Introduction

Understanding microaerobic growth is important in appreciating interactions in microbial communities, in the infection process of certain pathogens, and industrially in how we might better manipulate commercial large-scale cultures for efficient and reliable product formation. A number of articles have examined effects of microaerobic conditions on organisms from ecological and health perspectives [63, 79, 112], photosynthetic microbe biology [22, 132] and in pathogen invasion [108, 113] and infection specifically by Neisseria [4], and Campylobacter [73]. Many large-scale industrial microbiological processes operate under partial aerobic conditions. For example, the production of PHB [70, 116, 131], polysaccharide based biopolymers [2, 16, 43], lycopene and carotenoids [92], IPP-terpene processes for pharmaceuticals [35, 99, 114], and alcoholic biofuels [91, 117] [9, 15] have been discussed in reference to limited oxygen culture conditions. Other examples of use of microaerobic production conditions have appeared for lactate [39, 101], acetoin [28], succinate [87], 2,3-butanediol [102], and H2S removal [129]. The metabolic reprogramming occurring under microaerobic conditions has been considered along with the effect of several regulators FNR, ArcAB, NarL, NarP [140]. The screening of strains for high oxygen utilization has been emphasized for identification of industrial production organisms [176].

In industrial practice, to produce high levels of product, high cell density is required that further taxes the ability to properly aerate the culture. Inadequate mixing in large reactors coupled with limiting oxygen solubility and gas transfer obstacles are serious problems affecting reliability and optimal productivity. In these large-scale processes inefficient production by a fraction of cells can have serious economic consequences as the non-productive oxygen limited cells not only do not produce the desired product in the normal amount, but also their metabolic by-products can affect the other cells in the bioreactor, limiting overall performance. Efforts to control such processes by reactor design, gas transfer systems, and optimized electronic sensor-response systems have made technological advances for bioprocessing. However, the fundamental question of the physiology of microbial host and genetic modification of the host to optimize oxygen utilization has been less completely addressed.

There is a considerable background on factors affecting regulation of respiration. Too much respiration and oxidation of carbon substrates, while being beneficial for cell energetics and for growth rate, lead to loss of carbon that does not go into product, especially if a reduced product is desired. Many reduced products of commercial interest, such as fuel molecules, are compounds more reduced than glucose, and many chemical intermediates for pharmaceuticals, lactate or monomers for making polymers, fatty acids, etc. also require reduction reactions and similarly optimal formation of these molecules requires limiting oxidation of the feedstock for a high carbon atom yield process.

Respiration pathways

Oxygen is an effective electronic acceptor and can provide a significantly higher ATP/glucose yield (more than 30 ATP per glucose under aerobic conditions vs only 2 ATP from the glycolysis pathway under anaerobic conditions). Aerobic cultures are, therefore, in general more robust than their anaerobic counterpart. However, since NADH is being consumed in the electron transfer chain (ETC), the intracellular NADH/NAD+ ratio has been shown to decrease significantly with increasing culture dissolved oxygen levels [139]. Hence, these opposing trends impose conflicting demands: a robust culture for rapid cell growth to achieve high biomass and high cell energetics under aerobic metabolism, and the opposing demand for NADH in product formation most readily attained under anaerobic conditions. Microaerobic conditions have thus been shown experimentally and theoretically to improve performance of a number of bioproduction systems. However, it is difficult to maintain a set dissolved oxygen level in a large-scale production environment due to incomplete mixing and balance with the rate of oxygen uptake by culture.

