Transcriptomic and metabolomic profiling of Zymomonas mobilis during aerobic and anaerobic fermentations

Background Zymomonas mobilis ZM4 (ZM4) produces near theoretical yields of ethanol with high specific productivity and recombinant strains are able to ferment both C-5 and C-6 sugars. Z. mobilis performs best under anaerobic conditions, but is an aerotolerant organism. However, the genetic and physiological basis of ZM4's response to various stresses is understood poorly. Results In this study, transcriptomic and metabolomic profiles for ZM4 aerobic and anaerobic fermentations were elucidated by microarray analysis and by high-performance liquid chromatography (HPLC), gas chromatography (GC) and gas chromatography-mass spectrometry (GC-MS) analyses. In the absence of oxygen, ZM4 consumed glucose more rapidly, had a higher growth rate, and ethanol was the major end-product. Greater amounts of other end-products such as acetate, lactate, and acetoin were detected under aerobic conditions and at 26 h there was only 1.7% of the amount of ethanol present aerobically as there was anaerobically. In the early exponential growth phase, significant differences in gene expression were not observed between aerobic and anaerobic conditions via microarray analysis. HPLC and GC analyses revealed minor differences in extracellular metabolite profiles at the corresponding early exponential phase time point. Differences in extracellular metabolite profiles between conditions became greater as the fermentations progressed. GC-MS analysis of stationary phase intracellular metabolites indicated that ZM4 contained lower levels of amino acids such as alanine, valine and lysine, and other metabolites like lactate, ribitol, and 4-hydroxybutanoate under anaerobic conditions relative to aerobic conditions. Stationary phase microarray analysis revealed that 166 genes were significantly differentially expressed by more than two-fold. Transcripts for Entner-Doudoroff (ED) pathway genes (glk, zwf, pgl, pgk, and eno) and gene pdc, encoding a key enzyme leading to ethanol production, were at least 30-fold more abundant under anaerobic conditions in the stationary phase based on quantitative-PCR results. We also identified differentially expressed ZM4 genes predicted by The Institute for Genomic Research (TIGR) that were not predicted in the primary annotation. Conclusion High oxygen concentrations present during Z. mobilis fermentations negatively influence fermentation performance. The maximum specific growth rates were not dramatically different between aerobic and anaerobic conditions, yet oxygen did affect the physiology of the cells leading to the buildup of metabolic byproducts that ultimately led to greater differences in transcriptomic profiles in stationary phase.


Background
Recent high oil prices, concerns over energy security, and environmental goals have reawakened interest in producing alternative fuels via large-scale industrial fermentations. The potential and challenges involved in supplanting a substantial amount of petroleum derived transportation fuels with fuels derived from renewable resources such as ethanol from lignocellulosic materials has been the focus of several studies and reviews [1][2][3][4]. The development and deployment of ethanologenic microorganisms will be one critical component in the successful production of fuel ethanol in industrial-scale quantities.
Essential traits for an industrial microorganism include high ethanol yield, tolerance, and productivity (> 90% of theoretical, > 40 g L -1 , > 1 g L -1 h -1 , respectively); robust growth with simple, inexpensive growth requirements in conditions that retard contaminants (eg higher temperatures); and inhibitor tolerance, as reviewed previously [5]. Higher tolerance, productivity values and other positive industrial attributes have been reported for Z. mobilis, as reviewed previously [6]. Ethanol tolerance comparable up to 85 g L -1 (11% v/v) have been reported for Z. mobilis continuous culture and up to 127 g L -1 (16% v/v) in batch culture and productivities of 120-200 gL -1 h -1 in continuous processes with cell recycle [6]. Saccharomyces yeasts have been the preferred industrial biocatalyst for fuel ethanol production, although genetically engineered bacterial species such as Gram-negative bacteria Escherichia coli, Zymomonas mobilis, and Klebsiella oxytoca as well as Grampositive bacteria Bacillus subtilis and Corynebacterium glutamicum are in development to address commercially important inoculum requirements [5,7,8]. Indeed, a newly formed partnership between the DuPont and Broin companies will utilize recombinant strains of Z. mobilis for bio-ethanol fermentation from the lignocellulosic residues such as corn stover [9].
Z. mobilis ferments glucose, fructose, and sucrose producing ethanol and carbon dioxide via the Entner-Doudoroff (ED) pathway, utilizing pyruvate decarboxylase and alcohol dehydrogenase enzymes (see [6,[10][11][12] for reviews). Z. mobilis is not a classic facultative organism, rather it is aerotolerant, negating oxygen requirements in fermentations and the need for expensive oxygen transfer. The unusual physiology of Z. mobilis generates only one mole of ATP per mole of glucose, which results in low biomass production and greater carbon being available for fermentation products under anaerobic conditions. Its desirable ethanologenic attributes also include: high sugar uptake rates, near theoretical ethanol yields, high ethanol tolerance and generally regarded as safe (GRAS) status. Wildtype Z. mobilis can only utilize a limited range of substrates; however, it is amenable to genetic manipulation and recombinant strains have been developed to ferment pentose sugars such as xylose and arabinose [13][14][15]. Seo et al (2005) reported the first genome sequence for Z. mobilis ZM4 [16] and the U. S. Department of Energy's Joint Genome Institute (JGI) http://www.jgi.doe.gov/ has announced plans to sequence the genomes of additional Z. mobilis strains in the near future. The ZM4 genome sequence provides new opportunities for fundamental insight into the physiology and gene function and regulation of this unique microorganism and likely improvements in strain development [17]. The presence of oxygen during Z. mobilis fermentations has been observed to negatively affect cell and ethanol yields with acetic acid and acetaldehyde accumulating in the medium [12,[18][19][20]. However, there are also reports that respiration inhibition by cyanide stimulates Z. mobilis growth aerobically [21]. Many aspects of end-product inhibition and basic aspects of Z. mobilis physiology remain unexplored [11].
In this study, we combined transcriptomic profiling with metabolomic measurements of aerobic and anaerobic Z. mobilis fermentations to elucidate the molecular mechanisms of ZM4s growth with and without oxygen. Metabolomics is the study of the low molecular weight metabolites present in and around a biological organism at a given time [22,23]. To date, there are few reports of bacterial mRNA expression-profiling using whole genome microarray analysis conducted in conjunction with metabolomics [24]. We confirmed maintenance of anaerobiosis was important for maximizing ethanol yields since acetaldehyde, acetate, lactate and acetoin accumulated and ethanol was present in lower concentrations under controlled aerobic conditions. Our data revealed alterations in mRNA expression profiles, differences in intra-and extracellular metabolites, as well as identification of differential expression for coding sequences predicted by TIGR http://cmr.jcvi.org/tigr-scripts/CMR/ GenomePage.cgi?org=ntzm01 and not predicted in the primary annotation. These data provide global insight into potential molecular mechanisms of Z. mobilis aerobic and end-product stress responses.

