Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 7-Day Trial for You or Your Team.

Learn More →

DBD––taxonomically broad transcription factor predictions: new content and functionality

DBD––taxonomically broad transcription factor predictions: new content and functionality D88–D92 Nucleic Acids Research, 2008, Vol. 36, Database issue Published online 11 December 2007 doi:10.1093/nar/gkm964 DBD––taxonomically broad transcription factor predictions: new content and functionality 1, 1 2 Derek Wilson *, Varodom Charoensawan , Sarah K. Kummerfeld and Sarah A. Teichmann 1 2 MRC Laboratory of Molecular Biology, Hills Road, Cambridge, CB2 0QH, UK and Department of Developmental Biology, Stanford University Medical Center, 279 Campus Drive, Stanford, CA 94305-5329, USA Received September 14, 2007; Revised October 16, 2007; Accepted October 17, 2007 Like other proteins, TFs are composed of evolutionary ABSTRACT units called domains, which belong to families that can DNA-binding domain (DBD) is a database of pre- occur in many different proteins and various domain dicted sequence-specific DNA-binding transcription combinations. In the DBD database, we define TFs as factors (TFs) for all publicly available proteomes. proteins containing a sequence-specific DNA-binding The proteomes have increased from 150 in the initial domain (DBD). Other databases, such as TrSDB (3), or version of DBD to over 700 in the current version. data sets, such as Messina et al. (4), include both specific All predicted TFs must contain a significant match and general TFs. The precise description of TFs as to a hidden Markov model representing a sequence- sequence-specific DNA-binding we use is useful in a wide variety of studies. Examples include: improving specific DNA-binding domain family. Access to TF genome annotation; high-throughput experiments such as predictions is provided through http://transcription- ChIP–chip, protein chip or yeast one-hybrid (5); and factor.org, where new search options are now studies of the evolution of gene regulation comparing provided such as searching by gene names in multiple genomes (6), or gene regulation networks (7). The model organisms, searching for all proteins in a DBD database has been used as an annotation tool in the particular DBD family and specific organism. We context of the InterPro (8) and FlyTF (http://FlyTF.org) illustrate the application of this type of search (9) databases. facility by contrasting trends of DBD family occur- Access to the DBD database is via http://transcription rence throughout the tree of life, highlighting the factor.org, where all data is available for viewing and clear partition between eukaryotic and prokaryotic immediate download. The community can browse predic- DBD expansions. The website content has been tions for over 700 species (from Arabidopsis thaliana to expanded to include dedicated pages for each TF Zymomonas mobilis) or DBD family (including helix– containing domain assignment details, gene names, turn–helix, zinc-fingers, homeobox and many others); links to external databases and links to TFs with search predictions by sequence identifier or domain similar domain arrangements. We compare the family; receive classifications for submitted protein sequences, and download our domain assignments, as increase in number of predicted TFs with proteome well as our manually curated list of DBDs. size in eukaryotes and prokaryotes. Eukaryotes The prediction method in the DBD database (10) uses follow a slower rate of increase in TFs than hidden Markov models (HMMs) to identify domains prokaryotes, which could be due to the presence in proteins from two databases: SUPERFAMILY (11) of splice variants or an increase in combinatorial and Pfam (12). From DBD release 2.0 onwards, updated control. annotation resulted in 303 HMMs from SUPERFAMILY and 145 from Pfam compared to a total of 251 HMMs INTRODUCTION in the first version of DBD. The HMMs from SUPERFAMILY represent 37 superfamilies and 87 Sequence-specific DNA-binding transcription factors families according to the definitions in the SCOP database (TFs) each recognize a family of cis-regulatory DNA (13). This includes 98 new models representing 37 sequences described by a consensus motif (1) or position- sequence-specific DBD families. This resulted in an specific weight matrix (2). They regulate spatial and increase in additional TF predictions of 4.7%, for the temporal gene expression by binding to DNA and either activating or repressing action of an RNA polymerase. 