Metabolic Remodeling during Nitrogen Fixation in Zymomonas mobilis

ABSTRACT Zymomonas mobilis is an ethanologenic bacterium currently being developed for production of advanced biofuels. Recent studies have shown that Z. mobilis can fix dinitrogen gas (N2) as a sole nitrogen source. During N2 fixation, Z. mobilis exhibits increased biomass-specific rates of ethanol production. In order to better understand the physiology of Z. mobilis during N2 fixation and during changes in ammonium (NH4+) availability, we performed liquid chromatography-mass spectrometry (LC-MS)-based targeted metabolomics and shotgun proteomics under three regimes of nitrogen availability: continuous N2 fixation, gradual NH4+ depletion, and acute NH4+ addition to N2-fixing cells. We report dynamic changes in abundance of proteins and metabolites related to nitrogen fixation, motility, ammonium assimilation, amino acid biosynthesis, nucleotide biosynthesis, isoprenoid biosynthesis, and Entner-Doudoroff (ED) glycolysis, providing insight into the regulatory mechanisms that control these processes in Z. mobilis. Our analysis identified potential physiological mechanisms that may contribute to increased specific ethanol production during N2 fixation, including decreased activity of biosynthetic pathways, increased protein abundance of alcohol dehydrogenase (ADHI), and increased thermodynamic favorability of the ED pathway. Of particular relevance to advanced biofuel production, we found that intermediates in the methylerythritol phosphate (MEP) pathway for isoprenoid biosynthesis were depleted during N2 fixation, coinciding with decreased protein abundance of deoxyxylulose 5-phosphate synthase (DXS), the first enzyme in the pathway. This implies that DXS protein abundance serves as a native control point in regulating MEP pathway activity in Z. mobilis. The results of this study will inform metabolic engineering to further develop Z. mobilis as a platform organism for biofuel production. IMPORTANCE Biofuels and bioproducts have the potential to serve as environmentally sustainable replacements for petroleum-derived fuels and commodity molecules. Advanced fuels such as higher alcohols and isoprenoids are more suitable gasoline replacements than bioethanol. Developing microbial systems to generate advanced biofuels requires metabolic engineering to reroute carbon away from ethanol and other native products and toward desired pathways, such as the MEP pathway for isoprenoid biosynthesis. However, rational engineering of microbial metabolism relies on understanding metabolic control points, in terms of both enzyme activity and thermodynamic favorability. In Z. mobilis, the factors that control glycolytic rates, ethanol production, and isoprenoid production are still not fully understood. In this study, we performed metabolomic, proteomic, and thermodynamic analysis of Z. mobilis during N2 fixation. This analysis identified key changes in metabolite levels, enzyme abundance, and glycolytic thermodynamic favorability that occurred during changes in NH4+ availability, helping to inform future efforts in metabolic engineering.


RESULTS AND DISCUSSION
Experimental design and nitrogen availability regimes. We quantified relative metabolite and protein abundance under N 2 -fixing conditions in comparison to NH 4 1replete conditions and during transitions between these two growth conditions. Z. mobilis (ATCC 31821) was grown anaerobically using glucose as the sole carbon source (see Materials and Methods) (24). For conditions of replete NH 4 1 availability, 15 mM NH 4 1 was provided. For N 2 -fixing conditions, no NH 4 1 was added to the medium and the only available nitrogen source was N 2 gas (.90% N 2 in the anaerobic chamber).
We examined three separate regimes of nitrogen availability (Fig. 1A). (i) Continuous N 2 -fixing conditions were compared to continuous NH 4 1 -replete conditions; doubling times were approximately 3 h during N 2 fixation and 2 h under NH 4 1replete conditions (Fig. 1B). Samples were taken for targeted metabolomics at early, mid-, and late exponential phase and for shotgun proteomics at mid-exponential phase. (ii) NH 4 1 downshift; Z. mobilis was grown in medium containing limited NH 4 1 (,2 mM) such that growth stalled at mid-exponential phase. During the dynamic shift to N 2 fixation, a 6-h metabolomics and proteomics time course was conducted. Samples were collected for the initial (t = 0) time point during early exponential growth before a decrease in growth rate was observed, i.e., when the doubling time was still ;2 h ( Fig. 1C). Samples were also taken from NH 4 1 -replete controls which grew with a doubling time of 2 h (Fig. 1C). (iii) NH 4 1 upshift; Z. mobilis was grown under N 2 -fixing conditions until early exponential phase, at which point NH 4 Cl was added to the medium at a final concentration of 15 mM. During NH 4 1 upshift, a 2-h metabolomics and proteomics time course was conducted. Samples were collected for the initial (t = 0) time point immediately before addition of NH 4 Cl. Samples were also taken from N 2 -fixing controls. Both conditions grew with a consistent 3-h doubling time for the duration of the 2-h time course (Fig. 1D).
