|Home | About | Journals | Submit | Contact Us | Français|
The enzymes of bacteriophages and other viruses have been essential research tools since the first days of molecular biology. However, the current repertoire of viral enzymes only hints at their overall potential. The most commonly used enzymes are derived from a surprisingly small number of cultivated viruses, which is remarkable considering the extreme abundance and diversity of viruses revealed over the past decade by metagenomic analysis. To access the treasure trove of enzymes hidden in the global virosphere and develop them for research, therapeutic and diagnostic uses, improvements are needed in our ability to rapidly and efficiently discover, express and characterize viral genes to produce useful proteins. We discuss improvements to sampling and cloning methods, functional and genomics-based screens and expression systems that should accelerate discovery of new enzymes and other viral proteins for use in research and medicine.
Laboratory research in molecular biology has depended on a long list of enzymes for almost every manipulation used in amplification, detection, cloning, expression, mutagenesis and analysis of nucleic acids. While many cellular enzymes have been used in these methods, viruses, including bacteriophages, have been an especially rich source of useful enzymes. Phages — particularly T4, T7, lambda, M13 and phiX174 — were the first model systems of molecular biology , and numerous methods that use phage enzymes were developed based on this research. T4 phage is still the most prolific source of useful viral enzymes. Its 169 kb genome and estimated 300 genes include at least 85 genes involved in DNA replication, recombination and repair, nucleotide metabolism, transcription and translation . Enzymes from this and other phages have been instrumental in the development of the field of biotechnology (Table 1). One reason is the density of certain genes in viral genomes. For example, a typical bacterial genome of about 2 Mb contains only a single polI gene (coding for DNA polymerase I). In contrast, between 20 and 40 pol genes per 2 Mb might be found in viral metagenomic sequences (Table 2). Viral genomes are relatively simple compared to those of their hosts, and contain a comparatively high proportion of genes coding for structural proteins (e.g. for coat and tail) together with proteins involved in nucleic acid metabolism and lysis. Many of these genes have shown utility.
In contrast to the extreme diversity and abundance of viruses in the environment , the most common viral enzymes used today are derived primarily from a small number of phages, e.g. T4, T7, lambda, SP6 and phi29, and retroviruses Moloney murine leukemia virus (M-MLV) and avian myeloblastosis virus (AMV). Aquatic and soil environments consistently contain 107 to 108 viral particles per milliliter of water or gram of soil, respectively, and there are thousands of viral types in each sample [4-7]. However, efforts to access this diversity by systematic cultivation of new viruses are hampered by technical challenges, and few new viral enzymes have been introduced during the past few decades. Direct isolation and sequence analysis of uncultured viral assemblages, i.e. metagenomics, has provided insights into the composition and structure of environmental viral communities.
In this article, we briefly review the scientific, diagnostic, and therapeutic uses of viral enzymes. Other significant applications of viruses, including their use as cloning vectors [8-10], as phage display tools [11-13] and in phage therapy , have been reviewed elsewhere. We also discuss the use of viral metagenomics as a tool for the discovery of new enzymes.
Many research applications of viral enzymes are centered on nucleic acid metabolism (Table 1). DNA polymerases (Pols) have been the focus of most of our discovery efforts. These enzymes are essential for common molecular biology techniques including whole genome amplification, PCR, Sanger (dideoxy chain termination) DNA sequencing, and most methods for nucleic-acid-based detection of infectious agents, cancer and genetic variation. Most of the next generation sequencing platforms (e.g. Roche/454, Illumina, Helicos, Pacific BioSystems) [15, 16] use multiple microbial and viral DNA Pols both for template preparation and base discrimination.