150 organisms in the original version of DBD. *To whom correspondence should be addressed. Tel: +44 (0)1223 402479; Fax: +44 (0)1223 213556; Email: dbd@mrc-lmb.cam.ac.uk 2007 The Author(s) This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/ by-nc/2.0/uk/) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited. Nucleic Acids Research, 2008, Vol. 36, Database issue D89 The pipeline used to predict TFs begins with a domain NOVEL DEVELOPMENTS annotation of all proteins from completely sequenced Researchers can use the DBD database in several ways. For genomes with all HMMs from the SUPERFAMILY and instance, all TF predictions are available to download. Pfam databases (Supplementary Figure 1). A protein is However, most users are only interested in a small number classified as a TF if it has a significant match to a model of TFs, so we have expanded the website search options to we annotated as being a DBD, with the significance allow retrieval of individual TFs and subsets of TFs. New thresholds for HMM matches taken from the Pfam and search capabilities include: searching for gene names, for SUPERFAMILY databases. This results in an estimated example lacI or P53; listing all TFs that contain either a 1–5% of false-positive annotations. The TF predictions specified DBD or non-DBD family, for instance all TFs are limited to the families in our annotated collection, containing the bZIP (leucine zipper) family; retrieving all which means that the coverage is about two-thirds of TFs containing a specified DBD family, which occur in a known TFs. At the same time, up to an additional 50% of particular organism, e.g. all homoeodomain-containing proteins are predicted as TFs that have annotations such TFs in human (Figure 1a and b). as ‘hypothetical protein’, particularly in metazoan gen- We illustrate the TFs containing a specified DBD family omes. For details of benchmarking, please refer to (10). in a particular organism in Figure 1, where a hypothetical The prediction method is general and applicable to any researcher is interested in the Homeobox TFs. These TFs proteome or sequence set. In fact, the database has grown are known to regulate vertebrate limb formation amongst to encompass TF repertoires of over 700 publicly available other processes (16). Figure 1a depicts the search for TFs genomes. Predictions for newly sequenced genomes are in Homo sapiens containing the homoeobox domain. continuously added to the database. A subset of the results of this search are shown in The current DBD database contains information on Figure 1b. By selecting the HOXA9 TF from this result over 200 000 predicted TFs. These TFs are distributed set, the researcher can examine one of the new pages across the tree of life. It is not surprising that, we find a containing detailed information on each TF (Figure 1c). greater number of TFs in larger genomes. To investigate The detailed pages include the sequence of the TF, links to the relationship between TF abundance and proteome size external databases containing further information on the in different lineages we graph these variables on a log–log protein, domain assignment regions and an indication of plot as in Kummerfeld and Teichmann (10) (Supplemen- the quality of the domain assignment in the form of an tary Figure 2 in this paper). To illustrate the difference Evalue. Links to predicted TFs with similar domain between the eukaryotic and prokaryotic superkingdoms combinations are also provided on these pages. An we separately perform a model fitting for these lineages. example of predicted TFs with similar Pfam architectures From the linear relationship on the log–log scale a power to the HOXA9 TF (i.e. an N-terminal Hox9 activation law can be inferred. This power law could be due to the region and a C-terminal Homoeobox domain) is shown in underlying distribution of DBDs. A small number of Figure 1d. DBDs (such as helix–turn–helix and zinc-finger families) Using the data on DBD families in different organisms, occur in the majority of TFs. Whereas most DBDs occur we compare the occurrence of DBDs (from the Pfam in only a small number of TFs. In agreement with van project) across the tree of life. The heatmap in Figure 2 Nimwegen (14) and Ranea et al. (15), we find a higher demonstrates the lineage-specific DBD expansions and proportion of TFs are required to regulate larger contractions. The list of species and DBD lists are proteomes. We also find the TF abundance in archaea included in Supplementary Tables 1 and 2. We found and bacteria expands more rapidly than in eukaryotes. the number of occurrences of each DBD in each organism, Thus, in general, the same number of TFs regulate fewer and then normalized this number by the proteome size of prokaryotic genes than eukaryotic genes. The higher that organism. In order to represent both contractions and degree of combinatorial control, where gene expression is expansions, we calculated a Z-score for each of the regulated by not just one but by a group of TFs, may also normalized DBD occurrence values. The Z-score is contribute to the lower eukaryotic TF requirements. calculated from the distribution of normalized DBD Different combinations of TFs mean the number of gene occurrence across genomes for a particular DBD family, regulation modes can increase with a reduced increase in and has a mean of zero and a standard deviation of one. It TFs. Bacteria and archaea obey the same power law is negative when the normalized DBD occurrence is below in terms of number of TFs and number of proteins. This is the mean, and positive when above the mean. In Figure 2, in accordance with their shared repertoire of DBD DBD expansions (positive Z-scores) are represented using families, which we will return to below. red, and contractions (negative Z-scores) using green. Apicomplexa appear not to follow either the prokaryote Different sets of DBDs expand in different lineages. or typical eukaryote trends, perhaps because they are There is a clear separation between the DBD occurrence obligate parasites, and only