ZM4 grows and consumes glucose faster under anaerobic conditions
The presence of oxygen negatively affected glucose consumption and growth in Z. mobilis ZM4 fermentations (Fig. 1). Anaerobic fermentation led to a maximal culture density of 7.0 OD 600 units approximately 9 h post-inoculation, while Z. mobilis did not reach its highest culture density of 6.5 OD 600 units until 13 h post-inoculation under aerobic conditions despite initial inocula concentrations being slightly greater than the former condition. Z. mobilis also consumed glucose more slowly under aerobic conditions, with more than half of the initial aerobic glucose concentration (105 mM) remaining 9 h postinoculation. Under anaerobic conditions 99.5% of the glucose had been utilized at this time point. When Z. mobilis growth reached its peak after 13 h under aerobic conditions 14% of the glucose remained with the remainder consumed without cell growth (Fig. 1). Despite these differences and differences in extracellular metabolite production (below), the maximum specific growth rates were not dramatically different, which were estimated to be 0.45 h -1 and 0.55 h -1 between aerobic and anaerobic conditions, respectively. Fermentor pH, dissolved O 2 tension (DOT), and agitation speed were well-controlled, which was indicated by the mean fermentor pH, DOT, and agitation values for each condition (see Additional file 1), and DOT trend data for each condition (see Additional file 2).

Z. mobilis extracellular metabolite production
Previous reports have indicated that ethanol production was decreased and other end-products such as acetaldehyde, acetate and acetoin were increased under aerobic conditions [12,[18][19][20]25]. GC and HPLC were used to quantify and compare the kinetics of ethanol, acetate, acetaldehyde, lactate and acetoin production during aerobic and anaerobic fermentation processes and extracellular metabolites were often measured by more than one approach, which confirmed the observed trends. The more rapid production of ethanol under anaerobic conditions also corresponded with increased glucose uptake and growth under these conditions (Fig. 1). The ethanol concentration remained relatively stable post-peak production in anaerobic fermentations. In contrast, at 13 h the ethanol concentration dropped sharply from 52.7 mM to 3.2 mM at the end of the 26 h fermentation during aerobic fermentation. The decrease in ethanol concentration during this time was matched in nearly stoichiometric increases of acetate production, which went from 8.4 mM at 13 h to 72.1 mM at 26 h (Fig. 2). GC data also showed several other minor unidentified metabolites were being produced during aerobic fermentations relative to anaerobic conditions (data not shown). The concentration of acetaldehyde increased during the exponential growth and dropped appreciably during stationary phase, while acetoin was detected during stationary phase (Fig. 2). Lactate trended similarly to ethanol and acetoin profiles for respective conditions. The anaerobic ethanol yield at 13 h was 0.497 g/g of glucose or 97% theoretical and less than 1% of the yield went to solvent products other than ethanol or CO 2 (Fig. 1). In contrast, under aerobic conditions at 13 h 0.14 g/g of glucose or 27% theoretical yield was obtained for ethanol and by 26 h around 3.2 mM ethanol remained (Fig. 1). The total measured solvents produced aerobically was 101 g/L, which was approximately 50% of theoretical total solvent yield at 13 h and at 26 h these figures had dropped to 76 g/L or approximately 37% of theoretical total solvent yield. As cell biomass (as measured by optical density) was approximately equivalent under the two conditions more carbon went to maintenance energy under aerobic conditions.