Viral DNA Pols are functionally distinct from their cellular counterparts. All of the microbial DNA Pols used as reagents are derived from two families: bacterial Pol I and archaeal Pol II, which are highly similar cellular repair enzymes and not true replicases. In contrast, viral Pols are highly diverse in terms of primary amino acid sequence  and biochemical activities. As replicase enzymes, they have distinct properties. For instance, phi29 Pol has a processivity of >70,000 nucleotides  (i.e., it incorporates over 70,000 nucleotides before dissociating), far greater than that of Thermus aquaticus (Taq) Pol, with only 50-80 nucleotides . Additionally, phi29Pol has a strong strand displacement capability that together make it the polymerase of choice for whole genome amplification by multiple displacement amplification (MDA) . T7 phage Pol holoenzyme has a processivity of >10,000 nucleotides  and efficiently incorporates chain terminating nucleotide analogs; these attributes made it an optimal choice for Sanger sequencing until it was displaced by Thermosequenase, a Taq Pol derivative that was engineered based on sequence features in T7 DNA Pol that conferred efficient incorporation of dideoxynucleotides . T5 Pol has both high processivity and a potent strand displacement activity that are independent of additional host or viral proteins . The DNA Pols of T4-family phages have high proofreading activities that are commonly exploited for generating blunt ends, especially in physically sheared DNA . The replicases (i.e., reverse transcriptases) of retroviruses, especially M-MLV and AMV, are used for reverse transcription of RNA to form cDNA; they are indispensable for research on transcription processes and RNA viruses and for transcriptome analysis [25, 26]. Many uses for DNA Pols including PCR, RT PCR, thermocycled Sanger sequencing and certain whole genome amplification methods depend on thermostability up to 95°C. Before metagenomic screens were used to discover the so-called PyroPhage Pols (Box 1), no known viral Pol could withstand this temperature.
Other viral enzymes have distinct and useful properties. RNA polymerases (RNAPs) transcribe RNA from a DNA template. In contrast to their cellular counterparts, the RNAPs of phages T7, T3 and SP6 function as independent proteins that recognize short promoter elements (of 17 nucleotides in length) without the requirement of transcription factors or other cellular components . They can generate large amounts of RNA for direct use or for in vitro or in vivo protein synthesis. RNAPs are also key components of a number of transcription-mediated amplification approaches [28, 29]. Virtually all ligation methods used for cloning and linker attachment depend on T4 DNA ligase due to its relatively high activity on 5’- and 3’- extended and blunt DNA. The integrases and recombinases of various phages have been used to integrate genes into the genomes of a wide variety of bacterial and eukaryotic cells. The lambda-red-mediated system is used for stable transgene integration into the Escherichia coli genome , whereas the phage P1 cre/lox system functions in mammalian, fungal, plant and other eukaryotic cells. Resolvases (e.g. T4 endonuclease VII and T7 endonuclease I) have been used to detect single nucleotide polymorphisms (SNPs) .
Genes encoding useful enzymes have been discovered recently in viral genomes outside of the usual core group of phages and retroviruses. The RNA ligase from a Thermus scotoductus phage, for example, is ten times more efficient at joining single-strand DNA molecules than is the T4 RNA ligase . The thermophilic phage GBSV1 encodes a non-specific nuclease useful for degrading RNA and single- and double-stranded, circular or linear DNA . Phi6 replicase  is used for replicating RNA, particularly to generate small interfering RNA (siRNA) and micro RNA (miRNA) for RNA interference studies . However, the rate of discovery of new viral enzymes through traditional cultivation-based means is relatively slow.
In addition to research uses, several viral proteins might be useful in the clinic either as therapeutics or as diagnostic reagents.
Sensitivity to phage infection is often the only means of distinguishing between closely related bacterial strains. For instance, pathogenic Bacillus strains can be distinguished using a strain-typing phage, a time-consuming process . More direct and rapid tests should be possible based on direct detection of binding of differentially labeled phage tail proteins.
