Achievements and new knowledge unraveled by metagenomic approaches.

Metagenomics has paved the way for cultivation-independent assessment and exploitation of microbial communities present in complex ecosystems. In recent years, significant progress has been made in this research area. A major breakthrough was the improvement and development of high-throughput next-generation sequencing technologies. The application of these technologies resulted in the generation of large datasets derived from various environments such as soil and ocean water. The analyses of these datasets opened a window into the enormous phylogenetic and metabolic diversity of microbial communities living in a variety of ecosystems. In this way, structure, functions, and interactions of microbial communities were elucidated. Metagenomics has proven to be a powerful tool for the recovery of novel biomolecules. In most cases, functional metagenomics comprising construction and screening of complex metagenomic DNA libraries has been applied to isolate new enzymes and drugs of industrial importance. For this purpose, several novel and improved screening strategies that allow efficient screening of large collections of clones harboring metagenomes have been introduced.

Introduction

Metagenomics has been defined as function-based or sequence-based cultivation-independent analysis of the collective microbial genomes present in a given habitat (Riesenfeld et al. 2004b). This rapidly growing research area provided new insights into microbial life and access to novel biomolecules (Banik and Brady 2008; Edwards et al. 2006; Frias-Lopez et al. 2008; Venter et al. 2004). The developed metagenomic technologies are used to complement or replace culture-based approaches and bypass some of their inherent limitations. Metagenomics allows the assessment and exploitation of the taxonomic and metabolic diversity of microbial communities on an ecosystem level.

Recently, advances in throughput and cost-reduction of sequencing technologies have increased the number and size of metagenomic sequencing projects, such as the Sorcerer II Global Ocean Sampling (GOS) (Biers et al. 2009; Rusch et al. 2007), or the metagenomic comparison of 45 distinct microbiomes and 42 viromes (Dinsdale et al. 2008a). The analysis of the resulting large datasets allows exploration of biodiversity and performance of system biology in diverse ecosystems.

So far, the main application area of metagenomics is mining of metagenomes for genes encoding novel biocatalysts and drugs (Lorenz and Eck 2005). Correspondingly, new sensitive and efficient high-throughput screening techniques that allow fast and reliable identification of genes encoding suitable biocatalysts from complex metagenomes have been invented.

In this review, an overview of the recent developments and achievements of bioprospecting and metagenomic analyses of microbial communities derived from different environments is given. In addition, novel metagenomic approaches are briefly discussed.

Exploring the phylogenetic diversity

Metagenomics is a powerful tool for assessing the phylogenetic diversity of complex microbial assemblages present in environmental samples such as soil, sediment, or water. The total number of prokaryotic cells on earth has been estimated to be 4–6 × 1030 comprising 106 to 108 separate genospecies (Sleator et al. 2008). The majority of these microbes is uncharacterized and represents an enormous unexplored reservoir of genetic and metabolic diversity. In recent years, high-throughput metagenomic approaches produced millions of environmental gene sequences, thereby, providing access to the so far hidden phylogenetic composition of complex environmental microbial communities (Sjöling and Cowan 2008).

To explore the microbial diversity of environmental samples, also termed “taxonomical binning,” different approaches can be applied (Richter et al. 2008). Usually, phylogenetic relationships are determined by analysis of conserved ribosomal RNA (rRNA) gene sequences (Woese 1987). Extensive sequencing of ribosomal RNA genes resulted in generation of several large reference databases, such as the ribosomal database project (RDP) II (Cole et al. 2003), Greengenes (DeSantis et al. 2006), or SILVA (Ludwig et al. 2004). These comprehensive databases allow classification and comparison of environmental 16S rRNA gene sequences. Traditional surveys of environmental prokaryotic communities are based on amplification and cloning of 16S rRNA genes prior to sequence analysis. However, some inherent disadvantages such as PCR bias, instability of the recombinant plasmids in the host strain, or the varying number of gene copies between taxa are limitations of this approach (Biddle et al. 2008; Venter et al. 2004). More comprehensive views of prokaryotic communities can be achieved by use of high-throughput shotgun sequencing of environmental samples. Direct sequencing of metagenomic DNA has been proposed to be the most accurate approach for assessment of the taxonomic composition (von Mering et al. 2007). The major advantage of this cloning-independent approach is the avoidance of bias introduced by amplification of phylogenetic marker genes and cloning. In addition, Manichanh et al. (2008) showed that evaluation of a shotgun sequencing-derived dataset provides a reliable estimate of the microbial diversity stored in metagenomic libraries. Venter et al. (2004) were the first to apply whole genome shotgun sequencing to samples of the Sargasso Sea in order to characterize the microbial community and identify new genes and species. The dataset included 1.66 million sequences comprising 1.045 billion base pairs. The taxonomic composition was evaluated by 16S rRNA gene analysis and employment of alternative phylogenetic markers such as RecA/RadA, heat shock protein Hsp70, elongation factor Tu, and elongation factor G. The assignment to phylogenetic groups was consistent among the different markers but the abundance of the encountered phylogenetic groups varied (Venter et al. 2004).

