Metagenomics – a current snapshot and future forecast

July 22, 2020

Metagenomics is the term for applying untargeted next-generation sequencing (NGS) to specimens – sometimes from unlikely sources, such as swabs of cell phones or bus seats, as well as more traditional clinical sample types – with a goal to identify and enumerate to a species level all of the different organisms present.

This space has looked at this topic before in its use as a targeted means to probe for novel pathogen(s) in association with conditions of unknown etiology, but the uses of metagenomics go beyond this and are in some cases informing a much deeper understanding of previously unguessed human-microorganism interactions. In this episode of The Primer, we will touch on what some of these are and where this may lead in the future.

Metagenomics – the method

First, a very brief refresher on how the process works. A sample is obtained; DNA is extracted; depending on platform, some form of untargeted library preparation is done; NGS is performed on the library; and data is sent to a bioinformatics pipeline. Simplistically, this usually involves removal of all human-derived sequences, and then testing each remaining read against databases, such as Genbank, for identity with, well, anything non-human. Depending on sample type, this can include all sorts of environmental contact DNA (which can be interesting and likely has uses in applications such as forensics), but for more mundane clinical sample types, the majority of non-human DNA of interest is microbial.

Each read is either unassignable (no definitive homology found), or else is tallied as coming from a particular organism. Organisms that are high abundance in the sample tally up more “hits” than ones in low abundance (loosely; genome size has an impact here, too, but can be corrected). The end is a list of what organisms – and on a relative ratio, how much of each – were present in the input material. In a microbial context, this concept is referred to as the microbiome of the sample.

Technical variations sidebar

Metagenomics as described above, based on “shotgun sequencing” (random sampling and sequencing of all nucleic acids) by its nature, can provide the most widespread and comprehensive image of the microbiome, including bacteria, fungi, and (DNA) viral components (bear in mind this, for example, could include things such as bacteriophages, which can indirectly impact the human source, by modifying bacterial virulence).

If only bacteria, strictly, are of interest, then degenerate primers to a commonly shared but variable region, such as the 16S rRNA gene, can be used to PCR amplify a presumably not too badly biased library of fragments suitable for analysis. While narrower in focus, aspects of library preparation may be easier in this approach and truly miniscule sample inputs are required. In general, the applications of either method are similar, and we will not distinguish between these approaches further.

Gut feelings – more than just a term?

One of the most fascinating applications of metagenomics has been in the analysis of gut microbiomes, and what it has turned up. It’s easy to imagine how this might impact issues such as food tolerance, as microbial activity can produce secondary metabolites, which can be absorbed by the host and have biological activities (positive or negative, and with that distinction likely having some variation from host genetic factors). What might come as more of a surprise, however, is the now-overwhelming evidence that gut microbiome can have significant implications for neurological functions – what is known as the “gut-brain axis.” The list of neurological conditions which have demonstrated potentially causal correlates to variation in gut microbiota include Parkinson’s Disease, schizophrenia, autism spectrum disorders (ASD), and anxiety.

Some of this may be attributable to direct microbial impacts on serotonin metabolism. An important neurotransmitter, serotonin (aka 5-hydroxytryptamine or 5-HT) is primarily known for its induction of feelings of well-being or happiness. Perhaps unexpectedly, the majority of serotonin (estimated at ~95 percent) is produced by neurons in the gut, not the brain, and their synthetic activity for this compound is influenced by microbiome composition. Intriguingly, mouse models of depression have shown positive responses to increased Lactobacillus levels in the gut.

That this might translate to a basis for treatment of mood disorders in humans has not gone unnoticed, with some trials having been made with both prebiotic dietary supplements (food components which can bias gut microbiome composition by selectively supporting growth of some microbial types over others) and probiotics (basically, edible cultures of actively reproducing, presumed beneficial intestinal flora such as Lactobacillus and Bifidobacteria species), which can help to increase the number of these species in the microbiota.

Results of these trials are not completely clear-cut as multiple conditions (and perhaps even multiple underlying etiologies within those) have been examined, but at least some (e.g. generalized depression, but not schizophrenia) have shown statistically relevant improvements above both baseline or traditional antidepressive drug therapy.

A second mechanism whereby abnormal intestinal microbiota composition (dysbiosis) can influence neurological condition is by triggering chronic inflammatory responses. Cytokines produced locally in response to this circulate systemically, and either directly cross the blood-brain barrier, cross in the form of activated and secreting circulating immune cells, or otherwise indirectly influence inflammatory responses in the central nervous system (CNS).