survive in the nutrient-rich pattern in eukaryotes (in the top section of the heatmap) environment of their hosts. Thus, a different mode of gene regulation may be used by this lineage, or it is possible and prokaryotes. The DBD occurrence in prokaryotes is that their TFs are not well characterized by the current relatively diverse. For instance, there is a significant overlap model libraries. Below, we will illustrate in more detail between the DBD repertories of the actinobacteria, how the DBD database provides a consistent framework proteobacteria and firmicutes. This is almost certainly for comparison of the distribution of DBDs across the due to the ubiquitous horizontal gene transfer between tree of life. prokaryotes. The DBD expansion pattern in archaea is D90 Nucleic Acids Research, 2008, Vol. 36, Database issue Figure 1. Examples of new search capabilities and content. (a) Search for TFs from a particular organism containing a specified DBD. The example used here is TFs from Homo sapiens containing the homoeobox domain. (b) The search in (a) results in TF predictions from Homo sapiens containing the homoeobox DNA-binding domain. (c) Selection of HOXA9 from (b) results in a web page with detailed information on this particular TF. (d) Clicking on the Pfam domain combination link in (c) retrieves the subset of TF predictions, which have the same two-domain arrangement as the HOXA9 transcription factor. similar to that in bacteria, despite sharing conserved basal environmental conditions. Figure 2b clarifies the nature of the DBD expansions in the viridiplantae lineage. The transcriptional machinery with eukaryotes rather than with AP2 family is expanded throughout this lineage, but is bacteria. The majority of these prokaryotic DBDs have the believed to also occur in the apicomplexa (19). Figure 2c helix–turn–helix as part of their structure (17). shows the AP2 domain in complex with DNA. This family The eukaryote-specific DBD expansions have consider- is known to bind to the GCC-box pathogenesis-related ably greater variety than the prokaryotic expansions. An promoter element (20) and activate defence genes. Several increased DBD kingdom-specificity is found in the families are specifically expanded in the plant genomes of eukaryotes. The metazoan, fungal and plant kingdoms A. thaliana, Medicago truncatula and Oryza sativa (as are clearly distinguishable (Figure 2a). The fungal and opposed to the other viridiplantae, which are algae) metazoan kingdoms share more DBDs than the plant and including the family of ethylene insensitive 3 (EIN3) metazoan kingdoms, which reflects their closer phyloge- DBDs. This family regulates transcription in response to netic relationship (18). The metazoa, in the top right the chemically simplest plant hormone, ethylene (21). section of Figure 2a, have the largest kingdom-specific DBD repertoire. This is most likely due to the regulatory overhead of metazoan complexity in terms of cell types. FUTURE DIRECTIONS The significant plant-specific DBD expansion is possibly due to the regulation of a large defence system—which Above we described novel developments in the display plants have due to their inability to escape toxic facilities and search tools, as well as the content of the Nucleic Acids Research, 2008, Vol. 36, Database issue D91 Figure 2. (a) Expansion and contraction patterns of DBD occurrence across the tree of life. Each column corresponds to a Pfam DBD. Each row of the heatmap represents a genome, ordered using the NCBI taxonomy. The vertical coloured bars indicate superkingdoms, kingdoms or phyla to which genomes belong. Eukaryotes are indicated using a red bar, archaea using a green bar and bacteria using a blue bar. Other kingdoms are represented using white bars. DNA-binding domain families are clustered using the average linkage method with Pearson correlation distance. Red squares represent an expansion of a DBD family, green squares represent a contraction of that family in a genome relative to other genomes. (b) A zoom on DBD expansions in the viridiplantae lineage. (c) Illustration of the three-dimensional structure of one of the DBDs specifically expanded in the viridiplantae kingdom, the AP2 domain in complex with DNA. The AP2 family transcription factors are known to be involved in plant pathogen defence response processes. DBD database, with a few examples of the type of insight proteomes we hope to add soon include higher eukaryotes this provides. In the future, we will continue to update the such as orangutan, marmoset and wallaby, disease vector HMM libraries, which will result in improvements to the insects, additional nematodes and several plants. TF prediction coverage. When updating the Pfam HMMs We have eliminated several eukaryotic genomes we will make use of, and incorporate, the Pfam clan (Xenopus tropicalis, Apis melifera and Populus trichcarpa) information (12). We will also continue to add and update from our analysis of DBD occurrence due to the presence predictions for new proteomes. Exciting new eukaryotic of uncharacteristically high numbers of bacterial DBDs. D92 Nucleic Acids Research, 2008, Vol. 36, Database issue 7. Amoutzias,G.D., Robertson,D.L., Oliver,S.G. and Bornberg- This was a known problem in the X. tropicalis (frog) Bauer,E. (2004) Convergent evolution of gene networks by genome (22). The use of lineage-specific information on the single-gene duplications in higher eukaryotes. EMBO Rep., 5, occurrence of DBDs is a promising method for reducing 274–279. false-positive TF classifications in the eukaryotes. 