Before phage proteins were developed as laboratory reagents, an important rationale for studying phages was their potential therapeutic use in fighting bacterial infectious diseases . In North America and Western Europe, interest in phage therapy declined with the discovery and development of antibiotics; however, the emergence of multi-drug resistant bacterial pathogens is reviving interest in phage therapy for both humans and agricultural species. Originally this therapy used whole phage particles as the antimicrobial agent; however, isolated phage lytic enzymes are highly specific antimicrobial agents  that often target gram-positive pathogens without affecting beneficial co-occurring organisms. Some viral tail proteins also appear to have bacteriostatic activity and might be useful as antimicrobial compounds .
T4 endonuclease is a DNA repair enzyme with activity against ultraviolet-induced cyclobutane pyrimidine dimers. When applied to the skin, this enzyme has shown protective properties against damage from exposure to the sun, significantly reducing the incidence of basal cell carcinomas and actinic keratoses . Treatment with T4 endonuclease might be particularly useful for patients with DNA repair deficiencies such as xeroderma pigmentosa . Other viral enzymes that reverse alkylation of RNA bases might also be useful clinically .
The most intractable bacterial pathogens are present in the environment within biofilms adhered to a solid surface. Microbial biofilms in biomedical devices such as catheters and implants are particularly problematic as these biofilms can interfere with the device or cause sepsis and are highly resistant to antibiotics . Some phages express saccharolytic enzymes that degrade biofilm carbohydrates, and thus could be useful as part of a multipronged approach to treating pathogens present in biofilms on medical implants or responsible for recurrent infections [43-46].
The ability of viral coat proteins to self-assemble has been exploited to encapsulate imaging agents and drugs. For example, encapsulation of contrasting agents assists in magnetic resonance imaging [47-49]. Self-assembly around other molecules has allowed targeted delivery of anticancer and antimicrobial drugs [50, 51]. The ability to engineer binding domains allows for specific targeting of cancer or microbial cells. In addition, the coat protein of F-specific bacteriophage MS2 from E. coli has been developed into a system for producing nuclease-resistant RNA internal standards for RT-PCR .
Technical challenges related to the cultivation of new viral-host systems have been a primary impediment to the discovery of new viral enzymes. Traditional approaches require that both the host and the virus must be amenable to cultivation. Hosts that fail to form lawns and viruses that fail to form plaques can preclude isolation of the virus. Once isolated, significant empirical work is required to define parameters like multiplicity of infection (MOI), burst size and infection kinetics. These factors are important for detecting viral proteins that are induced at specific timepoints after infection. Even with optimization, it is often difficult to discern the viral proteins from those of the host cell.
Isolation of genes from metagenomic sequence libraries circumvents many of the limitations in cultivation-based discovery  and is a valuable alternative approach to the discovery of novel enzymes. Sequence analysis of viral metagenomes from a variety of environments, including marine, estuarine, soil, and hot springs [4, 7, 54] has begun to describe the genetic diversity within natural viral assemblages and suggests the value of this resource. However, attempts at mining functional enzymes using this approach are just beginning to be reported (Box 1).
As the focus of viral metagenomics shifts from purely ecological investigations of diversity to uncovering novel biological features of viruses and, ultimately, producing useful enzymes, a number of practical challenges arise. In this section, we suggest and discuss means to overcoming some of these challenges and point out those yet to be resolved.
Next-generation sequencing instruments, including Roche/454 FLX pyrosequencer, the Illumina Genome Analyzer and ABI Solid [15, 16], have exponentially increased the amount of information that can be extracted from a sample of DNA or RNA. However, with the recent introduction of the 454 Titanium pyrosequencer, sequence read lengths are just beginning to be adequate to allow occasional retrieval of entire genes from single reads. Although the avoidance of a clone library step in these next generation sequencing platforms has significant advantages in terms of cost and speed, the lack of a clone library make retrieval of specific sequences for follow-up functional characterization more difficult. Traditional random “shotgun” clone libraries combined with Sanger sequencing have advantages for functional metagenomic investigations. First, the longer reads often contain entire coding sequences. Second, the clones that are generated during the process can be fully sequenced or directly expressed to produce enzymes. In fact, with large enough inserts (3-5 kb), it has been possible to isolate multigene operons containing functionally related genes .
