Skip to main page content

• HOME
• Current Issue
• Archive
• Alerts
• About ASM
• Contact us
• Tech Support
• Journals.ASM.org

This Article Abstract Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Reprints and Permissions Copyright Information Books from ASM Press MicrobeWorld Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Singh, A. H. Articles by Bork, P. Search for Related Content PubMed PubMed Citation Articles by Singh, A. H. Articles by Bork, P.

 Previous Article  |  Next Article 

Journal of Bacteriology, January 2009, p. 32-41, Vol. 191, No. 1
0021-9193/09/$08.00+0     doi:10.1128/JB.01084-08
Copyright © 2009, American Society for Microbiology. All Rights Reserved.

Discovering Functional Novelty in Metagenomes: Examples from Light-Mediated Processes

Amoolya H. Singh, Tobias Doerks, Ivica Letunic, Jeroen Raes, and Peer Bork^*

Structural and Computational Biology Unit, European Molecular Biology Laboratory, Meyerhofstrasse 1, 69117 Heidelberg, Germany

Received 5 August 2008/ Accepted 1 October 2008

Top ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 
The emerging coverage of diverse habitats by metagenomic shotgun data opens new avenues of discovering functional novelty using computational tools. Here, we apply three different concepts for predicting novel functions within light-mediated microbial pathways in five diverse environments. Using phylogenetic approaches, we discovered two novel deep-branching subfamilies of photolyases (involved in light-mediated repair) distributed abundantly in high-UV environments. Using neighborhood approaches, we were able to assign seven novel functional partners in luciferase synthesis, nitrogen metabolism, and quorum sensing to BLUF domain-containing proteins (involved in light sensing). Finally, by domain analysis, for RcaE proteins (involved in chromatic adaptation), we predict 16 novel domain architectures that indicate novel functionalities in habitats with little or no light. Quantification of protein abundance in the various environments supports our findings that bacteria utilize light for sensing, repair, and adaptation far more widely than previously thought. While the discoveries illustrate the opportunities in function discovery, we also discuss the immense conceptual and practical challenges that come along with this new type of data.

Top ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 
One of the central questions in biology, starting from the time of Charles Darwin, has been the extent and distribution of biological diversity (68). The recent sequencing of several hundred bacterial and archaeal genomes and metagenomes, along with the discovery of large-scale lateral gene transfer (10) and recombination (25) in bacterial evolution, has not only renewed interest in the question of diversity but also confounded it. The sequencing projects reveal that contrary to previous estimates, it is microbes that account for the vast majority of diversity in phenotype and genotype on earth (44, 47). Underlying this dazzling diversity in species and habitat is molecular diversity. Indeed, we are just beginning to scratch the surface of this molecular diversity (50). Even though our understanding of how the living world functions at the molecular level is far from complete, the discovery of novel molecules has important applications to medicine, agriculture, industry, and environmental conservation and remediation.

But how are we to discover functional novelty in the exponentially increasing amounts of sequenced genes and habitats (Fig. 1)? The naïve method, which is to search for homology to known molecules and mark everything else as novel, is prone to errors due to the existence of paralogous sequences, i.e., homologs with likely different functionalities, as well as paralogous domains within an otherwise homologous sequence that may lead to divergent function (17). To address these challenges, three major, nonexclusive concepts have been successfully used to establish functional similarity and, conversely, to identify functional novelty: (i) operons and conserved gene neighborhoods, (ii) protein domain architectures, and (iii) protein subfamilies. Operon and gene neighborhood methods assume that if multiple genes are adjacent on a chromosome or contig, they are more likely to participate in the same cellular function (19, 48). The neighborhood approach is especially suitable when homology-based methods fail to detect sequences below the threshold for similarity (30). Domain-based methods infer the functions of similar segments within otherwise different sequences and are currently utilized by curated databases of known domains (see, e.g., reference 18). This approach is useful for the analysis of multidomain proteins that evolve in a modular fashion such that each domain may have high sequence similarity to a different gene and its evolution cannot be traced by homology alone (55). Finally, subfamilies can be identified within a family of homologous sequences by abstracting the information from the family's multiple sequence alignment into a generalized statistical profile (e.g., using hidden Markov models or support vector machines) (11) and then searching for shared properties (e.g., amino acids and hydrophobicity). This technique has been successful at identifying novel biological function (2, 29, 36) and even novel species (13).

View larger version (45K): [in this window] [in a new window]

FIG. 1. Trends in the increase of genomic data and represented habitats. The number of sequenced ORFs continues to increase exponentially, accompanied by an increase in the number and complexity of represented habitats. In 1995, two sequenced organisms (Haemophilus influenzae and Mycoplasma pneumoniae) contributed just a few thousand genes to the public databases and represented a single habitat (organism associated). By 2008, well over 10 million genes from over 150 distinct habitats have been sequenced. Raw habitat and sequence data were collected from the Genomes Online database (43), and habitats were classified into the categories mentioned above using the Habitat-Lite terms of Environment Ontology (32). When an organism was reported to have multiple habitats, the primary one was used. "Extreme environment" corresponds to the Environment Ontology categories "hot spring," "hydrothermal vent," and "extreme environment." "Sediment/sludge" corresponds to the Environment Ontology categories "sediment," "sludge," and "biofilm." Note that (i) dates reported for each sequence are publication dates, even if the genome was released to the public earlier in database form, and (ii) the numbers for 2008 represent the available data until June 2008.

Although these methods are conceptually straightforward, identifying novelty from environmental data remains difficult. The primary reason is the sheer volume of data, for which there is no centralized repository or standardized reporting of sampling conditions. The size of the data sets is further exacerbated by problems with the data itself, such as the presence of incomplete gene fragments, the uniformity of sequence coverage, and the use of shotgun sequencing. Incomplete gene fragments, an artifact of current environmental sequencing methodologies, limit the ability to correctly predict open reading frames (ORFs) and assign function. Uniform sequence coverage implies that if a protein family is rare in a particular environment, or belongs to a rare species, it might not be seen at all. Since functional novelty seems to be contributed primarily by rare families (30, 51, 70) that mediate unusual niche-specific adaptations, the inability to detect rare proteins fundamentally limits our ability to discover novelty, although this may change as technical advances enable targeted deep sequencing in high-diversity environments. Finally, the use of shotgun sequencing techniques (as opposed to, for example, single-cell sequencing) makes it tricky to unite species and function identification. The existing protocols to map taxonomy are either limited to querying against a small number (e.g., 30 to 40) of marker genes (66) or falling back on error-prone annotation transfer from homology (33).