8. Mulder,N.J., Apweiler,R., Attwood,T.K., Bairoch,A., Bateman,A., Binns,D., Bork,P., Buillard,V., Cerutti,L. et al. (2007) New We also plan to refine the TF prediction procedure by developments in the InterPro database. Nucleic Acids Res., 35, taking into account that DBDs have typical patterns of 224–228. domain repetition or combination with other DBDs or 9. Adryan,B. and Teichmann,S.A. (2006) FlyTF: a systematic review non-DBDs. It may be possible to make use of over- of site-specific transcription factors in the fruit fly drosophila represented domain combinations to further improve melanogaster. Bioinformatics, 22, 1532–1533. 10. Kummerfeld,S.K. and Teichmann,S.A. (2006) DBD: a transcription our predictions, for instance by including marginal DBD factor prediction database. Nucleic Acids Res., 34, 74–81. matches if they occur in common TF domain arrange- 11. Wilson,D., Madera,M., Vogel,C., Chothia,C. and Gough,J. (2007) ments as indicated by the statistical methods used in (23) The SUPERFAMILY database in 2007: families and functions. and (24). Nucleic Acids Res., 35, 308–313. 12. Finn,R.D., Mistry,J., Schuster-Bo¨ ckler,B., Griffiths-Jones,S., Hollich,V., Lassmann,T., Moxon,S., Marshall,M., Khanna,A. et al. (2006) Pfam: clans, web tools and services. Nucleic Acids Res., 34, SUPPLEMENTARY DATA 247–251. 13. Murzin,A.G., Brenner,S.E., Hubbard,T. and Chothia,C. (1995) Supplementary Data are available at NAR Online. SCOP: a structural classification of proteins database for the investigation of sequences and structures. J. Mol. Biol., 247, 536–540. ACKNOWLEDGEMENTS 14. van Nimwegen,E. (2003) Scaling laws in the functional content of genomes. Trends Genet., 19, 479–484. We gratefully acknowledge comments on the manuscript 15. Ranea,J.A., Buchan,D.W., Thornton,J.M. and Orengo,C.A. (2004) from Subhajyoti De and Siarhei Maslau. This work was Evolution of protein superfamilies and bacterial genome size. J. Mol. Biol., 336, 871–887. funded by Medical Research Council; Royal Thai 16. Cohn,M.J., Patel,K., Krumlauf,R., Wilkinson,D.G., Clarke,J.D. Government Scholarship to VC. Funding to pay the and Tickle,C. (1997) Hox9 genes and vertebrate limb specification. Open Access publication charges for this article was Nature, 387, 97–101. provided by Medical Research Council. 17. Pe´ rez-Rueda,E., Collado-Vides,J. and Segovia,L. (2004) Phylogenetic distribution of DNA-binding transcription factors in Conflict of interest statement. None declared. bacteria and archaea. Comput. Biol. Chem., 28, 341–350. 18. Baldauf,S.L., Roger,A.J., Wenk-Siefert,I. and Doolittle,W.F. (2000) A kingdom-level phylogeny of eukaryotes based on combined protein data. Science, 290, 972–977. REFERENCES 19. Balaji,S., Babu,M.M., Iyer,L.M. and Aravind,L. (2005) Discovery 1. Robertson,G., Bilenky,M., Lin,K., He,A., Yuen,W., Dagpinar,M., of the principal specific transcription factors of apicomplexa and Varhol,R., Teague,K., Griffith,O.L. et al. (2006) cisRED: a their implication for the evolution of the ap2-integrase DNA database system for genome-scale computational discovery of binding domains. Nucleic Acids Res., 33, 3994–4006. regulatory elements. Nucleic Acids Res., 34, 68–73. 20. Ohme-Takagi,M. and Shinshi,H. (1995) Ethylene-inducible DNA- 2. Bulyk,M.L. (2003) Computational prediction of transcription-factor binding proteins that interact with an ethylene-responsive element. binding site locations. Genome Biol., 5, 201–201. Plant Cell, 7, 173–182. 3. Hermoso,A., Aguilar,D., Aviles,F.X. and Querol,E. (2004) TrSDB: 21. Chao,Q., Rothenberg,M., Solano,R., Roman,G., Terzaghi,W. and a proteome database of transcription factors. Nucleic Acids Res., 32, Ecker,J.R. (1997) Activation of the ethylene gas response pathway 171–173. in arabidopsis by the nuclear protein ethylene-insensitive3 and 4. Messina,D.N., Glasscock,J., Gish,W. and Lovett,M. (2004) An related proteins. Cell, 89, 1133–1144. ORFeome-based analysis of human transcription factor genes and 22. Yang,S., Doolittle,R.F. and Bourne,P.E. (2005) Phylogeny deter- the construction of a microarray to interrogate their expression. mined by protein domain content. Proc. Natl Acad. Sci. USA, 102, Genome Res., 14, 2041–2047. 373–378. 5. Barrasa,M.I., Vaglio,P., Cavasino,F., Jacotot,L. and Walhout,A.J. 23. Mott,R., Schultz,J., Bork,P. and Ponting,C.P. (2002) Predicting (2007) EdgeDB: a transcription factor-DNA interaction database protein cellular localization using a domain projection method. for the analysis of C. elegans differential gene expression. BMC Genome Res., 12, 1168–1174. Genomics, 8, 21–21. 24. Coin,L., Bateman,A. and Durbin,R. (2003) Enhanced 6. Drosophila Comparative Genome Sequencing and Analysis protein domain discovery by using language modeling techniques Consortium. (2007) Evolution of genes and genomes in the context from speech recognition. Proc. Natl Acad. Sci. USA, 100, of the drosophila phylogeny. Nature, In press, Nov. 4516–4520. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Nucleic Acids Research Oxford University Press