Determination of the taxonomic diversity by analysis of pyrosequencing- or shotgun-derived datasets has been applied to various environments, including an acid mine biofilm (Tyson et al. 2004), seawater samples (Angly et al. 2006; DeLong et al. 2006), the Soudan mine (Edwards et al. 2006), the Peru Margin subsea floor (Biddle et al. 2008), honey bee colonies (Cox-Foster et al. 2007), and deep-sea sediments (Hallam et al. 2004). To date, the largest metagenomic dataset was generated within the framework of the GOS expedition (Rusch et al. 2007; Yooseph et al. 2007). The GOS dataset extends the previously published Sargasso Sea dataset (Venter et al. 2004). Random insert libraries were constructed from DNA isolated from bacterioplankton derived from 41 surface marine environments and a few nonmarine aquatic samples. The phylogenetic diversity stored in this dataset, which comprises 7.7 million sequences (6.3 billion bp), was assessed by analysis of the 16S rRNA gene sequences present in the metagenomic libraries (Biers et al. 2009; Rusch et al. 2007). In general, the alphaproteobacteria were the dominant phylogenetic group in ocean surface waters, whereas, the abundance of other phyla differed depending on the type of environment (Biers et al. 2009).

Due to the enormous quantity of short DNA fragments in large shotgun sequencing-derived or pyrosequencing-derived metagenomic datasets, methods have been developed that are more suitable for taxonomic binning than the analysis of highly conserved phylogenetic marker genes. Phylogenetic classification of metagenomic fragments can be based on sequence composition, i.e., oligonucleotide frequencies, which vary significantly among genomes and exhibit weak phylogenetic signals (Abe et al. 2003; Karlin and Burge 1995; Pride et al. 2003; Teeling et al. 2004a). For a phylogenetic classification of complex microbial communities based on oligonucleotide frequencies, bioinformatic software tools such as TETRA or PhyloPythia have been developed (McHardy et al. 2007; Teeling et al. 2004b). These tools require training, employing known genomic sequences of different taxonomic origin. The accuracy of the phylogenetic classification depends on different factors such as fragment length of the environmental DNA and amount or origin of the genomic sequences used for training. The above-mentioned tools have been successfully employed for characterization of several habitats such as the Sargasso Sea and sludge used in industrial wastewater processing (Abe et al. 2005; McHardy et al. 2007). Recently, other software tools such as the metagenome analyzer MEGAN (Huson et al. 2007), CARMA (Krause et al. 2008), and the sequence ortholog-based approach for binning and improved taxonomic estimation of metagenomic sequences Sort-ITEMS (Monzoorul et al. 2009) have been invented for taxonomic binning of large metagenomic datasets that consist of short environmental DNA fragments. The algorithms differ in the method for phylogenetic classification. MEGAN (Huson et al. 2007; Huson et al. 2009) compares metagenomic datasets with one or more sequence databases, i.e., NCBI-NR, NCBI-NT, NCBI-ENV-NR, or NCBI-ENV-NT (Benson et al. 2006). Subsequently, the reads are assigned to the lowest common ancestor of the nearest relatives in the reference databases. In order to validate the algorithm, the authors applied MEGAN to the Sargasso Sea dataset and deduced species distribution, which is similar to that reported by Venter et al. (2004). Additionally, Poinar et al. (2006) analyzed a dataset derived from a mammoth bone using MEGAN. Approximately 50% of the analyzed sequences were identified as mammoth DNA, whereas, the remaining sequences were derived from endogenous bacteria and nonelephantid environmental contaminants (Poinar et al. 2006).

