotes protein synthesis, and the TORC2 complex modulates various other cell biological processes. Homologs of some components of the TSC1/TSC2-TOR pathway have been reported from a range of organisms including animals, fungi, plants, and protozoa (710), leading to the view that those components and the pathway as a whole are present throughout the eukaryotes (11, 12). However, the distribution of components of the TSC1/TSC2-TOR pathway has not been surveyed systematically across eukaryotes, and nothing is known about how it evolved.

Fig. 1

The signaling pathway from AMPK to S6K and 4E-BP as described in mammalian systems. Elements in colored ellipses are inferred from this study to have been present in a primitive signaling pathway linking AMP concentration to cell growth. The latter pathway was present in the last common ancestor of present-day eukaryote groups. p38MAPK, p38 mitogen-activated protein kinase; MK2, MAPK-activated protein kinase 2.

Using the distribution of components of this important pathway as a phylogenetic marker should also allow us to draw conclusions about the early evolution of eukaryotes. It has been suggested that the eukaryotes can be divided into two groups, the “unikonts” and the “bikonts,” essentially according to whether they respectively have a single flagellum or two (13). The unikonts comprise the Opisthokonta (including animals and fungi) and the Amoebozoa (including Dictyostelium and Entamoeba). The bikonts comprise the Archaeplastida (including land plants and red and green algae), the Chromalveolata (including Phytophthora, Plasmodium, Toxoplasma, ciliates, and “brown” algae), the Excavata (including Euglena and pathogens such as Trypanosoma and Giardia), and a poorly characterized group, the Rhizaria (Table 1). It has been proposed, on the basis of gene organization data, that the root of the eukaryotic tree, representing the ancestor of all eukaryotes, lies between the unikonts and bikonts (14, 15), but this conclusion has been controversial in the light of ultrastructural and genetic data (1621). Analysis of the distribution of components of the TSC1/TSC2-TOR signaling pathway among the eukaryote groups might help in locating the root of the eukaryotic tree.

Table 1

Representatives of the supergroups (underlined) referred to here. Stramenopiles, Alveolates, and Rhizaria have also been proposed to form a larger grouping (27).

View this table:

We carried out a broad survey of the distribution of components of the TSC1/TSC2-TOR pathway across eukaryotes. Our analysis showed that the pathway evolved from a primitive bacterial energy–sensing system by successive addition of extra features. In addition, our study indicated that some organisms that are widely used as models for TSC1-TSC2 signaling may not be appropriate for this purpose. Our study also supported the view that the unikonts are monophyletic (sharing common ancestry to the exclusion of other groups), but demonstrated that the root of the eukaryotic tree lies within the bikonts, in contrast to earlier proposals.

Results

Search strategy

We searched genomic and expressed sequence tag (EST) databases using BLAST (22), PSI-BLAST (23), and HMMER, a tool based on hidden Markov models (24), for homologs of key components of the TSC1/TSC2-TOR signaling pathway, in particular adenosine 5′-monophosphate–activated protein kinase (AMPK), phosphatase and tensin homolog deleted from chromosome 10 (PTEN), phosphatidylinositol 3-kinase (PI3K), Akt, TSC1, TSC2, TOR, Rictor, Raptor, eukaryotic initiation factor 4E–binding protein (4E-BP), and S6 kinase (S6K). We excluded Rheb from the analysis because the large number of Rho homologs would make the results difficult to interpret (25). The genomic databases searched included representatives of the Amoebozoa, the Archaeplastida, the Chromalveolata, the Excavata (26), and the Opisthokonta. Only EST databases were available for Rhizaria, although it has been suggested that they form part of the Chromalveolata (16, 27) in which case genome searches in Rhizaria would not add substantially to this analysis. We also searched eubacterial and archaebacterial genomes. Individual domains of some proteins, such as the GAP domain of TSC2, showed sequence similarity to GAP domains in a number of proteins, in which case we repeated the searches excluding the widespread domains in an attempt to identify genuine orthologs. Putative orthologs were compared to mammalian sequences with MEME (28) to test whether they were genuine by looking for conservation in domain content and order. HMMER searches did not reveal the presence of components of the pathway in groups where BLAST searches had failed to identify them. The results are summarized in Table 2 and in the Supplementary Text.