DBD––taxonomically broad transcription factor predictions: new content and functionality

Download PDF
5 pages

Loading...

Page 2

Loading...

Page 3

Loading...

Page 4

Loading...

Page 5

 
/lp/oxford-university-press/dbd-taxonomically-broad-transcription-factor-predictions-new-content-UYItvUymvV

References (26)

Publisher
Oxford University Press
Copyright
© Published by Oxford University Press.
ISSN
0305-1048
eISSN
1362-4962
DOI
10.1093/nar/gkm964
pmid
18073188
Publisher site
See Article on Publisher Site

Abstract

D88–D92 Nucleic Acids Research, 2008, Vol. 36, Database issue Published online 11 December 2007 doi:10.1093/nar/gkm964 DBD––taxonomically broad transcription factor predictions: new content and functionality 1, 1 2 Derek Wilson *, Varodom Charoensawan , Sarah K. Kummerfeld and Sarah A. Teichmann 1 2 MRC Laboratory of Molecular Biology, Hills Road, Cambridge, CB2 0QH, UK and Department of Developmental Biology, Stanford University Medical Center, 279 Campus Drive, Stanford, CA 94305-5329, USA Received September 14, 2007; Revised October 16, 2007; Accepted October 17, 2007 Like other proteins, TFs are composed of evolutionary ABSTRACT units called domains, which belong to families that can DNA-binding domain (DBD) is a database of pre- occur in many different proteins and various domain dicted sequence-specific DNA-binding transcription combinations. In the DBD database, we define TFs as factors (TFs) for all publicly available proteomes. proteins containing a sequence-specific DNA-binding The proteomes have increased from 150 in the initial domain (DBD). Other databases, such as TrSDB (3), or version of DBD to over 700 in the current version. data sets, such as Messina et al. (4), include both specific All predicted TFs must contain a significant match and general TFs. The precise description of TFs as to a hidden Markov model representing a sequence- sequence-specific DNA-binding we use is useful in a wide variety of studies. Examples include: improving specific DNA-binding domain family. Access to TF genome annotation; high-throughput experiments such as predictions is provided through http://transcription- ChIP–chip, protein chip or yeast one-hybrid (5); and factor.org, where new search options are now studies of the evolution of gene regulation comparing provided such as searching by gene names in multiple genomes (6), or gene regulation networks (7). The model organisms, searching for all proteins in a DBD database has been used as an annotation tool in the particular DBD family and specific organism. We context of the InterPro (8) and FlyTF (http://FlyTF.org) illustrate the application of this type of search (9) databases. facility by contrasting trends of DBD family occur- Access to the DBD database is via http://transcription rence throughout the tree of life, highlighting the factor.org, where all data is available for viewing and clear partition between eukaryotic and prokaryotic immediate download. The community can browse predic- DBD expansions. The website content has been tions for over 700 species (from Arabidopsis thaliana to expanded to include dedicated pages for each TF Zymomonas mobilis) or DBD family (including helix– containing domain assignment details, gene names, turn–helix, zinc-fingers, homeobox and many others); links to external databases and links to TFs with search predictions by sequence identifier or domain similar domain arrangements. We compare the family; receive classifications for submitted protein sequences, and download our domain assignments, as increase in number of predicted TFs with proteome well as our manually curated list of DBDs. size in eukaryotes and prokaryotes. Eukaryotes The prediction method in the DBD database (10) uses follow a slower rate of increase in TFs than hidden Markov models (HMMs) to identify domains prokaryotes, which could be due to the presence in proteins from two databases: SUPERFAMILY (11) of splice variants or an increase in combinatorial and Pfam (12). From DBD release 2.0 onwards, updated control. annotation resulted in 303 HMMs from SUPERFAMILY and 145 from Pfam compared to a total of 251 HMMs INTRODUCTION in the first version of DBD. The HMMs from SUPERFAMILY represent 37 superfamilies and 87 Sequence-specific DNA-binding transcription factors families according to the definitions in the SCOP database (TFs) each recognize a family of cis-regulatory DNA (13). This includes 98 new models representing 37 sequences described by a consensus motif (1) or position- sequence-specific DBD families. This resulted in an specific weight matrix (2). They regulate spatial and increase in additional TF predictions of 4.7%, for the temporal gene expression by binding to DNA and either activating or repressing action of an RNA polymerase. 