Krause et al. (2008) introduced the CARMA algorithm, which uses conserved domains and protein families of the protein families (Pfam) database (Finn et al. 2006) as phylogenetic markers for taxonomic classification of the environmental DNA sequences. These environmental gene tags (EGTs) are identified by employing the Pfam profile hidden Markov models. Subsequently, for each matching Pfam family a phylogenetic tree is reconstructed, and the metagenomic sequences are, thereby, assigned to phylogenetic groups. In this way, EGTs as short as 27 amino acids can be classified (Krause et al. 2008). CARMA has been shown to provide accurate results, but it is computationally expensive (Diaz et al. 2009; Krause et al. 2008).

The most recent binning algorithm Sort-ITEMS utilizes the bit score and alignment parameters of the basic local alignment search tool BLAST (Altschul et al. 1990) for an initial taxonomic classification. Subsequently, a higher resolution is achieved by an orthology-based approach (Monzoorul et al. 2009).

Phylogenetic classification of the metagenomic datasets relies on the use of the above-mentioned reference databases that contain sequences of known origin and gene function. To date, the common databases are biased towards model organisms or readily cultivable microorganisms. This is a major limitation for taxonomic classification of microbial communities in ecosystems. According to Huson et al. (2009), up to 90% of the sequences of a metagenomic dataset may remain unidentified due to the lack of a reference sequence.

Connecting function to phylogeny

Exploring the phylogenetic diversity and population structure of environmental samples is essential for the reconstruction of the metabolic potential of individual organisms or phylogenetic groups and the discovery of their interactions. The employment of metagenomics allows the discovery of interactions between microorganisms and the environment and the assignment of ecosystem functions to microbial communities (Lopez-Garcia and Moreira 2008; Sjöling and Cowan 2008).

Linking functional genes of uncultured organisms to phylogenetic groups can be accomplished by cloning and sequencing of large genomic DNA fragments containing phylogenetic markers or by reconstruction of genomes from metagenomic datasets (Sjöling and Cowan 2008). An illustrative example is the discovery of rhodopsin-like photoreceptors and proteorhodopsin-dependent phototrophy in marine bacteria by analyzing large-insert metagenomic libraries (Béjà et al. 2000). The open reading frame coding for proteorhodopsin was located in the vicinity of a 16S rRNA gene, which originated from a member of the gammaproteobacteria (Béjà et al. 2000). In additional datasets derived from aquatic samples, new and diverse rhodopsin-like genes were identified and indicated a widespread abundance and importance of this light-driven way of energy conservation (Rusch et al. 2007; Venter et al. 2004). Reconstruction of near complete and complete genomes of individual microorganisms derived from metagenomic datasets is restricted to low-diversity habitats, since the species-richness of high-diversity habitats such as soil and sediment would require enormous sequencing and assembly efforts. Recently, this approach has been successfully applied for low-diversity samples from acid mines (Tyson et al. 2004), an anaerobic ammonium-oxidizing community (Strous et al. 2006), and enrichments (Hallam et al. 2004).

Functional diversity of microbial communities

Large-scale sequencing of metagenomic DNA permits the identification of the most frequently represented functional genes and metabolic pathways that are relevant in a given ecosystem. In this way, the dominant biosynthetic pathways and primary energy sources can be assessed. Edwards et al. (2006) conducted the first study in which metabolic profiles of whole microbial communities based on a pyrosequencing-derived dataset were analyzed. The authors compared two different sampling sites in the Soudan mine (Minnesota, USA). Significant differences in the use of substrates and metabolic pathways were established. In addition, the geochemical conditions of the two analyzed sites and the microbial metabolism correlated (Edwards et al. 2006). The rapid identification of the metabolic capacity and genetic diversity of this habitat indicated the significance of metagenomics for functional analysis of ecosystems. Other examples for identification of the functional diversity and profiles by analysis of pyrosequencing-derived datasets include an obesity-associated gut microbiome (Turnbaugh et al. 2006), a coral-associated microbial community (Wegley et al. 2007), a comparison of nine biomes (Dinsdale et al. 2008a), ocean surface waters (Frias-Lopez et al. 2008), the Peru Margin subseafloor (Biddle et al. 2008), coral atolls (Dinsdale et al. 2008b), and stressed coral holobionts (Thurber et al. 2009).

