GLUCONEOGENESIS AND COLONIZATION OF THE MOUSE INTESTINE BY E. COLI EDL933 AND E. COLI NISSLE 1917

Previous studies in this laboratory have indicated that E. coli Nissle 1917 can act as a probiotic and is able to prevent the pathogenic E. coli EDL933 from growing to high numbers in the mouse large intestine. This study evaluated the influence of gluconeogenesis in E. coli Nissle 1917 and E. coli EDL933 to compete against each other and for their maintenance in the intestine. Knockout mutants of both strains were created for the analysis of the importance of gluconeogenesis in colonization experiments in the mouse large intestine. To knock out the ability to use gluconeogenesis and therefore metabolizing substrates such as pyruvate, TCA cycle intermediates and amino acids, simultaneously both genes pckA, encoding for phosphoenolpyruvate carboxykinase, and ppsA, encoding for phosphoenolpyruvate synthetase, were replaced by a non-functional chloramphenicol cassette. It was shown that E. coli Nissle 1917 uses glycolysis and gluconeogenesis concurrently throughout colonization of the mouse large intestine. E. coli EDL933 utilized glycolysis exclusively when it was the only E. coli strain present, but switched to gluconeogenesis when it was in the presence of a competing E. coli strain. Knocking out the ability to use gluconeogenesis in E. coli EDL933, led to a severe disability to colonize the mouse large intestine when it competed against E. coli Nissle 1917 which used up most of the glycolytic substrates. By implication E. coli Nissle 1917 seems to use up most of the glycolytic substrates and some of the gluconeogenic substrates, E. coli EDL933 depends on utilizing gluconeogenic substrates to be able to maintain itself in the mouse large intestine. Phenotypic characterization of the usage of different substrates in E. coli Nissle 1917 and E. coli EDL933 strains revealed several differences between the strains.

viii List of Tables   Table 1:

Goals of this Study
Escherichia coli EDL933, an O157:H7 strain, is known to be a noninvasive enteric shiga toxin producing pathogen causing gastrointestinal illnesses including hemolytic uremic syndrome (HUS) and hemorrhagic colitis (HC) in humans [29,51] [26]. The colonization trials show that intentional colonization of the newborn's gut with E. coli Nissle 1917 or other commensal nonpathogenic E. coli strains results in early induction of the immunologic competence of the gutassociated lymphoid tissue (GALT). It also results in prevention of hospital-acquired infections and in prevention of colonization by unwanted bacteria such as multidrug resistant or pathogenic E. coli strains [26].
Based on findings like these I decided to focus on the interaction of probiotic E. coli Nissle 1917 and the pathogen E. coli EDL933 during their colonization of the mouse large intestine to better understand how E. coli Nissle 1917 acts as a probiotic.
When competing against E. coli EDL933, E. coli Nissle 1917 is not yet able to outcompete this pathogen completely from the mouse intestine in vivo [33]. Little is known about nutrients that are utilized by E. coli Nissle 1917 for growth in the intestine or its metabolic pathways involved in competing against E. coli EDL933. If these pathways were defined, it would help in developing more efficient treatments for patients infected with O157:H7 strains. Therefore my focus lies on 2 generating an E. coli Nissle 1917 ΔpckA ΔppsA knockout mutant, which grows normally on glycolytic substrates but is unable to utilize tricarboxylic acid (TCA) cycle intermediates and gluconeogenic substrates for growth. In addition ability to colonize the mouse intestine in the presence of E. coli Nissle 1917 wild type as well as E. coli EDL933 was analyzed. The focus lays also on the importance of the glyoxylate shunt, the shortcut of the TCA cycle, as I analyze an E.

General background of Escherichia coli.
Escherichia coli belongs to the family Enterobacteriaceae and is the only member of the genus Escherichia [58]. Its appearance is a short gram-negative, non-sporing and fimbriate bacillus [58], which grows as a facultative anaerobe readily on simple culture media and synthetic media with glycerol or glucose as the sole carbon and energy source [58]. The primary habitat of E. coli is the gastro-intestinal tract and the bowel of mammals and birds. Natural colonization of commensal E. coli strains is known to take place soon after birth [5] and its source is to be found in the mother and in the environment [5]. Once established in the gastro-intestinal tract, E. coli still remains a minority member of the fecal flora.