Metabolome analysis reveals global alterations in intracellular metabolite levels in response to changes in nitrogen availability. Metabolomic analysis using LC-MS produced relative intracellular abundance for 99 unique metabolites spanning central carbon metabolism (see Table S1 in the supplemental material). These included intermediates of Entner-Doudoroff (ED) glycolysis, the pentose phosphate pathway (PPP), the tricarboxylic acid (TCA) cycle, amino acid biosynthesis, nucleotide biosynthesis, isoprenoid biosynthesis, and peptidoglycan biosynthesis. Of the 99 detected metabolites, 79 were differentially abundant (fold change [FC] of .1.5 and false discovery rate [FDR]-adjusted P value of ,0.05) during at least one of the three conditions of nitrogen availability (Fig. 2). In general, amino acids and intermediates in amino acid biosynthesis were depleted during N 2 fixation and either increased or remained constant in response to NH 4 1 upshift. One notable exception was arginine, which increased during N 2 fixation and decreased after NH 4 1 addition. Intermediates of de novo nucleotide biosynthesis were severely depleted (.30-fold decrease) under N 2 -    4 1 controls (green circles) compared to NH 4 1 downshift (starting with ,5 mM NH 4 1 ) (blue squares). Arrows indicate start and end of time course. Times of metabolite and protein extraction are shown as red symbols. Data points are the average of 3 biological replicates for NH 4 1 downshift and 2 biological replicates for controls. Error bars showing standard deviation are smaller than the height of the symbols. (D) Continuous N 2 -fixing controls (blue squares) compared to NH 4 1 upshift (15 mM NH 4 Cl was added to N 2 -fixing cultures at t = 0) (green circles). Arrows and red symbols indicate metabolite and protein extractions taken at start and end of time course. See Materials and Methods for a full list of time points within this time course. Data points are the average of 3 biological replicates for NH 4 1 upshift and 2 biological replicates for controls. Error bars show the standard deviation.
fixing conditions. Nucleotide triphosphates (NTPs) were also less abundant during N 2 fixation, although to a lesser extent (;4-fold decrease). Conversely, intracellular concentrations of nucleosides and nucleotide monophosphates (NMPs) increased during the shift to N 2 fixation. There were dynamic changes in the PPP during shifts in NH 4 1 availability, which are likely linked to nucleotide biosynthesis. We observed depletion of intermediates in both the ED glycolytic pathway and the methylerythritol phosphate (MEP) pathway for isoprenoid biosynthesis during N 2 fixation. Overall, intracellular metabolite levels changed dramatically during the shift to N 2 fixation but remained much more consistent during NH 4 1 upshift, where the largest changes corresponded to increased amino acid abundance, particularly glutamine.
Proteome remodeling in response to changes in nitrogen availability. Proteomics analysis during NH 4 1 downshift and NH 4 1 upshift produced relative protein abundance for downshift (shift to N 2 -fixing conditions), and NH 4 1 upshift (from N 2 -fixing conditions), respectively. Rows are a single metabolite across all conditions, and columns are individual metabolomics samples, taken at the indicated times and conditions. Log 2 fold changes are relative to the first time point in the control condition for each experiment (indicated by an asterisk). Values are the averages of 2 biological replicates for time course controls, 3 biological replicates for time course treatments, and 5 biological replicates for both continuous conditions. Yellow corresponds to increased intracellular metabolite abundance compared to the control, and blue indicates depletion. Metabolites were manually arranged based on the biosynthetic pathway. 1,693 unique proteins (90% coverage of protein coding genes) (22,25,26). The comparison between continuous N 2 and NH 4 1 growth conditions was performed separately and yielded 1,429 proteins (75% coverage). Of the proteins detected, 615 changed significantly during at least one of the three conditions of N 2 availability (FC of .1.5 and FDR-adjusted P value of ,0.05) (Table S2). We identified 296 proteins that were affected during continuous N 2 fixation relative to continuous replete NH 4 1 , 467 proteins that were differentially abundant during the shift to N 2 fixation, and only 33 proteins that responded to the NH 4 1 upshift. We performed an overrepresentation analysis of gene ontology (GO) terms associated with the set of proteins found to be differentially abundant under each of the three conditions (Table S3) (27). All three conditions yielded "nitrogen fixation," "nitrogenase activity," and "iron-sulfur cluster binding" as terms that were enriched among affected proteins. Proteins that were differentially abundant during the N 2 versus NH 4 1 conditions were also enriched for "regulation of nitrogen utilization," "cellular iron ion homeostasis," "cysteine desulfurase activity," and "arginine biosynthetic process via ornithine" among other terms. Proteins that changed in abundance during NH 4 1 downshift were also enriched for GO terms such as "siderophore uptake transmembrane transporter activity," "bacterial-type flagellum," "oxidoreductase activity," and "response to oxidative stress." Products of the nitrogen fixation (nif) gene cluster were consistently among the most affected proteins under all three conditions of NH 4 1 availability (Fig. 3). During the transition to N 2 fixation, proteins involved in metal transport, electron transport, and oxidative stress increased in abundance, whereas proteins involved in translation and motility decreased in abundance ( Fig. S1 and S2). Our integrated metabolomics and proteomics analysis in Z. mobilis also revealed widespread and dynamic remodeling of metabolism in response to changes in nitrogen availability. In the following sections, we summarize a subset of the most significant alterations.