For functional binning of metagenomic datasets, sequences are compared to reference databases, such as the clusters of orthologous groups of proteins (Tatusov et al. 2003), the Kyoto encyclopedia of genes and genomes (Kanehisa et al. 2004), Pfam, SEED (Overbeek et al. 2005), the search tool for the retrieval of interacting genes/proteins STRING (Jensen et al. 2009), or TIGRFAM (Haft et al. 2003), which contain known protein functions, families, and pathways (Richter et al. 2008). Bioinformatic analyses are crucial for linking function to phylogenetic diversity of ecosystems. Recently, Meyer et al. (2008) introduced the metagenome rapid annotation using subsystem technology (MG-RAST) server for analysis of metagenomic datasets. The server provides annotation of sequence fragments, phylogenetic classification, and metabolic reconstruction by implementing the SEED, the national microbial pathogen database resource (McNeil et al. 2007), Greengenes, RDP-II, SILVA, and the European ribosomal RNA database (Wuyts et al. 2004). In addition, this open-source online tool allows the comparison of metagenomic datasets derived from different environments (Meyer et al. 2008). Comparative metagenomics is useful for identification of differences in the ability of microbial communities to adapt to changing environmental conditions. Tringe et al. (2005) analyzed and compared metagenomic datasets from various environments and deduced habitat-specific functions and profiles of the sampled environments. Thus, profiling of the functions encoded by a microbial community rather than the types of organisms producing them provides a means to distinguish samples on the basis of the functions selected by the local environment and reveals insights into features of that environment. This gene-centric approach to environmental sequencing suggests that the functional profile predicted from environmental sequences of a community is similar to that of other communities whose environments of origin pose similar demands.

Nevertheless, the analysis of the taxonomic diversity, functional binning, and profiling of metagenomic datasets bears several limitations. The reference databases used for functional annotation of the sequences are inherently incomplete. Therefore, metagenomic analyses can only be as good as the quality of the reference databases (Meyer et al. 2008). To cope with the increasing number and size of metagenomic sequencing projects, improvement and development of bioinformatic tools for metagenomic data analysis is still required (Meyer et al. 2008).

Metatranscriptomics

Recently, sequencing and characterization of metatranscriptomes have been employed to identify expressed biological signatures in complex ecosystems. Metagenomic complementary DNA (cDNA) libraries have been constructed from messenger RNA (mRNA) that has been isolated from environmental samples (Bailly et al. 2007; Frias-Lopez et al. 2008; Gilbert et al. 2008, 2009; Grant et al. 2006). In contrast to libraries constructed from environmental DNA, cDNA libraries reflect the active metabolic functions of a microbial community. However, due to difficulties associated with RNA isolation, separation of mRNA from other RNA species, and instability of mRNA, constructing libraries derived from environmental mRNA is more challenging than generation of metagenomic DNA libraries (Sjöling and Cowan 2008). Frias-Lopez et al. (2008) constructed cDNA libraries from metagenomic microbial mRNA derived from ocean surface water. The cDNA libraries were subjected to pyrosequencing, and the resulting dataset was compared to diverse databases. Many of the identified genes were highly similar to genes previously identified in ocean samples. Approximately 50% of all detected transcripts were unique, indicating that a large unknown metabolic diversity is present in the ocean. The few published metatranscriptomic studies were mainly performed with samples from marine environments and soils. The microbial community transcriptome analyses revealed that the identification of indigenous gene- and taxon-specific patterns, and the identification of key metabolic functions are feasible. In addition, when paired with metagenomic data, detailed analyses of both structure and function of microbial communities are provided (Frias-Lopez et al. 2008; Gilbert et al. 2008; Urich et al. 2008).