In addition to problems with the size and nature of metagenomic data, computational tools must be adapted for reproducibly handling gigabytes or terabytes of data, leading to constraints in memory, central processing unit (CPU), and network bandwidth at every level of analysis. Tools must be adapted to process, filter, assemble, and align the sequence data; identify genes; annotate the genes with function; map genes or sequences to taxonomy; estimate species evenness and richness; construct phylogenetic trees; perform multivariate analyses against ecological metrics; build and validate population or metabolic models where time series data are available; and visualize the results. Even as computational biologists adapt standard tools to complete these tasks, mathematicians and statisticians must rigorously reassess the suitabilities of different methods to large data sets, identify sources of analytical and numerical errors, and revise estimates of sensitivity and specificity.

Even if novel techniques such as single-cell sequencing reduce some of above-described problems in the future, certain challenges will remain unless our entire planet has been genetically explored in sufficient depth. This is because the remaining challenges are conceptual by nature. For example, the identification of orthology is already extremely difficult with complete genomes in hand due to chromosomal inversions, gene fusions, alternative splicing, retrotranscription, and a variety of genetic processes that dilute the necessary information. This genetic uncertainty is matched by a functional one: because the term "function" remains in use with an operational rather than absolute definition (8, 9, 30), annotation processes will remain of insufficient depth for quite some time.

Despite these limitations, the benefits of "bioprospecting" for natural and naturally derived products are considerable, with potential to cure genetic and infectious diseases, arrest environmental destruction, and offset global energy shortages. Here, we hope to raise awareness of the potentials and pitfalls of using environmental sequence data to discover novelty and illustrate the promise of our methods to discover novelty in light-mediated microbial pathways functioning in sensing, repair, and adaptation.

Top ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 
Collecting genome, metagenome, and habitat data. We collected raw data on habitat and number of sequenced ORFs (Fig. 1) from the Genomes Online database (43) on 12 June 2008, consisting of 802 genomes (669 species of bacteria, 53 archaea, and 80 eukaryotes) and 25 metagenomes. The date of publication was used as the sequencing date for a genome or metagenome. We classified the 86 distinct annotated habitats for these 827 genomes/metagenomes into 10 categories using the Habitat-Lite subset of terms (32) from the Environmental Ontology database (www.environmentontology.org). When an organism was reported to have multiple habitats, the primary one was used as input for data in Fig. 1. Primary habitats were checked against Bergey's Manual of Systematic Bacteriology (6), the online catalogs of the American Type Culture Collection (www.atcc.org) and the German Collection of Microorganisms and Cell Cultures (www.dsmz.de), and a previous large-scale description of bacterial phenotype and habitat (57). In the interest of clarity, for Fig. 1, we grouped certain habitats: "extreme environment" corresponds to the Habitat-Lite categories "hot spring," "hydrothermal vent," or "extreme environment." "Sediment/sludge" corresponds to the Habitat-Lite categories "sediment," "sludge," or "biofilm."

Calculating homologs and protein abundances in metagenomes. Metagenome sequence data from five metagenomes (6,109,937 ORFs from surface seawater from the Global Ocean Survey [70] including the Sargasso Sea [64], 46,771 ORFs from northern California acidic mine drainage [63], 121,927 ORFs from deep-sea Pacific whalefall [62], 183,159 ORFs from Minnesota farm soil [62], and 135,756 ORFs from a Mexican hypersaline microbial mat [39]) were BLASTed against the entire set of 1,510,991 proteins (representing 373 sequenced organisms) in the STRING 7.0 database (67) using wu-blastall with the parameters: –a 1 –p blastp –mformat 2 –filter seg –E 1 –V 17000000 –B 17000000. From this data set, the number of hits for each of the 20 query light-mediated proteins (Table 1) were counted, discarding hits less than 60 bits, which has been previously estimated to correspond roughly to an E value of <10⁸ (30). The abundances of genome orthologs were counted based on the size of clusters of orthologous groups and nonsupervised orthologous groups previously identified by the STRING database. Because each data set has a different total number of ORFs, protein abundances were normalized with Matlab (The Mathworks, Natick, MA) as follows. For each row (genome or metagenome) of Fig. 3, the absolute number of hits to each query protein was divided by the total number of predicted ORFs in that data set. This percentage is reported in Fig. 3B.

View this table: [in this window] [in a new window]

TABLE 1. Query genes used in metagenome searchesa

View larger version (46K): [in this window] [in a new window]

FIG. 3. Abundances of light-sensing proteins in metagenomes. (A) Total number of proteins orthologous to 20 query proteins (columns) in 373 sequenced genomes (top row) and five metagenomes (remaining rows). (B) Number of proteins as a percentage of the total number of predicted proteins per environment. Rows labeled in gray are subsamples. Columns are labeled as follows (see Table 1 for details): psa, photosystem I subunits ABC; psb, photosystem II subunits ABDEHIJKLF; pet, photosynthetic electron transfer subunits A123; apc, allophycocyanin; cpc, phycocyanin; kaiAB, circadian clock regulators; bluf, blue-light flavin adenine dinucleotide-binding domain-containing proteins; slr0359/plpA, blue-light-absorbing phototropins; cph1 and cph2, red- and far-red-absorbing phytochromes; taxD1, photoreceptor for phototaxis; cry, DNA photolyase and cryptochrome families; carot, water-soluble carotenoids as intracellular UV sunscreen; scyto, scytonemin as extracellular sunscreen; taxP1, phototaxis putative regulatory element; taxY1, phototaxis CheY-like protein; taxAY1, phototaxis histidine kinase; rcaE, complementary chromatic adaptation protein. Growth and repair proteins are more abundant in high-light environments than invariable-light ones, whereas sensing and adaptation proteins are more abundant in variable-light environments than in high-light ones. In particular, photolyase DNA repair proteins are overrepresented in the high-UV environment of surface seawater compared to all other environments. BLUF domain blue-light-sensing proteins are extremely rare in both genomes and environments, although the majority are found in surface rather than deep water. The red-light sensors Cph1 and Cph2 are overrepresented in deep water rather than primarily blue surface water. RcaE chromatic adaptation proteins are overrepresented in variable-light environments, such as the deep sea and lower (darker) layers of the microbial mat.