Concerted increase in the abundance of nitrogenase and nitrogenase-supporting enzymes during N 2 -fixing conditions. In Z. mobilis, as in other nitrogen-fixing bacteria, the genes required for N 2 fixation are clustered within a 30-kb genomic region called the nif cluster (28,29). The Z. mobilis nif cluster contains 31 genes: nifABZHDKENXQUSVW, fdxN, fixU, fdxB, modD, modE, the rnfABCDGEH operon, a gene (ZMO1832) encoding an iron-sulfur cluster assembly accessory protein, two genes (ZMO1815 and ZMO1822) encoding iron-associated TonB-dependent transporters (TBDT), an uncharacterized N 2 fixation gene (ZMO1829), and a gene (ZMO1821) encoding a hypothetical protein with an SIR2-like domain (30) (Fig. 3A). The nif cluster codes for the three proteins that form the active nitrogenase complex: nitrogenase reductase (NifH), also called the Fe protein, and the a and b subunits of nitrogenase (NifD and NifK), also called the MoFe protein. Several genes in the nif cluster (e.g., nifB, nifU nifE, nifN, and nifV) are involved in the biogenesis of iron-sulfur clusters required for N 2 fixation, including the [Fe 4 -S 4 ] cluster cofactor of NifH, the [Fe 8 -S 7 ] cluster (P-cluster) cofactor of NifDK, and the [Mo-Fe 7 -S 9 -C-homocitrate] molybdenum-iron cofactor (FeMo-co) at the active site of NifDK (31,32). The nif gene cluster also contains the rnf operon, whose products form a membrane-bound complex that couples ion translocation across the inner membrane to the transfer of electrons from NADH to ferredoxin (e.g., FdxN or FdxB), which then donates electrons to the nitrogenase complex (33,34). The nif cluster is regulated by the s 54 -dependent transcription factor NifA, which is also encoded within the nif cluster (35)(36)(37)(38)(39).
One of most affected proteins from the nif cluster was homocitrate synthase (NifV), which was 300-fold more abundant during continuous N 2 fixation and increased by 16-fold during the shift to N 2 -fixing conditions (Fig. 4B). NifV transfers an acetyl group from acetyl coenzyme A (acetyl-CoA) to a-ketoglutarate to produce homocitrate, which chelates the Mo atom in the FeMo-co at the active site of nitrogenase ( Fig. 4D) (40,41). Metabolomics analysis showed that intracellular homocitrate levels increased by over 60fold during N 2 fixation (both continuous N 2 fixation and transition to N 2 fixation), which was the largest increase in metabolite abundance observed in this study ( Fig. 2 and 4C). Asterisks indicate statistical significance (FC $ 1.5, FDR # 0.05) for NH 4 1 versus N 2 , NH 4 1 downshift, and NH 4 1 upshift experiments, respectively, from left to right. For example, RnfH protein abundance changed significantly for all three conditions, but changes in RnfE abundance were only significant for NH 4 1 versus N 2 and NH 4 1 downshift conditions. A gray tile indicates that protein was not detected under that condition. Proteins were manually arranged based on genomic location within the nif cluster.