Metagenomes as sources for novel biomolecules

Most biocatalysts employed for biotechnological or industrial purposes are of microbial origin. This reflects the fact that the broadest genetic variety in the biosphere can be found in the different microbial communities present in the various ecosystems on earth (Ferrer et al. 2009). The application of culture-independent metagenomic approaches allows exploiting this almost unlimited resource of novel biomolecules (Sjöling and Cowan 2008). The work published in this field showed that the cloning of metagenomic DNA and the subsequent screening of the constructed complex environmental libraries bear the potential to encounter entirely new classes of genes for new or known functions, including genes encoding, i.e., lipases, antibiotics, antibiotic resistance genes, oxidoreductases, catabolic enzymes, and biotin synthesis (see Table 1). Several techniques have been used to identify and retrieve genes and gene clusters from metagenomic libraries. Due to the complexity of metagenomic libraries, high-throughput and sensitive screening approaches have been employed. Screens have been based either on nucleotide sequence (sequence-driven approach) or on metabolic activity (function-driven approach) (Fig. 1).

Table 1

Recent examples for metagenome-derived biocatalysts and the employed screening strategy

Target	Source	Number of screened clones	Sampling site	Screening technique	Reference
Lipase	Fosmids	>7,000	Baltic sea sediment (Sweden)	Phenotypical detection	Hårdeman and Sjöling 2007
	Cosmids	10,000	Sequencing fed-batch reactor enriched with gelatin	Phenotypical detection	Meilleur et al. 2009
	Plasmids	Not mentioned	Soil samples from different altitudes of Taishan (China)	Phenotypical detection	Wei et al. 2009
	Cosmids	1,532	Soil from uncultivated field (Germany)	Phenotypical detection	Voget et al. 2003
	Fosmids	386,400	Tidal flat sediments (Korea)	Phenotypical detection	Lee et al. 2006b
Lipase/Esterase	Plasmids	1,016,000	Soil from a meadow, sugar beet field, and river valley (Germany)	Phenotypical detection	Henne et al. 2000
Esterase	Fosmids	5,000	Hot springs and mud holes in solfataric fields (Indonesia)	Phenotypical detection	Rhee et al. 2005
	Phagemids	385,000	Wadi Natrun (Egypt), Lake Nakuru, and Crater Lake (Kenya) and enrichments	Phenotypical detection	Rees et al. 2003
	Fosmids	100,000	Desert soil (Antarctica)	Phenotypical detection	Heath et al. 2009
	Plasmids	93,000	Vegetable soil	Phenotypical detection	Li et al. 2008
	BACs	8,000	Surface water microbes from Yangtze river (China)	Phenotypical detection	Wu and Sun 2009
Cellulase	Phagemids	385,000	Wadi Natrun (Egypt), Lake Nakuru, and Crater Lake (Kenya) and enrichments	Phenotypical detection	Rees et al. 2003
	Cosmids	1,700	Soil microbial consortia (Germany)	Phenotypical detection	Voget et al. 2006
	Cosmids	3,744	Aquatic community and soil Germany)	Phenotypical detection	Pottkämper et al. 2009
	Cosmids	15,000	Buffalo rumen	Phenotypical detection	Duan et al. 2009
	Cosmids	32,500	Rabbit cecum	Phenotypical detection	Feng et al. 2007
Protease	Plasmids	80,000	Compost soil (Germany), soil from mining shaft (Germany), and mixed soil sample (Germany, Israel, and Egypt)	Phenotypical detection	Waschkowitz et al. 2009
	Fosmids	30,000	Deep-sea sediment from a clam bed community (Korea)	Phenotypical detection	Lee et al. 2007
Agarase	Cosmids	1,532	Soil from uncultivated field (Germany)	Phenotypical detection	Voget et al. 2003
Oxidative coupling enzyme (OxyC)	Cosmids	10,000,000	Collection of soil samples (USA and Costa Rica)	Sequence-based	Banik and Brady 2008
Alcohol oxidoreductase	Plasmids	900,000 and 400,000	Soil and enrichment cultures from a sugar beet field (Germany), river sediment (Germany), sediment from Solar Lake (Egypt), and sediment from the Gulf of Eilat (Israel)	Sequence-based and phenotypical detection	Knietsch et al. 2003b, c
Amidase	Plasmids	193,000	Soil and enrichment cultures from marine sediment, goose pond, lakeshore, and an agricultural field (Netherlands)	Heterologous complementation	Gabor et al. 2004
Xylanase	Phagemids	5,000,000	Manure wastewater lagoon (USA)	Phenotypical detection	Lee et al. 2006a
Antibiotics	Cosmids	Not mentioned	Bromeliad tank water (Costa Rica)	Phenotypical detection	Brady and Clardy 2004
Glycerol dehydratase and diol dehydratase	Plasmids	158,000 and 560,000	Soil from a sugar beet field (Germany), river sediment (Germany), and sediment from Solar Lake (Egypt)	Sequence-based and heterologous complementation	Knietsch et al. 2003a
Magnetosome island gene clusters	Fosmids	5,823	Different aquatic sediments (Germany)	Sequence-based	Jogler et al. 2009
Benzoate 1,2-dioxygenase alpha subunit and chlorocatechol 1,2-dioxygenase	DNA	-	Soil from a conserved forest (Japan)	Sequence-based	Morimoto and Fuji 2009
DNA polymerase I	Plasmids and fosmids	230,000 and 4,000	Glacier ice (Germany)	Heterologous complementation	Simon et al. 2009
Multicopper oxidases	DNA	-	Not specified	Sequence-based	Meyer et al. 2007
Blue light photoreceptor	Cosmids	2,500	Soil from a botanical garden (Germany), enrichment	Sequence-based	Pathak et al. 2009
Na+/H+ antiporters	Plasmids	1,480,000	Soil from a meadow, sugar beet field, and river valley (Germany)	Heterologous complementation	Majernik et al. 2001
Antibiotic resistance	BACs and plasmids	28,200 and 1,158,000	Plano silt loam (USA)	Heterologous complementation	Riesenfeld et al. 2004a
Poly-3-hydroxybutyrate metabolism	Cosmids	45,630	Activated sludge and soil microbial communities (Canada)	Heterologous complementation	Wang et al. 2006
Lysine racemase	Plasmids	Not mentioned	Garden soil (Taiwan)	Heterologous complementation	Chen et al. 2009
Aromatic-hydrocarbon catabolic operon fragments	Plasmids	152,000	Crude-oil contaminated groundwater microbial flora (Japan)	SIGEX	Uchiyama et al. 2005
Quorum sensing inducer/inhibitor	BACs and fosmids	52,500 and 300	Soil on the floodplain of the Tanana River (Alaska)	METREX	Williamson et al. 2005
Beta-lactamase	Fosmids	8,823	Cold-seep sediments of Edison seamount (Papua New Guinea)	Phenotypical detection	Song et al. 2005
Chitinase	DNA	-	Water and sediment samples from aquatic environments (USA and Arctic ocean)	Sequence-based	LeCleir et al. 2004
Cyclodextrinase	Phagemids	200,000	Cow rumen	Phenotypical detection	Ferrer et al. 2005