E. coli O157:H7 pathogenesis.
Commensal E. coli strains are not harmful, however there are pathogenic E. coli strains, such as enterohemorrhagic Escherichia coli (EHEC) serotypes like E. coli EDL933 O157:H7 [45], a noninvasive enteric shiga toxin producing pathogen which is known to cause gastrointestinal illnesses like hemolytic uremic syndrome (HUS) and hemorrhagic colitis (HC) in humans [22,25,29,51] and is ingested by eating undercooked food. Serotype O157:H7 is the major source of reported E. coli food poisoning outbreaks in the United States (US) [28]. Shiga toxins produced by EHEC bacteria are thought to damage host endothelial cells in small vessels of the intestine, kidney and brain resulting in thrombotic microangiopathy [28].
They postulated that many bacterial pathogens cross the intestinal barrier through microfold cells which are known to transport organisms and particles from the gut lumen to immune cells across the epithelial barrier, and thus are important in stimulating mucosal immunity. Once passed through the microfold cells, EHEC cells are captured by mucosal macrophages. Since EHEC belongs to Shiga toxin (Stx) producing bacteria and is able to survive and to produce Stx within the macrophages, this translocation of Stx from the gut lumen to underlying tissues is a decisive step in the development of the infection. Following replication of bacteria in macrophages, their extensive Stx production induces macrophage cell death. Subsequently, released Stx can cross the downstream blood vessels to reach the kidneys, intestine, and the brain. Damage to these organs results in serious life-threatening complications in humans [13].
Njoroge et al. 2012 described a novel mechanism of regulation that links metabolism to pathogenesis [47]. The virulence strategy of EHEC to enter through enterocytes is to form attaching and effacing (AE) lesions on enterocytes (see Figure 1). The expression of most of the genes necessary for this enterocyte effacement requires the regulator Ler. Whereas growth within a glycolytic environment inhibits the expression of ler, growth within a gluconeogenic environment activates expression of these genes. This is achieved through two transcription factors calles KdpE and Cra, which directly bind to the ler promoter and are sugar-dependently regulated [47]. This means that a gluconeogenic environment is a signal for EHEC to turn on pathogenesis. [13]. It is illustrated how EHEC bacteria cross the intestinal epithelial barrier through M cells and reach the lamina propria underneath where they get phagocytosed by macrophages. EHEC's Stx production occurs in the intestine and continues within the macrophages as the bacteria survive their immune response. Stx production within the macrophages leads to apoptosis of macrophages, and thereby releases the Stx toxin which can cross the downstream blood vessels to reach the kidneys, intestine, and brain [13] of the human body.

Impact of diseases related to E. coli O157:H7 infections
Diseases caused by infection with E. coli O157:H7 range from vomiting, stomach cramps, bloody diarrhea to haemolytic-uraemic syndrome (HUS) or haemorrhagic colitis (HC). HUS was first described in 1955 and consists of an acute febrile illness followed by acute renal failure and intravascular haemolysis [58]. HC is characterized by sudden severe abdominal colic and grossly bloody diarrhea and was first described by Riley et al. 1983.
Currently for EHEC infections there is no treatment available [21]. It was found that the use of conventional antibiotics makes Shiga toxin-mediated cytotoxicity worse. In an epidemiology study conducted by the Centers for Disease Control and Prevention, patients treated with antibiotics for EHEC enteritis had a higher risk of developing HUS [56]. Additional studies support the contraindication of antibiotics in EHEC infection. Antibiotics promote Shiga toxin production by enhancing the replication and expression of stx genes that are encoded within a chromosomally integrated lambdoid prophage genome [65]. A promising way to treat EHEC infections is the usage of probiotics, such as E. coli Nissle 1917. To change the composition of the normal intestinal microbiota from a potentially harmful composition towards a microbiota that would be beneficial for the host has always been the original idea with probiotics [2], [32].

General background of probiotic Escherichia coli Nissle 1917
E. coli Nissle 1917 belongs to the non-pathogenic E. coli strains. It is one of the best analyzed E. coli strains so far and is used in the medical field to treat several intestinal or colon related diseases.
Several characteristics make E. coli Nissle 1917 such a candidate as a probiotic. It has been shown to produce no toxins like heat-labile or heat-stabile toxins and no Shiga-like toxins, as EDL933 does. E. coli Nissle 1917 is non-invasive and doesn't form hemolysins, doesn't possess any stx genes [26]. It expresses no CFA I or CFA II fimbriae and has no P-, M-, or S-fimbriae [26].
Focusing on its intestinal colonization ability, E. coli Nissle 1917 is generally able to colonize the human gut as well as the mouse intestine [49]. It is able to easily colonize newborns if administered within the first few days after birth, when the colonization resistance is not yet well established. This has been shown in a study in a neonatal unit of a hospital in Hagen as well as in a children's hospital in Prague [11]. Here they postulate that E. coli Nissle 1917's 7 colonization of premature infants stimulates significantly nonspecific natural immunity. To successfully colonize the intestine of adults, it has been shown that this is easily achieved after gut decontamination and/or lavage [32].