Construction of alignments and phylogenetic trees. All metagenome hits were then filtered for length (>250 amino acids) and diversity (<80% identity to any other hit in the orthologous family). Amino acid sequences were aligned using MUSCLE (21) with clustal-strict output, and the alignments were manually spot checked. Furthermore, 100 bootstrap replicates of each alignment were generated using seqboot from the Phylip package (23) with default parameters, and these replicates were used to estimate phylogenetic trees with PHYML (27) using eight gamma-estimated rate categories and the Jones-Taylor-Thornton rate matrix. A consensus of the 100 resulting trees was obtained with the consense package of Phylip (extended majority rule), and branch lengths for the consensus were calculated using tree-puzzle (54) with eight gamma-estimated rate categories and the Jones-Taylor-Thornton rate matrix. To check the taxonomic diversity of the metagenome hits and exclude the possibility that novel discoveries were the result of errors in sequence assembly, the phylogenetic placement of each of the metagenome fragments was calculated as follows. Metagenome fragments were used as query proteins and BLASTed against all STRING 7.0 genome proteins with a bit score cutoff of 60 bits. All hits within 5% of the maximum bit score were retained and then mapped to a phylogenetic tree of sequenced genomes (16) as previously described (66). Not all best hits were for genomes; some were at internal nodes of the tree. For cases where metagenome hits originated from closely related species, each gene was mapped to the assembled reads to conclusively exclude the possibility of assembly errors. All phylogenetic trees were visualized with iTOL (40).

Search for neighborhoods and domains. Gene neighborhoods for each of the metagenome hits to the 20 query proteins were calculated as previously described (30), counting genes as neighbors only if they were adjacent on the contig in the same transcription direction. We used cotranscribed gene neighbors (as opposed to bidirectional or convergently transcribed gene neighbors) because their existence was previously established to be most predictive of related function (38). Protein domains for metagenome hits were obtained by searching against the SMART, version 5, database (41) with default parameters.

All analyses were carried out on a dedicated 256-node supercomputing cluster with 1,320 CPU cores communicating via a Gigabit-Ethernet network, each running a 64-bit Linux operating system with 1 G of memory.

Top ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 
To illustrate the application of methods for discovering functional novelty, we focused on light-mediated microbial pathways. Although light is an important abiotic factor impacting all the major biological processes (growth, sensing, maintenance, and reproduction), its utilization by microbes remains poorly understood. We began by constructing a taxonomy of light-mediated processes (Fig. 2) guided by the Gene Ontology biological process ontology (31) with the broad categories growth (photosynthesis and circadian rhythms), sensing (phytochromes), maintenance (DNA repair and pigment synthesis), and adaptation (phototaxis, chromatic adaptation, and bioluminescence). We omitted photosynthetic bacterial pathways, whose evolution and variation were described extensively elsewhere previously (5, 14, 52, 69), from further analyses and instead focused on sensing, repair, and adaptation. Next, we conducted a literature search (1, 3, 14, 15, 20, 22, 45, 46, 58-60, 65) to identify representative bacterial proteins and their orthologs involved in those processes. We searched this candidate list of 20 proteins (Table 1) in five environmental metagenomes comprising 59 sample sites of surface seawater (64, 70), acidic mine runoff from an abandoned gold mine in northern California (63), three sample sites of deep-sea Pacific and Antarctic whalefall carcass (62), 5 g of Minnesota farm topsoil (62), and 10 layers of a 41.5-mm-thick Mexican hypersaline microbial mat (39). For the purpose of analysis, we denoted the surface seawater and top two layers of the microbial mat to be "high-light" environments and the others to be "variable-light" environments. We note here that our choice of proteins is by no means an exhaustive survey of all light-mediated proteins and pathways in bacteria but rather a selected list to illustrate metagenomic data-mining techniques.

View larger version (36K): [in this window] [in a new window]

FIG. 2. Overview of light-mediated processes in biology. Organisms sense visible and UV light that they use for growth, adaptation, and defense/repair. Light sensing is carried out by antenna molecules with a photoactive pigment, such as carotenoids, phycocyanin, phycoerythrin, or rhodopsins. Photosynthetic bacteria can process the light energy through a reaction center and store it as ATP via a proton gradient. Bacteria living in high-light environments must also protect against and repair UV damage. Extracellular and intracellular UV-absorbing compounds such as scytonemin and mycosporine-like amino acids act as a natural sunscreen, while photolyase enzymes reverse point mutations in UV-damaged DNA by using a photon of blue light to catalyze the repair reaction. Finally, bacteria living in variable-light environments can adapt to the changing light conditions in a number of ways, e.g., by moving to a more favorable environment via phototaxis, reconfiguring the wavelength specificity of light-sensing antennae via adaptation proteins, or providing their own light via luminescence.

Quantitative estimate of light-related proteins in the environment. We first counted both the absolute and relative amounts of 20 metagenome proteins (Fig. 3) functioning in light-mediated growth (seven protein families participating in photosynthesis and the circadian clock), sensing (six protein families of blue- and red-light sensors), repair and defense (three protein families consisting of photolyases, water-soluble carotenoids as intracellular UV sunscreens, and scytonemin as an extracellular sunscreen), and adaptation (four protein families participating in phototaxis and complementary chromatic adaptation). As expected, light-mediated growth and repair proteins are found predominantly in high-light environments in both absolute and relative terms (30,435, or 94%, of all growth-, defense-, or repair-related proteins) rather than variable-light environments (2,021 proteins, or 6% of all proteins). Interestingly, sensing and adaptation proteins are overrepresented in variable-light environments (13,786 proteins, or 57% of all sensing proteins, and 17,960 proteins, or 60% of all adaptation proteins) compared to high-light environments (10,273 proteins, or 43% of all sensing proteins, and 11,736 proteins, or 40% of all adaptation proteins). Furthermore, unlike sensing proteins, adaptation proteins are present in large amounts even in deeper (darker) layers of the salt mat. Below, we discuss three examples of novel light-mediated function in each of these categories discovered using analysis of gene neighborhood, protein domains, and protein subfamilies, respectively.

Novel light-mediated sensing. (i) Neighborhood approach. For a candidate sensing process mediated by light, we chose proteins containing the blue-light flavin adenine dinucleotide binding (BLUF) domain (26, 45), as it is rather rare in genomes, and we expected a limited variety in operon organization. BLUF domain proteins are part of the larger family of blue-light photosensors that use flavin chromophores, which together with the phytochromes, rhodopsins, and UV receptors make up the four major classes of bacterial light-sensing proteins (14). Proteins containing a BLUF domain have been shown to function as sensors upstream of phototaxis (24), nucleotide metabolism (35), and repression of anoxygenic photosynthesis (28). The domain is extremely well conserved among the Proteobacteria and Cyanobacteria, absent from the Archaea, and absent from eukaryotes except for the protist Euglena gracilis. As expected, BLUF domain-containing proteins are relatively rare not only in the genomes (34 instances, or 0.002% of all proteins) (Fig. 3) but also in the metagenomes (73 instances, 46 of which are from surface seawater, accounting for 0.0008% of that data set) (Fig. 3).

