Periodic Reporting for period 2 - EcoBox (Ecosystem in a box: Dissecting the dynamics of a defined microbial community in vitro)
Reporting period: 2020-07-01 to 2021-12-31
The overall goal of the EcoBox project is to improve our understanding of microbial community dynamics. Although we understand quite well the behavior of bacterial species growing in monoculture in a bioreactor, it is still challenging to predict what happens if several interacting species are grown together. Will they all survive or will some of them go extinct? Which bacteria will dominate the community – or will they all take an even share of the available resources? Will their abundances fluctuate wildly, or do they reach a stable state? If the latter, will it always be the same state in a constant environment, or will small variations in initial conditions lead to big differences? How will the community respond to perturbations, such as a change in pH or irregular feeding intervals? Can we predict what is going to happen in these different scenarios?
The answers to these research questions have important practical implications. In bioengineering, bacterial consortia are optimized to carry out specific tasks such as biodegradation or biofuel production. In the context of human health, the gut microbiome is known to be involved in the development of several diseases, such as inflammatory bowel disease and metabolic syndrome. The development of therapeutic consortia (next-generation probiotics) to treat or prevent these diseases is an active field of research. A better understanding of bacterial interactions and their impact on the community dynamics is an important step towards the rational design of such bacterial consortia.
In the EcoBox project, we plan to answer the fundamental questions raised above for an example community of human gut bacteria studied in vitro and in silico. These bacteria have the advantage that their genomes and main metabolic strategies are known, that their metabolism is comparatively well resolved and that they are relevant to human health.
The answers to these research questions have important practical implications. In bioengineering, bacterial consortia are optimized to carry out specific tasks such as biodegradation or biofuel production. In the context of human health, the gut microbiome is known to be involved in the development of several diseases, such as inflammatory bowel disease and metabolic syndrome. The development of therapeutic consortia (next-generation probiotics) to treat or prevent these diseases is an active field of research. A better understanding of bacterial interactions and their impact on the community dynamics is an important step towards the rational design of such bacterial consortia.
In the EcoBox project, we plan to answer the fundamental questions raised above for an example community of human gut bacteria studied in vitro and in silico. These bacteria have the advantage that their genomes and main metabolic strategies are known, that their metabolism is comparatively well resolved and that they are relevant to human health.
In the first half of the EcoBox project, we had to address three major experimental challenges: find a medium in which all selected species grow, set up parallel bioreactors for controlled high-throughput experiments and count bacteria efficiently in a species-specific manner. After testing several media, we chose Wilkins-Chalgren Anaerobe Broth, which is easy to make, does not contain major carbon sources that are hard to measure and supports the growth of all our selected species. Given the difficulties posed by the medium and the counting technique, we reduced our species number to five, selecting species that are highly abundant and prevalent in the human gut and that are suspected of being drivers in different gut community types (enterotypes). Next, we tested growing our bacteria in a parallel automated fermenter set-up with 24 vessels, where pH, temperature and atmosphere are controlled, and samples are taken automatically at defined intervals. We successfully implemented a chemostat through a constant feed of fresh medium and regular automated removal of old liquid from the vessels. Achieving this was crucial since a bacterial community cannot reach a stable state without the constant supply of fresh nutrients. Chemostats are often emulated through serial transfers, but these come with fluctuations in nutrient concentrations, which are less pronounced in our set-up. In addition, we can adjust the pH and flow rate, which are known to be important parameters for gut microbiota. The last challenge was counting our bacteria. The classical way to do this is to plate bacteria on agar and count resulting colonies, but this is a labour-intensive technique that also requires the colony morphology of the species to be sufficiently different. Alternative approaches to bacterial counting such as qPCR with species-specific primers, FISH and others are even more work-intensive and do not scale well with species number. The current standard for quantifying bacterial abundance is to sequence marker genes such as the 16S rRNA gene. On its own, 16S rRNA sequencing only gives relative abundances (that is percentages), but it can be combined with cell counts from flow cytometry to yield absolute abundances, which recently has become the state of the art. However, there are several drawbacks to 16S rRNA sequencing. First, it is also a work-intensive technique that requires DNA extraction, amplification and several more steps (known as library preparation) before the DNA can be sequenced. In addition, we need to know the copy number of the 16S rRNA gene to correctly determine bacterial percentages. Finally, its technical variability is known to be high, which we also confirmed in our own experiments. For this reason, we have the ambition to sequence only for contamination checks and to establish an alternative counting technique that is fast, delivers absolute abundances and has a low technical error. For this, we planned to introduce fluorescent labels in gut bacteria. However, despite ongoing efforts, we successfully labelled only a single gut bacterial species so far, using a plasmid published previously. These difficulties were expected and identified as one of the major risks for the EcoBox project. The fallback option was to continue with 16S rRNA sequencing despite its limitations. However, we came up with an alternative approach. Inspired by previous attempts to distinguish different species in flow cytometry data with machine learning, we systematically tested several supervised clustering techniques on flow cytometry data. We were able to distinguish some of our species well, achieving accuracies as good as or better than 16S rRNA sequencing. Although this approach rapidly drops in accuracy with increasing species number, we were still able to reproduce trends seen in communities resolved with 16S rRNA. We also made unexpected observations, for instance that phenotypic variation varies across gut species and that it increases over time in growth experiments. Since this manner of species quantification is based on flow cytometry alone, it is rapid, gives absolute abundances and has a low technical error. In future, we hope to combine it with fluorescence labels to reach a sufficiently high accuracy for routine measurements.
Counting gut bacteria by combining flow cytometry data with machine learning goes beyond the state of the art. In addition, we have now collected first data sets on gut bacterial communities in several biological replicates, with technical replicates for 16S rRNA sequencing, as well as flow cytometry data and metabolite measurements. These data sets are unique and led to interesting findings. For instance, we saw that the dynamics of our defined gut community is highly reproducible and stable, and that all species co-exist. In addition, we found that one dominant species is gradually replaced by several others before a stable state is reached. In the next steps, we will complete our metabolic community model and test to what extent it can predict community behaviour in different scenarios, including perturbations. By the end of the project, we hope to make three main contributions: i) improve bacterial counting methods, ii) increase our basic knowledge of gut bacteria (their response to different conditions, metabolic preferences, production and consumption rates and refined metabolic maps) and iii) gain a thorough understanding of the dynamics of our synthetic gut community, encoded in a predictive community model.