As part of its ability to adapt to different growth conditions, E. coli alters the composition of its respiratory system [159]. Important factors influencing aerobic respiration are listed in Table 1. The three types of respiratory components are (1) dehydrogenases, which carry out the oxidation of organic substrates and feed electrons into the mobile quinone pool, (2) quinones, which deliver reducing equivalents to the terminal oxidoreductases, and (3) oxidoreductases, which reduce the terminal electron acceptors [44]. The ETC of E. coli is composed of membrane-anchored dehydrogenases that reduce the quinone pool (ubiquinone-8, Q8) under respiratory conditions. Of these, the nuo and ndh pathways are most important in aerobic conditions. The quinone redox state is sensed by the ArcB protein and relayed through phosphorylation of the transcriptional regulator ArcA, to affect expression of genes of the TCA cycle and the electron transport chain; these levels of functional components are then adjusted to modify the cell’s respiration vs fermentative metabolism in an aerobic–anaerobic transition. The amount of each component is strictly regulated to optimize the respiratory chain according to the substrates present and the physiological needs of the cell. One important function of the respiratory chain is the maintenance of redox balance and the regeneration of NAD+ from NADH. Under aerobic growth E. coli normally makes two different NADH dehydrogenases, NAD I and NAD II, and two different terminal oxidases, cytochrome bo3 and cytochrome bd. The electron flux through these enzymes is dependent on the concentrations of the enzyme in the membrane, the NADH, quinone, and oxygen concentrations, and the steady-state characteristics of the enzymes (i.e. V max and K m values for NADH, quinone and oxygen).

Table 1 Factors influencing respiration
Full size table

E. coli cells regenerate NAD+ and generate proton motive force for ATP production through the respiratory chain. One way to reduce the activity of ETC and thus the amount of oxygen used is reducing the cytochrome protein levels [61, 78, 127, 128] or affecting assembly of the protein–membrane complex [95] and both of these ways affect the proteins responsible for the utilization of NADH. Another way is to control the level of quinone electron carrier involved in the transfer of electrons from NADH to essential membrane proteins during the process by inactivating its biosynthetic pathway [147] and then adding different amounts of an analog of quinone back to the culture of cells that are unable to synthesize the quinone, such as coenzyme Q1 [172] and noting the change in metabolites formed.

Regulation of proteins involved in the aerobic–anaerobic transition

The regulation of the aerobic–anaerobic shift in E. coli has been extensively studied. E. coli possesses a number of sensing/regulation systems for the response to availability of oxygen [3, 12, 30, 53, 65, 72, 90, 96, 122, 133, 134, 136, 157, 158] and channels electrons from donor to terminal acceptors such that the overall redox potential difference is maximized for any given growth condition. The adaptive responses are coordinated by a group of global regulators, which includes Fnr (fumarate, nitrate reduction) protein, and the two-component Arc (aerobic respiration control) system. With the initial onset of anaerobiosis, ArcA is activated, and if these conditions persist or become more anaerobic, Fnr is activated leading to the upregulation of ArcA amplifying its effect [72]. The oxygen levels (percent oxygen) that gave rise to half-maximal synthesis were reported as 0.02–0.04% for ethanol, acetate, and succinate, and 0.1% for formate [12]. The pO0.5, which is defined as the pO2 value resulting in a half-maximal expression, for expression of the adhE gene encoding alcohol dehydrogenase was ~0.08% oxygen [12] and for expression of cytochrome oxidases was measured [154] for cyd (maximal at 1–1.4%) and cyo shut-off at 1–4%. The global gene expression effects of Fnr and ArcA have been reported [11, 133, 134] with the Fnr transition occurring at 0.1–2% and Arc at 1–4%. The Arc system is a two-component regulatory system composed of ArcA, the cytosolic response regulator, and ArcB, the transmembrane histidine kinase sensor. ArcB is activated during the transition from aerobic to microanaerobic growth [66, 67] in response to the redox composition of the quinone pool [46] including menaquinone or dimethylmenaquinone [1, 13, 141] whose synthesis does not require oxygen [94, 103, 143, 159]. ArcB undergoes autophosphorylation, and the ~P group is transferred to ArcA by a His → Asp → His → Asp phosphorelay [45, 47, 84]. Consequently, the increased level of phosphorylated ArcA represses the synthesis of the citric acid cycle enzymes, while it activates the expression of cytochrome d oxidase and enzymes involved in fermentative metabolism [90, 96, 158]. The Fnr protein, [157, 160] is a transcription factor that coordinates the switch between aerobic and anaerobic metabolism at low oxygen levels. Fnr contains a Fe-S cluster that serves as a redox sensor. In the presence of oxygen Fnr is converted to a non-DNA binding monomer form [11, 53]. The Fnr system induces the expression of genes that permit anaerobically growing E. coli to transfer electrons to alternative terminal acceptors [51, 90]. Moreover, active Fnr elevates expression of arcA in anaerobic cells. Fnr also affects a number of genes in E. coli whose function has not been determined [75].

