1. Introduction
The nonpathogenic Escherichia coli strain Nissle 1917 (EcN) is among the most studied probiotic bacteria [1,2]. Over 100 years, EcN has been actively used as a pharmaceutical product (trade name: Mutaflor) for the treatment of intestinal disorders, such as inflammatory bowel disease, ulcerative colitis, and diarrhea [2]. EcN is a successful colonizer of the human gut, possessing strong antagonistic activity against entero-pathogens and immunomodulatory properties. The persistent colonization of EcN confers an advantage over Gram-positive probiotic strains, such as lactic acid bacteria, which transiently colonize the gut mucosa [1].
EcN has been genetically engineered to diagnose, prevent, and treat diseases owing to its facile genetics and biosafety profile [3]. Recombinant EcN was constructed to secrete human epidermal growth factors to heal wounds in human intestinal epithelial cells [4]. It was engineered to sense and kill Pseudomonas aeruginosa in Caenorhabditis elegans and mouse infection models [5].
EcN can selectively colonize and replicate in solid tumors, and its tumor-targeting ability has been exploited to bind to cancer cells and produce cytotoxic compounds in tumor-bearing mice [6–8]. EcN was developed as a live microbial therapeutic to treat metabolic disorders, such as phenylketonuria [9] and hyperammonemia [10]. With advances in synthetic and systems biology, EcN is expected to be engineered to develop safer, cheaper, and more effective therapeutics for a wide range of diseases [3].
A genome-scale metabolic network model is a computational framework that allows the prediction of metabolic flux values for a whole set of metabolic reactions using optimization techniques, such as flux balance analysis (FBA) [11,12]. Metabolic models of diverse organisms are used for various purposes, including the prediction of possible functions and phenotypes, understanding genotype–phenotype relationships, and development of metabolically engineered microbes [6]. Moreover, metabolic models of gut microbes can provide insights into the interactions of microbes with each other as well as their host [13]. Strain-specific metabolic models have been reconstructed for nonpathogenic and pathogenic E. coli strains [14]. In particular, highly curated and fine-tuned metabolic network models have been developed for the laboratory strains of K-12 MG1655 [15] and BL21(DE3) [16].
Although genome-scale metabolic models were reconstructed for 773 strains of the human gut microbiota [17], the probiotic EcN model has not been developed. Considering the clinical and pharmaceutical importance of EcN, an accurate and comprehensive metabolic model needs to be developed for the systems and synthetic biology of EcN. In this study, we reconstructed a comprehensive and highly curated metabolic network model of EcN based on comparative genome and phenome analyses. A phenotype microarray (PM) test was performed to validate and update the metabolic model, and the reconstruction was extensively manually curated. Combined with results from the phenome analysis, metabolic simulation using the model identified genetic determinants of EcN-specific phenotypes. Further, genetic and metabolic factors responsible for intestinal colonization of EcN were identified.
3. Discussion
Differences in E. coli strains of Nissle and K-12 in the utilization of diverse nutrient substrates were identified (Figure 2) and could, in many cases, be correlated with genotypic differences (Figure 3). In general, genome sequence data alone do not allow the identification of the genetic determinants of a given phenotypic variation, and the number of genetic differences even between closely related strains is often too large to be evaluated by gene deletion and complementation tests [59]. The metabolic simulations for nutrient utilization described in this study outline a general strategy for identifying the genetic basis of phenotypic divergence based on clues from comparative genome and phenome analyses. Experimental evidence for the phenotypic differences between the two organisms was provided by the PM tests, and the genetic bases responsible for the phenotypic differences were corroborated by metabolic simulation. Previously, we used this approach to reveal the genotype–phenotype associations between closely related E. coli strains B and K-12 [16,24]. Our approach forms a general strategy to understand the genetic basis of phenotypic functions of naturally occurring strains and laboratory-evolved strains. The spectrum of nutrient utilization and antibiotic resistance is the first step toward a comprehensive understanding of the mechanisms by which commensals and pathogens colonize and persist in the gut in both normal and inflammatory conditions. This is in line with the previous finding that the human commensal E. coli K-12 MG1655 and enterohemorrhagic E. coli EDL933 grew primarily as single cells dispersed in the mouse cecal mucus layer [60]. In this regard, PM data in this study serve as a compendium of nearly 2000 monoculture phenotypes of EcN over a wide range of digestible substrates and inhibitory substrates. The phenotypic profiles in this study provide insights into the colonizing capabilities of EcN. Nutrient sources supporting cell growth are important for E. coli to colonize the intestine [53]. EcN can use more diverse carbon (87 vs. 80) and nitrogen (57 vs. 53) sources than K-12 (Figure 1). EcN can aerobically grow on 13 sugars (fructose, fucose, galacturonate, gluconate, glucose, glucuronate, lactose, maltose, mannose, N-acetyl-D-glucosamine, N-acetylneuraminate, ribose, and xylose) and two carboxylic acids (acetate and malate), which were reported to be present in the gastrointestinal tract [52,53]. Commensal E. coli strains grow predominantly in the mucus layer [61]. They cannot degrade mucus polysaccharides and rely on other anaerobes for the release of metabolizable sugars. Thus, metabolic interactions with anaerobes and utilization of mucus-derived sugars is important for intestinal colonization by EcN [61,62]. Of the 15 nutrients, K-12 could not grow on the mucus-derived amino sugar of N-acetyl-D-galactosamine. Furthermore, it could not grow on 2-deoxy-D-ribose, which promotes pathogenic E. coli strains during host colonization because deoxyribose is derived from DNA degradation and is abundant in the intestine [63]