344 volume 27 number 4 april 2009 nature biotechnology

to create similar libraries for other organisms. Indeed, in light of recent advances in bacte-rial chemical genetics, adapting this approach to Escherichia coli would be timely and highly beneficial to antibacterial discovery.

1. Hoon, S. et al. Nat. Chem. Biol. 4, 498–506 (2008).2. Ho, C.H. et al. Nat. Biotechnol. 27, 369–377

(2009).3. Broach, J.R. et al. Genes 8, 121–133 (1979).4. Heitman, J. et al. Science 253, 905–909 (1991).5. Butcher, R.A. et al. Nat. Chem. Biol. 2, 103–109

(2006).6. Parsons, A.B. et al. Cell 126, 611–625 (2006).7. Lum, P.Y. et al. Cell 116, 121–137 (2004).8. Giaever, G. et al. Proc. Natl. Acad. Sci. USA 101,

793–798 (2004). 9. Xu, D. et al. PLoS Pathog. 3, e92 (2007).10. Jiang, B. et al. Chem. Biol. 15, 363–374 (2008).

pump or are otherwise hypersensitized. The amount of drug required for cloning using MoBY-ORF could be reduced in these back-grounds, and/or by using smaller culture volumes, making the method even more attractive. Moreover, as each plasmid con-tains mating-assisted genetically integrated cloning (MAGIC) recombination sites that flank the barcoded gene, the entire MoBY-ORF library can be readily switched to another vector, such as a high-copy vector, which could enable systematic analysis of the gene dosage effect on drug susceptibility in a format similar to that used by Ho et al.2. An intriguing extension of the method would be

previous approach based on a high-copy sup-pression strategy yielded multiple genes that confer drug resistance when their copy num-ber is increased, making it difficult to identify the correct drug targets without prior knowl-edge5. Ho et al.2 also identify the gene mutated in a yeast isolate that is resistant to theopalau-amide, a novel bicyclic peptide with antifungal and other activities6. The resistance mutation enables them to characterize theopalauamide’s mechanism of action and to determine that it is a member of a novel class of sterol-bind-ing antifungal compounds, a property not detected by genomic profiling experiments using the yeast deletion strains6.

Complementation cloning with MoBY-ORF dovetails nicely with other genomic technologies. Ho et al.2 show that the muta-tions they identify are corroborated by data from yeast tiling microarrays, suggesting that by combining the two technologies, bona fide drug-resistant mutations can be resolved with great precision and confidence. Another chemical-genomics method, haploinsuf-ficiency profiling (also known as the fitness test), has been shown to have pharmaceuti-cal utility in both baker’s yeast7,8 and Candida albicans9, an opportunistic pathogen that is the leading cause of fungal infections in hos-pitals worldwide. In haploinsufficiency profil-ing, an antiproliferative compound is tested on a pool of barcoded yeast strains that are each heterozygous for a single gene deletion. Usually, only a few strains display significant growth deviations from the population, and these are likely to contain a heterozygous deletion of a gene that encodes a target of the compound. Haploinsufficiency profiling pro-vides insights into the mechanisms of action of antiproliferative compounds with minimal quantity of material, even in mixtures such as natural product extracts6,10. However, some drugs, such as amphotericin B, do not gener-ate informative haploinsufficiency profiles8, and others, like theopalauamide, yield profiles that are difficult to decipher6. Characterization of drug-resistant mutations using the MoBY-ORF library provides a complementary strategy when haploinsufficiency profiling is inadequate. By expending resources upfront on such chemical-genomics studies, research-ers are empowered to make informed decisions on subsequent discovery efforts.

The MoBY-ORF library and method will have diverse applications in both drug dis-covery and basic research. The library is a highly portable biological resource, in con-trast to the deletion strains. It can be applied in different genetic backgrounds commonly used in drug discovery, including in strains that have a mutator, a defective drug efflux

Getting to the core of the gut microbiomeMatthias H Tschöp, Philip Hugenholtz & Christopher L Karp

Metagenomic analysis of gastrointestinal bacteria sheds light on obesity.

Many may be surprised to learn that the bac-terial cells in and on their bodies outnumber the human cells by about a factor of ten. As metagenomics research extends its reach to the microbial populations colonizing the gut, skin and other tissues, it is beginning to generate results that may be of interest from a therapeutic perspective. A case in point is a recent study in Nature by Gordon and colleagues1, which exam-ined the role of gut bacteria in the development of obesity.

