Escherichia coli Nissle 1917 (EcN) is probably the most intensively investigated bacterial strain today [1]. Despite the fact that the EcN strain is widely used as a probiotic, a lot of questions remain. Hence, we decided to dive deep into the characterization of EcN. Our goal was to find out more about EcN itself and to provide crucial information to include EcN as a platform organism for other iGEM Teams and researchers.

The first step in the characterization of a bacterium is conducting growth experiments in different media and under different conditions. This helps to get a better understanding of beneficial conditions for growth of the culture, which will improve folliwing experiments. We chose to perform half of our experiments under aerobic and the other half under anaerobic conditions. As a control, we grew EcN in LB medium at 37°C and pH 7 under aerobic conditions. We tested growth at a variety of different temperatures (25°C, 8°C, 4°C and 0°C) showing that EcN growth is inhibited at 8°C, the temperature often used for cold shock (Fig. 1).

Moreover, we evaluated the growth in the pH range from pH 6 to pH 1, since in our application EcN will have to pass the acidity of the stomach. Here, at pH 4 we discovered an interesting recovery of EcN growth after three hours (Fig. 2).

To look into inflammatory stress, we subjected EcN to up to 100 µM hydrogen peroxide [2] and showed that the growth is not substantially influenced by this chemical (Fig. 4).

Lastly, we performed dryfreezing and recovered EcN afterwards under control conditions [3][4][5]. The experiment showed that dryfreezing within skim milk protects the bacteria from substantial death (Fig. 5).

Anaerobic conditions were used to characterize EcN growth under the circumstances provided in the human intestines. Thus, as a control we compared EcN growth in LB medium versus mGAM medium under anaerobic conditions. The experiment showed that mGAM, a medium designed for anaerobic application, is in fact a more suitable medium (Fig. 6). Then, we subjected the culture to metformin treatment, since it is one of the most common treatments of Diabetes and is known to accumulate to 720 µM to 7.2 mM within the intestines due to low bioavailability [6][7]. We showed that EcN growth is not substantially influenced by 1.3 mM of metformin into mGAM medium (Fig. 6). Moreover, we tested for EcN’s resistance against cholic acid, since it is commonly secreted into the ileum. Here, we added 0.25 mM cholic acid to mGAM medium, considering that the concentration of free bile salts may rise to 0.25 mM to 1 mM of total bile salts in the Ileum, with cholic acid only being one component [8][9]. Our results suggested that EcN growth is not greatly influenced by cholic acid (Fig. 6).

Finally, we tested EcN interaction with other bacteria by adding bacterial supernatant to our medium. The supernatant was sterilized and provided by Dr. Lisa Maier. We chose Bacteroides thetaiotaomicron, Prevotella copri and Ruminococcus gnavus since they are commonly found within the human microbiome [10]. Bifidobacterium adolescentis was chosen due to its probiotic nature and Clostridium difficile supernatant was used, since it is often found in the microbiome of people with chronic inflammation [10][11]. The results suggest that EcN grows under all additions, however was initially inhibited in its growth by the addition of Bacteroides spp., thus we used these samples for RNA-Seq (Fig. 7).

The transcriptome describes the set of all RNA molecules in a population of cells and is subject to continuous changes. Understanding the complete transcriptome, the expressed genes, post-transcriptional modifications and additional properties of interest is imperative towards understanding genetic cause, disease and possible treatment strategies.

The inherent complexity of the transcriptome requires precise and scaling analysis techniques. RNA sequencing (RNA-Seq), also known as whole transcriptome shotgun sequencing (WTSS), is most commonly used for this purpose, today. It uses next-generation sequencing (NGS) to detect and quantify RNA in biological samples.

The generated RNA-Seq read data is then analyzed according to a sample RNA-Seq work-flow (Fig. 7).

Next, a statistical analysis yields all differentially expressed genes, which are used for Gene Ontology (GO) enrichment, as well as pathway analysis.

We set ourselves the goal of understanding the transcriptomic changes that EcN undergoes under various stress conditions to gain a deeper insight into its responses. Understanding these stress responses of E. coli Nissle 1917 could lead to the development of more robust strains, which not only our project would benefit from, but also scientists working on probiotic drugs in general.

We divided our experimental design into two parts. One part was to investigate the effect of environmental stress on E. coli Nissle 1917 under aerobic conditions, the other part was to examine the responses under anaerobic conditions. Hence, our experimental design looks as follows:

The respective temperature, pH values, and doses were determined by growth curves (Figures 1-6). To get the most out of RNA-Seq applied to stress factors, it is important to find the cutoff values, where E. coli Nissle 1917 is put under stress the most, but still survives.

Sequencing and library preparation was conducted on two sites. All aerobic samples were prepared and sequenced at the NGS Competence Center Tübingen (NCCT), whereas all anaerobic samples were prepared and sequenced at the European Molecular Biology Laboratory (EMBL) in Heidelberg. For more details please visit our Notebook and Outlook.

The following data analysis was, therefore solely conducted on the EMBL dataset.