The vast majority of BLUF-containing proteins in the metagenomes do not contain additional domains, which precludes a domain-based analysis as described above. Furthermore, the BLUF domain is short (98 amino acids) and highly conserved in sequence (70% identity of the multiple sequence alignment) so that constructing phylogenetic trees with robust statistical support is practically impossible. Thus, a tree- or subfamily-based analysis is ruled out as well. However, since BLUF domain proteins are known to function in sensing and the stress response, we surmised that either their expression or the expression of their functional partners would be inducible and thus correlated with the expression of nearby genes on the chromosome. This made it a good candidate for gene neighborhood analysis.

For the 73 environmental BLUF domain proteins, we identified 36 functionally characterizable neighborhoods (32 neighborhoods from surface seawater and 4 from deep-sea whalefall) (Fig. 4). We rediscovered the known functions of BLUF in phototaxis (two neighborhoods), nucleotide metabolism (five neighborhoods), and the repression of anoxygenic photosynthesis (five neighborhoods). Interestingly, we also discovered neighborhoods of BLUF with novel function, including luciferase synthesis (four neighborhoods), nitrate metabolism (three neighborhoods), and quorum sensing (three neighborhoods). These neighbors are promising candidates for the experimental elucidation of BLUF's cellular role.

View larger version (42K): [in this window] [in a new window]

FIG. 4. BLUF operons from genomes and metagenomes. BLUF domain proteins are shown in blue (center), with none containing additional domains. Genome neighbors include genes that function in phototaxis, nucleotide metabolism, repression of anoxygenic photosynthesis, and virulence, primarily from the Alphaproteobacteria (Rhodopseudomonas and Rhodobacter), Betaproteobacteria (Ralstonia and Chromobacterium), and Gammaproteobacteria (Shewanella and Psychrobacter). Novel metagenome neighbors include genes that function in luciferase synthesis, nitrate metabolism, and quorum sensing, primarily from Rhodopseudomonas and Comamonaceae.

(ii) Domain approach. For a candidate adaptation process mediated by light, we chose the RcaE (regulator of chromatic adaptation) protein, which is best characterized in the filamentous cyanobacterium Fremyella diplosiphon (7, 37). This protein regulates the ability to radically alter cell pigmentation in response to changes in ambient-light wavelength, particularly across the green-red range, and has been shown to optimize light antennae for photosynthesis (65). In the sequenced genomes, RcaE is a relatively rare protein (212 homologs, primarily in cyanobacteria and plants, accounting for 0.01% of total proteins). In the metagenomes, however, RcaE orthologs are overrepresented in variable-light environments (723 proteins, or 0.53%, in the microbial mat and 1,155 proteins, or 0.97%, in deep-sea whalefall) (Fig. 3) and underrepresented in high-light environments (5,434 proteins, or 0.08%, in surface seawater) (Fig. 3). Because the known cyanobacterial homologs of this protein have an unusual domain composition (GAF-PAS-PAC-HisKA-HATPase-REC) that has been modified among plants and nonphotosynthetic bacteria (Fig. 5), we hypothesized that additional modular arrangements of this protein must exist in the wild. Furthermore, we expected that these novel domain arrangements, especially the associated receiver/output domains, would provide clues as to the downstream cellular function being adapted.

View larger version (82K): [in this window] [in a new window]

FIG. 5. RcaE domain variations in sequenced bacteria and plants. Whereas the majority of bacterial proteins have the conserved domain architecture of GAF-PAS-PAC-HisKA-HATPase-REC, many additional architectures with different signal transduction domains, multiple sensing domains, and multiple receiver domains exist.

Finding "true" domain hits within the metagenomes, however, proved to be less than straightforward. Several of RcaE's domains, such as the kinase (HisKA-HAPTase), redox-sensing (PAS-PAC) (61), and response regulator receiver (REC) (49) domains, are extremely widespread and promiscuous, and the metagenome data sets typically contain fragments of genes without the key light-sensing GAF domain, together leading to many spurious hits. Of the 11,456 environmental sequences initially obtained at a >60-bit BLAST score (corresponding roughly to a stringent E value of <10^–8) (30), only 762 sequences were longer than 250 amino acids and less than 80% identical to one another. Of these sequences, 112 contained the GAF domain and aligned to the query at greater than 60% of their length, a typical cutoff for excluding single-domain hits (34). However, since this cutoff entirely eliminated proteins from the hypersaline microbial mat, we relaxed it to 40% alignment length, which yielded 650 environmental sequences (632 surface seawater, 18 deep-sea whalefall, 26 soil, 11 salt mat, and 2 acid mine sequences).

This sample of 650 environmental sequences contained 50 unique domain arrangements, 16 of which were novel and not seen before in any genome. All 16 novel arrangements preserve the pattern of "specific sensing domain(s)-PAS-PAC-kinase-receiver" but vary in the numbers and types of domain repeats. Two arrangements include repeats in PBPb (periplasmic solute binding) as one of the sensing domains and another has PBPp with PAS repeats without a PAC domain. Another seven arrangements include three to six repeats of PAS-PAC; three arrangements have duplicated REC domains. This is a surprising result because domain repeats are generally less common in bacteria than in eukaryotes, where they are thought to encode increased variability to compensate for longer eukaryotic generation times (4). However, the conservation of the overall domain pattern of the protein, together with the remarkable number of PAS-PAC repeats, allows us to speculate that this domain architecture provides increased substrate affinity and a tuning switch for the sensitivity of the response.

(iii) Subfamily approach. For a candidate repair process mediated by light, we focused on photolyases, an intriguing family of light-activated DNA repair enzymes that are virtually ubiquitous in bacterial species. Photolyases reverse T<>T cyclobutane dipyrimidine dimers (CPDs) formed by UV damage to DNA using a photon of light to transfer electrons from a catalytic flavin chromophore to the damaged DNA (53). While the structure and function of photolyases were being characterized, an additional family of homologs, the cryptochromes, were discovered (12, 42, 53, 56). Cryptochromes are similar to photolyases in sequence and three-dimensional structure but lack catalytic activity for DNA repair and have unclear function. To date, two kinds of photolyases (CPD-I and CPD-II) and three kinds of cryptochromes have been identified (plant cryptochromes, animal cryptochromes, and CRY-DASH proteins, which are named after the representative four genera in which they were identified, Drosophila, Arabidopsis, Synechocystis, and Homo). Thus, the photolyase-cryptochrome family in sequenced genomes is quite large, spanning the inclusive gene family COG0415 (328 proteins in 209 species) but also including COG3046 (56 genes in 53 species), COG4338 (35 genes in 33 species), and NOG16378 (22 proteins in 19 species). In the metagenomes, the photolyase-cryptochrome homologs are overrepresented in surface seawater (9,703 proteins, or 0.1% of the total) and the top two layers of the microbial mat (8 proteins, or 0.6% of the total) compared to all other environments together (84 proteins). This is consistent with the large amount of UV radiation incident on surface waters of the open ocean or the top layers of the microbial mat but does not account for the other possible functions of cryptochromes in remaining niches.