Z. mobilis intracellular metabolomic profiles
The physiological status of ZM4 was investigated further by GC-MS analysis of stationary phase intracellular metabolomic profiles for each respective condition. GC-MS identified a large number of metabolites, however, we present those that showed differences in relative abundance and showed some statistical significance. We observed differences in relative abundance profiles for metabolites related to central carbon metabolism, although only several were considered significantly different during the stationary phase time point we examined. We present a relative comparison of the metabolite difference from two conditions in keeping with current practices, while recognizing accessing metabolites like ATP will require modified procedures [22]. Metabolite identification and analysis gave 20 metabolites that were different between anaerobic and aerobic cultures, with five out of the 20 less abundant in anaerobic Z. mobilis. Twelve metabolites were considered significantly different with a p-value less than 0.1 ( Table 1). The inclusion of a greater number of replicates would have increased analysis power and possibly increased the number of metabolites considered significantly different between anaerobic cells and those exposed to oxygen.
Glucose-6-phosphate and 2-phosphoglycerate were 2.7 and 3.8 fold more abundant under anaerobic conditions (p = 0.089 and 0.053, respectively), while others in the ED pathway may have been at greater levels under aerobic conditions. The ability to detect labile phosphorylated intermediates such as these, in the absence of accumulation phosphate was indicative of adequate sample preparation for GC-MS analysis of the metabolites analyzed in the present study. Aerobic Z. mobilis contained more of the glucose substrate, as well as glycerate, and gluconate as compared to the anaerobic cells, while other intermediate sugar metabolites that included glucose-6-phosphate, 2phosphoglycerate and mannose-6-phosphate were more abundant within the anaerobic fermenting Z. mobilis ( Table 1). The transcriptomic profiles of Entner-Dondoroff and pyruvate metabolic pathways showed higher expression values for these genes under anaerobic conditions even though extracellular glucose had been consumed by 26 h (Fig. 1, see Additional file 3). All of the differentially detected amino acids were at lower concentrations within anaerobic fermenting Z. mobilis ( Table 1). The other metabolites showing differences included lactate, 4-hydroxybutanoate, ribitol, trehalose and unknown metabolites. Trehalose was 2.9-fold more abundant within aerobic Z. mobilis compared to anaerobic cultures, and 4-hydroxybutanoate was the most abundant metabo-Z. mobilis extracellular fermentation product analysis lite detected within anaerobic condition at 7.6-fold higher levels compared to aerobic conditions (Table 1).

Transcriptome comparison of Z. mobilis aerobic and anaerobic fermentations
The global transcriptional response of Z. mobilis ZM4 to aerobic stress was examined in a time series DNA microarray experiment for early exponential phase (3 h postinoculation) and stationary phase (26 h post-inoculation) under aerobic and anaerobic conditions with wholegenome microarrays. In this study, we have presented experimental data of 166 differentially expressed genes at statistically significant values (-log 10 (p) = 5.43) with a number of 11 putative protein-coding sequences in strain ZM4 (ATCC31821) that were not originally described in the primary genome annotation (see Additional file 4, 5).
We have deposited the entire microarray dataset including exponential and stationary phase at Gene Expression Omnibus (GEO, http://www.ncbi.nlm.nih.gov/geo/) database with the accession number of GSE10302 so interested parties can conduct their analyses.
Gene expression during the early exponential phase was surprisingly similar between anaerobic and aerobic conditions used in the present study (Fig. 3A). Eight genes were discovered that were considered differentially expressed at a significant level utilizing the ANOVA model using the False Discovery Rate (data not shown). Of these, only one gene (ZMO1752, encoding a hypothetical protein) showed a greater than a two-fold difference in relative expression levels. To confirm the microarray results, seventeen genes involving in ED and pyruvate pathways and different cellular functions were chosen for the qPCR analysis (see Additional file 6, 7). The data showed that qPCR was more sensitive and showed greater fold change differences compared to the microarray analysis, which was in keeping with previous reports [26]. We describe here only a rigorous analysis of stationary phase genes in this study and allow interested parties to conduct their own analyses on the entire dataset, which is available publicly through the GEO database. In the stationary phase 166 genes were significantly differentially expressed between anaerobic and aerobic conditions ( In the presence of oxygen, genes related to regulation, cell processes, transport, and unknown function showed greater expression as compared to anaerobic conditions. Nearly half of the genes showing greater expression aerobically remain uncharacterized (see Additional file 5).
In the stationary phase, ED pathway mRNAs such as glk, zwf, pgl, pgk, and eno as well as ethanol fermentation gene transcripts like pdc and adhB were shown to be more abundant (at least two-fold) under anaerobic conditions by microarray analysis (see Additional file 3, 4, 7). In this study we observed that one arginyl-tRNA synthetase involved in ribosome-mediated polypeptide synthesis, eight genes related to ribosomal protein synthesis and amino acid and co-factor biosynthetic genes such as leuC, trpB, argC, ilvI, ilvC, thrC, thiC, and ribC showed greater expression under anaerobic conditions (see Additional file 4). Metabolomics data showed that a number of detected amino acids were generally less abundant in anaerobic fermenting cells compared to aerobic cells (Table 1).
Our microarray data identified a cysteine desulfurase (ZMO1022), thiosulfate sulfurtransferase (ZMO1460) and a [2Fe-2S] binding domain family protein (NT01ZM1467) as being up-regulated under aerobic conditions (see Additional file 5) and suggested the metabolism of sulfur compounds was impacted by aerobic conditions. However, we did not observe any trends related to sulfur-containing metabolites in our metabolomics dataset. Genes related to sensing and responding to environmental signals including three chemotaxis genes cheX (ZMO0084), motD (ZMO0641) and fliD (ZMO0651), six transcriptional regulators including a two-component signal transduction (TCSTS) histidine kinase (ZMO1216) and response regulator (ZMO1387) were up-regulated during aerobic fermentation (see Additional file 5). We also observed that several ATP synthase subunit genes (alpha and beta) were expressed less under aerobic conditions. The expression of NAD synthetase gene (nadE), involved in nicotinamide adenine dinucleotide de novo biosynthesis and salvage pathways was approximately 9-fold greater in the presence of oxygen (see Additional file 5). Several flavoprotein transcripts, nitroreductase (tdsD) and flavodoxin (nifF), were approximately 9-fold more abundant under aerobic conditions.
Expression of a number of stress response genes was found to be greater in the presence of oxygen or by-products in the stationary phase. Under aerobic conditions ZMO1097 encoding a thioredoxin was induced approximately four-fold, ZMO1830 (fdxB) encoding a ferredoxin showed six-fold induction, ZMO1732 (ahpC) encoding an alkyl hydroperoxide reductase showed 18-fold greater expression, ZMO0279 encoding a cold-shock protein was induced two-fold, and a glutathione S-transferase family protein encoded by ZMO1118 was induced eight-fold. The E. coli alternative sigma factor rpoH plays an important role in overcoming oxidative stress responses [27] and rpoH (ZMO0749) was induced approximately 32-fold under aerobic conditions via q-PCR in our study (see Additional file 7). Other regulators with greater expression levels under aerobic condition include ZMO1121 encoding a MerR family regulator, ZMO1216 encoding a two-component signal transduction histidine kinase, ZMO1387 encoding a two-component response regulator, and ZMO1063 (pspA) encoding a sigma 54-dependent transcription suppressor (see Additional file 5). A number of these microarray expression values were confirmed by qPCR (see Additional file 5, 7).
The microarray data also indicated that a number of CDS predicted by TIGR but not present in the primary annotation [16] were differentially expressed between the two conditions. Transcripts for NT01ZM1467 encoding a Fe-S binding domain family protein and nine other ORFs with unknown functions were more abundant under aerobic conditions representing approximately 9% of all the genes up-regulated under these conditions (see Additional file 5). The only example of an up-regulated gene (~six-fold) without primary locus identification during anaerobic fermentation was NT01ZM0869, which encodes a putative arginyl-tRNA synthetase (see Additional file 4).