Since we received our sequencing data from the NCCT on the 18.10.2019, we weren’t able to analyze the data according to our standards in this short time frame. However, as noted in our Attributions, we will analyze both datasets in depth and are planning to publish the results next year

Workflow

Our complex and cutting-edge data analysis is based on several workflows and tools. The initial quality control and feature counting was conducted using the highly sophisticated nf-core/rnaseq pipeline. The pipeline is implemented in nextflow and runs singularity containers to absolutely ensure reproducibility. The workflow processes raw data from FastQ inputs (FastQC, Trim Galore!), aligns the reads (STAR), generates counts relative to genes (featureCounts) or transcripts (Salmon, tximport), and performs extensive quality-control on the results (RSeQC, Qualimap, dupRadar, PreseqPreseq, edgeR, MultiQC).

The differential gene expression analysis was conducted using qbic/rnadeseq, a nextflow based pipeline, which was also run with a singularity container for reproducibilities sake. The workflow runs DESeq2 for the detection of differentially expressed genes. The pathway analysis was conducted manually using clusterprofiler.

Quality Control

The following is only a small selection of all quality control metrics. A focus is set on the most interesting and important quality control metrics. We highly recommend visiting our downloads page to download the full interactive multiqc report. It features short explanations of the quality metrics and lots of useful plots.

To verify that our replicates are indeed replicates of the same condition, and also to explore whether or not our various conditions indeed have a biological effect and differ from the anaerobic ground truth, we created a Multidimensional scaling (MDS) plot using edgeR. MDS visualizes the level of similarity of individual cases of a dataset.

It is used to translate "information about the pairwise 'distances' among a set of n objects or individuals" into a configuration of n points mapped into an abstract Cartesian space. MDS is a form of non-linear dimensionality reduction.

It can be observed, that the majority of the reads are protein coding, with some mapping to sRNA and rRNA. Sample four shows significant rRNA contamination, but still in tolerable numbers.

All in all, our extensive quality control suggests that the libraries were well prepared and rRNA or adapter contamination is very low with anaerob sample four posing a small exception. However, since the quality was high enough according to our standards, it was still included in all downstream analyses. Our initial quality control already suggests, that biological differences between the samples are very likely to exist. These will be examined more in depth in the first step of our downstream analysis: Differential expression analysis.

Differential Expression Analysis

To evaluate the results of our RNA-Seq experiments, we performed a differential expression analysis using DESeq2. This allows identifying the genes which are differentially expressed between the anaerobe control and the respective treatment conditions. Looking at the list of differentially expressed genes allows first interpretations of the results and give a good starting point for more in-depth analysis.

The conditions bile acid, metformin, bacterial supernatant, and mGAM all share common genes among the Top 10 differentially expressed genes. The anaerobic sn-glycerol-3-phosphate dehydrogenase is the most differentially expressed gene in all those conditions, with a log2Fold change greater than 8. Other prominent gens are the succinate-semialdehyde dehydrogenase (NADP+) which is highly upregulated in the mGAM, metformin and bile acid sample and the Zn-dependent periplasmic chaperone, which is downregulated in all samples except the supernatant treatment.

In general, the most differentially expressed gene set of the aerobe treatment against the anaerobe control is more dissimilar to the other conditions. A more in-depth functional analysis can be performed using GO term enrichment or pathway analysis.

Enrichment

To identify groups of genes that are overrepresented under treatment conditions, gene set enrichments for common gene ontology (GO) terms and pathways were performed. GO terms are standardized functions, processes or components defined by the GO initiative across all species, which facilitate functional interpretation of experimental data. The enrichment allows summarizing the large list of differentially expressed genes into a smaller list of affected biological processes that changed most on the treatment. We highly recommend to download our Parts / Downloads site to download far more plots of the enrichment analysis, than we can show on this page.

The GO enrichment funneled hundreds of differentially expressed genes into a few easily interpretable categories for the aerobic, bile acid, mGAM, and Metformin samples. However, for the samples treated with the bacterial supernatant, no GO terms were enriched. This indicates no significant changes in gene expression of related genes with similar functions.

Summary

In general, a clear difference in expression profiles of samples cultivated under aerobic and anaerobic conditions can be observed. This result was expected since the availability of oxygen largely influences the whole metabolism of the bacterium. This is a crucial aspect that has to be considered when working with E. Coli Nissle 1917 and in general all bacteria, which are meant to be used as probiotic. In particular, ATP-biosynthesis related genes were upregulated for the samples cultivated under aerobic conditions.

Generally, treatment with bile acid, metformin, the bacterial supernatant, and the mGAM medium resulted in more similar expression profiles than compared to the aerobic samples. Our results indicate a potential influence on cell respiration by Metformin, a frequently used drug for diabetes patients, and by cultivation mGAM medium, which represents the environment of the gut more precisely than other culture media like the typical LB medium. However, further experiments are required to draw a conclusion.

Metabolic Model

Our E. coli Nissle 1917 characterization includes our metabolic modeling. Please visit our model subpage for more details.