An inherent difficulty in viral metagenomics is isolating adequate amounts of genomic material for large-insert library construction. Although viruses are quite abundant, their genomes are generally less than 50 kb in length. Consequently, the 109 to 1011 viral particles in a typical liter of water or kilogram of soil often yield subnanogram quantities of genomic DNA or RNA. Because a typical microbial genome can be equivalent in size to hundreds of viral genomes combined, a few microbial cells or low amounts of free nonviral DNA  can represent significant contamination in a viral nucleic acid preparation. Contamination issues have been successfully addressed through a combination of filtration, centrifugation and nuclease treatments [7, 56] (Figure 1). Non-sequence-specific amplification then allows production of metagenomic libraries from the resulting viral DNA. The quality of both the amplification and the library construction are crucial to the success of a metagenomic expression screen. In our experience, the difficulty of library preparation increases with longer average insert sizes and, as a result, most viral metagenomic shotgun libraries have insert sizes of less than 3 kb. Cloning artifacts or chimeras complicate expression of authentic genes, and sequence stacking due to amplification or cloning biases reduces the comprehensiveness of the screen. Viral genomes in particular can be difficult to clone without significant bias using standard high-copy-number vectors with leaky promoters (e.g. pUC19), compared to transcription-free circular or linear cloning vectors . Most work has focused on DNA viruses because the genomes of RNA viruses must be reverse transcribed before being cloned, which further complicates the procedure. Nevertheless, RNA viral libraries have been constructed and sequenced  and they might be a valuable source of enzymes such as reverse transcriptases, RNA replicases and proteases.
Among next generation DNA sequencing instruments, the Roche 454 pyrosequencer has been the favored platform for sequencing viral metagenomes due to its relatively long read lengths (currently ca. 400 nt). Viral DNA is amplified prior to sequencing because the technique requires microgram quantities of DNA. Several studies have amplified viral DNA using phi29 Pol in a method called multiple displacement amplification (MDA) [6, 59]. While this approach is rapid, MDA is known to have significant amplification bias when applied to small starting amounts of bacterial genomic DNA [60-62] and will preferentially amplify circular DNA . The latter phenomena were evident in the high frequency of single-stranded DNA viral sequences within 454-pyrosequencing libraries of multiple-displacement amplified viral DNA from several marine samples . Because of these biases, the preferred method for amplification of environmental viral DNA prior to next-generation sequencing involves the addition of oligonucleotide linker-adapters to randomly fragmented viral DNA, followed by PCR-amplification using primers homologous to the adapter sequences .
In theory, very large scale sequencing of viral metagenomes should permit the assembly of large contiguous stretches of DNA and potentially entire genomes. In practice, however, the high degree of sequence polymorphism within viral populations has largely confounded attempts to assemble large contigs with high confidence. In viral metagenomic studies thus far, sequence assembly has been used as a tool for predicting the potential genotypic diversity within a given environmental viral assemblage [54, 64]. Towards this end, estimates of genotypic diversity have relied on assemblies of 95% match over at least a 20 bp overlap. Such high stringency tends to prevent misassembling noncontiguous parts of the genome. However, these assemblies likely over-estimate the amount of ecologically-meaningful population diversity within a given viral assemblage. Alterations in the assumed size of the average phage genome within a viral assemblage  or changes in the percent match used in the assembly  can have dramatic effects on resultant diversity estimates. Moreover, the genomes of closely related viral types can diverge significantly  and assembly at high stringencies fails to associate reads from related, but genetically distinct viral types. Such stringent assembly criteria can lead to over estimation of unique viral types and can prevent discovery of genes, enzymes and genomes. For example, lowering the assembly stringency of two high-temperature viral communities from 95 to 50% resulted in the assembly of operons and potentially entire genomes from an environmental viral sample . Although these lower stringency assemblies were likely composites of partial sequences with inherent microheterogeneity, at least they provided insight on the possibly dominant viral populations within a given hot-spring environment. Unfortunately, synthesis of a gene from such a low stringency assembly is probably not a viable strategy for accessing new enzymes because the inherent polymorphism would likely prevent accurate translation and ultimately proper folding of an expressed protein. This caveat can be circumvented by accessing the full insert of the original composite clones or by developing PCR primers to selectively amplify, clone and sequence the putative genes from the original viral DNA preparation.