During NH 4 1 upshift, fewer changes in nif cluster protein abundance occurred (Fig. 3B). Only 14 of the 31 nif cluster proteins decreased significantly relative to N 2 -fixing controls. The largest decreases were in NifZ, NifU, NifB, NifW, NifV, NifE, NifN, and the iron-sulfur cluster assembly accessory protein ZMO1832, all of which decreased by over 2-fold ( Fig. 3B and 4B). Given that less than one cell doubling occurred during the NH 4 1 upshift time course experiment, decreases over 2-fold imply active protein degradation rather than dilution by cell division. Many of the proteins that decreased in abundance are involved in the biogenesis of FeS or FeMo cofactors, including NifV, whose depletion coincided with an 8-fold drop in intracellular levels of homocitrate (Fig. 4C). Protein levels of nitrogenase proteins NifH, NifD, and NifK all showed decreasing trends during NH 4 1 upshift, but only NifK met our criteria (FC . 1.5, FDR , 0.05) for differential expression and all three remained within 2-fold of the N 2 -fixing baseline   (Fig. 4A). Together, these results suggest that biogenesis of FeS and FeMo cofactors is rapidly repressed via protein degradation during an acute increase in NH 4 1 availability, even while NifHDK protein abundance remains relatively constant. Delaying degradation of nitrogenase may provide a fitness advantage in environments where bioavailable nitrogen is transiently available. Modulation of iron-sulfur cluster biogenesis and intracellular homocitrate levels may contribute to the regulation of nitrogenase activity during changes in NH 4 1 availability.
Decreased abundance of flagellar and chemotaxis proteins during N 2 fixation. In the Z. mobilis genome, the genes involved in flagellar motility (ZMO0602-ZMO0652) and several genes related to chemotaxis signal transduction (ZMO0078-ZMO0085) are each organized within their own gene cluster (30,42). Several flagellar proteins (FlgL, FlgK, FlgI, FlgD, FlgC, FlgB, FliF, FliI, FliK, FliO, FliD, and FliK) and chemotaxis proteins (CheY, CheD, CheA, and CheX) from these two gene clusters were significantly depleted (FC . 1.5, FDR , 0.05) during the transition to N 2 -fixing conditions (Fig. S2). Decreases in abundance of motility proteins was seen as early as 1 h into the NH 4 1 downshift. FlgI, FliF, FliI, CheR, CheA, and McpJ were also less abundant during continuous N 2 -fixing conditions (FC . 1.5, FDR , 0.05). No significant trends were observed in proteins from either gene cluster in response to NH 4 1 upshift. Bacterial flagellar motors are powered by the translocation of ions across the inner membrane (43,44). Others have observed decreased expression of flagellar genes in Z. mobilis under stress conditions that disrupt the maintenance of an electrochemical gradient across the membrane, such as low pH, osmotic stress, and high ethanol concentrations (45)(46)(47). This is in contrast to other environmental inhibitors such as oxygen and lignocellulosic toxins, which have been shown to increase expression of motility-related genes in Z. mobilis (48)(49)(50). Previous transcriptomics analysis of Z. mobilis both during adaptation to high glucose concentrations and in the presence of a quorum sensing autoinducer each showed that expression of flagellar genes decreased concomitantly with increased expression of genes from the nif cluster (20,46). Given that the Rnf complex utilizes energy stored in the electrochemical gradient to drive production of reduced ferredoxin (the electron donor to nitrogenase reductase), decreased expression of flagellar proteins (and, presumably, decreased energy consumption by the flagellar motor) may be important for maintaining a robust electrochemical gradient during N 2 -fixing conditions.
Dynamics of NH 4 + assimilation via the GS-GOGAT cycle. There are two major pathways for NH 4 1 assimilation in bacteria (51). One of these pathways is comprised of glutamine synthetase (GS) and glutamine oxoglutarate aminotransferase (GOGAT), which together form the GS-GOGAT cycle (Fig. 5A). In this pathway, GS catalyzes the condensation of glutamate and ammonia to form glutamine, converting one molecule of ATP to ADP in the process (52). Next, GOGAT transfers the amino group from glutamine to a-ketoglutarate (aKG), forming two molecules of glutamate and converting one molecule of NAD(P)H to NAD(P) 1 (53). The other bacterial route for ammonia assimilation is via glutamate dehydrogenase (GDH), which directly converts aKG to glutamate by reductive amination of aKG, also consuming one molecule of NAD(P)H (54). The two pathways differ in their energy (ATP) consumption and affinity for NH 4 1 (54-56). In Escherichia coli, it is generally understood that the GS-GOGAT cycle is employed during low NH 4 1 availability while GDH is active during high NH 4 1 availability or energy limitation (56,57). In Z. mobilis, no gdh gene has been identified based on sequence homology (58). However, genes encoding GS (glnA) and the large and small subunits of GOGAT (gltB and gtlD) have been annotated (30) (Fig. 5A).