The mammalian large intestine
The large intestine is the terminal part of the digestive system after the stomach and small intestine. The small intestine absorbs most of the nutrients. The large intestine consists of three main sections: the cecum, the colon and the rectum [61]. Its main function is to reabsorb water and inorganic salts [63]. The epithelial surface of the large intestine is unlike the small intestine, rather smooth without intestinal villi and consists of goblet cells and enterocytes [61]. The mucus layer overlies the epithelial surface and is relatively thick. It consists of mucin, a 2-MDagelforming glycoprotein and a large number of smaller glycoproteins, proteins, glycolipids, lipids and sugars [3], [50], [60]. The mucus layer is in a dynamic state meaning it is constantly being synthesized and secreted by the mucin-secreting goblet cells and is degraded to a larger extent by the indigenous intestinal microbes [63], [41]. Degraded mucus components are shed into the intestinal lumen, forming a part of the luminal content that is excreted in the feces [41].
To digest and uptake all necessary nutrients the intestine is supported by a huge variety of bacteria colonizing the inner surface and interacting with each other. This is a symbiosis between host intestine and bacteria, since food is presented to bacteria of the digestive system and they in return provide breakdown products of molecules which cannot be broken down by the intestinal enzymes alone. There are about 500 major bacterial strains in the intestine, which interact with each other symbiotically [15]. The microbiota in the intestine of humans and animals consists of a variety of different bacteria including obligate anaerobes such as Bacteriodes, Bifidobacterium, Clostridium, Eubacterium, Fusobacterium, Lactobacillus, Peptococcus, Peptostreptococcus and Veillonella [35,54]. The obligate anaerobes in the mammalian intestine make up greater than 99.9% of the cultivable bacteria [44]. The predominant facultative anaerobe in the gastrointestinal tract is Escherichia coli [15].
According to Freter et al. 1983 there are four microhabitats within each section of the intestine [52]. There is the surface of the epithelial cells, the deep mucus layer of the crypts in the ileum, cecum and colon, the mucus layer that covers the epithelial cells throughout the intestinal tract in the cecum and colon of the mammalian intestine and finally the lumen [52]. Commensal E.
coli strains colonize the mucus layer covering the epithelial cells [43], whereas especially pathogenic E. coli strains colonize the surface of the epithelial cells via attachment to specific receptors [9,31].

Mouse models used in colonization studies
The animal model of a conventional laboratory mouse would be ideal to study intestinal colonization of E. coli, since its microbiota and biofilm interactions are congruent with those of a conventional human patient. The major problem with this model is to introduce and colonize an invader strain in an established microbiota, meaning to overcome colonization resistance [63].
To prevent this difficulty there are other animal models, such as the germfree animal which doesn't contain an established microbiota, since it has never been exposed to a natural germcontaining environment. This animal model is not only very expensive it also doesn't reflect natural intestinal colonization [63]. The streptomycin-treated mouse model was established in 1954 by Bohnhoff et al. [7]. The advantage of this animal model is that the intestine can be cleared of native facultative anaerobe bacteria [23], such as facultative anaerobic enterobacocci, streptococci, lactobacilli, anaerobic lactobacilli and bifidobacteria purely by adding the antibiotic streptomycin in the drinking water (5g/L). However, besides changes in the concentration of 9 volatile fatty acids and a decrease of pH (6.42 to 6.73), the numbers of strict anaerobes and populations of the genera Bacteriodes and Eubacterium remain unchanged [23]. The advantages of this animal model are its relatively low cost and a relatively simple colonization procedure by providing streptomycin in drinking water and feeding streptomycin resistant bacteria to the mice. The streptomycin-treated mouse model is used during this study.

Nutritional aspects of colonizing the mouse large intestine
The dominant glycoprotein of the intestinal mucus is mucin which consists of about 80% polysaccharide and 20% protein [1]. The five major sugars in mucin are N-acetylgalactosamine, N-acetylglucosamine, N-acetylneuraminic acid, L-fucose and D-galactose [1]. The mucus layer in the mouse intestine also contains smaller proteins, glycoproteins, lipids and glycolipids, and sugars like arabinose, mannose and ribose [1], [50].
As about 500 different cultivable bacteria provide a very diverse microbiota, each organism is best served to cultivate a specific ecological niche in order to be maintained in the intestine.
Those ecological niches can be defined by nutrient availability according to Freter et al. 1983 [52], such that the population size of an individual species in the large intestine depends on the available concentration of its preferred nutrient. The population of E. coli in the intestine is relatively small, despite its rapid growth rate, which indicates that the concentration of its preferred monosaccharide(s) is low. To survive in the intestine without being physically attached to the host intestinal wall, a bacterial strain would have to out-compete all other organisms for a particular nutrient or nutrient mixture, since failure to do so would lead to displacement from the host. To be able to make predictions about nutritional competition between simultaneously colonizing E. coli strains in the mouse large intestine, it is necessary to understand which pathways are utilized by each strain while growing in the intestine.