Owing to the high diversity of viral genes and the relatively low numbers of viral genomes in public sequence databases, most viral coding sequences (even for well-studied phages such as coliphages T4 or T7) have no significant similarity to any known genes. For most viral metagenomic data with relatively long reads, 30% or fewer of the sequences show homology to a previously identified gene, and many of these genes are not associated with an identifiable function. The homology rate is substantially lower for short read sequences [66, 67]. The likelihood of finding genes by similarity also depends on the evolutionary conservation of the genes. For example, genes coding for lysozymes are highly conserved and commonly detected, whereas those coding for holins are highly diverse  and are virtually undetectable in viral metagenomes.
In some cases, a sizable proportion of viral metagenomic sequences (up to 60%) show homology to environmental sequences, indicating some level of prevalence and persistence of viral genes across a variety of environmental contexts. Thus, improved understanding of the phylogeny and ecology of viral genes will come from a more thorough bioinformatic exploration of predicted viral proteins encoded by metagenomic libraries. Nevertheless, better informatics will not directly solve a more fundamental problem: the accumulation of sequence data has outpaced the basic research into virus biology that is critical to functionally annotate the sequence data. Ultimately, enhancing our understanding of viral biology will improve annotation of viral genes and the utility of metagenomics as a predictive tool for the potential influence of viral processes within microbial communities, as well as enhance enzyme discovery.
Similarity-based detection is only useful for enzyme discovery if the respective genes can be expressed in a suitable host. This depends, in part, on accurate determination of the start and stop sites, which is hindered by characteristics unique to metagenomes in general and to viral metagenomes in particular. A basic problem is that most ORF prediction programs were written for complete cellular genomes. Even after assembly, metagenomic sequences tend to be short compared to the whole genomes for which programs like Glimmer  and GeneMark  were developed. The flood of metagenome sequence data has spurred the development of new ORF calling algorithms, such as MetaGene Annotator , designed specifically for ORF prediction from fragmentary sequence data. However, the applicability of these newer ORF prediction algorithms for downstream expression studies has not been experimentally tested.
Other challenges in expressing viral genes derive from unique characteristics of viral biology. For example, in vivo expression of M-MLV reverse transcriptase involves recoding of the cellular translation system to read though a stop codon and produce a polyprotein (Gag-Pol fusion) that must be post-translationally processed to form an active enzyme . These processes highlight several of the problems commonly faced in producing functional proteins from viral metagenomes (i.e., nonstandard codon usage, involvement of cellular proteins and post-translational processing) that complicate production of enzymes discovered by metagenomics. In addition, overlapping genes are found throughout the viral world, especially in RNA and single-stranded DNA viruses. Nonstandard (non-ATG) start codons create another challenge. Research on T5 phage DNA Pol was hampered for years by difficulties in expression due to the erroneous assignment of the start codon of the phage gene, which turned out to be a rare TTG start codon . This mistake was discovered by determining the amino terminal sequence of the phage T5 native protein; however, such an approach would be nearly impossible for proteins discovered through metagenomic analysis because phage isolates are not available for purification of native proteins. A final challenge is the codon usage bias that can occur in viral genes and prevent adequate expression in common laboratory expression systems .