150 organisms in the original version of DBD. *To whom correspondence should be addressed. Tel: +44 (0)1223 402479; Fax: +44 (0)1223 213556; Email: dbd@mrc-lmb.cam.ac.uk 2007 The Author(s) This is an Open Access article distributed under the terms of the Creative Commons Attribution Non-Commercial License (http://creativecommons.org/licenses/ by-nc/2.0/uk/) which permits unrestricted non-commercial use, distribution, and reproduction in any medium, provided the original work is properly cited. Nucleic Acids Research, 2008, Vol. 36, Database issue D89 The pipeline used to predict TFs begins with a domain NOVEL DEVELOPMENTS annotation of all proteins from completely sequenced Researchers can use the DBD database in several ways. For genomes with all HMMs from the SUPERFAMILY and instance, all TF predictions are available to download. Pfam databases (Supplementary Figure 1). A protein is However, most users are only interested in a small number classified as a TF if it has a significant match to a model of TFs, so we have expanded the website search options to we annotated as being a DBD, with the significance allow retrieval of individual TFs and subsets of TFs. New thresholds for HMM matches taken from the Pfam and search capabilities include: searching for gene names, for SUPERFAMILY databases. This results in an estimated example lacI or P53; listing all TFs that contain either a 1–5% of false-positive annotations. The TF predictions specified DBD or non-DBD family, for instance all TFs are limited to the families in our annotated collection, containing the bZIP (leucine zipper) family; retrieving all which means that the coverage is about two-thirds of TFs containing a specified DBD family, which occur in a known TFs. At the same time, up to an additional 50% of particular organism, e.g. all homoeodomain-containing proteins are predicted as TFs that have annotations such TFs in human (Figure 1a and b). as ‘hypothetical protein’, particularly in metazoan gen- We illustrate the TFs containing a specified DBD family omes. For details of benchmarking, please refer to (10). in a particular organism in Figure 1, where a hypothetical The prediction method is general and applicable to any researcher is interested in the Homeobox TFs. These TFs proteome or sequence set. In fact, the database has grown are known to regulate vertebrate limb formation amongst to encompass TF repertoires of over 700 publicly available other processes (16). Figure 1a depicts the search for TFs genomes. Predictions for newly sequenced genomes are in Homo sapiens containing the homoeobox domain. continuously added to the database. A subset of the results of this search are shown in The current DBD database contains information on Figure 1b. By selecting the HOXA9 TF from this result over 200 000 predicted TFs. These TFs are distributed set, the researcher can examine one of the new pages across the tree of life. It is not surprising that, we find a containing detailed information on each TF (Figure 1c). greater number of TFs in larger genomes. To investigate The detailed pages include the sequence of the TF, links to the relationship between TF abundance and proteome size external databases containing further information on the in different lineages we graph these variables on a log–log protein, domain assignment regions and an indication of plot as in Kummerfeld and Teichmann (10) (Supplemen- the quality of the domain assignment in the form of an tary Figure 2 in this paper). To illustrate the difference Evalue. Links to predicted TFs with similar domain between the eukaryotic and prokaryotic superkingdoms combinations are also provided on these pages. An we separately perform a model fitting for these lineages. example of predicted TFs with similar Pfam architectures From the linear relationship on the log–log scale a power to the HOXA9 TF (i.e. an N-terminal Hox9 activation law can be inferred. This power law could be due to the region and a C-terminal Homoeobox domain) is shown in underlying distribution of DBDs. A small number of Figure 1d. DBDs (such as helix–turn–helix and zinc-finger families) Using the data on DBD families in different organisms, occur in the majority of TFs. Whereas most DBDs occur we compare the occurrence of DBDs (from the Pfam in only a small number of TFs. In agreement with van project) across the tree of life. The heatmap in Figure 2 Nimwegen (14) and Ranea et al. (15), we find a higher demonstrates the lineage-specific DBD expansions and proportion of TFs are required to regulate larger contractions. The list of species and DBD lists are proteomes. We also find the TF abundance in archaea included in Supplementary Tables 1 and 2. We found and bacteria expands more rapidly than in eukaryotes. the number of occurrences of each DBD in each organism, Thus, in general, the same number of TFs regulate fewer and then normalized this number by the proteome size of prokaryotic genes than eukaryotic genes. The higher that organism. In order to represent both contractions and degree of combinatorial control, where gene expression is expansions, we calculated a Z-score for each of the regulated by not just one but by a group of TFs, may also normalized DBD occurrence values. The Z-score is contribute to the lower eukaryotic TF requirements. calculated from the distribution of normalized DBD Different combinations of TFs mean the number of gene occurrence across genomes for a particular DBD family, regulation modes can increase with a reduced increase in and has a mean of zero and a standard deviation of one. It TFs. Bacteria and archaea obey the same power law is negative when the normalized DBD occurrence is below in terms of number of TFs and number of proteins. This is the mean, and positive when above the mean. In Figure 2, in accordance with their shared repertoire of DBD DBD expansions (positive Z-scores) are represented using families, which we will return to below. red, and contractions (negative Z-scores) using green. Apicomplexa appear not to follow either the