Advancedadvanced biology

Metagenomics: Sequencing Microbial Communities Without Culturing

Learning objectives

  • Contrast amplicon (16S) and shotgun metagenomic sequencing in resolution, cost, and bias.
  • Explain metagenomic assembly, binning, and metagenome-assembled genomes (MAGs).
  • Describe functional profiling and how it complements taxonomic profiling.
  • Evaluate major findings about the human and environmental microbiomes.
  • Critique limitations: contamination, database bias, and correlation versus causation.

The uncultured majority

The great plate count anomaly -- that microscopy reveals far more microbes than ever grow in culture -- meant microbiology studied perhaps 1% of microbial life. Metagenomics bypasses culturing entirely by sequencing DNA directly from environmental samples. Two strategies dominate. Amplicon sequencing targets one marker gene, usually the 16S ribosomal RNA gene in bacteria and archaea (ITS in fungi), clustering reads into operational taxonomic units or exact amplicon sequence variants to profile who is there -- cheap and standardized, but limited to genus-level resolution and blind to function. Shotgun metagenomics sequences all DNA in the sample, capturing bacteria, archaea, viruses, and eukaryotes plus their genes -- higher resolution and functional insight at higher cost and analytical complexity. Norman Pace's pioneering 16S surveys of Yellowstone hot springs and marine picoplankton in the 1980s-90s first revealed entire uncultured phyla; the Human Microbiome Project and Earth Microbiome Project then scaled the approach globally.

From reads to genomes: assembly and binning

Shotgun reads are fragments; reconstructing genomes requires assembly -- overlapping reads into contigs -- complicated by strain-level diversity, repeats, and wildly uneven coverage across community members. Binning then groups contigs into metagenome-assembled genomes (MAGs) using two signals: similar coverage depth across samples (contigs from one genome rise and fall together) and similar tetranucleotide composition (genomic signatures). CheckM estimates completeness and contamination from single-copy marker genes; high-quality MAGs exceed 90% completeness with under 5% contamination. This pipeline has reconstructed tens of thousands of genomes for organisms never seen under a microscope, including the Asgard archaea -- the archaeal lineage most closely related to eukaryotes, whose discovery reshaped theories of eukaryogenesis. Long-read sequencing (PacBio, Nanopore) now yields complete circular MAGs, resolving plasmids and strain variation that short reads scramble.

Functional profiling: what the community can do

Taxonomy tells who is there; functional profiling tells what they can do. Reads or MAGs are annotated against databases (KEGG, UniRef, CAZy for carbohydrate enzymes), quantifying pathways for carbon fixation, nitrogen cycling, antibiotic resistance, and secondary metabolism. This revealed that functional profiles are often more conserved than taxonomy across healthy human guts -- different species assemblages performing similar jobs, functional redundancy buffering compositional change. In oceans, metagenomics mapped the global distribution of photosynthesis genes and nitrogen-cycling pathways, turning biogeochemistry into a genomic science. Viral metagenomics (viromics) uncovered vast unknown viral diversity -- most ocean viral sequences match nothing in databases, the famous viral dark matter. A key insight: gene presence does not equal activity, which is why metatranscriptomics (sequencing RNA) and metaproteomics increasingly complement DNA-based surveys.

What we have learned -- and what remains correlation

Metagenomics rewrote several textbooks. The human gut harbors roughly as many microbial cells as human cells (revising the old 10:1 myth), dominated by Bacteroidetes and Firmicutes, with diversity linked to health across dozens of conditions. Environmental surveys revealed that Candidate Phyla Radiation bacteria -- tiny genomes, symbiotic lifestyles -- may comprise a large fraction of bacterial diversity. Yet the field's central weakness is causal inference: nearly all human associations are correlational, confounded by diet, drugs, geography, and disease itself altering the microbiome. Fecal transplant cures recurrent C. difficile infection -- the causal gold standard -- but most claimed links (obesity, autism, depression) lack such proof. Germ-free mouse experiments help but translate imperfectly. The maturation of the field is visible in its standards: pre-registration, confounder control, strain-level resolution, and interventional trials are replacing the early era of underpowered association studies.