The Emden-Meyerhof Parnas pathway (glycolysis)
A major metabolic pathway for glucose utilization is the Entner-Doudoroff pathway, which is present in bacteria but not eukaryotes and is needed to break down gluconate, a sugar also present in the mucus layer of the intestine [59]. The Pentose-Phosphate pathway is needed to produce pentose phosphates and the Emden-Meyerhof Parnas pathway provides glucose for glycolysis into pyruvate.
Glycolysis degrades one molecule of glucose (six-carbon molecule) in a series of enzymedependent reactions and yield two molecules of pyruvate (three-carbon molecule) [55]. During several enzyme-catalysed steps, energy is released and conserved in the form of ATP and NADH.
Each molecule of pyruvate can then be decarboxylized to yield acetyl-CoA which enters the TCA cycle and serves to conserve more energy [55].
The breakdown of glucose into two molecules of pyruvate occurs during 10 reaction steps [36].
Glucose is first phosphorylated at the hydroxyl group of C-6 (by hexokinase) to D-glucose 6phosphate. This is further converted into D-fructose 6-phosphate (by phosphohexose isomerase). D-fructose 6-phosphate is phosphorylated at its C-1 leading to D-fructose 1,6bisphosphate (by phosphofructokinase-1). ATP is for both phosphorylation reactions the phosphoryl group donor. D-fructose 1,6-bisphosphate is further split into dihydroxyacetone phosphate (DHAP) and into glyceraldehyde 3-phosphate (by aldolase). DHAP and glycerol 3phosphate are isomerized by triose phosphate isomerase. During the next reaction step, two molecules of glyceraldehyde 3-phosphate are phosphorylated by inorganic phosphate to two molecules of 1,3-bisphosphoglycerate (by glyceraldehyde 3-phosphate dehydrogenase), while two molecules of NADH are created. Converting two molecules of 1,3-bisphosphoglycerate by substrate-level phosphorylation into two molecules of 3-phosphoglycerate (by phosphoglycerate kinase) energy is released in form of two molecules of ATP. The enzyme phosphoglycerate mutase converts two molecules of 3-phosphoglycerate each into 2phosphoglycerate. Dehydration of two molecules of 2-phosphoglycerate (by enolase) leads to two molecules of phosphoenolpyruvate (PEP). Finally, two molecules of PEP are converted into two molecules of pyruvate via substrate-level phosphorylation (by pyruvate) kinase, releasing energy in the form of two molecules of ATP [46]. The net yield of energy conversion is two molecules of ATP and two molecules of NADH per molecule of glucose. To assure a steady supply of ATP, glycolysis is tightly regulated in coordination with other energy-yielding pathways [55].

The tricarboxylic acid (TCA) cycle
The TCA cycle can be used both under aerobic and anaerobic conditions. It is the second major stage of respiration which enzymatically further oxidizes the pyruvate produced by glycolysis into H 2 O and CO 2 [64]. The energy released during these oxidation reactions is conserved in the reduced coenzymes and electron carriers NADH and FADH 2 . They themselves undergo oxidation during the third stage of respiration, and transfer their electrons to O 2 , producing ATP by the electron flow [46].
The TCA cycle is comprised of eight steps [39]. Pyruvate, the product of glycolysis is converted into acetyl-CoA by pyruvate dehydrogenase, and the acetyl-CoA molecule enters the TCA cycle.
First citrate is formed by condensation of the incoming acetyl-CoA with oxaloacetate, catalyzed by citrate synthase. Second isocitrate is formed via cis-aconitate by adding a H 2 O molecule catalyzed by aconitase. Third oxidative decarboxylation of isocitrate forms α-ketoglutarate and one molecule of CO 2 . This reaction is catalyzed by isocitrate dehydrogenase and also forms a molecule NADH by electron transfer. Fourth another oxidative decarboxylation reaction produces succinyl-CoA and CO 2 . The reaction is catalyzed by α-ketoglutarate dehydrogenase and transfers an electron to NAD + , leading to NADH. Fifth succinyl-CoA is converted into succinate by succinyl-CoA synthetase and thereby produces GTP. Sixth succinate is oxidized to fumarate catalyzed by succinate dehydrogenase. This reaction also produces FADH 2 . In a seventh step fumarate is hydrated to malate by fumarase. And in the last eighth step malate dehydrogenase catalyzes the oxidation of malate to oxaloacetate and thereby producing NADH 2 [46]. This pathway is cyclic which means that the intermediates of the cycle are not used up. For one molecule of oxaloacetate consumed in the TCA cycle, one is produced. The energy gain per acetyl-CoA oxidized by the TCA cycle are three molecules of NADH, one molecule of FADH 2 and one molecule of nucleoside triphosphate (ATP or GTP) [64].
The TCA cycle is not only dependent on entering molecules of acetyl-CoA. Any four or fivecarbon intermediate of the TCA cycle can be oxidized further. These intermediates can be products of amino acid breakdowns [36].