Table 2

Distribution of orthologs of TSC1/TSC2-TOR signaling pathway proteins across the eukaryotes. Tick with asterisk indicates a rudimentary form of TSC2 (see text).

View this table:

Distribution of PI3K, AMPK, TOR, PTEN, Raptor, and S6K

In all of the eukaryotic groups sampled, we found sequences encoding PI3K; the α, β, and γ subunits of AMPK; TOR; PTEN; Raptor; and S6K. The α subunit of AMPK comprises an N-terminal serine-threonine kinase domain, with the rest of the protein having an autoregulatory function (29). We found homologs of the N-terminal kinase domain, but not the rest of the protein, in eubacteria and archaebacteria. Similarly, we identified homologs of the CBS (cystathionine β-synthase) domains found in the eukaryote γ subunit in the eubacteria and archaebacteria. Some eubacteria and archaebacteria proteins contained four CBS domains, as with the γ subunit of AMPK in eukaryotes. These results indicate that AMPK evolved early in the eukaryotes by recruitment of preexisting domains, including kinase and CBS domains, into a multisubunit complex. PTEN includes a phosphatase domain and a C-terminal domain (C2) that binds phospholipid membranes (30). Homologs containing both domains were found in all the eukaryotes screened. Raptor homologs were also identified in all eukaryote groups, whether or not its WD40 domain was excluded. No homologs of TOR, PTEN, PI3K, or Raptor were identified in eubacteria or archaebacteria, suggesting that these proteins evolved early in the eukaryotes, before the main groups diverged from the last common ancestor. Given the lack of distinctive features of S6K and the large number of kinases in bacteria, it is difficult to determine whether S6K is restricted to eukaryotes.

Distribution of Akt and 4E-BP

Akt contains kinase and pleckstrin domains (31). Although sequences encoding homologs of either the kinase or the pleckstrin domains were present in all the eukaryote groups sampled (with homologs of the kinase domain also present in the prokaryotes), orthologs containing both domains in the same protein were found only in Opisthokonta, Amoebozoa, and Excavata (Figs. 2 and 3). Homologs of 4E-BP had a distribution similar to those of Akt and were identified in Opisthokonta and Amoebozoa, as well as in the excavate amoeboflagellate Naegleria. No homologs of 4E-BP or Akt were identified among the chromalveolates or the Archaeplastida.

Fig. 2

Examples of proteins containing Akt kinase domains identified across the eukaryotes. The figure shows the distribution of pleckstrin and protein kinase domains in the various proteins. Predicted numbers of amino acids in the proteins are indicated. Orthologs containing both the pleckstrin and protein kinase domains were identified only in the unikonts (Homo and Dictyostelium) and Excavata (Trichomonas and Trypanosoma).

Fig. 3

The evolution of the TSC1/TSC2-TOR signaling pathway. The figure indicates the stages at which particular components of the pathway evolved, superimposed on a phylogenetic tree of the major eukaryote groups inferred from the distribution of components of the pathway. The inferred position of the root is indicated. TSC2* refers to the rudimentary form of TSC2 found in Dictyostelium.

Distribution of Rictor

Sequences encoding homologs of Rictor were found in all the Opisthokonta and in the Amoebozoa, but the pattern in other groups was less consistent. Putative Rictor homologs were identified in Trypanosoma, Leishmania, and Trichomonas, but not in the other Excavata. The trypanosomal protein has previously been shown to have TOR-binding activity (10). In the Chromalveolata, putative homologs were found in Tetrahymena, Paramecium, and Phytophthora, but not in any of the photosynthetic chromalveolates, or the Apicomplexa, which have recent common ancestry with photosynthetic chromalveolates [the dinoflagellates and diatoms (32)]. No Rictor homologs were found in any of the Archaeplastida.