Some of these problems can be addressed more easily than others. Newer gene finding programs have been developed that take into account the unique properties of viral metagenomic sequences [71, 74-78]. None of the informatic solutions solves the problem of post-translationally processed polyproteins that are encoded by viral genes. In these polyproteins, the amino and/or carboxy terminus of the functional protein is determined by protease recognition sequences within the protein sequence or by site-specific autolytic cleavage and not by the start or stop codons of the genes. The problem of post-translational modification can sometimes be addressed. Some polyproteins appear to be capable of autoproteolytic cleavage in vitro to form functional proteins, whereas in other cases, it is possible to insert artificial start codons based on alignment to known genes and produce functional proteins (T. Schoenfeld, unpublished). The problem of codon bias can be overcome by expressing genes in cells that supply tRNAs for rare codons (e.g. Rosetta cells, EMD BioSciences) or by resynthesizing the gene with optimal codons for expression in E. coli.
Functional screening is an alternative approach to sequence-based discovery that is not dependent on initial DNA sequencing. Rather, clones of environmental DNA are screened directly for enzymatic activity. By its nature, this approach selects for genes that can be detectably expressed in the host, usually E. coli. However, there are challenges to functional screening. First, it requires an assay capable of screening a large number of clones with relatively few false positives or false negatives. Plate-based colony assays are typically the simplest screens, but they are not amenable to detection of certain useful proteins. Alternatively, colonies can be picked and screened by high-throughput robotic methods. The random nature of shotgun libraries means the coding sequence might not necessarily be in proximity to a vector-based promoter, especially if a transcription-free vector is used. Therefore, a gene of interest must express from its own promoter, which might not be active in the host cell at a detectable level. Also, target genes might have unstable secondary structures when cloned, or their encoded proteins might be toxic to the host cells, preventing their isolation. Despite these challenges, two significant advantages make functional screens worth considering. First, this approach can identify genes that are too divergent from known genes to be identified by sequence similarity. Second, once clones are detected by function, the genes can be expressed immediately (at least at low levels) to produce proteins that can be further characterized.
A final caveat in metagenomics-based work is that discovery is often only the beginning of an enzyme improvement project. Most of the proteins discovered in metagenomic libraries require some form of modification to make them more suitable for their intended applications. For this reason, discovery of novel genes and gene products from metagenome libraries provides an initial pool of useful sequence diversity to an enzyme engineering project, but is not necessarily the end of the process.
As we add new and unique enzyme activities to the “tool chest” of molecular biology, several sources should be explored. The hundreds of viral and bacterial genomes already deposited in sequence databases represent a ready source of molecular diversity to be used for enzyme discovery. However, as viral metagenomic studies have shown, the pool of sequenced viral genomes is woefully unrepresentative of extant viral diversity. Thus, viral metagenomics offers a means of exploring genetic diversity within the vast uncultivated portion of the virosphere. However, practical application of functional viral metagenomics as an approach to new enzyme development is in its infancy. The work discussed in Box 1 represents the first demonstration that enzymes discovered in viral metagenomic libraries can be expressed, characterized and put to use. Until now, scientists have relied on the inherent characteristics of a very few available enzymes from cultivated strains of bacteria and viruses. Functional metagenomics promises to provide a wealth of starting information for the development of enzymes applicable to a broad range of industrial, biomedical and research applications.
The focus of this article has been on viral Pols, but several newer applications might be improved by substituting viral enzymes for the currently used cellular enzymes. For instance, helicases separate DNA strands during replication and can facilitate DNA sequencing and amplification . To date, no published sequencing protocol uses viral helicases, but genes encoding helicases are highly abundant in viral genomes (Table 2). Dual-function helicase-primase proteins like that of T7 phage are particularly promising [80, 81] as the natural integration of these activities in a single protein suggests that it might augment the already useful T7 Pol. DNA methylases are gaining significance in studies of epigenetics ; most reported methods use bacterial methylases, but genes encoding other methylases are common in viral genomes . Transposases are used for insertional mutagenesis and nested deletions, and as an alternative to subcloning or primer walking for sequencing longer templates . Although bacterial transposases are more commonly used, phage Mu transposase has also been used for insertional mutagenesis and sequencing .