Gluconeogenesis
Gluconeogenesis ("new formation of sugar") is needed as a method for synthesizing glucose from non-carbohydrate precursors by E. coli when glucose is depleted. Gluconeogenesis in E.
coli (see figure 1) starts from simple organic compounds of two or three carbons, such as lactate, acetate and propionate, in their environment or growth medium [46] and it is needed to convert pyruvate and related three-and four-carbon compounds (including TCA cycle intermediates) to glucose [46]. Seven of the reaction steps in gluconeogenesis are catalyzed by the same enzymes as in glycolysis (see glycolysis). Different enzymes are used to reverse three irreversible reaction steps of glycolysis. These reaction steps are first of all the conversion of pyruvate to PEP via oxaloacetate (catalyzed by pyruvate carboxylase and PEP carboxykinase).
Second it is the dephosphorylation of fructose 1,6-bisphosphate by FBPase-1. And third it is the dephosphorylation of glucose 6-phosphate by glucose 6-phosphatase [36]. Gluconeogenesis is energetically expensive as the formation of one molecule of glucose from two molecules of pyruvate requires two molecules of NADH and six high-energy phosphate groups coming from four molecules of ATP and two molecules of GTP [55]. Gluconeogenesis and glycolysis have to be reciprocally regulated to prevent wasteful operation of both pathways at the same time. Some or all of the carbon atoms of most amino acids derived from proteins are ultimately catabolized to pyruvate or to intermediates of the TCA cycle. Such amino acids can therefore undergo net conversion to glucose and are said to be glucogenic [46].

Glyoxylate shunt
Bacteria like E. coli contain the full repertoire of enzymes needed for the glyoxylate cycle as well as those for the TCA cycle in the cytosol. This way they are able to grow on acetate as their sole carbon and energy source [36]. Organisms that lack the glyoxylate shunt cannot synthesize glucose from acetate or fatty acids that give rise to acetyl-CoA. In the glyoxylate cycle , acetyl-CoA is condensed together with oxaloacetate to form citrate. Citrate is further converted into isocitrate, just like in the TCA cycle. Starting from isocitrate both pathway differ. In the glyoxylate cycle isocitrate is cleaved into succinate and glyoxylate by the enzyme isocitrate dehydrogenase. Glyoxylate together with an additional acetyl-CoA forms malate. This reaction is catalyzed by the enzyme called malate synthase. Malate is finally converted into oxaloacetate, which can either condense with another acetyl-CoA and start another turn of the cycle or it can undergo gluconeogenesis and generate glucose. Each turn of the glyoxylate cycle converts two molecules of acetyl-CoA into one molecule of succinate and one molecule of malate. Succinate can also undergo another turn of the TCA cycle or it can be converted into fumarate, then malate, then oxaloacetate and follow the steps of gluconeogenesis to PEP and glucose [46].
Since the carbon atoms of acetate molecules are converted in 8 steps to oxaloacetate in the TCA cycle, the TCA cycle itself should be able to generate PEP for gluconeogenesis from acetate. But looking at the stoichiometry of the TCA cycle, there is no net conversion of acetate to oxaloacetate. For every two carbons which enter the cycle as acetyl-CoA, two carbons leave as CO 2 which makes it impossible to generate PEP out of it. Looking at the glyoxylate cycle, the two steps of decarboxylation of the TCA cycle are bypassed and with this the net formation of succinate, oxaloactetate and other cycle intermediates from acetyl-CoA becomes possible [46].
As there are many common intermediates being shared between these pathways, they are coordinately regulated. This is accomplished by a covalent but reversible modification of the enzyme isocitrate dehydrogenase. The level of isocitrate dehydrogenase is dependent on its reversible phosphorylation and regulates the partitioning of isocitrate between the TCA cycle and the glyoxylate cycle [39].