Distribution of TSC2 and TSC1

A number of sequences encoding possible TSC2 homologs were identified outside the Opisthokonta with the full-length query sequence. However, TSC2 includes a GAP domain, which is likely to be present in paralogous proteins. We therefore repeated the search with the GAP domain excluded, and saw little similarity between TSC2 and proteins outside the Opisthokonta. We identified a putative TSC2 homolog in Dictyostelium discoideum, although it lacked many of the motifs present in the opisthokont protein. Within the Opisthokonta, the Schizosaccharomyces pombe sequence (33) showed only 15% identity at the amino acid sequence level to the human protein and lacked a number of motifs that occur outside the GAP domain in mammalian proteins. Although the putative S. pombe homolog is involved in signaling in response to nutrient status (34), a number of the residues implicated in signaling in mammalian systems, such as Ser664, which is phosphorylated by extracellular signal–regulated kinase, or Ser981, which is phosphorylated by Akt, do not appear to be conserved (fig. S1) (35). This suggests that TSC2 may have evolved progressively from a rudimentary precursor GAP protein occurring in the ancestor of Opisthokonta and Amoebozoa by recruitment of a number of input signaling motifs within the opisthokont lineage.

We identified a number of possible TSC1 homologs outside of the Opisthokonta with the full-length sequence as query. However, TSC1 includes a coiled-coil domain, which is a widespread motif. When the region of the protein including this domain was excluded from the search sequence, no clear homologs were apparent outside the Opisthokonta. Alignment within the Opisthokonta was similarly difficult and the reported S. pombe homolog showed only 13% identity to the human protein at the amino acid sequence level. As with TSC2, a number of phosphorylation sites implicated in the regulation of the function of the TSC1-TSC2 complex (35), such as Ser584, which is phosphorylated by cyclin-dependent kinase 1, and Ser487 and Ser511, which are phosphorylated by IKKβ [IκB (inhibitor of nuclear factor κB) kinase β], were not conserved (fig. S2). These observations indicate that TSC1 evolved within the Opisthokonta from preexisting coiled-coil domain proteins by acquisition of additional features, including residues required for inputs from signaling pathways.

Discussion

Primitive signaling pathway

Our analysis shows that components of the TSC1/TSC2-TOR pathway that are found among all the eukaryote groups (Fig. 1) could provide a mechanism for linking the relative concentrations of adenine nucleotides to protein synthesis and growth through the TOR-Raptor complex. This is consistent with the observation that AMPK can signal directly to the TOR-Raptor complex (12). The first eukaryotic organisms had just acquired a system allowing them to generate high relative concentrations of adenosine triphosphate under particular conditions in a separate compartment (the mitochondrion) from the protein synthetic machinery (17). It seems likely that a signaling pathway allowing cell growth to be linked to adenine nucleotide concentration would have offered a selective advantage. Although we cannot eliminate the possibility that all components of the pathway were present in the ancestral eukaryote and then lost in some lineages, the latter is a less parsimonious model because it would require loss of a number of components of the pathway in entire groups shortly after they had been developed in the ancestral eukaryote.

After the divergence of the main eukaryote lineages, the pathway became more sophisticated in some lineages through the incorporation of additional input (such as TSC1-TSC2) and output (such as 4E-BP, and Rictor to Akt) elements. Some of the additional components were generated by attaching new functional elements to existing proteins (as in the evolution of TSC2 through the addition of new regulatory features to a GAP protein) or by making new combinations of existing domains (as in Akt). A number of TSC1 and TSC2 residues implicated in signaling in mammalian systems are not conserved in the S. pombe proteins, calling into question the value of the latter as models for signaling in the mammalian pathway. With the acquisition of Akt and TSC1-TSC2, PI3K (already regulating other functions) began signaling to the TOR pathway. This model of evolution of a signaling pathway, in which input and output elements are progressively added to a central core, contrasts with the simple “retrograde” model for the evolution of metabolic pathways (36). In the latter model, a pathway is built up sequentially from the endpoint by addition of individual steps to the beginning of the pathway. It will be interesting to see whether other signaling pathways also evolved by the progressive addition of input and output elements to a central core rather than according to a retrograde model.

