The role of host and gut microbiome metabolism in human disease

This PhD thesis consists of three scientific studies (I-III) and a methodological study (IV). The
scientific studies are examples of how metabolomics and microbiomics data can be used to
explore disease phenotypes in clinical cohorts in collaboration with clinical experts. Each of
these studies focused on a separate disease, but all with the aim of exploring the intersection
between host metabolism and disease. The methodological study describes the setup and use
of a data quality framework. This framework was implemented at the data collection stage of a
large observational cohort study to create less biased and error-prone data for future research.
All studies are an example of bringing in bioinformatics expertise early on in studies and
bridging the gap between data science and clinical area experts.
Haematopoietic stem cell transplant recipients are at high risk of cytomegalovirus (CMV)
infection following the transplantation. In Study I we assessed twelve previously identified
metabolites for association with CMV infection and found only trimethylamine N-oxide (TMAO)
to be associated, although with a correlation in the opposite direction than what was previously
found. Using data-driven network analysis of metabolomics data, a group of metabolites
(n=44), including TMAO, which showed an inverse association with CMV infection was
identified. TMAO is a gut microbiome-derived metabolite affected by diet and therefore it may
be beneficial to include microbiome data in future studies of CMV risk in HSCT populations.
Patients hospitalised with influenza or SARS-CoV-2 infection display a broad range of symptoms
where some progress to severe respiratory distress and others can be managed with room
oxygen. In Study II we investigated association of tryptophan, kynurenine and the
kynurenine/tryptophan ratio (representative of the indoleamine-2,3-dioxygenase mediated
depletion of tryptophan pathway). Here we found higher levels of kynurenine and the
kynurenine/tryptophan ratio associated with disease progression, and no associations of
tryptophan. In network analyses of metabolomics data, we identified two groups of metabolites
(n=51 and n=345) positively correlated with disease progression as well as death, higher age,
and chronic kidney disease. The larger group contained kynurenine and other downstream
metabolites of the tryptophan pathway. The metabolites found in these two clusters should be
further investigated to ascertain their relationship with respiratory disease progression. Further,
kynurenine and the kynurenine/tryptophan ratio should be investigated as clinical biomarkers
for disease progression.
Higher levels of cholesterol in blood is a risk factor for cardiovascular disease. A recent study
has identified the bacterial intestinal sterol metabolism A (ismA) gene family which is
responsible for the first and last step of the conversion of cholesterol to coprostanol,
constituting a way for cholesterol to be removed from circulation. In Study III we investigated
the variance and abundance of ismA genes in faecal samples collected from an hospital patient
cohort. By searching for the gene in metagenome-assembled genomes (MAGs), we found 270
ismA variants in genomes taxonomically classified as close relatives to Eubacterium
coprostanoligenes or Fimenecus lineages. Further, we found the presence of an ismA variant in
the microbiome to be associated with lower mean levels of cholesterol compared to those
without the gene. These results present a potential way for modulating circulating lipids
through the gut microbiome and expand on the current knowledge of ismA carrying bacteria.
The collection of high-quality clinical and demographic data is crucial for downstream data
analyses. In Study IV we created and implemented a data quality framework for the collection
of questionnaire data and sample tracking in a large, international, observational, multi-centre
cohort study. The framework was set up as an interactive HTML report which could be used to
get an overview of the collected data, assess typing errors and discrepancies between different
data sources, and keep track of samples – all in real time. Using the data quality framework, we
minimised errors and missingness in the data which will be used for research, while saving time
and limiting manual work.