Discussion
Z. mobilis has a number of positive attributes as an ethanogen and is a leading current generation candidate microorganism for use in commercial scale fermentations to produce fuel ethanol [6,10]. In the present study, we confirmed oxygen levels were important in Z. mobilis

A B
Hierarchical cluster analysis of significantly differentially expressed ZM4 genes for aerobic and anaerobic fermentations at 3 and 26 h Anaerobic (26h) Anaerobic (3h) Aerobic (3h) Aerobic (26h) growth rates, ethanol production and altered metabolite pools ( Fig. 1, 2, Table 1). Anaerobic Z. mobilis fermentations utilized glucose more rapidly and grew more quickly with concomitant increases in ethanol productivity and yield as compared to aerobic Z. mobilis cultures (Fig. 1). Both intra-and extracellular lactate was identified as being more abundant in aerobic stationary phase Z. mobilis (Table 1, Fig. 2). Our data were also consistent with previous reports [28,29] showing metabolites such as lactate, acetate, and acetoin were more abundant during aerobic fermentation ( Table 1, Fig. 2). Ishikawa et al. [19] proposed a model to explain the effect of oxygen on the NADH reducing power pool for the conversion of glucose into ethanol as the major end-product. They suggest that under aerobic conditions NADH is limiting due to it being oxidized by the NADH oxidase and therefore unavailable for reduction of acetaldehyde to ethanol. These data combined with lower intracellular glucose and ED pathway intermediates (Table 1) would agree with such a model and previous reports [30,31]. While we did not obtain the data for NAD + , NADH, ATP metabolites, altered redox balance with concomitant increases in other metabolites produced such as acetate, lactate, acetoin, and acetaldehyde in the presence of oxygen may account for lower ethanol production (Table 1, Fig. 2). We were able to observe differences in gene expression between exponential phase conditions by microarray, but not at levels considered significant using the Bonferroni multiple testing method (p < 0.05). Quantitative PCR assays showed greater sensitivity than array analyses but the majority of genes showed little difference in expression levels between aerobic and anaerobic conditions in early exponential phase. Each culture had more than doubled by the early exponential growth phase sampling point and this time point should have mitigated any potential inocula effects. The observation that there was an absence of a lag phase indirectly supports the idea that there was neither a large perturbation nor large regulatory response by the microbe upon inoculation into the fermentors (Fig. 1). The large differences in exponential and stationary phase transcriptomic profiles imply ZM4 is aerotolerant and oxygen affected the physiology of the cells leading to the buildup of metabolic byproducts, which ultimately led to greater expression differences in stationary phase. Seo et al (2005) describe the Z. mobilis ZM4 (ATCC31821) genome as consisting of a single chromosome [16], which we utilized for probe design and microarray fabrication. The same Z. mobilis ZM4 (ATCC31821) strain utilized in this study contained plasmid DNA, which was in keeping with an earlier report describing the nomenclature and derivation of Z. mobilis strains [32]. Therefore, the array data in the present study may not fully represent the differences between aerobic and anaerobic conditions since these plasmid DNA sequences were unavailable. We used a multiplex array format with an average probe length of 36 nucleotides and were able to detect significantly differentially expressed genes. However, the use of shorter probes likely affected the sensitivity of the microarray analyses [33]. The proper identification of putative coding sequences and the use of systems biology tools will be important in identifying genome features that can be altered for improved ethanol production. The JGI's plans to sequence the genomes for additional Zymomonas mobilis strains, including strain ZM4 will likely improve gene prediction models and elucidate potential coding sequences present on plasmid DNA.
The effects of oxygen on Z. mobilis fermentation performance parameters and physiology have been examined in a number of previous studies [18][19][20][21]25,31,[34][35][36][37]. However, this is the first report we are aware of that examines its effects on whole genome transcriptomic and metabolomic profiles of the bacterium under rigorously controlled conditions. The Z. mobilis strain CP4 glf, zwf, edd and glk genes for glucose uptake and utilization form an operon and coordinate gene expression appears to result from a complex pattern of mRNA degradation [38][39][40]. The ZM4 glf, zwf, edd and glk genes are similarly organized and all showed greater expression in stationary phase under anaerobic conditions, which was about 20, 9, 2, and 10-fold respectively more than that of aerobic condition, respectively. However edd differential expression did not meet our stringent significance criterion with alog 10 (p) = 5.2. The glf gene encodes a glucose-facilitated diffusion transporter protein, which showed greater expression (approximately 20-fold) under anaerobic conditions even in the absence of detectable amounts of glucose in the medium at 26 h (Fig. 1). ZMO1859 showed 5fold greater expression in the absence of oxygen. ZMO1859 is a putative OprB family carbohydrate-selective porin containing a pfam04966 (OprB family) conserved domain with an E-value score of 5e -64 and 27% identity to Pseudomonas aeruginosa outer membrane protein OprB [41]. P. aeruginosa OprB transports glucose, mannitol, glycerol, and fructose. The E. coli expression of pathways for hexose uptake and metabolism has been shown to be up-regulated in the response to oxygen deprivation [30]. A number of ED and pyruvate biosynthetic pathway genes (gntK, edd, eda, gap) were showed higher expression levels under anaerobic conditions at 26 h when compared to the equivalent aerobic time point, but at levels not considered significant (see Additional file 3). Previous physiological studies have indicated that enzymatic activities for glucokinase, glucose-6-phosphate dehydrogenase, glyceraldehyde-3-phosphate dehydrogenase, ATPase, pyruvate kinase and alcohol dehydrogenase are decreased when oxygen partial pressure increases [31]. The intracellular metabolomic data suggests glucose present in anaerobic cells rapidly becomes phosphor-ylated, providing substrate for further glucose utilization (see Additional file 3). Expression of the pdc and adhB genes involved in Z. mobilis ethanol production was increased at least two-fold during anaerobic fermentation and was consistent with fermentation product and growth data (Fig. 1, 2, see Additional file 4).
These data suggest oxygen affects how carbon is utilized and partitioned in Z. mobilis under highly aerobic fermentations leading to the buildup of inhibitors such as acetaldehyde and acetate and to the accumulation of nonreducing sugars like trehalose that contribute to overall lower ethanol yields. The accumulation of acetate at the end of the aerobic fermentation was much more pronounced than seen in a number of previous studies, although the highly aerated fermentors with DOT measurement used in this study likely meant the culture conditions were quite different from previous batch fermentations without DOT measurement. In this study, we used an agitation rate of 700 rpm in conjunction with sparging air at 2.5 L/min to maintain fully aerobic conditions as previous experiments had demonstrated a lower agitation rate and this airflow rate was insufficient to maintain fully aerobic conditions as cell density increased (data not shown). Putative aldehyde dehydrogenase genes such as ssdA (ZMO1754) did not show differential expression in our study, but may potentially play a role in conversion of acetaldehyde to acetate in conjunction with possible reverse alcohol dehydrogenase reactions. However, much more detailed studies are required to understand the large accumulation of acetate under highly aerobic conditions. Kalnenieks [11] reviewed the physiology of Z. mobilis, identified a number of unanswered questions about its energy metabolism, and determined that the structure and physiological role of the Z. mobilis respiratory chain remains to be fully elucidated. Kalnenieks et al [37] concluded that Z. mobilis contains a single NAD(P)H dehydrogenase encoded by ndh gene, and that its inhibition results in decreased respiration and improved aerobic growth. While we were not able to measure NAD, we did observe that ndh (ZMO1113) encoding NADH dehydrogenase was down-regulated in aerobic stationary phase cultures (see Additional file 4). A number of other respiratory genes were also down-regulated under these conditions. These included ZMO0022, ZMO1571, ZMO1572 and ZMO1844 encoding a putative Fe-S oxidoreductase, NADH dehydrogenase, cytochrome bd-type quinol oxidase subunits 1 and 2, and oxidoreductase genes, respectively (see Additional file 4). The stimulation of aerobic growth is proposed to be due to redirection of the NADH flux from respiration to ethanol synthesis resulting in less accumulation of toxic intermediates [11]. Several models have been proposed to account for the distribution of reducing equivalents between putative respiratory gene products and alcohol dehydrogenase reactions [11]. In one model both alcohol dehydrogenase isozymes catalyze ethanol synthesis and oxidation back to acetaldehyde and in another, the two isozymes operate in opposite directions. Under aerobic conditions, ethanol yield peaked around 13 h followed by net ethanol and acetaldehyde oxidation with concomitant increases in acetate and acetoin concentrations while approximately 0.66 mM of glucose remained in the medium at 26 h (Fig. 1, 2). We observed that expression of adhB (alcohol dehydrogenase II) in anaerobically cultured cells was approximately three-fold greater than when compared to aerobic cells at 26 h (see Additional file 4). adhA (alcohol dehydrogenase I) showed approximately 1.5-fold higher expression aerobically when compared to anaerobic cells, but did not meet our stringent criterion of a -log 10 (p-value) of 5.4. Z. mobilis