Viral metagenomics promises to feed an almost unlimited diversity of enzymes into screens that can be tailored to the practical needs of a variety of applications. More efficient approaches based on advances in genomics and screening technology will accelerate this work (Box 2). An advantage of functional metagenomics-based enzyme discovery is that metagenomic screens are well suited to focusing on specific environments, which might result in discovery of enzymes with desirable attributes. For instance, hot springs proved a fertile source of thermostable enzymes (Box 1), whereas sewage effluents would seem likely sources of phage-derived enzymes specific for lysis of human enteric pathogens. Screening viromes from a broad range of environments will certainly provide a vast reservoir of genetic diversity for discovery of the next generation of molecular tools and medicines.
The power of viral metagenomics is illustrated by the discovery of novel classes of thermostable viral DNA polymerases (T. Schoenfeld, unpublished). Viral particles were isolated from hundreds of liters of water from two Yellowstone hot springs (74°C and 93°C) and separated from microbial cells by tangential flow filtration  (Figure 1). Less than 100 ng of viral nucleic acid was extracted from each sample, sheared, amplified and cloned. In collaboration with the US Department of Energy Joint Genome Institute, about 29,000 sequence reads were determined (approximately 28 Mb in total). The sequences contained several hundreds of apparent pol genes, including one from every known pol family. Only 59 of these genes were full length, and ten genes were expressed to produce thermostable DNA Pols, or “PyroPhage Pols”. The predicted amino acid sequences of seven PyroPhage Pols were compared with representatives of known viral and cellular Pol families and with the commonly used thermostable Pols (Figure I) . By this analysis, the PyroPhage Pols fall into two groups distinct from other known viral and microbial Pols, one group much more diverse than the other.
Consistent with the molecular diversity, the biochemical activities of the PyroPhage enzymes appear to be unique . For instance, PyroPhage 3173 Pol possesses inherent reverse transcriptase activity and the highest thermostability among known viral Pols, which will allow its use in conventional PCR and single-enzyme RT PCR. It also shows promise in whole genome and single cell amplification .
Figure I (in Box 1). Dendrogram showing sequence similarity relationships among PyroPhage enzymes and representative viral and microbial DNA polymerases. Comparisons are based on full-length amino acid sequence aligned by ClustalW. Scale bar represents 10% sequence divergence.
The short read lengths of ultrahigh-throughput next-generation sequencing technologies are currently problematic for viral metagenomics-based enzyme discovery . However, these problems should be surmountable and the advantages that these methods offer in throughput should make them a logical choice for functional metagenomics. Read lengths from the current generation of 454 pyrosequencing instruments are approaching those of Sanger shotgun sequencing [15, 16]. In the near future, single molecule methods might allow sequencing of entire or nearly entire viral genomes in a single read , obviating the need for assembly of sequences or recovery of clones. Alternatively, some of the single-cell genomics methods  might be applicable to viral genomes. Using either approach, the ability to determine entire sequences of genes will allow direct to synthesis of genes. Moreover, such long read sequencing will allow for the isolation of entire coding sequences including complete multigene operons. Another promising technology is micro- and nano-fluidics , which should improve our ability to perform ultrahigh-throughput functional screens. With the availability of these technologies, remaining challenges will be: (i) improving gene predictions to relate evolutionarily distant proteins to a possible function, (ii) developing the informatics to sift through the huge amounts of data to find these genes, and (iii) improving our ability to rapidly express and characterize the new enzymes.
The authors gratefully acknowledge the contributions of numerous colleagues, especially the technical contributions of Vinay Dhodda, Darby Renneckar and Nicholas Hermersmann. This work was supported by grants 0839404 from National Science Foundation and 5 R44 HG002714-03 from National Human Genome Research Institute to TS. The contribution of K.E.W. and S.W.P. was supported through NSF grant MCB-0731916.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.