Obesity has reached epidemic proportions in Westernized cultures, and the diseases associ-ated with it—including insulin resistance and type II diabetes mellitus, hepatic steatosis and steatohepatitis, dyslipidemia and atherosclerotic cardiovascular disease—have become major public health problems. Traditional obesity research has focused on environmental and host genetic factors, including descriptive studies in

humans and mechanistic studies using geneti-cally altered rodent models. More recently, research over the last five years using culture-independent methods and high-throughput sequencing has suggested that the pathogenesis of obesity may be influenced by our endoge-nous gastrointestinal microbiota2–5.

The Gordon group first reported that germ-free mice gained weight after colonization with bacteria from the lower gut of conventionally raised mice2. They also showed that the devel-opment of obesity in leptin-deficient mice coincided with broad, phylum-level changes in the gut community structure, with obese mice having a reduced fraction of bacteria belonging to the Bacteroidetes phylum and a proportional increase in Firmicutes bacteria6 (although it should be noted that these patterns were appar-ent only in data averaged over many specimens and could be poor indicators for individual mice). Mechanistically, such microbiota have been implicated in both increasing energy-har-vesting efficiency and altering host genes that regulate fat storage.

Subsequent publications demonstrated that obese rodents and humans have a significantly lower percentage of Bacteroidetes in their gut microbiome compared with their lean coun-terparts and that weight loss by obese indi-viduals corrected this disproportion (again observed in averaged data). It was also shown that the microbiome associated with obesity3 is more efficient at harvesting dietary energy4,

Mathias H. Tschöp is at the Obesity Research Center and Genome Research Institute, Depts. of Psychiatry & Medicine, University of Cincinnati College of Medicine, Cincinnati, Ohio, USA; Philip Hugenholtz is at DOE Joint Genome Institute, Walnut Creek, California, USA; and Christopher L. Karp is at the Division of Molecular Immunology, Cincinnati Children’s Hospital Research Foundation and the University of Cincinnati College of Medicine, Cincinnati, Ohio, USA. e-mail: tschoemh@ucmail.uc.edu

NE wS AND v IE wS©

2009

Nat

ure

Am

eric

a, In

c. A

ll ri

gh

ts r

eser

ved

.

nature biotechnology volume 27 number 4 april 2009 345

and, most importantly, that transfer of gut microbiota from obese mice to germ-free mice caused more weight gain than similar transfers from lean mice4.

The new study by Gordon and colleagues1 provides further insight. The authors used pyrosequencing of 16S rRNA genes to deeply interrogate community structure (phylo-type abundance distribution) and randomly sequenced fecal microbial communities (metagenomics) to obtain functional profiles (Fig. 1a). The individuals studied were obese and nonobese adult human twins, and their mothers. Family members had considerable overlap in their gut microbial communities, with the degree of variation being similar between monozygotic and dizygotic twin pairs. Notably, however, obese individuals showed an impressive overall reduction in microbial diver-sity. The authors liken this reduced diversity to a fertilizer runoff, in which a subset of the

microbial community blooms in response to abnormally high energy input, as opposed to the rainforest- or reef-like community of the lean gut, which displays high species diversity in the face of high energy flux.

Most notably, Gordon and colleagues1 identi-fied a core gut microbiome in unrelated indi-viduals (both lean and obese), which occurs at the level of shared bacterial gene families but not at the level of shared abundant bacterial phylotypes (>0.5% of the community). This suggests the presence of guilds—that is, sets of microbial species sharing a common ecological niche—and implies that different combinations of species could fulfill the same functional roles required by the host (Fig. 1a). An interesting follow-up question is whether the lack of shared phylotypes is still as pronounced if rare popu-lations are taken into account. It may be that each individual harbors many low abundance ‘fail-safe’ populations, in common with other

individuals, for critical functional roles in case the dominant population filling that niche col-lapses. Such a collapse could be caused by phage predation, allowing a rare population to bloom and substitute for the loss (monocultures by contrast lack this type of functional redundancy and are vulnerable to catastrophic collapse). The Gordon study hints at this possibility as the rela-tive abundance of phylotypes (measured at the phylum level only) within an individual varied between two sampling times even though the specific phylotypes found remained consistent.