Proteomics analysis determined relative abundance of GS (GlnA) and the large and small subunits of GOGAT (GltB and GltD) during continuous N 2 fixation, NH 4 1 downshift, and NH 4 1 upshift (Fig. 5B). Interestingly, GS levels did not change significantly for any of the three conditions of NH 4 1 availability (FC , 1.3, FDR . 0.2) (Fig. 5B). However, both subunits of GOGAT were 2-fold lower during continuous N 2 fixation than during NH 4 1 -replete conditions (Fig. 5B). During NH 4 1 downshift, GOGAT levels were already lower than in NH 4 1 -replete controls at the initial time point (before changes in growth rate were observed), and they continued to fall for the duration of the NH 4 1 downshift time course, reaching 2-fold below controls at the 6-h time point (Fig. 5C). There was an increasing trend for both subunits of GOGAT during NH 4 1 upshift (FDR = 0.072 and 0.029, respectively), but levels remained within 1.5-fold of controls (Fig. 5B). Overall, GS enzyme levels remained constant and GOGAT enzymes levels decreased during N 2 fixation. Regulation of GS via posttranslational adenylylation by GS adenylyltransferase/adenylyl-removing enzyme (AT/AR) is common in other proteobacteria (59). However, no AT/AR gene has been annotated in the Z. mobilis genome. On the other hand, differential phosphorylation of both GS and GOGAT have been observed during N 2 fixation in Z. mobilis, likely contributing to the regulation of GS-GOGAT cycle activity (60). Metabolomics analysis produced relative intracellular abundance of glutamine, glutamate, and aKG during the three conditions of NH 4 1 availability (Fig. 5C). Under continuous N 2 -fixing conditions, intracellular glutamine levels were 3-fold lower, glutamate levels were 2-fold higher, and aKG levels were 8-fold higher than under NH 4 1replete conditions (Fig. 5C). During NH 4 1 downshift, glutamine levels were already 3fold lower than those of NH 4 1 -replete controls at the initial time point and dropped to 12-fold below controls after 1 h, when growth began to stall. From the 1-h time point to the 6-h time point, intracellular glutamine levels rose, tracking with the rise in nitrogenase protein levels ( Fig. 5C and 4A). Glutamate levels remained within 2-fold of control levels for the entire downshift time course but matched the trends seen in glutamine, dropping in the first hour and then increasing from 1 to 6 h. Levels of aKG increased for the duration of the downshift time course, reaching 5-fold above baseline after 6 h (Fig. 5C).
During NH 4 1 upshift, intracellular glutamine levels immediately increased, reaching 6fold above the N 2 -fixing control within 1 min of NH 4 1 addition. Glutamine levels continued to rise for 30 min following NH 4 1 addition, reaching 20-fold above baseline before gradually decreasing to 13-fold above baseline by the 2-h time point. Glutamate levels trended in the opposite direction, first dropping to 2.5-fold below N 2 -fixing controls within 1 min and then increasing for the remainder of the time course, ending within 1.5-fold of the controls. aKG levels did not change within the first 15 min of the NH 4 1 upshift time course but decreased by around 2-fold from 15 min to 2 h (Fig. 5C).
Previous studies in E. coli and Rhodospirillum rubrum have found that upon addition of NH 4 1 to nitrogen-limited cultures, intracellular glutamine levels increased rapidly but then quickly declined within 5 to 10 min (61,62). The decline in glutamine levels following their accumulation was associated with fast inactivation of GS via adenylylation by AT/AR (61,63,64). In Z. mobilis, glutamine levels rapidly increased following NH 4 1 upshift but did not decline in the first 30 min and remained substantially elevated compared to steady-state levels for over 2 h, suggesting that GS activity is not rapidly inhibited in response to increased NH 4 1 availability. The fact that the Z. mobilis genome apparently does not encode AT/AR may help explain the persistence of elevated glutamine levels following NH 4 1 addition in Z. mobilis. More research should be conducted to investigate the regulatory mechanisms that control GS activity in Z. mobilis, as they appear to be distinct from the classical model that has been well studied in other proteobacteria (65,66).