contains two alcohol dehydrogenases, AdhB, an iron-containing alcohol dehydrogenase inactivated by oxygen and AdhA, a zinc-containing alcohol dehydrogenase that is resistant to oxidation [36]. Our adhB expression data were consistent with potential iron limitation and/or differential inactivation of AdhB under aerobic conditions. We identified, for the first time to our knowledge, transcripts for the putative respiratory gene ZMO1814 (rnfA), encoding a putative NADH:ubiquinone oxidoreductase subunit that were expressed more greatly (3.3-fold) under aerobic conditions. The Rhodobacter capsulatus rnf gene products form a membrane complex involved in electron transfer to nitrogenase, however no physiological information is available for the rnf gene products of Z. mobilis, which does not fix nitrogen [42]. Although Sootsuwan et al. [42] demonstrate Z. mobilis has a cyanide-resistant terminal oxidase and could detect little difference between aerobic and anaerobic exponential phase ubiquinol oxidase activities, they documented significantly more ubiquinol oxidase activity in stationary phase cell membrane prepared anaerobically. We observed that expression of the Z. mobilis cytochrome bd-type quinol oxidase subunits genes (cydA/B) was slightly greater in stationary phase under anaerobic conditions. Trace amounts of oxygen, below detection limits, might have been present in anaerobic fermentors during stationary phase (see Additional file 1, 2). However, E. coli cydAB genes show greater expression under anaerobic conditions due to the change in negative supercoiling status and it has been suggested this may provide a mechanism for increasing cytochrome bd levels in response to environmental stress [43]. Sootsuwan et al. [42] proposed a number of possible physiological roles for a Z. mobilis respiratory chain including: controlling NADH levels to control ethanol production, reducing intracellular oxygen concentration and maintenance of NADH levels inside cells. The present study iden-tified a number of putative genes involved in respiration as being expressed differentially depending upon conditions. However, it is clear much remains to be done to fully elucidate their function in redox balance and overall regulation. The contribution of individual genes in electron transfer and metabolic processes need to be followed up with targeted deletion mutant studies. The most surprising finding was the minimal impact of oxygen on the transcriptomic profiles in early exponential. Mutant studies would also be useful in assessing the physiology of early exponential growth phase aerobic Z. mobilis cells.
Ribitol was the most abundant differentially detected intracellular metabolite by GC-MS, at approximately 17fold greater levels within aerobic cells compared to anaerobic cells ( Table 1). The addition of various straight-chain alditols including ribitol has been reported to lead to the formation of novel E. coli phospholipids [44]. However, we did not conduct detailed phospholipid fatty acid (PLFA) analysis, nor target intact membranes. Ribitol and adenosine (latter was not considered significantly differentially expressed) are also both flavin adenine dinucleotide (FAD) components, which may reflect differences in redox potential and electron transfer between the two different conditions. Further studies are required to elucidate the role of ribitol and other metabolites like 4-hydroxybutanoate in Z. mobilis physiology and identify the unknown metabolites we detected. GC-MS based metabolomics provides a useful platform to examine a broad range of metabolites such as sugars, sugar alcohols, sugar acids, organic acids, amino acids, fatty acids, sterols, secondary metabolites, phenolic glycosides, alkaloids, purines, pyrimidines, and nucleosides [22,23,45,46]. However, a number of classes of metabolites cannot be detected by GC-MS and ribitol may not be the largest metabolite difference between these states. The differences of intracellular lactate indicated this metabolite was more abundant under aerobic conditions (Table 1), which was consistent with levels detected in the medium supernatant (Fig. 2).
Several ZM4 genes related to oxidative stress responses were up-regulated under aerobic, stationary phase conditions, including ahpC encoding alkyl hydroperoxide reductase, ZMO1097 encoding thioredoxin, and ZMO1118 encoding a glutathione S-transferase family protein (see Additional file 4). However, many systems involved in oxygen detoxification and redox homeostasis were not differentially expressed. There were no significant gene expression differences for classical oxidative stress response genes such as peroxidase (ZMO1573), catalase (ZMO0928), and iron-dependent superoxide dismutase (ZMO1060) when aerobic cultures were compared to anaerobic cultures. The glutathione system related genes such as gor (ZMO1211), gshB (ZMO1913) and ZMO1556 were also not affected significantly (data not shown). Pyrococcus furiosus exposed to gamma irradiation, which generates hydroxyl radicals and oxidative stress showed many systems involved in oxygen detoxification and redox homeostasis appeared to be constitutively expressed [47]. Furthermore, genes encoding superoxide dismutase, catalase, nonspecific peroxidases or thioredoxin reductase were not significantly expressed in response to air exposure in the obligate anaerobe Methanosarcina barkeri [48]. The Z. mobilis transcriptome analyses also suggested a relationship between aerobic conditions (or metabolic end-products) and cellular iron requirement (or limitation) as indicated by higher levels of expression of genes involved in iron sequestration and uptake. Transcripts for Z. mobilis iron-uptake and scavenging genes including pbuA (ZMO0188), feoB (ZMO1541), ZMO1463, and ZMO1847 (see Additional file 5) were all more abundant under aerobic conditions in stationary phase. A number of other studies such as those described by Brown et al [26] and Williams et al [47] have shown linkages between oxidative stress and expression of genes related to iron-uptake and storage.
The Z. mobilis ZMO1107 and ZMO0347 genes were upregulated under anaerobic conditions (see Additional file 4) and are annotated as transcriptional regulator and hypothetical proteins, respectively [16]. BLASTP analyses indicated that ZMO1107 is similar to the E. coli global regulator leucine-responsive regulatory protein (Lrp) and ZMO0347 is similar to the E. coli global regulator Hfq, sharing about 40% and 60% identity to the E. coli Lrp and Hfq proteins, respectively. E. coli Lrp regulates transcription of many Lrp regulon genes with leucine as a co-regulator and also acts as a determinant of chromosome structure to coordinate cellular metabolism with the environmental nutritional state [49,50]. We observed different levels of intracellular leucine in the present study, but not at highly significant levels ( Table 1). The best-studied Lrp family regulator is the E. coli Lrp, which affects the expression of at least 10% of all E. coli genes. Microarray analysis of gene expression profiles for E. coli Lrp + and Lrpstrains indicates more than 400 genes are significantly Lrp-responsive with expression for 147 genes lower in Lrp + compared to Lrpcells [50,51]. In addition, Lrp has been suggested to play an important role in E. coli stationary phase regulation. Out of the 200 genes induced upon entrance into stationary phase, Lrp affects nearly threequarters of them in E. coli [50]. It is well known that E. coli RpoS alternative sigma factor controls stationary phase gene expression [52,53]. The genome sequence shows that Z. mobilis has rpoH (ZMO0749), rpoE (ZMO1404), rpoN (ZMO0274), and fliA (ZMO0626, sigma F) genes but no identifiable rpoS gene [16]. ZMO1107 may play a role in stationary phase and stress regulation in Z. mobilis, but further studies are required to determine what func-tional overlap, if any, this protein has with RpoS-like functions in this bacterium.
A putative Z. mobilis Hfq global regulator was induced during anaerobic fermentation ( Table 2). Hfq is a bacterial member of the Sm family of RNA-binding proteins, which acts by base-pairing with target mRNAs and functions as a chaperone for non-coding small RNA (sRNA) in E. coli [54][55][56]. Hfq is involved in regulating various processes and deletion of hfq has pleiotropic phenotypes, including slow growth, osmosensitivity, increased oxidation of carbon sources, and altered patterns of protein synthesis in E. coli [54,57]. E. coli Hfq has also been reported to affect genes involved in amino acid biosynthesis, sugar uptake, metabolism and energetics [58]. The expression of thirteen ribosomal genes was down-regulated in hfq mutant background in E. coli [58]. Hfq also up-regulated sugar uptake transporters and enzymes involved in glycolysis and fermentation such as pgk and pykA, and adhE [58]. Hfq is also involved in regulation of general stress responses that are mediated by alternative sigma factors such as RpoS, RpoE and RpoH. An hfq null mutant of E. coli exhibits strongly reduced RpoS levels by impairing rpoS translation [59]. Cells lacking Hfq induce the RpoEmediated envelope stress response and rpoH is also induced in cells lacking Hfq in E. coli [58]. In our study, Hfq was less abundant in aerobic fermentation condition in ZM4 at 26 h post-inoculation (see Additional file 4) with eight ribosomal protein genes including rplF, rplY, and rpmE down-regulated (see Additional file 4). Meanwhile, glucose uptake genes such as glf and oprB, ED pathway genes such as glk, zwf, pgl, eno, and ethanol fermentation genes pdc and adhB were less abundant under aerobic condition in Z. mobilis (see Additional file 4). In addition, our microarray and qPCR data indicate that rpoH is also induced in aerobic condition (see Additional file 6) with low abundance of hfq transcripts (see Additional file 4). The association of Hfq with the expression of alternative sigma factor RpoH, ribosomal proteins and glycolytic enzyme in Z. mobilis is similar to that of E. coli as discussed above, which may indicate a similar conserved role of Hfq in both E. coli and Z. mobilis. However, further studies such as mutagenesis are required to elucidate the Z. mobilis Hfq regulon.