Consistent with the idea that the gut micro-biota influence the efficiency of energy harvest from the diet, the core microbiome was enriched in genes involved in fat, sugar and protein metabolism. The majority of the potentially relevant genes enriched in obesity was derived from Actinobacteria (75%) and Firmicutes (25%), whereas genes enriched in lean individuals were predominantly from

Caloricintake

Gut bacterial samplefrom obese individuals

Organism-levelprofiling:

16S rRNA genesequencing

Gene-levelprofiling:

metagenomicsequencing

Comparative analysisof sequence data

Core microbiome

Nutrientcomposition

Antibiotics

Geneticinfluence

Bacterialexposure

GI surgery

Neuroendocrineappetite regulation

GI peptidehormones

Gut glucoseproduction

GI nutrient digestion/absorption

Gut lipidproduction

Inflammatorysignals

b

a

Gut bacterial samplefrom lean individuals

Figure 1 The core gut microbiome and energy metabolism. (a) The microbial communities of fecal material from lean and obese individuals were profiled at the organism and gene level by 16S rRNA gene and random shotgun sequencing (metagenomics), respectively. After comparative analysis of the data, a key finding was that all individuals share a core microbiome at the level of gene families but not at the level of organisms (abundant populations). This suggests guild structure in the gut where different functionally-related species (different colored pigeons in the cartoon) can occupy the same niche (same pigeon hole location in the cartoon) in different individuals. The central and lower-right pigeon holes denote nonessential functions in the human gut and hence are empty in this schematic. (b) Knowledge of the gut microbiome adds to our understanding of energy balance regulation and the pathophysiology of obesity. Recent studies suggest that gut bacteria modulate the digestibility and absorbability of ingested nutrients, thereby increasing energy-harvesting efficiency. Multiple factors affect the composition and function of the gut microbiome, including dietary, antibiotic, surgical and genetic influences. Gut microbiota also contribute to host metabolism and energy homeostasis in ways beyond gastrointestinal nutrient processing, including modulation of gut-derived endocrine, neuronal and inflammatory signals.

Kat

ie v

icar

i

NE wS AND v IE wS©

2009

Nat

ure

Am

eric

a, In

c. A

ll ri

gh

ts r

eser

ved

.

346 volume 27 number 4 april 2009 nature biotechnology

and obese gut microbiota gene pools. As the cost of sequencing continues to fall, the next decade should be a productive one for gastrointestinal biology in which micro-bial-community analysis will be expanded to thousands of individuals and along the complete length of the gastrointestinal tract, and fully integrated with other analytical approaches.

1. Turnbaugh, P.J. et al. Nature 457, 480–484 (2009).2. Bäckhed, F. et al. Proc. Natl. Acad. Sci. USA 101,

15718–15723 (2004).3. Ley, R.E., Turnbaugh, P.J., Klein, S. & Gordon, J.I.

Nature 444, 1022–1023 (2006).4. Turnbaugh, P.J. et al. Nature 444, 1027–1031

(2006).5. Zhang, H. et al. Proc. Natl. Acad. Sci. USA 106, 2365–

2370 (2009).6. Ley, R.E. et al. Proc. Natl. Acad. Sci. USA 102, 11070–

11075 (2005).7. Samuel, B.S. & Gordon, J.I. Proc. Natl. Acad. Sci. USA

103, 10011–10016 (2006).8. Stunkard, A.J., Foch, T.T. & Hrubec, Z. J. Am. Med.

Assoc. 256, 51–54 (1986).9. Troy, S. et al. Cell Metab. 8, 201–211 (2008).10. Gillum, M.P. et al. Cell 135, 813–824 (2008).11. Cani, P.D. et al. Diabetes 56, 1761–1772 (2007).12. Samuel, B.S. et al. Proc. Natl. Acad. Sci. USA 105,

16767–16772 (2008).

To test the mechanistic hypotheses arising from the results of Gordon and colleagues1 and to develop preventive and therapeutic strategies, we will need to go beyond analysis of the gut microbiome to interdisciplinary collaborations among genome researchers, microbial ecologists, immunologists and obe-sity experts. An impressive example of such teamwork, also out of the Gordon group12, who identified short-chain fatty acid prod-ucts of microbial polysaccharide fermenta-tion as ligands of Gpr41, a gut epithelial cell G protein–coupled receptor that regulates pep-tide hormone secretion, and went on to show that conventionally raised and gnotobiotic (but not germ free) Gpr41−/− mice showed a leaner phenotype compared with similarly raised Gpr41+/+ controls. These results high-light a potentially relevant mechanistic con-nection between gut microbial function and endogenous molecular pathways controlling energy balance.