Fig. 1

Strategies for recovery of novel biomolecules

Sequence-based screening

The sequence-based screening approach is limited to the identification of new members of known gene families. In general, target genes are identified either by PCR-based or hybridization-based approaches employing primers and probes derived from conserved regions of known genes and gene products (Daniel 2005; Handelsman 2004). Thus, only genes harboring regions with similarity to the sequences of the probes and primers can be recovered by this approach. In addition, sequence-driven screening is not selective for full-length genes and functional gene products. The advantage of this screening strategy is the independence on gene expression and production of foreign genes in the library host (Lorenz et al. 2002). Several novel functional enzymes such as chitinases, alcohol oxidoreductases, diol dehydratases, and enzymes conferring antibiotic resistance have been recovered by employing sequence-driven approaches (see Table 1). For example, Banik and Brady (2008) isolated two novel glycopeptide-encoding gene clusters from a large-insert megalibrary, which comprised 10,000,000 cosmid-containing clones derived from desert soil by a PCR-based screen. Degenerate primers were employed, which were deduced from OxyC, an oxidative coupling enzyme encoded by glycopeptide biosynthetic clusters. The isolation of these biosynthetic clusters is important for the development of novel glycopeptides analogs, which can serve as substitutes of currently used antibiotics such as vancomycin.