Addition of 15 NH 4 + to N 2 -fixing cultures shows immediate assimilation of exogenous NH 4 + and possible residual nitrogenase activity. The immediate increase in intracellular glutamine levels following NH 4 1 addition indicates that Z. mobilis is poised to rapidly incorporate exogenous NH 4 1 as soon as it becomes available, even in an N 2 -fixing state. This was expected given that NH 4 1 is a product of nitrogenase, and at high (mM) extracellular concentrations, NH 4 1 can passively diffuse across the membrane to support growth in other bacteria (67). To examine the dynamics of NH 4 1 assimilation and incorporation more closely, we performed a separate iteration of the NH 4 1 upshift experiment using 15 NH 4 Cl (see Materials and Methods). This allowed us to trace the incorporation of exogenous NH 4 1 into metabolite pools, providing insight into the turnover rate of nitrogen-containing metabolites during NH 4 1 upshift (Fig. 6). Within 1 min after addition of 15 NH 4 1 , glutamine levels were 99% 15 N labeled (i.e., containing at least one 15 N atom). Similar trends were observed for glutamate, which was 90% 15 N labeled after 5 min. Most amino acids followed these trends, with a few exceptions. Most notably, asparagine labeled much more slowly and was still over 50% unlabeled after 5 min. This suggests that either rates of asparagine biosynthesis are low compared to other amino acids, intracellular pools of asparagine are high compared to other amino acids, or there is some compartmentalized pool of asparagine (e.g., periplasmic or extracellular) that is subject to a lower turnover rate. The rate of 15 N incorporation into nucleotide biosynthetic intermediates was similar to that of glutamine and glutamate, showing that exogenously supplied NH 4 1 was rapidly utilized for de novo synthesis of both purines and pyrimidines. Nucleotide triphosphates such as ATP labeled more slowly than their upstream intermediates, as would be expected, but after 2 h, less than 5% of the NTP pool remained unlabeled.
Interestingly, from 15 min to 2 h, we observed a small but significant (P , 0.01) loss of 15 N label in glutamine. The fully labeled fraction (i.e., containing two 15 N atoms) of glutamine went from 95% after 15 min to 88% after 2 h, implying increasing assimilation of unlabeled ( 14 N) nitrogen. Loss of 15 N label was also seen in other metabolites, including intermediates in amino acid and nucleotide biosynthesis. This loss of 15 N label might be the result of protein degradation, which could liberate amino acids and NH 4 1 from proteins that had been translated during 14 N 2 fixation. However, given the observation that nitrogenase protein levels did not decrease by more than 2-fold during NH 4 1 upshift, loss of 15 N label could also be caused by residual nitrogenase activity following increased availability of NH 4 1 . In some diazotrophic alphaproteobacteria (e.g., R. rubrum), NH 4 1 upshift induces rapid but reversible inactivation of NifH via ADP-ribosylation by DraT/DraG (68)(69)(70). However, neither draT nor draG homologues have been identified in the Z. mobilis genome. Control samples extracted from cells grown with replete 15 NH 4 1 for over 6 generations were .99% fully 15 N labeled for all nitrogen-containing metabolites detected (Fig. 6). This indicates that long-term replete NH 4 1 availability does result in complete repression of nitrogenase activity.
Effect of nitrogen availability on amino acid abundance. We measured the relative intracellular abundance of 17 amino acids (Fig. 7). Levels of glutamine, asparagine, aspartic acid, isoleucine, methionine, lysine, and aromatic amino acids (phenylalanine, tryptophan, and tyrosine) were all significantly depleted (FC . 1.5, FDR , 0.05) during N 2 fixation compared to NH 4 1 -replete conditions. Glutamate, leucine, and arginine were the only amino acids whose levels were significantly elevated during continuous N 2 fixation. During the dynamic shift to N 2 fixation, all measured amino acids other than lysine, arginine, and leucine decreased significantly. Leucine and arginine levels both increased by over 3-fold, but lysine did not display any significant trends. During NH 4 1 upshift, the largest change in amino acid abundance was the 20-fold increase in glutamine levels, but asparagine, isoleucine, leucine, valine, methionine, tyrosine, and phenylalanine levels also increased significantly compared to the N 2 -fixing control.  4 1 versus N 2 , NH 4 1 downshift, and NH 4 1 upshift conditions, respectively, from left to right. For example, intracellular levels of alanine changed significantly during NH 4 1 downshift and NH 4 1 upshift but not during continuous NH 4 1 versus N 2 conditions. Metabolites were manually arranged based on chemical properties of the side chain for amino acids and by metabolic pathway order for the shikimate pathway. Abbreviations: PEP, phosphoenolpyruvate; E4P, erythrose 4-phosphate; DAHP, 2-dehydro-3-deoxy-D-arabino-heptonate 7-phosphate.