Pitfalls: contamination, databases, and dark matter

Metagenomics has characteristic failure modes every practitioner must know. Contamination -- from reagents (the kitome), lab workers, and cross-sample bleed -- plagues low-biomass samples like placenta or tumor microbiomes, where claimed findings have repeatedly collapsed under scrutiny; negative controls are non-negotiable. Database bias skews everything: reference databases overrepresent pathogens and model organisms, so novel lineages are misclassified or invisible, and functional annotation transfers functions by homology that may not hold. The taxonomic resolution limit means 16S cannot distinguish pathogenic from harmless E. coli strains. Batch effects -- DNA extraction kits, sequencing runs -- can outweigh biological signals. And statistical traps abound: microbiome data are compositional (relative abundances sum to 100%), so standard correlations produce spurious results without compositional methods. Honest metagenomics pairs powerful sequencing with skeptical analysis -- the sequence is easy; the inference is hard.

Correlation is not causation

Most microbiome-disease associations are correlational. Disease, diet, and drugs all reshape the microbiome, so arrows of causation run both ways. Interventional evidence -- transplants, defined consortia, germ-free models -- is the standard for causal claims.

Key points

  • Metagenomics sequences community DNA directly, bypassing the unculturability of most microbes.
  • 16S amplicon profiling is cheap but genus-level; shotgun sequencing gives strain resolution plus function.
  • Assembly plus binning reconstructs MAGs, including entirely uncultured lineages like Asgard archaea.
  • Functional profiles are often more conserved than taxonomy -- redundancy buffers compositional change.
  • Human microbiome associations are mostly correlational; causal proof requires interventions.
  • Contamination, database bias, and compositional statistics are the field's characteristic pitfalls.

Common misconceptions

✗ Myth: We have 10 times more bacterial cells than human cells.

✓ Truth: Revised estimates put the ratio near 1:1. The 10:1 figure was a decades-old guess; careful cell counting by Sender, Fuchs, and Milo (2016) showed roughly equal numbers, with microbes concentrated in the colon.

✗ Myth: Metagenomics identifies every microbe in a sample.

✓ Truth: Detection depends on sequencing depth, reference databases, and DNA extraction biases. Rare taxa are missed, novel lineages may be unclassifiable, and viruses and eukaryotes need tailored approaches.

✗ Myth: Finding a microbe enriched in disease means it causes the disease.

✓ Truth: Disease states alter the gut environment -- inflammation, drugs, diet changes -- reshaping the microbiome secondarily. Enrichment is a starting hypothesis demanding interventional testing, not a conclusion.

Frequently asked questions

What is a MAG and how is it validated?

A metagenome-assembled genome is a draft genome binned from environmental sequencing without culturing. CheckM estimates completeness and contamination using single-copy marker genes; high-quality drafts exceed 90% complete with under 5% contamination, and long reads now yield finished circular genomes.

When should I use 16S versus shotgun sequencing?

Use 16S for cheap, standardized bacterial community profiling across many samples when genus-level taxonomy suffices. Use shotgun when you need strain resolution, viruses and eukaryotes, functional genes, or genome reconstruction -- at roughly 5-10x the cost and analytical burden.

What are Asgard archaea and why do they matter?

An archaeal superphylum reconstructed entirely from metagenomes, encoding eukaryotic-like proteins for membrane trafficking and cytoskeleton. As the closest known archaeal relatives of eukaryotes, they transformed theories of eukaryogenesis toward an archaeal host engulfing the mitochondrial ancestor.

Why are low-biomass microbiome studies so controversial?

Because reagent and environmental contamination (the kitome) can dominate the signal when true microbial DNA is scarce. Several high-profile claims about placental or tumor microbiomes collapsed when proper negative controls revealed the signal was contamination.