prokaryote Different sets of DBDs expand in different lineages. or typical eukaryote trends, perhaps because they are There is a clear separation between the DBD occurrence obligate parasites, and only survive in the nutrient-rich pattern in eukaryotes (in the top section of the heatmap) environment of their hosts. Thus, a different mode of gene regulation may be used by this lineage, or it is possible and prokaryotes. The DBD occurrence in prokaryotes is that their TFs are not well characterized by the current relatively diverse. For instance, there is a significant overlap model libraries. Below, we will illustrate in more detail between the DBD repertories of the actinobacteria, how the DBD database provides a consistent framework proteobacteria and firmicutes. This is almost certainly for comparison of the distribution of DBDs across the due to the ubiquitous horizontal gene transfer between tree of life. prokaryotes. The DBD expansion pattern in archaea is D90 Nucleic Acids Research, 2008, Vol. 36, Database issue Figure 1. Examples of new search capabilities and content. (a) Search for TFs from a particular organism containing a specified DBD. The example used here is TFs from Homo sapiens containing the homoeobox domain. (b) The search in (a) results in TF predictions from Homo sapiens containing the homoeobox DNA-binding domain. (c) Selection of HOXA9 from (b) results in a web page with detailed information on this particular TF. (d) Clicking on the Pfam domain combination link in (c) retrieves the subset of TF predictions, which have the same two-domain arrangement as the HOXA9 transcription factor. similar to that in bacteria, despite sharing conserved basal environmental conditions. Figure 2b clarifies the nature of the DBD expansions in the viridiplantae lineage. The transcriptional machinery with eukaryotes rather than with AP2 family is expanded throughout this lineage, but is bacteria. The majority of these prokaryotic DBDs have the believed to also occur in the apicomplexa (19). Figure 2c helix–turn–helix as part of their structure (17). shows the AP2 domain in complex with DNA. This family The eukaryote-specific DBD expansions have consider- is known to bind to the GCC-box pathogenesis-related ably greater variety than the prokaryotic expansions. An promoter element (20) and activate defence genes. Several increased DBD kingdom-specificity is found in the families are specifically expanded in the plant genomes of eukaryotes. The metazoan, fungal and plant kingdoms A. thaliana, Medicago truncatula and Oryza sativa (as are clearly distinguishable (Figure 2a). The fungal and opposed to the other viridiplantae, which are algae) metazoan kingdoms share more DBDs than the plant and including the family of ethylene insensitive 3 (EIN3) metazoan kingdoms, which reflects their closer phyloge- DBDs. This family regulates transcription in response to netic relationship (18). The metazoa, in the top right the chemically simplest plant hormone, ethylene (21). section of Figure 2a, have the largest kingdom-specific DBD repertoire. This is most likely due to the regulatory overhead of metazoan complexity in terms of cell types. FUTURE DIRECTIONS The significant plant-specific DBD expansion is possibly due to the regulation of a large defence system—which Above we described novel developments in the display plants have due to their inability to escape toxic facilities and search tools, as well as the content of the Nucleic Acids Research, 2008, Vol. 36, Database issue D91 Figure 2. (a) Expansion and contraction patterns of DBD occurrence across the tree of life. Each column corresponds to a Pfam DBD. Each row of the heatmap represents a genome, ordered using the NCBI taxonomy. The vertical coloured bars indicate superkingdoms, kingdoms or phyla to which genomes belong. Eukaryotes are indicated using a red bar, archaea using a green bar and bacteria using a blue bar. Other kingdoms are represented using white bars. DNA-binding domain families are clustered using the average linkage method with Pearson correlation distance. Red squares represent an expansion of a DBD family, green squares represent a contraction of that family in a genome relative to other genomes. (b) A zoom on DBD expansions in the viridiplantae lineage. (c) Illustration of the three-dimensional structure of one of the DBDs specifically expanded in the viridiplantae kingdom, the AP2 domain in complex with DNA. The AP2 family transcription factors are known to be involved in plant pathogen defence response processes. DBD database, with a few examples of the type of insight proteomes we hope to add soon include higher eukaryotes this provides. In the future, we will continue to update the such as orangutan, marmoset and wallaby, disease vector HMM libraries, which will result in improvements to the insects, additional nematodes and several plants. TF prediction coverage. When updating the Pfam HMMs We have eliminated several eukaryotic genomes we will make use of, and incorporate, the Pfam clan (Xenopus tropicalis, Apis melifera and Populus trichcarpa) information (12). We will also continue to add and update from our analysis of DBD occurrence due to the presence predictions for new proteomes. Exciting new eukaryotic of uncharacteristically high numbers of bacterial DBDs. D92 Nucleic Acids Research, 2008, Vol. 36, Database issue 7. Amoutzias,G.D., Robertson,D.L., Oliver,S.G. and Bornberg- This was a known problem in the X. tropicalis (frog) Bauer,E. (2004) Convergent evolution of gene networks by genome (22). The use of lineage-specific information on the single-gene duplications in higher eukaryotes. EMBO Rep., 5, occurrence of DBDs is a promising method for reducing 274–279. false-positive TF classifications in the eukaryotes. 