The concentration of certain anaerobic metabolites also affects ArcB function [96]. A mutant of Fnr that is active aerobically [68] has been used in metabolic experiments of the aerobic–anaerobic transition [137–139] and these studies suggested the recycling of Fnr to the active form was an important aspect. The Fnr protein cycle has been examined theoretically [152, 153] to prepare a more complete kinetic model of Fnr activity that included assembly of the iron-sulfur cluster and degradation by ClpXP protease [153]. The simulation of E. coli response to sudden oxygen starvation has been reported with software for analyzing and modeling the process [135].

A kinetic model of oxygen regulation of cytochrome production has been developed [124], and the metabolic flux analysis of various arc, fnr or other strains using C-13 labeling technique has been reported [118, 173, 174]. The new findings of additional factors affecting the anaerobic–aerobic transition such as small RNAs [48, 49, 97] add complexity to the models of the transition. For example, ArcZ directly represses arcB, and is itself repressed by the ArcBA system, providing a negative feedback loop [97]. Another example is FnrS, a highly conserved, anaerobically induced small sRNA, whose expression is strictly dependent on Fnr and negatively regulates many genes encoding enzymes with aerobic functions [19, 37]. These studies add to the wide impact of processes regulated by the Arc and Fnr systems under microaerobic conditions [13, 19, 37, 125]. The discovery of new proteins that influence components of the transition, for example, the SixA phosphatase [55, 100, 119], FrsA (fermentation protein) [81], and the major anaerobic nucleoid protein, Dan, that acts in DNA filament formation [89, 151] indicate there are further characters to consider in the aerobic–anaerobic transition in more detail.

Correlating the metabolic consequences of low oxygen with gene expression studies

The effects of low oxygen on gene expression and cell metabolism have been studied using C-13 labeling and flux analysis [144]. Studies have examined dynamics [34] and correlated models with experiments [23]. The general foundation concerning genes, regulators, and physiological implications of low oxygen have been reviewed [13, 17, 54, 64, 154]. As oxygen levels are reduced the cell will shift from the use of the Cyo pathway to the Cyd pathway, then at lower oxygen levels the fermentative pathway engages, oxygen using pathways are shut down and NADH formed in glycolytic processes is recycled by forming reduced products from pyruvate, e.g. lactate and ethanol. The change in metabolite pattern in response to availability of oxygen also modulates gene expression as well as enzyme activity. The flow through the citric acid cycle is inhibited by product accumulation. Acetyl-CoA is produced in oxygen-rich environments by pyruvate dehydrogenase complex (Pdh) or anaerobically by pyruvate formate lyase (Pfl). Pdh is inhibited by ATP, acetyl-CoA, and NADH [41, 50, 76]. Active Pfl is a radical form protein inactivated by oxygen [85] but the YfiD protein can reactivate Pfl [162, 169]. Reduced flux of pyruvate through Pfl and Pdh results in pyruvate accumulation that activates lactate dehydrogenase (Ldh) [164]. These soluble metabolites exhibit effects on gene expression [77] and can affect physiological properties of surrounding cells. Useful recent models of the central metabolic pathway [27, 74], the response of the Fnr system [152, 153], and the integration of “omic” and signal transduction data [31] have been published and serve as a framework in further modeling the transition.