Abundance of glutamate, arginine, alanine, serine, and aspartic acid all decreased significantly during the NH 4 1 upshift. Depletion of aromatic amino acids and shikimate pathway intermediates during N 2 fixation. All three aromatic amino acids were significantly depleted under continuous N 2 -fixing conditions and during the dynamic shift to N 2 -fixing conditions. Intermediates of aromatic amino acid biosynthesis were also depleted under these conditions (Fig. 7). Of the seven intermediates of the shikimate pathway that we detected, all were found to be significantly depleted during continuous N 2 fixation (FC . 1.5, FDR , 0.05). Additionally, levels of 2-dehydro-3-deoxy-D-arabino-heptonate 7-phosphate (DAHP), shikimate 3-phosphate, and (to a lesser extent) phosphoenolpyruvate (PEP) all decreased significantly during the shift to N 2 -fixing conditions. Shikimate 3-phosphate showed dramatic changes, dropping to 24-fold below NH 4 1replete controls within the first hour. No changes were observed in levels of shikimate pathway intermediates during the NH 4 1 upshift. Arginine levels are elevated during N 2 fixation despite depletion of intermediates in arginine biosynthesis. Of the 17 measured amino acids, arginine was the only amino acid that increased during NH 4 1 downshift and decreased during NH 4 1 upshift ( Fig. 7 and 8C). Arginine was also the only amino acid with levels over 4-fold higher during continuous N 2 fixation than during NH 4 1 -replete conditions. These results were surprising given that biosynthesis of arginine requires more nitrogen than any other amino acid and may therefore be expected to be depleted under conditions of nitrogen limitation. However, examination of intracellular levels of intermediates in arginine biosynthesis provides a potential explanation for this apparent contradiction. Intracellular levels of all intermediates of arginine biosynthesis downstream of N-acetyl glutamate 5-semialdehyde (i.e., N-acetyl ornithine, ornithine, citrulline, and argininosuccinate) were depleted during both continuous N 2 fixation and the dynamic shift to N 2 fixation ( Fig. 8A and B). During NH 4 1 upshift, intracellular concentrations of N-acetyl ornithine, ornithine, and citrulline all increased by over 1.5-fold (Fig. 8B). During continuous N 2 fixation, protein abundance of two enzymes in the arginine biosynthetic pathway (acetylglutamate kinase and argininosuccinate lyase) was significantly lower than in NH 4 1replete controls (see Fig. S3 in the supplemental material). However, no significant changes were observed in protein abundance of arginine biosynthetic enzymes during NH 4 1 downshift or upshift. Taken together, these data suggest that arginine biosynthesis does decrease during N 2 fixation. The fact that arginine still accumulates during N 2 fixation implies a decreased rate of arginine consumption.
A potential pathway for sym-homospermidine biosynthesis from arginine. Examination of proteomics data identified a candidate enzyme potentially responsible for differential arginine consumption during N 2 fixation. Among the top 5 most differentially expressed proteins during the NH 4 1 downshift was a group IV decarboxylase, encoded by ZMO1020. Proteins of this family act as ornithine, diaminopimelate (DAP), or arginine decarboxylases (71). The ZMO1020 protein product decreased in abundance by 64-fold during the shift to N 2 -fixing conditions and remained severely depleted for 1 h (Fig. 8D). This severe depletion was somewhat transient, and after 6 h, protein levels were only 4-fold below the NH 4 1 -replete baseline. During NH 4 1 upshift, ZMO1020 protein levels increased by over 2-fold within 15 min of NH 4 1 addition (Fig. 8B). Under continuous N 2 -fixing conditions, the ZMO1020 protein was 2-fold less abundant than under NH 4 1 -replete conditions (Fig. 8D).
ZMO1020 is predicted to be in an operon with ZMO1019, which codes for a deoxyhypusine synthase-like protein (72). The ZMO1019 protein product was also depleted by around 2-fold during continuous N 2 fixation, although it did not change significantly during dynamic shifts in NH 4 1 availability (Fig. 8D). Another operon in Z. mobilis is predicted to encode agmatine deiminase (ZMO1369) and N-carbamoyl putrescine amidase (ZMO1370). It has been demonstrated that the major polyamine produced in Z. mobilis is sym-homospermidine (73). The enzymes encoded by ZMO1020, ZMO1369, ZMO1370, and ZMO1019 form a feasible biosynthetic route from arginine to symhomospermidine (Fig. 8E) (74)(75)(76)(77). In this case, the deoxyhypusine synthase-like protein would function as a bacterial homospermidine synthase, combining two molecules of putrescine to generate sym-homospermidine. A similar pathway for symhomospermidine production from arginine was identified in nitrogen-fixing cyanobacteria and was found to be required for robust diazotrophic growth (78). The negative correlation of arginine levels with ZMO1020 protein abundance, the organization of ZMO1020 and ZMO1019 genes within an operon, and the polyamine profile of Z. mobilis suggest that the protein encoded by ZMO1020 may function as an arginine decarboxylase, constituting the first step in sym-homospermidine biosynthesis in Z. mobilis. It is plausible that a shutdown of sym-homospermidine production during N 2 fixation caused the observed accumulation of intracellular arginine. Accumulation of arginine may have then triggered the depletion of arginine biosynthetic intermediates via feedback inhibition, as has been well documented in other bacteria (79). More research is needed to confirm the activity of the proposed biosynthetic pathway for sym-homospermidine in Z. mobilis.
