I’ve been thinking recently about how we (the scientific community) are going to translate our knowledge of the human microbiome into practical therapies that can cure important diseases. This question used to be entirely hypothetical, but now we can see some examples of microbiome-based therapeutics which have strong evidential support and could plausibly make it into the clinic. Instead of presenting a thoroughly cited review, I’m going to keep the discussion very high-level because I want to focus on the concept of one particular experimental approach which I think is crucially important — the microbiome survey.
In experiment after experiment, in lab after lab, we can see over and over that the biggest differences between people’s microbiome are at the level of the microbial strain. Broad groups of bacteria may boom or bust depending on what you eat, but the differences that stick around for months and years are the specific *strain* of a given species which is present. In other words, I have E. coli in my gut, and so do you, but my E. coli is different from yours and will stick around inside me for a long time. Moreover, the differences between strains can have a huge impact on human health. We like to think of a bacterial species as being a cohesive type, but in fact the diversity of strains within a species can encompass a range of phenotypes for things like bile acid metabolism, polysaccharide utilization, nitrogen capture, and many more pieces of biology that we haven’t begun to understand.
Find the Bacteria, find the Cure
The phrase “microbiome-based therapeutic” covers all sorts of possibilities, but the lowest hanging fruit for therapeutic development is still an intact bacterial cell. There is a chance that our microbiome studies will uncover a new biological mechanism for human disease which can be manipulated directly with small molecules or biologics, but that presupposes a high degree of biological understanding. As this field continues to emerge, it is far more likely that our first successes will be with mixtures of bacteria which are observed to have a specific effect, even if we don’t quite know why. This argument from first principles is also supported by where we see the most recent advances coming from across the field.
Too Many Fish in the Sea
I used to study the diversity and dynamics of phages (bacterial viruses) in the human gut, and the nice thing was that every phage you found was a new phage. Sure, there is at least one family of phages that seem to be found broadly across people (crAssphage), but we still find that each new isolate is unique and different. Unsurprisingly, this is also true of bacteria. But, you may ask, how different are they? Incredibly different. A newly discovered isolate of E. coli may only share 2,500 out of 3,000 genes with any other known isolate, with every single one of those new genes potentially having a huge impact on human health. When you multiply this across all of the people in the world, with all the different strains of bacteria in their gut, with the rapid evolution of wholly new strains over short periods of time, the total number is astronomical. Far too many for us to check each one individually. If we were looking for a new planet to live on, we wouldn’t travel to each star in turn until we found a good planet. Similarly, if we want to find a therapeutic bacteria we can’t just test all of the bacteria until we find one. It would take too long and cost too much.
Introducing the Microbiome Survey
There is an argument for microbiome surveys as being the best approach for finding therapeutic candidates in the human microbiome:
Everybody has a unique collection of strains in their gut microbiome.
Some people experience health conditions which are influenced by the bacteria in their gut.
We have a technology for measuring what bacteria are present in a single stool microbiome sample.
If we compare the stool microbiome of people with and without a disease, we can identify those strains which are most strongly associated with that disease.
Those strains which are most strongly associated with a human disease may have some therapeutic value for preventing or treating that disease.
I’ll be the first to note that this argument is not 100% airtight. Notably, we have no way of knowing whether what we can measure (3) has a strong biological link to therapeutic impact (5). However, I feel strongly that metagenomic sequencing, when used to detect and characterize the specific genes present in a given stool microbiome sample, is likely to be a good candidate for this approach. The other point to note is that the assertion in (4) has a strong caveat based on the complexity of the microbiome and the size of a given microbiome survey. Because the microbiome is incredibly complex and diverse (remember that most people have different strains, and that each person has thousands of strains), we correspondingly need to conduct surveys over the largest possible number of people. Nobody really knows what number of people provides the best cost-benefit tradeoff, but existing metagenomic microbiome surveys have been limited to hundreds, and in some few cases thousands, of people. Hopefully the upper limit of survey size will increase as costs come down, and as we generate more evidence for the claim that this is a fruitful avenue to pursue.
In Closing, Consider the Source
As we start to venture into a world in which microbiome-based therapeutics are coming to market, think hard about where these strains come from. If there is a cocktail of 10 bacteria which really helps with disease X, and they were found by screening 500 random isolates in vitro, then consider how much of an improvement we could make if we used a microbiome survey to pick the best possible set of 500 isolates for that screen. We all dream of a world where we find microbiome-based therapeutics for colorectal cancer, inflammatory bowel disease, obesity, and type I diabetes, and I think that microbiome surveys will help make that world a reality.