Gene knockout mutants used in colonization experiments
Since the effect of gluconeogenesis during colonization of the mouse large intestine is analyzed and the effect of the glyoxylate shunt is discussed during this study, knockout mutants of the genes which are necessary to utilize both metabolic pathways are generated.
Two required genes involved in gluconeogenesis are pckA and ppsA. The pckA gene encodes for phosphoenolpyruvate (PEP) carboxykinase (EC 4.1.1.49) which is an enzyme catalyzing the first step of gluconeogenesis in E. coli [38]. It phosphorylates and decarboxylates the TCA cycle fourcarbon intermediate oxaloacetic acid to PEP [53]. PEP carboxykinase is activated by calcium [19] and the pckA gene's expression is regulated by cyclic AMP [20].
The ppsA gene encodes for PEP synthetase (EC 2.7.9.2), an enzyme which catalyzes the phosphorylation of pyruvate to PEP, being activated by a Pi-dependent pyrophosphorylation (ATP) and inactivated by an ADP-dependent phosphorylation on a regulatory threonine [8].
During growth on three-carbon substrates that require the gluconeogenesis pathway (such as lactate or pyruvate), phosphoenolpyruvate (PEP) synthetase provides the ability to generate phosphoenolpyruvate, which is required for the synthesis of precursor metabolites for cellular carbon compounds [10].
A ΔpckA ΔppsA double mutant does not grow on acetate [48] or succinate [20] as the sole source of carbon, but is able to grow on carbon sources that enter central carbon metabolism "above" PEP such as glycerol.
A required gene involved in the glyoxylate shunt is the gene aceA. The aceA gene encodes for the enzyme isocitrate lyase (EC 4.1.3.1) which catalyzes the cleavage of isocitrate, forming succinate and glyoxylate, a key step of the glyoxylate shunt. It is activated by phosphorylation on histidine and is inhibited by PEP, 3-phosphoglycerate and succinate [24]. The second reaction of the glyoxylate shunt is catalyzed by an enzyme called malate synthase and catalyses the condensation of glyoxylate with a second molecule of acetyl-CoA, forming malate [34]. The malate is then oxidized to oxaloacetate, which can either start another turn of the cycle or can be converted to PEP by PEP carboxykinase and undergo gluconeogenesis to form glucose [46].
The glyoxylate shunt bypasses two CO 2 -evolving steps of the TCA cycle, allowing the net accumulation of carbon from acetyl-CoA [34]. To be able to grow on substrates such as acetate or fatty acids, E. coli needs operation of the glyoxylate shunt [30]. This indicates that an E. coli ΔaceA knockout mutant does not grow on acetate or fatty acids.
In Figure 2 the major metabolic pathways of E. coli are shown. Arrows indicate the physiological directions of the reactions. Genes encoding the enzymes for each reaction are listed beside each reaction [41]. The Embden-Meyerhof-Parnas pathway starts at the six carbon sugar glucose and breaks it down to two molecules of pyruvate (C-3). Pyruvate further enters the TCA cycle and energy in form of ATP, GTP, NADH is released as well as CO 2 as a one carbon molecule. The reverse pathway which is important when there is a depletion of sugar carbon sources in the environment, is the gluconeogenesis pathway. It allows the organism to regenerate glucose out of gluconeogenic substrates such as TCA cycle intermediates, amino acids and other molecules which enter the TCA cycle. Glycolysis yields two high energy phosphate bonds of ATP, gluconeogenesis expends six high energy phosphate bonds of ATP. A futile cycle consisting of both pathways would waste four high energy phosphate bonds of ATP. To prevent this waste glycolysis and gluconeogenesis are reciprocally regulated. To be able to metabolize acetate or fatty acids, the organism has to be able to use the glyoxylate shunt. Both substrates can enter the TCA cycle at acetyl-CoA (C2) and be modified to isocitrate. Starting from isocitrate the TCA cycle is not used anymore, since two molecules of CO 2 would be lost, each during the step leading to alpha-ketoglutarate and the step leading to succinyl-CoA. Since two carbons would be lost, another round of the TCA cycle would be impossible. The only possibilities are to use the glyoxylate shunt avoiding the loss of two molecules of CO 2 as well as using so called anapleurotic reactions such as from PEP to oxaloacetate (C4) which provide two carbons and the TCA can continue further.

Agarose gel electrophoresis
This technique is used to separate DNA fragments based on their size and charge. The

Bacterial strains and plasmids
The bacterial strains as well as the plasmids used in this study are listed in Table 2. E. coli Nissle 1917 is a human commensal strain. It has been used as a probiotic agent since the early 1920s.
Originally the strain was isolated during World War I from a soldier who escaped a severe outbreak of diarrhea affecting his regiment [57]. Since the 1920s E. coli Nissle 1917 has been marketed as a probiotic remedy against intestinal disorders in several European countries. E. coli EDL933 is an O157:H7 strain isolated from an outbreak caused by contaminated beef in 1982 [51]. It is known to be a noninvasive enteric shiga toxin producing pathogen causing 18 gastrointestinal illnesses including hemolytic uremic syndrome (HUS) and hemorrhagic colitis (HC) in humans [29,51] and is ingested by eating undercooked food.

Biolog
The Biolog is a phenotype microarray based on metabolic capabilities. For purposes of this study GN2 MicroPlate TM (Gram negative carbon nutrition) plates are chosen. 95 different nutrients and biochemicals are dried into the wells of each plate (see Table 1).

Construction and characterization of Escherichia coli mutants
Gene deletion mutants are constructed by allelic exchange mutagenesis originally described by Datsenko and Wanner [12] and are shown in Table 2.  [42] pKD3 Template plasmid, contains chloramphenicol resistance cassette flanked by FRT (FLP recombinase) sites [12] pKD46 Temperature-sensitive plasmid, contains Larabinose-inducible λ-red recombinase gene for homologous recombination [12] pCP20 Temperature-sensitive plasmid, contains FLP recombinase gene for removal of antibiotic resistance genes [12]    Since genes pckA and ppsA have to be knocked out in the same organism to asure that its ability to use gluconeogenesis is lost completely, the protocol of constructing a second gene deletion is followed. To be able to use the same chloramphenicol cassette with different homologous regions of the corresponding gene, the already inserted chloramphenicol cassette is removed by electroporation with pCP20. This plasmid contains a FLP recombinase to remove the sequence between both FRT sites. The procedure is repeated as already described except for growing and recovering the cells in LB broth instead of SOB medium and for plating transformants on LB plates containing ampicillin (100 µg/ml). To lose pCP20 again, the cell culture is grown in LB broth at 43 °C for a maximum of 5 h and plated on LB plates at 37 °C. The final selection is done by toothpicking on LB, chloramphenicol and ampicillin plates to select for Amp S and Cam S mutants. These mutants have lost the chloramphenicol cassette and are prepared to generate the second mutation. To knockout the second gene, electroporation with the appropriate PCR amplicon is done as previously described for the first gene knockout. Each mutant is verified via PCR and DNA sequencing.