Conclusion
Our study has provided insights into transcriptomic and metabolic profiles of the model ethanogenic bacterium Z. mobilis during aerobic and anaerobic fermentation under controlled fermentation conditions for the first time. It is unlikely the oxygen levels we used in the present study would be as high in an industrial setting, even in a major perturbation to the fermentation. However, our study showed oxygen affects maintenance energy, how carbon is utilized and partitioned in Z. mobilis fermentations and can lead to the buildup of inhibitors such as acetaldehyde and acetate under highly aerobic conditions, which contributed to lower ethanol yields. The most surprising finding was that while high oxygen concentrations resulted in slower growth, there seemed to be minimal impact on early exponential growth phase transcriptomic profiles. However, differences in energy-spilling pathways due to oxygen and uncoupled growth later led to large differences in ethanol yield and transcriptomic profiles between anaerobic and aerobic stationary phase cultures. Our study also identified a range of genes such as cydA/B, rpoH or ZMO0347 that could be targeted for deletion to better understand Z. mobilis physiology and coordinate regulation. Future studies examining the transcriptomic profiles for aerobic, microoxic and anaerobic steady states, the dynamics of transition between steady states and the use of mutant strains are warranted to provide greater insights into Z. mobilis physiology and gene regulation. Finally, our microarray based on the TIGR annotation has identified a number of genes that were not originally annotated as coding sequences in 2005 when the first Z. mobilis genome sequence was published [16], which included about 9% of the aerobic up-regulated genes (see Additional file 4,5). The rapid accumulation of genome sequence information and sequencing technology advances make improving gene coding models and annotations even more important.