Root of the eukaryote tree

The distribution of a number of the components of the TSC1/TSC2-TOR pathway provides information on the phylogenetic relationships of the different eukaryotic groups. The distribution of TSC2 supports the existence of Opisthokonta and Amoebozoa as a monophyletic group, the unikonts. The presence of an Akt homolog containing both kinase and pleckstrin domains in Excavates, as well as the unikonts (Opisthokonta and Amoebozoa), indicates that these all form a monophyletic group, supporting the recent proposal that the other eukaryotic groups form a “megagroup” that excludes the unikonts and Excavata (26, 37). The presence of 4E-BP in Naegleria is also consistent with a unikont-excavate group. Although there are no clear homologs of 4E-BP in the other excavates searched, this may reflect the fact that Naegleria is a free-living organism, whereas the others are not, and generally have smaller coding contents (38).

The presence of Rictor in some Excavates and many of the Chromalveolates and its absence from all of the Archaeplastida suggest that the Chromalveolates form a monophyletic group with the unikonts and Excavata. The Chromalveolates that lack Rictor are all photosynthetic (such as Emiliania or Thalassiosira) or have a recent photosynthetic ancestry (such as Plasmodium). Given this observation and the absence of Rictor from any of the Archaeplastida examined, we suggest that the function of Rictor is unnecessary in photosynthetic organisms, as a consequence of their phototrophic rather than heterotrophic metabolism. Rictor became redundant in those Chromalveolates that acquired a chloroplast through endosymbiotic uptake of a photosynthetic eukaryote (39) and was lost.

The most parsimonious model for the relationships between the eukaryote groups according to these results is shown in Fig. 3. On this basis, the root of the eukaryote tree lies between the Archaeplastida and the other groups, making the acquisition of a chloroplast the first evolutionary step that distinguished one eukaryote group from the others. An alternative model, with the root between a group containing Archaeplastida and Chromalveolata and a group containing the remainder, would be less parsimonious, requiring an ancestrally present Rictor to have been lost independently in Archaeplastida, as well as some of the Chromalveolata (and Excavata). In view of the apparent tendency for loss of Rictor, we cannot exclude this model. However, a root between Archaeplastida and other groups is consistent with phylogenetic trees based on rare genomic changes (21). Models with the root of the eukaryote tree between bikonts and unikonts would be even less parsimonious than either of these models, requiring, in addition to multiple independent losses of Rictor, losses of Akt and 4E-BP.

Analysis of the distribution of components of the TSC1/TSC2-TOR signaling pathway therefore indicates that, although the entire pathway is not present throughout eukaryotes, elements of it are, and the present pathway was built up from an ancestral one linking growth to energy supply. The analysis argues against the previously proposed root of the eukaryotic tree between the bikonts and unikonts (14, 15) and suggests that the root may lie between those organisms containing a primary chloroplast (the Archaeplastida) and the others. Our work shows that simplified forms of the TSC1/TSC2-TOR signaling pathway may be biologically effective, and has implications for our understanding of the mechanisms and consequences of defects in parts of this pathway in animals. The work also provides new insights into the evolutionary relationship between eukaryotic groups and has implications for our attempts to identify the root of the eukaryotic tree of life.