DNA Sequencing
Sequencing is done at the University of Rhode Island Genomic and Sequencing Center, University of Rhode Island, Kingston, using the Appled Biosystems 3130xl Genetic Analyzer (Applied Biosystems, Foster City CA).

Growth of bacterial cultures
To determine bacterial growth the cell density can be spectrometrically measured using a Pharmacia Biotech Ultrospec 2000 UV/visible spectrometer, since the light beam which passes through the cuvette is scattered depending on the cell density. The optical density (OD) is proportional to the cell density. The OD measurement occurs at a wavelength of 600 nm. The accurate detection range lies between the OD of 0.1 and 0.6 which requires appropriate culture dilutions. A growth curve can be made by plotting the OD 600 value versus time.

Streptomycin-treated mouse model used for colonization experiments
The streptomycin-treated mouse model is used to investigate the in vivo carbon nutrition of E.
coli strains and to study the competition in the intestine between streptomycin-resistant E. coli strains. Since the numbers of an E. coli strain in the mouse large intestine is reflected by their numbers in mouse feces [27], fecal counts are used to judge the relative colonizing abilities of E.
Treatment of mice with streptomycin during the whole time of colonization is needed to overcome colonization resistance on one hand by the loss of facultative anaerobes from the microbiota and on the other hand by the observed decrease in the concentrations of short-chain fatty acids and hydrogen sulfide [23]. The antibiotic streptomycin is a protein synthesis inhibitor which prevents binding of the tRNA fMet to the bacterial ribosome and therefore leads to codon misreading, protein synthesis inhibition and finally to cell death [66]. E. coli Nissle 1917 Str R , and E. coli EDL933 Str R contain the same point mutation in rpsL which makes them resistant to greater than 2 mg/ml of streptomycin sulfate [16]. Some of the introduced E. coli strains used in the colonization experiments are, in addition to being streptomycin resistant, also resistant to either chloramphenicol, nalidixic acid, or rifampin, genetic markers that have no effect on their colonization abilities [9,42,43].
For colonization experiments mice in the animal facilities in Morrill Hall are given streptomycin sulfate in their drinking water (5 g/liter) over the entire course of these experiments, which selectively removes facultative anaerobic E. coli, enterococci, streptococci, lactobacilli, and anaerobic lactobacilli and bifidobacteria [23]. It is important to note that the overall populations of anaerobes, including Bacteroides and Eubacterium, are unchanged in the cecal contents following streptomycin treatment [23].

The streptomycin-treated mouse model is well established for colonization experiments of
Escherichia coli in the mouse large intestine and has been extensively used by Dr. Cohen's laboratory group [9,14,17,18,43,49,59,62]. The optimal colonization length has been

Sequencing and growth of mutants in vitro.
All mutants used in this study were sequenced at the University of Rhode Island Genomic and ΔpckA::Cam was co-colonized with its parent strain (Figure 3). E. coli Nissle 1917 Str R Nal R was simultaneously fed to mice at a low concentration (

E. coli EDL933 utilizes only glycolytic substrates when it is the only strain present during colonization of the mouse large intestine.
Co-colonization of E. coli EDL933 Str R Rif R and E. coli EDL933 Str R ΔppsA ΔpckA::Cam in the mouse large intestine shows that the colonization ability of both strains remains the same throughout the colonization when they are the only E. coli strain present ( Figure 4). The wild type strain colonizes the mouse large intestine beginning from day one with a concentration of 1.9 x 10 9 CFU/gram of feces, only tenfold higher than its mutant (1.9 x 10 8 CFU/gram of feces).
Towards the end of the colonization both strains colonize the intestine with equal concentrations of 3.2 x 10 7 CFU/gram of feces ( Figure 4). This result confirms the published data by Miranda et al. [42]. Since the gluconeogenic knockout mutant colonizes the intestine about as well as the wild type, it appears that E. coli EDL933 uses mainly glycolysis to maintain itself in the intestine and is not dependent on gluconeogenesis, when it is the only E. coli strain present.