Another recent example for a screening based on sequence similarity was published by Jogler et al. (2009). After selective enrichment of magnetotactic bacteria (MTB), large DNA fragments from uncultured MTB derived from various aquatic habitats were cloned into fosmid vectors. Four fosmid libraries comprising 5,823 clones were screened by hybridization using mam genes of known magnetotactic alphaproteobacteria as probes. Two fosmids, which contain operons with similarity to magnetosome islands of cultured MTB, were detected, and the organization of the magnetosome island of uncultured MTB was elucidated.

A new approach to retrieve complete functional genes is PCR-denaturing gradient gel electrophoresis (DGGE) followed by metagenomic walking. Morimoto and Fujii (2009) conducted a PCR-DGGE targeting benA and tfdC, which encode the alpha subunits of benzoate 1,2-dioxygenase and chlorocatechol 1,2-dioxygenase, respectively. Two DGGE bands, which appeared after addition of 3-chlorobenzoate to the samples, were chosen for further analysis. The complete functional genes were recovered by metagenome walking (Morimoto and Fujii 2009).

Recently, Meyer et al. (2007) introduced subtractive hybridization magnetic bead capture as approach for recovery of multicopper oxidases from metagenomic DNA. Conserved regions of the target genes are amplified from a metagenomic DNA sample by PCR using biotinylated degenerated primers. Subsequently, the resulting amplified target gene fragments are immobilized on streptavidin-covered magnetic beads, which are then used as probes for capturing the full-length genes from metagenomic DNA by hybridization. In contrast to previously published PCR-based techniques, the subtractive hybridization approach allows the recovery of multiple gene targets in a single reaction. According to Meyer et al. (2007), the employment of immobilized large gene fragments as probes results in specificity, which is higher than that of other PCR-based approaches (Meyer et al. 2007).

In a few cases, microarray technology has been employed for sequence-driven screening of metagenomic DNA and libraries. A recent example is the recovery of genes encoding blue light-sensitive proteins (Pathak et al. 2009).

Function-based screening

Function-driven screening of metagenomic libraries is not dependent on sequence information or sequence similarity to known genes. Thus, this is the only approach that bears the potential to discover new classes of genes that encode either known or new functions (Heath et al. 2009; Rees et al. 2003). A significant limitation of this technique is the dependence on expression of the target genes and production of functional gene products in a foreign host, which is in most studies Escherichia coli. Thus, the incapability to discover functional gene products or a low detection frequency during function-based screens of metagenomic libraries might be a result of the inability of the host to express the foreign genes and to form active recombinant proteins. In addition, function-driven screening often requires the analysis of more clones than sequence-based screening for the recovery of a few positive clones (Daniel 2005). The major advantage of a function-based screening approach is that only full-length genes and functional gene products are detected. The following three different types of function-driven approaches have been employed for screening of metagenomic libraries: (1) direct detection of specific phenotypes of individual clones; (2) heterologous complementation of host strains or mutants; (3) induced gene expression (Fig. 1 and Table 1).

To identify enzymatic functions of individual clones, chemical dyes and insoluble or chromophore-containing derivatives of enzyme substrates can be incorporated into the growth medium (Daniel 2005; Ferrer et al. 2009; Handelsman 2004). Examples for this simple activity-based approach are the detection of recombinant E. coli clones exhibiting protease activity on indicator agar containing skimmed milk as protease substrate (Lee et al. 2007; Waschkowitz et al. 2009) or the detection of lipolytic activity by employing indicator agar containing tributyrin or tricaprylin as enzyme substrates (Hårdeman and Sjöling 2007; Heath et al. 2009; Lee et al. 2006b). Clones with proteolytic or lipolytic activity are identified by halo formation on solidified indicator medium.

A different approach is the use of host strains that require heterologous complementation by foreign genes for growth under selective conditions. Only recombinant clones harboring the targeted gene and producing the corresponding gene product in an active form are able to grow. In this way, a high selectivity of the screen is achieved. One recent example is the identification of DNA polymerase-encoding genes from metagenomic libraries derived from microbial communities present in glacier ice (Simon et al. 2009). An E. coli mutant, which carries a cold-sensitive lethal mutation in the 5′-3′ exonuclease domain of the DNA polymerase I, was employed as host for the metagenomic libraries. At a growth temperature of 20 °C only recombinant E. coli strains complemented by a gene conferring DNA polymerase-activity are able to grow. In this way, novel genes encoding DNA polymerases were recovered and almost no false positive clones were obtained (Simon et al. 2009). Further examples for this screening approach are the detection of genes encoding Na+/H+ antiporters (Majernik et al. 2001), antibiotic resistance (Riesenfeld et al. 2004a), enzymes involved in poly-3-hydroxybutyrate metabolism (Wang et al. 2006), and lysine racemases (Chen et al. 2009).