Bacterial strains and fermentation conditions
Z. mobilis ZM4 was obtained from the American Type Culture Collection (ATCC31821) and cultured in RM medium [60] at 30°C. For the inoculum preparation a single colony of ZM4 was added to a test tube containing 5 mL RM broth and cultured aerobically at 30°C until it reached late exponential or early stationary phase. A 1/ 100 dilution was added into the pre-warmed RM broth (10 mL culture into 1000 mL RM), which was then cultured aerobically at 30°C with shaking at 150 rpm for approximately 12 h. The optical density was measured with a spectrophotometer at 600 nm and the inoculum was added to each fermentor so that the initial OD600 nm was approximately 0.17 in each fermentor. Batch fermentations were conducted in approximately 2.5 L of RM medium in 7.5-L BioFlo110 bioreactors (New Brunswick Scientific, Edison, NJ) fitted with agitation, pH, temperature and DOT probes and controls. Culture pH was monitored using a pH electrode (Mettler-Toledo, Columbus, OH) and the pH control set point was maintained at 6.0 by automatic titration with 3 N KOH. Temperature was maintained automatically at 30°C and the vented gases exiting fermentors were passed through condenser units, chilled by a NESLAB Merlin M-150 refrigerated recirculator (Thermo Fisher Scientific, Newington, NH) to a vented hood via a water trap. DOT was monitored by using InPro 6800 series polarographic O 2 sensors (Mettler-Toledo). Three anaerobic fermentors were sparged overnight with filter-sterilized N 2 gas and for approximately one hour post-inoculation and the three aerobic fermentors were continually sparged with filter-sterilized air at 2.5 L/min to maintain fully aerobic conditions. The agitation rate was 700 rpm in each vessel.