In the presence of precolonized E. coli Nissle 1917, E. coli EDL933 uses both glycolytic substrates and gluconeogenic substrates in order to colonize.
It has been shown [42] that E. coli EDL933 switches from glycolytic substrates to gluconeogenic substrates when competing against E. coli MG1655 which only uses glycolytic substrates in order to maintain itself in the mouse large intestine.
To answer the question as to whether E. coli Nissle 1917 uses up both glycolytic as well as gluconeogenic substrates and therefore is able to limit the colonization of E. coli EDL933 in the intestine, both E. coli EDL933 strains, wild type simultaneously together with its gluconeogenic knockout mutant, were fed to mice precolonized with E. coli Nissle 1917 wild type ( Figure 5 Figure   7) was fed at day ten (3.8 The question now was whether or not E. coli Nissle 1917 alone was responsible for using up all glycolytic and much of the gluconeogenic substrates, which keeps E. coli EDL933 down or does the anaerobic microbiota contribute? The colonization of E. coli EDL933 and E. coli EDL933 Str R ΔpckA ΔppsA::Cam being fed at day ten without precolonizing another E. strain (Figure 9), but providing streptomycin containing drinking water throughout the whole colonization, creates clarity. The antibiotic streptomycin in the drinking water (5g/L) of the mice clears out all facultative anaerobes, which leaves a lot of nutrients behind, which can be used for incoming E.
coli strains to get established in the intestine. The idea of this colonization was to test the effect of the rest of the microbiota in using up all nutrients during the first ten days of the colonization and then analyzing the reaction of both E. coli EDL933 and E. coli EDL933 Str R ΔpckA ΔppsA::Cam without the presence of anther E. coli strain.
As shown in Figure 9 both E. coli EDL933 strains grew up to high numbers around 10 8 -10 9 CFU/gram of feces and remained on top of each other throughout the whole colonization experiment. Since wild type and mutant colonize all mice equally well, the microbiota doesn't use up the nutrients available to E. coli in the intestine. Important in both colonizations is the fact that both E. coli EDL933 strains decreased in number, but the wild type always remains two or more orders of magnitude higher than the mutant. This suggests that E. coli EDL933 has an advantage in surviving in the mouse large intestine while competing against E. coli Nissle 1917. This advantage is the usage of gluconeogenesis. Therefore E. coli EDL933 definitely switches to using gluconeogenesis in the presence of another E. coli strain in the intestine.

E. coli Nissle 1917 relies on the usage of the glyoxylate shunt to colonize the mouse large intestine
Knowing that E. coli EDL933 switches to gluconeogenesis in the presence of E. coli Nissle 1917, a further focus lies on acetate or fatty acids as gluconeogenic substrates, which are dependent on a functioning glyoxylate shunt. Therefore the aceA gene was knocked out using the gene 35 replacement method by Datsenko and Wanner to generate mutants which are not able to metabolize acetate or fatty acids. type. This colonization shows that being able to utilize the glyoxylate shunt seems to be an advantage for E. coli Nissle 1917 in order to maintain itself in the intestine.

Phenotypic characterization of the usage of different substrates in E. coli Nissle 1917 and E. coli EDL933 strains
The phenotypic metabolic GN2 MicroPlates were used to analyze the ability of a gram-negative

According to the sequencing results of the University of Rhode Island Genomic and Sequencing
Center all generated deletions were in the expected places in the genome (data not shown

Discussion
Analysis of all generated mutants showed that all gene deletions are in the expected places in the genome (data not shown) according to the sequencing results of the University of Rhode Island Genomic and Sequencing Center. Therefore the created E. coli mutants are in fact the mutants which they are supposed to be. In vitro experiments on the mutants' ability to grow on certain carbon sources show that gluconeogenic knockout mutants fail to grow on gluconeogenic substrates and glyoxylate shunt knockout mutants fail to grow on acetate as the sole carbon source. This proves that the phenotypes of all mutants are correct as expected.
In the gastrointestinal tract of mammals, Escherichia coli is the predominant facultative anaerobe [15], but makes up at most only 1 % of the microbiota [44]. to grow up to high numbers in the intestine, but compared to E. coli EDL933 Str R ΔpckA ΔppsA::Cam it is growing at a significantly higher concentration ( Figure 10, Figure 11, Figure 5 and Figure 6). Therefore E. coli EDL933 needs the ability to utilize gluconeogenesis in order to be able to maintain itself in the intestine next to E.  Figure 5, Figure 6, Figure 7, Figure   8), and based on the finding that E. coli EDL933 turns on virulence factors in a gluconeogenic environment [47], this leads to an hypothetical approach of defeating E. coli EDL933 in the intestine by finding a way to knockout the ability to use gluconeogenesis distictively in E. coli 44 EDL933. If this ability is taken away, E. coli EDL933 struggles to maintain in the mouse large intestine and therefore this might become a promising and effective treatment of EHEC related diseases. Based on several studies in which the intestine of newborns are colonized with E. coli Nissle 1917 and finding that this colonization supports their immune response to intestinal pathogens [26], this method of treatment right after birth might be a promising way of supporting colonization resistance against incursion of new and harmful microorganisms.