The third function-driven approach is based on induced gene expression. Uchiyama et al. (2005) introduced a substrate-induced gene expression screening system (SIGEX) for the identification of novel catabolic genes. An operon-trap expression vector, which contains the gene for a promoterless green fluorescent protein (gfp), was employed for cloning of environmental DNA. Catabolic operons are often adjacent to cognate transcriptional regulators and promoters that are induced by the substrate. If expression of a target gene is induced by the substrate, the gfp gene is coexpressed, and positive clones can rapidly be separated from other clones by fluorescent-activated cell sorting (Handelsman 2005; Uchiyama et al. 2005). This method was validated by the screening of a metagenomic library derived from groundwater microbial flora. Regulated by the induction substrates benzoate and naphthalene 58 and 4 positive clones, respectively, were identified. The major drawback of this high-throughput screening approach is the possible activation of transcriptional regulators by other effectors than the specific substrates. This may lead to the recovery of false-positives (Galvao et al. 2005). A similar screening strategy termed metabolite-regulated expression has been published by Williamson et al. (2005). In contrast to SIGEX, metagenomic clones producing small molecules are identified. A biosensor that detects small diffusible signal molecules, which induce quorum sensing, is inside the same cell as the vector harboring a metagenomic DNA fragment. When a threshold concentration of the signal molecule is exceeded, green fluorescent protein is produced. Subsequently, positive fluorescent clones are identified by fluorescence microscopy (Williamson et al. 2005).

Metagenomics of extreme environments with low microbial community size

Physicochemical extreme environments such as ice (Simon et al. 2009), highly polluted environments (Abulencia et al. 2006), or deep hypersaline anoxic basins (van der Wielen et al. 2005) contain a low microbial community size. These habitats represent a widely unexplored ecological niche with a vast potential of novel biocatalysts of industrial use (Abulencia et al. 2006; Sjöling and Cowan 2008). Microbes that are capable of living in these hostile environments have evolved special mechanisms for survival. Due to the low community size and biomass of these ecosystems, these habitats are not as easily accessible as other environments by metagenomic approaches (Ferrer et al. 2009). The major challenge is to extract a sufficient amount of high-quality DNA. To overcome this limitation, whole genomic amplification of environmental DNA using the φ29 polymerase can be applied. In this way, high-throughput metagenomic approaches from small quantities of DNA as starting material are feasible. Drawbacks of whole genome amplification are the formation of chimeric artifacts and amplification bias, which is a result of template inaccessibility or low priming efficiency (Abulencia et al. 2006). Nevertheless, this approach has been successfully employed in several metagenomic studies of different environments, including contaminated sediments (Abulencia et al. 2006), the Soudan mine (Edwards et al. 2006), scleratinian corals (Yokouchi et al. 2006), the marine viral metagenomes of four oceanic regions (Angly et al. 2006), and glacier ice (Simon et al. 2009).

Conclusions

Metagenomics is an important and indispensable tool for the identification of novel biomolecules and analysis of the genetic diversity and metabolic potential of microbial communities. New and efficient high-throughput screening techniques have been developed, which facilitated the recovery of a high amount of new biocatalysts and small molecules. One of the main hurdles with respect to bioprospecting is the limited production of active recombinant proteins in heterologous hosts. Progress in metagenomic sequence analysis has been driven by the development of next-generation sequencing technologies, which permit cloning-independent and low-cost sequencing analyses of metagenomes. The rapid development of high-throughput DNA sequencing technologies and the corresponding increase in large and complex environmental require permanent development of appropriate bioinformatic tools for their analysis. A combination of metagenomics, metatranscriptomics, and metaproteomics is necessary for a comprehensive understanding of complex microbial communities. In this way, the structure and function of microbial communities in complex environments can be unraveled, and the monitoring of in situ responses and activities of microbes on an ecosystem level is feasible.