Evidence suggests this can have adverse effects on memory and cognitive abilities, as well as being linked with specific diseases. For example, current hypotheses suggest that gut dysbiosis may be a key triggering event in Parkinson’s disease, and there is evidence that routine use of non-steroidal anti-inflammatory agents (NSAIDs such as acetylsalicylic acid “Aspirin” or ibuprofen) correlates with reduced progression of Alzheimer’s disease.

On the surface

Another major site for human/microbiome interaction should come as no surprise – the skin. What is perhaps less expected though is that the skin microbiome differs across the body (for a fascinating figure depicting this, and a broader discussion in context, see [1]). Variations between individuals is also quite broad, although Staphylococcus, Propionibacterium, and Corynebacterium species generally make up the major constituents of most locations on most individuals; variability is primarily observed in distribution of less-common constituent organisms. Aspects including geographic locale, occupation and related environmental exposure, age, and perhaps even time of year might be expected to also play a role in this. Many of these organisms may be considered commensal, as they either act to out-compete and displace more potentially harmful microorganisms or directly influence the chemical microenvironment of skin in ways beneficial to the human host.

From a dermatological disease perspective, P. acnes subspecies present (but not total number) are associated with severe acne; increased number and subspecies diversity of Staphylococcus has been found in atopic dermatitis; and dysbiosis (although exactly of what form, or whether it’s related to a single pattern of dysbiosis or just dysbiosis in general is unclear) appears to be a component of psoriasis. More broadly, there is evidence for bidirectional interaction between the human immune system and skin microbiota, with immune system deficits shown linked to alternate and more diverse microbiota profiles. Conversely, skin microbiota, either directly or through the activity of released metabolites, have been found to modulate inflammation and immunity in ways thought to be beneficial.

As with all correlations, drawing assumptions of causal links in absence of Koch’s Postulate being fulfilled is unwise, but at least worthy of consideration as a possibility in lack of contradictory evidence. With this in mind, it’s notable that as with gut microbiomes above, intentional manipulation of skin microbiome through application of both pre- and probiotics continues to be examined as a means to treat various conditions. More exotic applications for manipulation of skin microbiota, such as doing so with an aim to repel insect disease vectors, have also been postulated.

Back to diagnostics

So where does all of this impact the clinical molecular lab? Right now, probably only on the bleeding edge of research. As such research starts to produce supported correlates of microbiome composition to prophylaxis, diagnoses, or actionable medical interventions, however, it will likely start to become more common and perhaps even part of a routine clinical workup in some settings. That such methods can have clinical utility has long ago been shown in a context predating NGS: analysis of airway microbial composition in cystic fibrosis (CF) patients by traditional culture has long been a part of routine care. Supplanting culture with NGS just allows for a vastly more detailed picture to emerge than that available by culture.

The likelihood of increased clinical future application of metagenomics is supported by the technical side considerations that microbiome metagenomics can be done very cheaply in comparison to other NGS applications. Massive read depth is absolutely not needed in a microbiome – as opposed to specific pathogen detection context – because it’s unlikely a species present at 0.01 percent of the population is very important. Samples from multiple patients can be readily multiplexed on a single instrument run; during each library preparation, a sample-specific “barcode” in the form of a short unique DNA sequence is appended to each captured random-template DNA.

Samples on a shared instrument run spread out the reagents, consumables, and labor costs, and may be demultiplexed bioinformatically post-run (reads are grouped by shared barcode). Subsequent bioinformatics workflow – identification and classification of microbiome constituents – is about as close to turnkey for the non-bioinformatics specialist as is possible, making meaningful downstream analysis potentially accessible to a wide range of laboratorians and clinicians. Finally, these processes are readily viable across all NGS platforms of either long- or short-read technology; whatever instrument(s) are available to your lab can be put to use for this. A combination of utility plus low barriers to implementation bodes well for increased use in the near future.

What such application would look like remains to be seen, but in broad strokes, it would likely consist of a surveillance microbiome taken from a relevant sample; a comparison of its composition in terms of relative microbial species, abundance, and diversity against a composite “normal” microbiome for the sample, based on population studies; and interpretation of any observed dysbiosis based, again, on preexisting associative study data.


  1. Grice EA. The skin microbiome: potential for novel diagnostic and therapeutic approaches to cutaneous disease. Semin Cutan Med Surg. 2014;33(2):98-103. doi:10.12788/j.sder.0087