To tease apart the functional diversity of this protein family, we undertook a subfamily analysis by constructing high-quality alignments, feeding them into a hidden Markov model, and using the resulting hidden Markov model profile to refine the alignment and construct a phylogenetic tree. Although this approach is now standard practice when small gene families are analyzed, it foundered when fed with roughly 10,000 sequences, and our phylogenetic tools of choice (phyml [27 and tree-puzzle [54]) often took weeks to estimate a tree when running on dedicated supercomputing clusters, even without statistical bootstraps. We therefore added several filtering steps to our protocol. First, we removed sequences shorter than 250 amino acids as well as sequences that were >80% identical to any other sequence in the data set. This approximately halved the number of photolyase hits from 9,703 to 4,828. Next, we constructed phylogenetic trees by randomly subsampling the 4,828 sequences in batches of 1,000 sequences and compared the resulting trees for topology and grouping. Finally, we combined the genomic and metagenomic sequences, constructed trees again, and checked whether the same groupings resulted.

Because bootstraps on the photolyase trees could not be calculated for the entire data set of approximately 10,000 sequences, we report here 1,196 photolyase-cryptochrome orthologs from the sequenced genomes and four metagenomes: surface seawater from Sargasso sea samples 1 to 4, farm soil, acidic mine runoff, and deep-sea whalefall (Fig. 6). Although the bootstrap at the deeper branches is somewhat low (<25%), it is consistently high near the leaves (>80%), indicating that the relationships between the subfamilies are poorly resolved but that the clustering within subfamilies is strong. Most notably, our tree recovers the four known groups of photolyases (CPD-I, CPD-II, DASH cryptochromes, and animal cryptochromes) and additionally identifies two novel deep-branching groups of photolyases/cryptochromes. The deepest-branching "novel family I" represents a new family of 34 photolyases/cryptochromes of unknown function never seen before in the genomes. Because the tree covers photolyases from all known species from all three domains of life, the novel family must include newly detected enzymes of unknown function related to the cryptochrome superfamily. The taxonomic origins of these enzymes are a mixture of Pelagibacter/SAR11-like species and other Alphaproteobacteria (69%) and Cyanobacteria dominated by Prochlorococcus (31%). "Novel family II," which is clearly grouped between CPD-II photolyases and animal DASH cryptochromes, is an additional uncharacterized diverse subfamily with 54 sequences from Alphaproteobacteria (80%) and Cyanobacteria (20%). The species compositions are as expected, since both Alphaproteobacteria and Cyanobacteria are the dominant marine microbial species. Both newly discovered photolyase/cryptochrome families are exciting candidates for further computational and experimental characterization.

View larger version (65K): [in this window] [in a new window]

FIG. 6. Photolyase/cryptochrome subfamilies representing 1,196 sequences from sequenced genomes and four metagenomes. The tree recovers the known groupings of CPD-I and CPD-II photolyases and animal and DASH cryptochromes (plant cryptochromes, the fifth known group, are not shown here). In addition, we discovered two novel subfamilies of photolyases with 38 (family I) and 54 members (family II) that appear to originate from diverse members of the Alphaproteobacteria and Cyanobacteria.

Top ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 
We have analyzed the distribution and molecular diversity of light-mediated proteins from five diverse environments receiving varying ambient light. Instead of assigning genes to functions as is usually done with current metagenomics data sets using BLAST-like procedures, we sought to identify novel protein functions. Using gene neighborhood, domain, and subfamily analyses, we have attempted to characterize functional novelty in proteins sensing light, adapting to changes in light color, and repairing UV-damaged DNA. We found new functional partners for blue-light sensors, new domain architectures of chromatic adaptation proteins, and new subfamilies of DNA repair enzymes. Our results represent the first quantification of these cellular processes and provide an early insight into their spectacular diversity.

While these results serve as a proof of principle for the possibility to infer novel functionality by using the three different concepts described above and represent the opportunities inherent in those huge data sets, they also implicitly illustrate the challenges of mining environmental sequence data to discover novel functions. The difficulty is due to the nature of the data itself (vast amount, fragmented, uniform coverage, and shotgun sequence); the lack of appropriate methods and analysis tools together with bottlenecks in CPU, memory, and network bandwidth; and ongoing conceptual difficulties with defining homology/paralogy and novel function. Indeed, while the sequencing of environment after environment continues to generate gigabytes of data, there has been little corresponding investment in the analysis of these data, pointing to an urgent and immediate need for methods and tool development. For example, we would have been unable to derive bootstrap values for some of the phylogenetic trees had we included more environments, not to mention the enormous challenges for the CPU to compute all the data. Our previous work demonstrated that even a slightly better function assignment protocol could lead to a near doubling of the number of functional annotations for gene fragments, from 40% to 70% (30), suggesting that with improved analysis, perhaps only half the sequence data are really needed. The saved effort could be redirected at gathering time series and spatial data, which would help to interpret functional novelty and allow the development of dynamic models to explore larger concepts in ecology and evolution such as species succession, pathway evolution, or metabolic flux.

In summary, we have demonstrated the use of computational analysis techniques for discovering molecular functional novelty in environmental snapshots of bacterial communities. Our results indicate that information on gene neighborhood, protein domains, and subfamilies can be successfully used to discover functional novelty, although various challenges hamper the analysis considerably and will continue to do so as more data are generated in the future.

 
We thank Chris Creevey and Jean Muller for suggesting alternate phylogenetic analysis methods when the existing tools crashed, Mani Arumugam for helping with domain analysis, and Yan Yuan for excellent technical assistance. We thank members of the Bork group for useful discussions and feedback.

 
* Corresponding author. Mailing address: Structural and Computational Biology Unit, European Molecular Biology Laboratory, Meyerhofstrasse 1, 69117 Heidelberg, Germany. Phone: 49-6221-387-8526. Fax: 49-6221-387517. E-mail: bork{at}embl.de

Published ahead of print on 10 October 2008.