Interesting as the new molecular data are, they represent only a dip into the lean

Bacteroidetes (42%). Specifically, the authors identified almost 400 genes of metabolic path-ways that were enriched or depleted in the gut microbiome of obese individuals compared with lean controls, with an interesting example being enrichment of phosphotransferase sys-tems responsible for microbial processing of carbohydrates. Despite the apparent absence of a core microbial community in the human gastrointestinal tract, another recent study did find enrichment of particular microbial groups in the gut of obese individuals: the H2-producing Prevotellaceae (ironically, a family belonging to the Bacteroidetes) and the H2-using Methanobacteriales (an order of methanogenic archaea)5. Methanogens increase the extraction of energy by the host from otherwise indigestible polysaccharides7.

The finding by Gordon and colleagues1 of gene-level core microbiomes may have impor-tant implications. With regard to the Human Microbiome Project, it raises the question of whether community profiling using 16S rRNA genes should be used to select samples for meta-genomic sequencing as unifying functional pat-terns may be missed in samples with variable community profiles. It is also possible that drug targets or drug candidates for the treatment of obesity could be identified from the obesity-associated microbiome.

The sudden obesity epidemic is likely to be the result of changes upstream of the human gastrointestinal microbiome (Fig. 1b). The role of host genetics, for example, is underscored by the fact that monozygotic twins exhibit a higher degree of co-variation in body adipos-ity compared with dizygotic twins8, despite the similar microbial community structures of monozygotic and dizygotic twins demon-strated by Gordon and colleagues1. Future progress toward understanding the role of the gut microbiome in body-fat regulation should include collaborative approaches aimed at linking microbial genetic studies with established models of molecular body-weight regulation. Different gut microbiota may well have differential impacts on (i) afferent gastrointestinal peptide hormones that regulate appetite, energy metabolism and body weight; (ii) portal vein–sensed gut glucose production9; and (iii) gut-derived lipid signals to the brain10. Changes in the microbiome may also affect what has been termed “metabolic endotoxemia”—increases in plasma lipopolysaccharide (and, presum-ably, bacterial lipopeptide) concentrations in mice and humans exposed to high-fat diets11. In this context, it should be noted that mice with genetic deletions in a variety of pro-inflammatory mediators exhibit exacerbated diet-induced obesity.

Missing lincs in the transcriptomeThomas Gingeras

Are long, intervening noncoding (linc) RNAs a new class of functional transcripts?

Ask scientists at an RNA meeting whether most of the recently observed non-protein-coding RNAs are functional, and the group is apt to be divided. Many would acknowledge that these RNAs form the bulk of transcrip-tomes and that they are characterized by remarkable complexity, but skeptics would cite their lack of sequence conservation and the meager results from functional studies using forward and reverse genetics1. A recent report by Guttman et al.2 helps to shed new light on this debate by identifying a large number of non-protein-coding RNAs that are enriched in evolutionarily conserved sequences and map to intergenic regions.

The complexity of non-protein-coding RNAs is evidenced by their broad incor-poration of genome sequences, their inter-leaved organization and the variety of types detected, including long and short polyade-nylated and nonpolyadenlyated RNAs3–5.

Until recently, however, the extent of this complexity in cells from yeast6 to humans7 has been underappreciated.

Interestingly, Guttman et al.2 began not with experimental discovery of novel RNAs but with a computational approach in which they searched outside of protein-coding sequences for chromatin signatures of actively transcribed genes (Fig. 1a). This chromatin signature was defined as tracts of trimethy-lated lysine 4 of histone H3 (indicative of tran-scriptional initiation at promoters) adjacent to tracts of trimethylated lysine 36 of histone H3 (indicative of elongation of transcribed regions). From an analysis of four mouse cell types, the authors identified 1,250 unanno-tated intergenic regions at least 5 kb in size.

Subsequent screening of these regions for evolutionarily conserved sequences provided a prioritized list of candidate non-protein-coding RNAs. Expression levels for 350 of these were measured using custom tiling arrays, northern hybridizations and RT-PCR (Fig. 1b). Similar to what was previously reported for intergenic non-protein-coding

Thomas Gingeras is at Cold Spring Harbor Laboratory, Cold Spring Harbor, New York, USA. e-mail: gingeras@cshl.edu

NE wS AND v IE wS©

2009

Nat

ure

Am

eric

a, In

c. A

ll ri

gh

ts r

eser

ved

.