The electron carrier quinones of the electron transfer chain

The lipid-soluble quinones that carry electrons within the membrane between electron input protein complexes and the cytochrome oxidases for reaction with the electron acceptor (oxygen) are generally ubiquinone with an isoprene tail of varying length in different organisms and this biochemistry and the methods for production of Coenzyme Q or menaquinone (vitamin K) for health have been reviewed [5, 6, 29, 69, 80, 103, 104, 146, 163]. Under more anaerobic conditions or when using a different electron acceptor, menaquinone or dimethylmenaquinone acts as the carrier [142, 166]. The composition and level of the quinone carrier pool is altered under differing conditions of electron acceptors [143, 165].

The biosynthetic pathways of the quinone carriers have been determined and the enzymes and corresponding genes of E. coli have been identified. Studies of the growth of various mutants and the levels of ubiquinone have been reported. There is a low amount of ubiquinone in the cell under normal aerobic growth conditions. However, this low concentration is more than sufficient since mutations resulting in a moderately reduced level do not have drastic effects on cell growth physiology but grow more slowly [115] and have hypersensitivity to thiols [170]. For example, ispA mutants disrupted for farnesyl diphosphate synthase have lower levels of ubiquinone-8 and menaquinone-8 (less than 13 and 18%), respectively [42]. UbiX mutants (an alternative to UbiD activity) have lower ubiquinone levels [52, 171]. The ability of ubiquinone-8 to serve in the electron transfer chain with type II dehydrogenases was somewhat limited in membranes and cells that did not contain phosphatidylethanolamine in the phospholipid [107].

Genetic approaches to control respiration and oxygen utilization

Recent work on regulating respiration and generating altered E. coli that could form lactate or other reduced compounds under aerobic conditions has progressed based on the literature of genes required for respiration and newly appreciated genes that have a role in redox transfer. This avenue was advanced by Palsson’s group in articles demonstrating aerobic fermentation properties on E. coli with several mutations [127, 128]. In that work the terminal cytochrome oxidase genes (cydAB, cyoABCD, and cbdAB) were inactivated and subsequent selection allowed a strain to be obtained that could carry out mixed acid or primarily lactate production (0.8 g/g lactate from glucose) under aerobic conditions. Additional removal of a quinol monooxygenase gene (ygiN) led to activation of ArcA aerobically. This strain formed D-lactate as a sole by-product under both oxic and anoxic conditions. The quinone pool changed from ubiquinones in normal cells to menaquinones in the mutant strain [128]. This physiological situation led to activation of the ArcB/ArcA system and altered the metabolic flux pattern with the flux through the tricarboxylic acid (TCA) cycle being greatly reduced, while glycolysis and formation of oxaloacetate were enhanced. The finding that transcriptomic results could be correlated with the in vivo function was encouraging for future gene manipulation strategies in the redox arena. In studies of E. coli mutants with deletions of three cytochrome terminal oxidases the results showed a fermentative metabolic pattern in the presence of oxygen [14]. The electron flux through cytochrome bd-II oxidase did not form a proton motive force for ATP generation and cells then conducted ATP synthesis only by substrate level phosphorylation reactions. Thus the P/O ratios can vary severalfold depending on the electron flux distribution through the respiratory chain. An interesting hypothesis based on a genome-scale metabolic model of E. coli was proposed [175]. The authors proposed that bacterial cells managed the protein composition of the cytoplasmic membrane for optimal ATP production under the growth condition by regulating the production of ATP by either the coupled ETC or substrate-level reactions and that membrane occupancy and total allowable protein level in the membrane may constrain cell metabolism, phenotypic properties and oxygen-energy physiology. The study of gene regulation and coordination of nitrogen and carbon metabolism by analysis of transcript levels of metabolic pathway genes and metabolic gene regulators in E. coli with mutations in cytochrome oxidases, and regulatory genes, fnr, and fur and others, also demonstrated the value of transcriptomics in metabolic analysis [82, 83]. The examination of the coordination of metabolic consequences of such genetic changes suggested an interconnection between cAMP and PII-Nitrogen regulatory systems and illustrates the connections of more complex networks that should be taken into account in practical applications related to metabolic engineering and defining optimal culture conditions.