Materials and Methods

Bioinformatic analyses were carried out with published EST (complementary DNA) and genomic databases for a range of members of the Amoebozoa (Dictyostelium and Entamoeba), the Archaeplastida (Arabidopsis, Chlamydomonas, Cyanidioschyzon, Oryza, Populus, Ostreococcus, Physcomitrella, and Selaginella), the Chromalveolata (Cryptosporidium, Emiliania, Paramecium, Phaeodactylum, Phytophthora, Plasmodium, Tetrahymena, Theileria, Toxoplasma, and Thalassiosira), the Excavata (Giardia, Leishmania, Naegleria, Trichomonas, and Trypanosoma), and the Opisthokonta (Anopheles, Apis, Aspergillus, Batrachochytrium, Caenorhabditis, Cryptococcus, Danio, Drosophila, Homo, Monosiga, Mus, Nematostella, Rattus, Rhizopus, Saccharomyces, Schizosaccharomyces, and Strongylocentrotus). Sufficient sequence data from the Rhizaria were not available to draw reliable conclusions. Databases used included the Broad Institute database for Aspergillus nidulans, Cryptococcus neoformans, Rhizopus oryzae, and Batrachochytrium dendrobatidis; http://dictybase.org for D. discoideum; the Cyanidioschyzon merolae Genome Project (http://merolae.biol.s.u-tokyo.ac.jp) for Cyanidioschyzon merolae; the J. Craig Venter Institute Database for Trichomonas vaginalis, Trypanosoma cruzi, and Tetrahymena thermophila; Panthema-Entamoeba for Entamoeba species; the Joint Genome Institute Eukaryotic Genomics databases (http://genome.jgi-psf.org) for Thalassiosira pseudonana, Phytophthora ramorum, Ostreococcus tauri, Selaginella moellendorfii, Physcomitrella patens, Chlamydomonas reinhardtii, Naegleria gruberi, Populus trichocarpa, and Emiliania huxleyi; The Sanger Institute Database (http://www.sanger.ac.uk) for Leishmania major; ParameciumDB (http://paramecium.cgm.cnrs-gif.fr) for Paramecium tetraurelia; PlasmoDB (http://plasmodb.org) for Plasmodium species; ToxoDB (http://toxodb.org) for Toxoplasma gondii; and CryptoDB (http://cryptodb.org) for Cryptosporidium parvum. Analysis of published ESTs was performed with TBestDB. The National Center for Biotechnology Information (NCBI) database was used for all other searches. The reference numbers given for homologs in the Supplementary Text are NCBI GI (GenInfo Identifier) reference numbers for all Opisthokonta, Arabidopsis thaliana, and Oryza saliva and the corresponding reference number for the databases searched for all other species. Databases were searched with default settings in BLAST (basic local alignment search tool) and PSI-BLAST (22, 23) with human TSC1 (gi 4507693); TSC2 (gi 116256352); mammalian TOR (mTOR) (gi 1169735); Raptor (gi 22094987); Rictor (gi 46093886); AMPK subunits α (gi 46877068), β (gi 14194425), and γ (gi 4506061); Akt1 (gi 18027298); S6K (gi 4506737); PI3K (gi 34761064); PTEN (gi 73765544); and 4E-BP (gi 4758258) as the query sequences. To exclude the GAP domain of TSC2, we omitted residues 1500 to 1650, and to exclude the coiled-coil domain of TSC1, we omitted residues 706 to 961. To confirm the BLAST results, we repeated searches with HMMER (24) on a proteome data set constructed by A. O’Reilly (University of Cambridge), including representative organisms from the five main eukaryotic groups, as listed in the Supplementary Text.

Multiple alignments were constructed with the ClustalX algorithm (40). Default settings for protein alignments in FASTA format were used (41). MEME (28) was used to identify and compare motifs in sequences identified by BLAST to test assignment as homologs. MEME was set to identify any number of sequence repetitions for sequences ranging from 6 to 300 amino acids in size. Functional domains were identified with pfam (http://pfam.sanger.ac.uk/). Conclusions were confirmed with the identified putative homologs as queries back against the relevant databases. Where BLAST failed to detect homologs in particular groups with human sequences as the query, searches were repeated with homologs identified in other groups using the human sequences.

Acknowledgments

Acknowledgments: We thank A. O’Reilly (University of Cambridge) for providing her eukaryotic proteome data set for analysis with HMMER and G. Walker (University of Cambridge) for helpful discussions. Funding: Supported in part by a Tuberous Sclerosis Alliance grant (to P.J.d.V.) and the Cambridgeshire and Peterborough NHS Foundation Trust (to J.S. and P.J.d.V.). Author contributions: J.S., R.E.R.N., C.J.H., and P.J.d.V. designed the experiments, analyzed the data, and wrote the paper. J.S., R.E.R.N., and C.J.H. performed computational work. Competing interests: The authors declare no competing interests.

Supplementary Material

www.sciencesignaling.org/cgi/content/full/3/128/ra49/DC1

Supplementary Text. List of inferred protein sequences searched using HMMER, and homologs found using BLAST, PSI-BLAST or HMMER.

Fig. S1. Alignment of part of TSC2 sequences.

Fig. S2. Alignment of part of TSC1 sequences.

References and Notes

View Abstract

Stay Connected to Science Signaling

Navigate This Article