Top ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERENCES

 

1
1. Ashby, M. K., and J. Houmard. 2006. Cyanobacterial two-component proteins: structure, diversity, distribution, and evolution. Microbiol. Mol. Biol. Rev. 70:472-509.[Abstract/Free Full Text]
2
2. Beja, O., L. Aravind, E. V. Koonin, M. T. Suzuki, A. Hadd, L. P. Nguyen, S. B. Jovanovich, C. M. Gates, R. A. Feldman, J. L. Spudich, E. N. Spudich, and E. F. DeLong. 2000. Bacterial rhodopsin: evidence for a new type of phototrophy in the sea. Science 289:1902-1906.[Abstract/Free Full Text]
3
3. Bhaya, D. 2004. Light matters: phototaxis and signal transduction in unicellular cyanobacteria. Mol. Microbiol. 53:745-754.[CrossRef][Medline]
4
4. Bjorklund, A. K., D. Ekman, and A. Elofsson. 2006. Expansion of protein domain repeats. PLoS Comput. Biol. 2:e114.[CrossRef][Medline]
5
5. Blankenship, R. E. 1992. Origin and early evolution of photosynthesis. Photosynth. Res. 33:91-111.[CrossRef][Medline]
6
6. Boone, D. R., R. W. Castenholz, and G. M. Garrity. 2001. Bergey's manual of systematic bacteriology, 2nd ed. Springer, New York, NY.
7
7. Bordowitz, J. R., and B. L. Montgomery. 2008. Photoregulation of cellular morphology during complementary chromatic adaptation requires sensor-kinase-class protein RcaE in Fremyella diplosiphon. J. Bacteriol. 190:4069-4074.[Abstract/Free Full Text]
8
8. Bork, P., T. Dandekar, Y. Diaz-Lazcoz, F. Eisenhaber, M. Huynen, and Y. Yuan. 1998. Predicting function: from genes to genomes and back. J. Mol. Biol. 283:707-725.[CrossRef][Medline]
9
9. Bork, P., and L. Serrano. 2005. Towards cellular systems in 4D. Cell 121:507-509.[CrossRef][Medline]
10
10. Boucher, Y., C. J. Douady, R. T. Papke, D. A. Walsh, M. E. Boudreau, C. L. Nesbo, R. J. Case, and W. F. Doolittle. 2003. Lateral gene transfer and the origins of prokaryotic groups. Annu. Rev. Genet. 37:283-328.[CrossRef][Medline]
11
11. Brown, D. P., N. Krishnamurthy, and K. Sjolander. 2007. Automated protein subfamily identification and classification. PLoS Comput. Biol. 3:e160.[CrossRef][Medline]
12
12. Brudler, R., K. Hitomi, H. Daiyasu, H. Toh, K. Kucho, M. Ishiura, M. Kanehisa, V. A. Roberts, T. Todo, J. A. Tainer, and E. D. Getzoff. 2003. Identification of a new cryptochrome class. Structure, function, and evolution. Mol. Cell 11:59-67.[CrossRef][Medline]
13
13. Bryant, D. A., A. M. Costas, J. A. Maresca, A. G. Chew, C. G. Klatt, M. M. Bateson, L. J. Tallon, J. Hostetler, W. C. Nelson, J. F. Heidelberg, and D. M. Ward. 2007. Candidatus Chloracidobacterium thermophilum: an aerobic phototrophic acidobacterium. Science 317:523-526.[Abstract/Free Full Text]
14
14. Bryant, D. A., and N. U. Frigaard. 2006. Prokaryotic photosynthesis and phototrophy illuminated. Trends Microbiol. 14:488-496.[Medline]
15
15. Buchanan, B. B., and Y. Balmer. 2005. Redox regulation: a broadening horizon. Annu. Rev. Plant Biol. 56:187-220.[CrossRef][Medline]
16
16. Ciccarelli, F. D., T. Doerks, C. von Mering, C. J. Creevey, B. Snel, and P. Bork. 2006. Toward automatic reconstruction of a highly resolved tree of life. Science 311:1283-1287.[Abstract/Free Full Text]
17
17. Ciccarelli, F. D., C. von Mering, M. Suyama, E. D. Harrington, E. Izaurralde, and P. Bork. 2005. Complex genomic rearrangements lead to novel primate gene function. Genome Res. 15:343-351.[Abstract/Free Full Text]
18
18. Copley, R. R., T. Doerks, I. Letunic, and P. Bork. 2002. Protein domain analysis in the era of complete genomes. FEBS Lett. 513:129-134.[CrossRef][Medline]
19
19. Dandekar, T., B. Snel, M. Huynen, and P. Bork. 1998. Conservation of gene order: a fingerprint of proteins that physically interact. Trends Biochem. Sci. 23:324-328.[CrossRef][Medline]
20
20. Dvornyk, V., O. Vinogradova, and E. Nevo. 2003. Origin and evolution of circadian clock genes in prokaryotes. Proc. Natl. Acad. Sci. USA 100:2495-2500.[Abstract/Free Full Text]
21
21. Edgar, R. C. 2004. MUSCLE: multiple sequence alignment with high accuracy and high throughput. Nucleic Acids Res. 32:1792-1797.[Abstract/Free Full Text]
22
22. Ehling-Schulz, M., W. Bilger, and S. Scherer. 1997. UV-B-induced synthesis of photoprotective pigments and extracellular polysaccharides in the terrestrial cyanobacterium Nostoc commune. J. Bacteriol. 179:1940-1945.[Abstract/Free Full Text]
23
23. Felsenstein, J. 1993. PHYLIP—Phylogeny Inference Package version 3.2. Cladistics 5:164-166.
24
24. Fiedler, B., T. Borner, and A. Wilde. 2005. Phototaxis in the cyanobacterium Synechocystis sp. PCC 6803: role of different photoreceptors. Photochem. Photobiol. 81:1481-1488.[CrossRef][Medline]
25
25. Fraser, C., W. P. Hanage, and B. G. Spratt. 2007. Recombination and the nature of bacterial speciation. Science 315:476-480.[Abstract/Free Full Text]
26
26. Gomelsky, M., and G. Klug. 2002. BLUF: a novel FAD-binding domain involved in sensory transduction in microorganisms. Trends Biochem. Sci. 27:497-500.[CrossRef][Medline]
27
27. Guindon, S., and O. Gascuel. 2003. A simple, fast, and accurate algorithm to estimate large phylogenies by maximum likelihood. Syst. Biol. 52:696-704.[Abstract/Free Full Text]
28
28. Han, Y., S. Braatsch, L. Osterloh, and G. Klug. 2004. A eukaryotic BLUF domain mediates light-dependent gene expression in the purple bacterium Rhodobacter sphaeroides 2.4.1. Proc. Natl. Acad. Sci. USA 101:12306-12311.[Abstract/Free Full Text]
