Scientific discovery from a clinical study: surprises from the lung and stomach microbiomes
Behind the paper: "Aerodigestive sampling reveals altered microbial exchange between lung, oropharyngeal, and gastric microbiomes in children with impaired swallow function." PLOS ONE, 2019. doi: 10.1371/journal.pone.0216453
This post is about our recent paper, "Aerodigestive sampling reveals altered microbial exchange between lung, oropharyngeal, and gastric microbiomes in children with impaired swallow function", published in PLOS ONE recently. In the course of analyzing samples from a clinical study, we discovered some pretty interesting things about the aerodigestive microbiome (lung, stomach, and oropharynx) that I hope will be of interest to the Nature Microbiology community.
This study was led by Dr. Rachel Rosen at Boston Children's Hospital. She collected bronchoalveolar lavage fluid (i.e. lung), oropharyngeal swab (behind the oral cavity and above the epiglottis), and gastric fluid samples from over 200 kids with a variety of aerodigestive symptoms (e.g. unexplained coughing or wheezing, swallowing dysfunction, gastroesophageal reflux, etc). This paper specifically focused on kids with and without impaired swallowing function (oropharyngeal dysphagia). The clinical literature explaining oropharyngeal dysphagia is pretty complicated, but for us microbiome scientists it's pretty simple: kids with oropharyngeal dysphagia can't swallow right, and so food or liquids often "go down the wrong tube." These children are also at higher risk for respiratory infections, and so we wanted to know whether the lung microbiome could provide some insight into why that is.
Specifically, we wanted to know whether changes in the lung microbiomes of these kids were more likely to be coming from the oropharynx or the stomach. There's a clear clinical motivation for this question: children with swallow dysfunction often undergo anti-reflux surgery such as Nissen fundoplication. Again, the clinical literature is complicated but the gist of the surgery is that the top of the stomach gets wrapped around the bottom of the esophagus to prevent stomach contents from coming up into the mouth and lungs. Kids with impaired swallow sometimes have this surgery even if they don't have obvious reflux symptoms because clinicians think that their eventual lung complications are caused from inhaling bacteria, food, or acid from their stomachs. So if we could show that aspirators have more 'stomach bacteria' in their lungs than non-aspirators, that would provide strong support for this clinical hypothesis and motivation for the anti-reflux surgery.
However, we actually found that changes in the lung microbiomes of aspirators are probably coming from aspiration of contents from the mouth rather than the stomach. Specifically, we showed that lung and oropharyngeal communities within aspirators were more similar to each other than these communities were in non-aspirators. In the paper, we also identified bacteria that we think might be directly exchanging between the oropharynx, stomach, and/or lungs. We showed that aspirators tended to have the lung-oropharynx bacteria in both their lungs and oropharynxes much more frequently than non-aspirators. We didn't see these similar trends when we compared the lung and stomach. Clinically, these findings are really exciting because they suggest that, unlike the fundoplication surgery, interventions to reduce aspiration-related respiratory microbial-related complications should target aspiration during swallowing rather than from the stomach.
There was also a lot of exciting microbiome science in this paper that I hope doesn't get lost in the clinical story. The dataset Dr. Rosen generated is really special: it's a really large cohort (about 200 patients) and includes very precious aerodigestive tract samples: lung and stomach samples. In fact, neither lung nor gastric samples were included in the Human Microbiome Project, which actually made this project really difficult since we know so little about the "normal" lung or gastric fluid microbiomes. Furthermore, we have samples from multiple sites in the aerodigestive tract from the same patients, so we can directly compare communities within and across patients.
When looking at these within-patient relationships, we saw that lung and gastric communities were more similar to each other within patients than lungs were to other people's lungs and stomachs were to other people's stomachs. This is exciting because it challenges the prevailing hypothesis that human-associated microbial communities are primarily driven by body habitat and instead suggest that patient-specific relationships may be equally, if not more, important in determining community structure in the aerodigestive microbiome. It also gets at a question that the lung microbiome community hasn't been able to agree on: does there exist a "core" lung microbiome, or are lung microbiomes just a result of stochastic environmental contributions? While we can't definitively lay this question to rest with our data, my hypothesis is that it's a bit more complicated than this binary: there probably isn't a "core" lung microbiome across people, but the lung microbiome is a biologically meaningful community resulting from more than just environmental contributions. Maybe, instead, it's that each person has a "core" aerodigestive microbiome, which is distributed across the different sites differently.
In retrospect, the road to this finding started right at the beginning of the project, from the question of whether aspirators have more "stomach bacteria" or "oropharyngeal bacteria" in their lungs than non-aspirators. When I originally tried to define "stomach" or "oropharyngeal" bacteria based on whether they were present in many different people above a certain abundance threshold, my results didn't make any sense - no matter what thresholds I used, the majority of the site-specific OTUs were also present in the other aerodigestive sites. I eventually realized that a better way to answer our question was to look at the overall similarities between microbial communities within each patient (i.e. beta diversities), which led to the main clinical finding in the paper. But when I zoomed out and just looked at the relationships between sites across the aerodigestive tract, regardless of aspiration status, something seemed strange. It looked like the oropharynx and stomach were as similar to each other as the stomach and the lungs. Now, I expected the oropharynx and stomach to be very similar because there's frequent, direct exchange of material between those two sites from swallowing. But the fact that the oropharynx and stomach were as similar as the *lungs* and stomach was quite surprising!
Then, we finally took one last step back and connected this with something else that had been confusing me about the data. It seemed like all the lung samples from different patients were super different from each other, and the stomach samples were also fairly different from other stomach samples. In one of our last read-throughs before submitting, my advisor was flipping through Figure 2, where I compared similarities between the same body site across different people, and Figure 3, where I compared similarities between different sites in the same person, and pointed out that it looked like lungs across people were more different than lungs and stomachs within people were. Whoa, weird!
As usual, he was right: I came up with the statistic to test this and showed that lungs and stomachs within people are more similar than lungs or stomachs across people. In our field, it's pretty well-accepted that body site is a primary driver in microbial community composition (Costello et al, Science 2009; HMP Consortium, Nature 2012; Lozupone et al, Genome Research 2013). It looks like in the case of the aerodigestive tract, especially in the lungs and stomach, that may not be true.
There's lots of reasons why this makes sense (both communities are highly selective, have low biomass and probably few actively replicating bacteria, and have bi-directional movement of material), and I'm excited to see more research looking into these interesting body sites, where the intuition that we've built from projects which did not include these sites may not apply!