Growth, glucose and fermentation product analysis
Growth was monitored turbidometrically by measuring optical density at 600 nm with a model 8453 spectrophotometer (Hewlett-Packard, Palo Alto, CA.). Fermentation media and fermentation products from filter-sterilized

RNA isolation and preparation of fluorescein-labeled cDNA
RNA was isolated essentially described previously [26]. Briefly, samples from aerobic and anaerobic fermentors were harvested by centrifugation and the TRIzol reagent (Invitrogen, Carlsbad, CA) was used to extract total cellular RNA. Each total RNA preparation was treated with RNase-free DNase I (Ambion, Austin, TX) to digest residual chromosomal DNA and subsequently purified with the Qiagen RNeasy Mini kit in accordance with the instructions from the manufacturer. Total cellular RNA was quantified at OD 260 and OD 280 with a NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies, Wilmington, DE). The purified RNA from each sample was used as the template to generate cDNA copies labeled with either Cy3-dUTP or Cy5-dUTP (GE Healthcare Bio-Sciences Corp, Piscataway, NJ). In a duplicate set of cDNA synthesis reactions the fluorescent dyes were reversed for each sample so that the effects of a specific dye were minimized. Labeling reaction components and incubation conditions as well as cDNA probe purification and concentration determination have been described previously [26].

Microarray hybridization, scanning, image quantification, and data analyses
Z. mobilis microarrays were constructed by CombiMatrix Corporation (Mukilteo, WA) using coding sequences predicted by The Institute for Genomic Research (TIGR, http://www.tigr.org/) as 101 more genes were predicted in this later genome annotation. Both the primary and TIGR annotations are presented in the present study. Microarray hybridizations were performed according to the manufacturer's instructions. Briefly, gene expression analysis was performed using six independent microarray experiments (two dye reversal reactions × three biological replicates) with each microarray containing one to two probes per predicted coding sequence each. The two separately labeled cDNA pools (i.e., the aerobic and the corresponding anaerobic time point) to be compared were mixed together in a hybridization solution containing 50% (v/v) formamide, applied to microarrays and incubated overnight (16 h) at 50°C. Microarrays were washed using buffers of increasing stringency according to the manufacturer's instructions, scanned with a ScanArray 5000 Microarray Analysis System (PerkinElmer Life Sciences Inc, Boston, MA) and the images were quantified using Microarray Imager software (CombiMatrix). Raw data was log 2 transformed and imported into the statistical analysis software JMP Genomics 3.0 (SAS Institute, Cary, NC). A distribution analysis and data correlation analysis were done as a quality control step. The overlayed kernel density estimates derived from the distribution analysis allowed the visualization of sources of variation based on strain, as well as variation attributed to technical factors such as array and dye. The data were subsequently normalized using the standard normalization algorithm within JMP Genomics. An analysis of variance (ANOVA) was performed to determine differential expression levels between conditions and time points using the Bonferroni multiple testing method (p < 0.05).

Quantitative-PCR (qPCR) analysis
Microarray data were validated using real-time qPCR as described previously [26]. Seventeen genes representing different functional categories and the range gene expression values, based on microarray hybridizations were analyzed using qPCR from cDNA derived from stationary phase samples. Primer pairs were designed as described previously, and the oligonucleotide sequences of the seventeen genes selected for qPCR analysis are listed in Table  2.

Gas chromatography-mass spectrometry (GC-MS) metabolite analysis
Culture samples were rapidly pelleted by centrifugation, supernatants removed, cell pellets snap-frozen in liquid nitrogen and then stored at -80°C until analysis. Analyses were performed on microbial pellets collected in the stationary phase. A 5 mL aliquot of 80% ethanol (aqueous) was added to each pellet and cells disrupted using a sonicator 3000 (Misonix, Inc., Farmingdale, NY) operated at power level 5 for 6 times with a 1 min processing time and 1 min interval among each processing. An internal standard of 200 μL of sorbitol (1 mg/mL aqueous solution) was then added to each tube and 2 mL aliquots then dried in a helium stream. The internal standard was added to correct for differences in derivatization efficiency and changes in sample volume during heating. Dried exudates were dissolved in 500 μL of silylation-grade acetonitrile followed by the addition of 500 μL N-methyl-N-trimethylsilyltrifluoroacetamide (MSTFA) with 1% trimethylchlorosilane (TMCS) (Pierce Chemical Co., Rockford, IL), and samples then heated for 1 h at 70°C to generate trimethylsilyl (TMS) derivatives. After 1 day, 1-μL aliquots were injected into a ThermoFisher DSQII GC-MS, fitted with an Rtx-5MS (crosslinked 5% PH ME Siloxane) 30 m × 0.25 mm × 0.25 μm film thickness capillary column (Restek, Bellefonte, PA). The standard quadrupole GC-MS was operated in electron impact (70 eV) ionization mode, with 6 full-spectrum (70-650 Da) scans per second. Gas (helium) flow was set at 1.1 mL per minute with in injection port configured in the splitless mode. The injection port and detector temperatures were set to 220°C and 300°C, respectively. The initial oven temperature was held at 50°C for 2 min and was programmed to increase at 20°C per min to 325°C and held for another 11.25 min, before cycling back to the initial conditions. Quantified metabolites of interest were extracted using a key selected m/z that was characteristic for each metabolite, rather than the total ion chromatogram, to minimize integration of co-eluting metabolites. Peaks were quantified by area integration and the concentrations were normalized to the quantity of the internal standard (sorbitol) recovered, amount of sample extracted, derivitized, and injected. Two technical replicates were analyzed for two biological samples from each condition. Metabolite data of ZM4 under aerobic and anaerobic condition were averaged and presented as relative responses between ZM4 under aerobic fermentation versus ZM4 under anaerobic fermentation.