29
29. Hannenhalli, S. S., and R. B. Russell. 2000. Analysis and prediction of functional sub-types from protein sequence alignments. J. Mol. Biol. 303:61-76.[CrossRef][Medline]
30
30. Harrington, E. D., A. H. Singh, T. Doerks, I. Letunic, C. von Mering, L. J. Jensen, J. Raes, and P. Bork. 2007. Quantitative assessment of protein function prediction from metagenomics shotgun sequences. Proc. Natl. Acad. Sci. USA 104:13913-13918.[Abstract/Free Full Text]
31
31. Harris, M. A., J. Clark, A. Ireland, J. Lomax, M. Ashburner, R. Foulger, K. Eilbeck, S. Lewis, B. Marshall, C. Mungall, J. Richter, G. M. Rubin, J. A. Blake, C. Bult, M. Dolan, H. Drabkin, J. T. Eppig, D. P. Hill, L. Ni, M. Ringwald, R. Balakrishnan, J. M. Cherry, K. R. Christie, M. C. Costanzo, S. S. Dwight, S. Engel, D. G. Fisk, J. E. Hirschman, E. L. Hong, R. S. Nash, A. Sethuraman, C. L. Theesfeld, D. Botstein, K. Dolinski, B. Feierbach, T. Berardini, S. Mundodi, S. Y. Rhee, R. Apweiler, D. Barrell, E. Camon, E. Dimmer, V. Lee, R. Chisholm, P. Gaudet, W. Kibbe, R. Kishore, E. M. Schwarz, P. Sternberg, M. Gwinn, L. Hannick, J. Wortman, M. Berriman, V. Wood, N. de la Cruz, P. Tonellato, P. Jaiswal, T. Seigfried, R. White, et al. 2004. The Gene Ontology (GO) database and informatics resource. Nucleic Acids Res. 32:D258.[Abstract/Free Full Text]
32
32. Hirschman, L., C. Clark, K. B. Cohen, S. Mardis, J. Luciano, R. Kottmann, J. Cole, V. Markowitz, N. Kyrpides, and D. Field. 2008. Habitat-Lite: a GSC case study based on free text terms for environmental metadata. Omics 12:129-136.[CrossRef][Medline]
33
33. Huson, D. H., A. F. Auch, J. Qi, and S. C. Schuster. 2007. MEGAN analysis of metagenomic data. Genome Res. 17:377-386.[Abstract/Free Full Text]
34
34. Huynen, M., T. Dandekar, and P. Bork. 1998. Differential genome analysis applied to the species-specific features of Helicobacter pylori. FEBS Lett. 426:1-5.[CrossRef][Medline]
35
35. Jung, A., T. Domratcheva, M. Tarutina, Q. Wu, W. H. Ko, R. L. Shoeman, M. Gomelsky, K. H. Gardner, and I. Schlichting. 2005. Structure of a bacterial BLUF photoreceptor: insights into blue light-mediated signal transduction. Proc. Natl. Acad. Sci. USA 102:12350-12355.[Abstract/Free Full Text]
36
36. Kannan, N., S. S. Taylor, Y. Zhai, J. C. Venter, and G. Manning. 2007. Structural and functional diversity of the microbial kinome. PLoS Biol. 5:e17.[CrossRef][Medline]
37
37. Kehoe, D. M., and A. R. Grossman. 1996. Similarity of a chromatic adaptation sensor to phytochrome and ethylene receptors. Science 273:1409-1412.[Abstract]
38
38. Korbel, J. O., L. J. Jensen, C. von Mering, and P. Bork. 2004. Analysis of genomic context: prediction of functional associations from conserved bidirectionally transcribed gene pairs. Nat. Biotechnol. 22:911.[CrossRef][Medline]
39
39. Kunin, V., J. Raes, J. K. Harris, J. R. Spear, J. J. Walker, N. Ivanova, C. von Mering, B. M. Bebout, N. R. Pace, P. Bork, and P. Hugenholtz. 2008. Millimeter-scale genetic gradients and community-level molecular convergence in a hypersaline microbial mat. Mol. Syst. Biol. 4:198.[Medline]
40
40. Letunic, I., and P. Bork. 2007. Interactive Tree Of Life (iTOL): an online tool for phylogenetic tree display and annotation. Bioinformatics 23:127-128.[Abstract/Free Full Text]
41
41. Letunic, I., R. R. Copley, B. Pils, S. Pinkert, J. Schultz, and P. Bork. 2006. SMART 5: domains in the context of genomes and networks. Nucleic Acids Res. 34:D257-D260.[Abstract/Free Full Text]
42
42. Lin, C., and T. Todo. 2005. The cryptochromes. Genome Biol. 6:220.[CrossRef][Medline]
43
43. Liolios, K., K. Mavromatis, N. Tavernarakis, and N. C. Kyrpides. 2008. The Genomes On Line Database (GOLD) in 2007: status of genomic and metagenomic projects and their associated metadata. Nucleic Acids Res. 36:D475-D479.[Abstract/Free Full Text]
44
44. Lozupone, C. A., and R. Knight. 2007. Global patterns in bacterial diversity. Proc. Natl. Acad. Sci. USA 104:11436-11440.[Abstract/Free Full Text]
45
45. Montgomery, B. L. 2007. Sensing the light: photoreceptive systems and signal transduction in cyanobacteria. Mol. Microbiol. 64:16-27.[CrossRef][Medline]
46
46. Moran, M. A., and W. L. Miller. 2007. Resourceful heterotrophs make the most of light in the coastal ocean. Nat. Rev. Microbiol. 5:792-800.[CrossRef][Medline]
47
47. Oren, A. 2004. Prokaryote diversity and taxonomy: current status and future challenges. Philos. Trans. R. Soc. Lond. B Biol. Sci. 359:623-638.[Abstract/Free Full Text]
48
48. Overbeek, R., M. Fonstein, M. D'Souza, G. D. Pusch, and N. Maltsev. 1999. The use of gene clusters to infer functional coupling. Proc. Natl. Acad. Sci. USA 96:2896-2901.[Abstract/Free Full Text]
49
49. Pao, G. M., and M. H. Saier, Jr. 1995. Response regulators of bacterial signal transduction systems: selective domain shuffling during evolution. J. Mol. Evol. 40:136-154.[CrossRef][Medline]
50
50. Pignatelli, M., G. Aparicio, I. Blanquer, V. Hernandez, A. Moya, and J. Tamames. 2008. Metagenomics reveals our incomplete knowledge of global diversity. Bioinformatics 24:2124-2125.[Abstract/Free Full Text]
51
51. Raes, J., E. D. Harrington, A. H. Singh, and P. Bork. 2007. Protein function space: viewing the limits or limited by our view? Curr. Opin. Struct. Biol. 17:362-369.[CrossRef][Medline]
52