8. Mulder,N.J., Apweiler,R., Attwood,T.K., Bairoch,A., Bateman,A., Binns,D., Bork,P., Buillard,V., Cerutti,L. et al. (2007) New We also plan to refine the TF prediction procedure by developments in the InterPro database. Nucleic Acids Res., 35, taking into account that DBDs have typical patterns of 224–228. domain repetition or combination with other DBDs or 9. Adryan,B. and Teichmann,S.A. (2006) FlyTF: a systematic review non-DBDs. It may be possible to make use of over- of site-specific transcription factors in the fruit fly drosophila represented domain combinations to further improve melanogaster. Bioinformatics, 22, 1532–1533. 10. Kummerfeld,S.K. and Teichmann,S.A. (2006) DBD: a transcription our predictions, for instance by including marginal DBD factor prediction database. Nucleic Acids Res., 34, 74–81. matches if they occur in common TF domain arrange- 11. Wilson,D., Madera,M., Vogel,C., Chothia,C. and Gough,J. (2007) ments as indicated by the statistical methods used in (23) The SUPERFAMILY database in 2007: families and functions. and (24). Nucleic Acids Res., 35, 308–313. 12. Finn,R.D., Mistry,J., Schuster-Bo¨ ckler,B., Griffiths-Jones,S., Hollich,V., Lassmann,T., Moxon,S., Marshall,M., Khanna,A. et al. (2006) Pfam: clans, web tools and services. Nucleic Acids Res., 34, SUPPLEMENTARY DATA 247–251. 13. Murzin,A.G., Brenner,S.E., Hubbard,T. and Chothia,C. (1995) Supplementary Data are available at NAR Online. SCOP: a structural classification of proteins database for the investigation of sequences and structures. J. Mol. Biol., 247, 536–540. ACKNOWLEDGEMENTS 14. van Nimwegen,E. (2003) Scaling laws in the functional content of genomes. Trends Genet., 19, 479–484. We gratefully acknowledge comments on the manuscript 15. Ranea,J.A., Buchan,D.W., Thornton,J.M. and Orengo,C.A. (2004) from Subhajyoti De and Siarhei Maslau. This work was Evolution of protein superfamilies and bacterial genome size. J. Mol. Biol., 336, 871–887. funded by Medical Research Council; Royal Thai 16. Cohn,M.J., Patel,K., Krumlauf,R., Wilkinson,D.G., Clarke,J.D. Government Scholarship to VC. Funding to pay the and Tickle,C. (1997) Hox9 genes and vertebrate limb specification. Open Access publication charges for this article was Nature, 387, 97–101. provided by Medical Research Council. 17. Pe´ rez-Rueda,E., Collado-Vides,J. and Segovia,L. (2004) Phylogenetic distribution of DNA-binding transcription factors in Conflict of interest statement. None declared. bacteria and archaea. Comput. Biol. Chem., 28, 341–350. 18. Baldauf,S.L., Roger,A.J., Wenk-Siefert,I. and Doolittle,W.F. (2000) A kingdom-level phylogeny of eukaryotes based on combined protein data. Science, 290, 972–977. REFERENCES 19. Balaji,S., Babu,M.M., Iyer,L.M. and Aravind,L. (2005) Discovery 1. Robertson,G., Bilenky,M., Lin,K., He,A., Yuen,W., Dagpinar,M., of the principal specific transcription factors of apicomplexa and Varhol,R., Teague,K., Griffith,O.L. et al. (2006) cisRED: a their implication for the evolution of the ap2-integrase DNA database system for genome-scale computational discovery of binding domains. Nucleic Acids Res., 33, 3994–4006. regulatory elements. Nucleic Acids Res., 34, 68–73. 20. Ohme-Takagi,M. and Shinshi,H. (1995) Ethylene-inducible DNA- 2. Bulyk,M.L. (2003) Computational prediction of transcription-factor binding proteins that interact with an ethylene-responsive element. binding site locations. Genome Biol., 5, 201–201. Plant Cell, 7, 173–182. 3. Hermoso,A., Aguilar,D., Aviles,F.X. and Querol,E. (2004) TrSDB: 21. Chao,Q., Rothenberg,M., Solano,R., Roman,G., Terzaghi,W. and a proteome database of transcription factors. Nucleic Acids Res., 32, Ecker,J.R. (1997) Activation of the ethylene gas response pathway 171–173. in arabidopsis by the nuclear protein ethylene-insensitive3 and 4. Messina,D.N., Glasscock,J., Gish,W. and Lovett,M. (2004) An related proteins. Cell, 89, 1133–1144. ORFeome-based analysis of human transcription factor genes and 22. Yang,S., Doolittle,R.F. and Bourne,P.E. (2005) Phylogeny deter- the construction of a microarray to interrogate their expression. mined by protein domain content. Proc. Natl Acad. Sci. USA, 102, Genome Res., 14, 2041–2047. 373–378. 5. Barrasa,M.I., Vaglio,P., Cavasino,F., Jacotot,L. and Walhout,A.J. 23. Mott,R., Schultz,J., Bork,P. and Ponting,C.P. (2002) Predicting (2007) EdgeDB: a transcription factor-DNA interaction database protein cellular localization using a domain projection method. for the analysis of C. elegans differential gene expression. BMC Genome Res., 12, 1168–1174. Genomics, 8, 21–21. 24. Coin,L., Bateman,A. and Durbin,R. (2003) Enhanced 6. Drosophila Comparative Genome Sequencing and Analysis protein domain discovery by using language modeling techniques Consortium. (2007) Evolution of genes and genomes in the context from speech recognition. Proc. Natl Acad. Sci. USA, 100, of the drosophila phylogeny. Nature, In press, Nov. 4516–4520.

Journal

Nucleic Acids Research – Oxford University Press

Published: Jan 11, 2008

There are no references for this article.