The idea of control of respiration and of avoiding excessive loss of carbon to the formation of undesired products such as CO2 has been taken to the industrial organisms, Corynebacterium glutamicum [78] and Zymomonas mobilis [60]. A C. glutamicum strain that had a completely inactivated aerobic respiratory chain (cydAB qcr) was able to grow aerobically in complex medium but had a 70% reduced biomass yield and could also grow in glucose minimal medium after supplementation with peptone. The modified C. glutamicum strain displayed a fermentative metabolism with L-lactate as the major metabolite and succinate and acetate formed in lesser quantity. The observation that phosphofructose kinase overexpression is critical for high production of lactate in C. glutamicum under oxygen deprivation has been reported [155]. The use of the normally aerobic organism, C. glutamicum, under oxygen deprivation has been examined for high yield production of several commercial products [56, 57, 71, 106, 120, 121]. In the case of Z. mobilis, mutation of genes leading to respiratory deficiency generated strains with higher ethanol fermentation under aerobic conditions. The strains were also more thermotolerant and it was considered that the strains benefited from the lower amount of reactive oxygen species and reduced stress associated with oxygen respiration.

The use of oxygen in the electron transport chain has been studied in E. coli and many organisms. This membrane system contains various cytochromes and electron carriers, such as the quinone, ubiquinone. One way to try to control the extent of electron transfer chain (ETC) activity and thus the amount of oxygen used would be to control the cytochrome protein levels by controlling their expression [10, 109, 110]. This strategy exists in the native cyo and cyd systems that respond to different levels of oxygen [17, 109, 154]. Another viewpoint is to consider the level of quinone [5]; Zhu et al. [172] have shown by knockout of the biosynthesis pathway and adding small amounts of ubiquinone back to the culture, respiration could be controlled. Another approach is to modify the biosynthesis of the quinone electron carrier by altering expression of that operon encoding the enzymes of the biosynthetic pathway. Some of these alternatives have been explored previously; however, it is experimentally difficult to control the exact level of a partially-on system by these means, and generally it was found that cells behaved either as unaffected wild type or were completely inactive for the pathway, enhancing interest in other ways to finely control levels of such key molecules by genetic control of the network (Bennett and San, unpublished data). In general, control can be difficult since the level of the factor may be very low, for example, the quinone pool is only about 1 nmol/mg DCW [38, 52, 170, 171] so changes in expression of the operon encoding the biosynthetic pathway for the quinone can overshoot or undershoot the transition threshold level and make the ETC always off or always on.

The concept of manipulation of a key cofactor to control a large flux pathway

In cell metabolism, it is desired to control a large flux using a controller that can be regulated at an appropriate level, either at a defined fixed level or at a feedback controlled response level. At the genetic level, synthetic biology approaches of “gene circuits” have allowed control of cell responses [58, 59, 98, 130, 148, 150] by modifications in promoter strength [20, 33, 105], subjecting expression to exogenous parameters such as inducer or light [86, 149] or cell formed substances in feedback loops [21, 26, 36, 62, 93] using repressors and antirepressors [24], activators and layered circuits [111], inverters [40, 123], or RNA responsive elements that act primarily at the level of transcription or translation [7, 8, 18, 25, 32, 88, 126, 145] which affect the level of the appropriate enzyme(s) of a biosynthetic pathway.