52. Raymond, J., O. Zhaxybayeva, J. P. Gogarten, S. Y. Gerdes, and R. E. Blankenship. 2002. Whole-genome analysis of photosynthetic prokaryotes. Science 298:1616-1620.[Abstract/Free Full Text]
53
53. Sancar, A. 2004. Photolyase and cryptochrome blue-light photoreceptors. Adv. Protein Chem. 69:73-100.[CrossRef][Medline]
54
54. Schmidt, H. A., K. Strimmer, M. Vingron, and A. von Haeseler. 2002. TREE-PUZZLE: maximum likelihood phylogenetic analysis using quartets and parallel computing. Bioinformatics 18:502-504.[Abstract/Free Full Text]
55
55. Schultz, J., F. Milpetz, P. Bork, and C. P. Ponting. 1998. SMART, a simple modular architecture research tool: identification of signaling domains. Proc. Natl. Acad. Sci. USA 95:5857-5864.[Abstract/Free Full Text]
56
56. Selby, C. P., and A. Sancar. 2006. A cryptochrome/photolyase class of enzymes with single-stranded DNA-specific photolyase activity. Proc. Natl. Acad. Sci. USA 103:17696-17700.[Abstract/Free Full Text]
57
57. Singh, A. H., D. M. Wolf, P. Wang, and A. P. Arkin. 2008. Modularity of stress response evolution. Proc. Natl. Acad. Sci. USA 107:7500-7505.
58
58. Singh, S. P., M. Klisch, R. P. Sinha, and D. P. Häder. 13 June 2008, posting date. Effects of abiotic stressors on synthesis of the mycosporine-like amino acid shinorine in the cyanobacterium Anabaena variabilis PCC 7937. Photochem. Photobiol. [Epub ahead of print.]
59
59. Sinha, R. P., N. K. Ambasht, J. P. Sinha, M. Klisch, and D. P. Hader. 2003. UV-B-induced synthesis of mycosporine-like amino acids in three strains of Nodularia (cyanobacteria). J. Photochem. Photobiol. B 71:51-58.[CrossRef][Medline]
60
60. Soule, T., V. Stout, W. D. Swingley, J. C. Meeks, and F. Garcia-Pichel. 2007. Molecular genetics and genomic analysis of scytonemin biosynthesis in Nostoc punctiforme ATCC 29133. J. Bacteriol. 189:4465-4472.[Abstract/Free Full Text]
61
61. Taylor, B. L., and I. B. Zhulin. 1999. PAS domains: internal sensors of oxygen, redox potential, and light. Microbiol. Mol. Biol. Rev. 63:479-506.[Abstract/Free Full Text]
62
62. Tringe, S. G., C. von Mering, A. Kobayashi, A. A. Salamov, K. Chen, H. W. Chang, M. Podar, J. M. Short, E. J. Mathur, J. C. Detter, P. Bork, P. Hugenholtz, and E. M. Rubin. 2005. Comparative metagenomics of microbial communities. Science 308:554-557.[Abstract/Free Full Text]
63
63. Tyson, G. W., J. Chapman, P. Hugenholtz, E. E. Allen, R. J. Ram, P. M. Richardson, V. V. Solovyev, E. M. Rubin, D. S. Rokhsar, and J. F. Banfield. 2004. Community structure and metabolism through reconstruction of microbial genomes from the environment. Nature 428:37-43.[CrossRef][Medline]
64
64. Venter, J. C., K. Remington, J. F. Heidelberg, A. L. Halpern, D. Rusch, J. A. Eisen, D. Wu, I. Paulsen, K. E. Nelson, W. Nelson, D. E. Fouts, S. Levy, A. H. Knap, M. W. Lomas, K. Nealson, O. White, J. Peterson, J. Hoffman, R. Parsons, H. Baden-Tillson, C. Pfannkoch, Y. H. Rogers, and H. O. Smith. 2004. Environmental genome shotgun sequencing of the Sargasso Sea. Science 304:66-74.[Abstract/Free Full Text]
65
65. Vierstra, R. D., and S. J. Davis. 2000. Bacteriophytochromes: new tools for understanding phytochrome signal transduction. Semin. Cell Dev. Biol. 11:511-521.[CrossRef][Medline]
66
66. von Mering, C., P. Hugenholtz, J. Raes, S. G. Tringe, T. Doerks, L. J. Jensen, N. Ward, and P. Bork. 2007. Quantitative phylogenetic assessment of microbial communities in diverse environments. Science 315:1126-1130.[Abstract/Free Full Text]
67
67. von Mering, C., L. J. Jensen, M. Kuhn, S. Chaffron, T. Doerks, B. Kruger, B. Snel, and P. Bork. 2007. STRING 7—recent developments in the integration and prediction of protein interactions. Nucleic Acids Res. 35:D358-D362.[Abstract/Free Full Text]
68
68. Wilson, E. O. 1999. The diversity of life. W. W. Norton, New York, NY.
69
69. Xiong, J., K. Inoue, and C. E. Bauer. 1998. Tracking molecular evolution of photosynthesis by characterization of a major photosynthesis gene cluster from Heliobacillus mobilis. Proc. Natl. Acad. Sci. USA 95:14851-14856.[Abstract/Free Full Text]
70
70. Yooseph, S., G. Sutton, D. B. Rusch, A. L. Halpern, S. J. Williamson, K. Remington, J. A. Eisen, K. B. Heidelberg, G. Manning, W. Li, L. Jaroszewski, P. Cieplak, C. S. Miller, H. Li, S. T. Mashiyama, M. P. Joachimiak, C. van Belle, J. M. Chandonia, D. A. Soergel, Y. Zhai, K. Natarajan, S. Lee, B. J. Raphael, V. Bafna, R. Friedman, S. E. Brenner, A. Godzik, D. Eisenberg, J. E. Dixon, S. S. Taylor, R. L. Strausberg, M. Frazier, and J. C. Venter. 2007. The Sorcerer II Global Ocean Sampling expedition: expanding the universe of protein families. PLoS Biol. 5:e16.[CrossRef][Medline]

---

Journal of Bacteriology, January 2009, p. 32-41, Vol. 191, No. 1
0021-9193/09/$08.00+0     doi:10.1128/JB.01084-08
Copyright © 2009, American Society for Microbiology. All Rights Reserved.

This article has been cited by other articles:

• Zhulin, I. B. (2009). It Is Computation Time for Bacteriology!. J. Bacteriol. 191: 20-22 [Full Text]  

This Article Abstract Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Reprints and Permissions Copyright Information Books from ASM Press MicrobeWorld Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Singh, A. H. Articles by Bork, P. Search for Related Content PubMed PubMed Citation Articles by Singh, A. H. Articles by Bork, P.