An additional strategy to finely control a large metabolic flux by using a “metabolic transistor” approach has been proposed and its implementation feasibility has been examined [168]. In this approach a small change in the level or availability of a key participant in generating the large flux is controlled and the change in this basal level, considered as a base current in a transistor, has a large effect on the flux through the major pathway analogous to the change in current in an electrical transistor [156, 161]. There are a number of ways in which the level of a small molecule in the cell can be manipulated. Among these are the direct alteration of the expression of the biosynthetic pathway via promoter or translation control; another means to affect the level could be to degrade, excrete or otherwise bind the small molecule so it cannot participate in the reaction where it is required for the major flux.

In the “metabolic transistor” approach, the precursors leading to the biosynthetic pathway of the small molecule were being manipulated by effective partitioning of these precursor intermediates at the introduced node. This precursor control approach is particularly effective for a pathway network that consists of irreversible or close to irreversible reactions since the first step of the pathway network exerts the most control on its metabolic flux according to metabolic control theory. Wu et al. [168] has successfully demonstrated the control the ETC in wild-type E. coli by fine-turning the expression of geranyl diphosphate:4-hydroxybenzoate geranyltransferase from Lithospermum erythrorhizon (lePGT-1). The reaction catalyzed by lePGT-1 plays as a competing pathway on the substrates of the ubiquinone-8 (Q8) synthesis pathway, namely at IPP and 4-HB, and introduction of this new node is effective in affecting the flow through the Q8 biosynthetic pathway and subsequently the level of Q8. The Q8, which serves to deliver reducing equivalents between electron donors, such as NADH dehydrogenase (NDH), succinate dehydrogenase (SDH), and terminal electron acceptors, such as cytochrome oxidases or reductases, is an essential element under aerobic respiratory conditions [6] (Fig. 1). Hence, controlling the level of intracellular Q8 has a direct influence on the ETC activity.

Fig. 1
figure 1

A schematic showing the interplay/competition for NADH between the electron transfer chain (ETC) and a reduction reaction leading to reduced product formation. The metabolic transistor approach to control the ETC activity through Q8 manipulations is also shown. The intracellular ubiquinone Q8 concentration can be adjusted by dialing lePGT-1 expression up and down. Higher lePGT-1 expression increases the drainage of two key Q8 biosynthetic precursors, GPP and 4-HB, leading to a lower intracellular Q8 concentration. In addition, controlling the level of intracellular Q8 has a direct influence on the ETC activity. Lower the ubiquinone Q8 concentration will reduce ETC activity and increase NADH availability. Increased NADH availability will provide more reducing equivalence for reduced product formation. Gpp geranyl pyrophosphate, 4-HB 4-hydrobenzoate, G-4-HB geranyl-4-hydrobenzoate, lePGT-1 geranyl diphosphate:4-hydroxybenzoate geranyltransferase from Lithospermum erythrorhizon

Full size image

Using this approach, Wu et al. [167, 168] showed that it is possible to control the activity of the electron transfer chain in a tunable manner and manipulate the production of reduced products while limiting consumption of oxygen to a defined amount even under fully aerobic conditions. The intracellular Q8 concentrations showed a graded response to the lePGT-1 induction levels (IPTG concentrations). Furthermore, the authors have shown that it is possible to achieve a maximum theoretical yield of lactate production under fully aerobic conditions and minimize the “carbon burnt” (carbon lost as carbon dioxide) by limiting the ETC activity via lowering the intracellular Q8 concentrations.

Conclusions and future prospects

The complexity of respiratory metabolism and its control has been explored through biochemical and genetic experiments. Routes for the more precise control of oxygen utilization under conditions providing microaerobic culture conditions have been proposed and demonstrated. With the advent of synthetic biology and the ability to test numerous regulatory designs, particular applications useful to industrial processes that allow more reliable and controllable and optimized operations of large-scale cultures are likely to be implemented. The potential application of a “metabolic transistor” approach to control respiration, such as situations where biosynthesis or availability of a limiting factor is crucial for a biological process, is appealing. With further study of the variety of respiratory systems existing in wide diversity of microbes, the understanding of the controls that make an organism well suited to an ecological niche will become more apparent.