While the precise physiological role of polyamines in bacteria remains an open area of discovery, polyamine production has been implicated in the response to a variety of environmental stressors, including high temperatures, oxidative stress, and acidic conditions (80)(81)(82). In our previous study of oxygen exposure in Z. mobilis, intracellular levels of arginine, acetyl-ornithine, and ornithine followed trends during the shift to aerobic conditions very similar to those during the shift to N 2 -fixing conditions (Fig. S4) (48). Additionally, protein abundance of the potential arginine decarboxylase encoded by ZMO1020 dropped sharply in response to oxygen exposure, as it did in response to NH 4 1 depletion (Fig. S4). These observations suggest that the hypothesized production of sym-homospermidine from arginine may be regulated in response to a variety of environmental stressors.
Depletion of intermediates in nucleotide biosynthesis during N 2 fixation. Intermediates of de novo purine biosynthesis and de novo pyrimidine biosynthesis were depleted during both continuous N 2 fixation and the shift to N 2 -fixing conditions (Fig. 9). Of the two pathways, purine biosynthesis was more severely impacted, with purine biosynthetic intermediates such as 5-phosphoribosylamine (5PRA), phosphoribosyl-N-formylglycineamide (FGAR), and phosphoribosylaminoimidazole-succinocarboxamide (SAICAR) reaching over 180-fold below NH 4 1 -replete controls during the shift to N 2 fixation. The largest change observed in intermediates of pyrimidine biosynthesis was in carbamoyl aspartate, which dropped to 30-fold below baseline during the shift to N 2 fixation. During N 2 fixation, there was a general depletion of nucleotide triphosphates and nucleotide diphosphates but an accumulation of nucleotide monophosphates and nucleosides. During the NH 4 1 upshift, both 5PRA and carbamoyl aspartate increased significantly but nucleotide levels remained unchanged. The 5-fold increase in intracellular 5PRA was accompanied by a 5-fold decrease in phosphoribosyl diphosphate (PRPP) levels within 15 min of addition of NH 4 1 . This implies that amidophosphoribosyltransferase (PurF), which catalyzes the committed step in purine biosynthesis by incorporating an amino group from glutamine into PRPP to form 5PRA, is poised for rapid incorporation of nitrogen as soon as it becomes available (Fig. 9). This was consistent with rapid incorporation of 15 N into FGAR following 15 NH 4 1 addition (Fig. 6). We also observed dynamic changes in the PPP pathway following NH 4 1 upshift. In particular, ribose 5-phosphate (R5P) levels exhibited a sharp spike, reaching 5-fold above N 2 -fixing controls within 5 min of NH 4 Cl addition, only to drop back down to baseline by 15 min (Fig. S3). These trends highlight the tight regulation of PPP activity required to tune the supply of 5-carbon sugars for nucleotide biosynthesis during changes in nitrogen availability.
Despite the dramatic changes in intracellular metabolite levels of nucleotide biosynthetic intermediates, there were few significant changes in protein abundance of nucleotide biosynthetic enzymes during changes in NH 4 1 availability (Fig. S3). The only dynamic trend was a 1.7-fold decrease in PurF levels during the shift to N 2 fixation. The metabolic activity of nucleotide biosynthesis therefore appears to be regulated via the availability of nitrogen-containing metabolic precursors (e.g., amino acids) rather than the abundance of metabolic enzymes.
Depletion of MEP pathway intermediates and decreased abundance of DXS during N 2 fixation. Z. mobilis exclusively utilizes the MEP pathway for the biosynthesis of isoprenoid precursors isopentenyl diphosphate (IDP) and dimethylallyl diphosphate (DMADP), which are required to produce biological compounds such as quinones and carotenoids (83). The MEP pathway starts with the condensation of pyruvate and glyceraldehyde 3-phosphate (GAP) to form 1-deoxy-D-xylulose 5-phosphate (DXP), catalyzed by DXP synthase (DXS). DXP is then converted to either IDP or DMADP via six enzymatic reactions carried out by IspC, IspD, IspE, IspF, IspG, and IspH (Fig. 10A). In Z. mobilis, there are two copies of the DXS enzyme (DXS1 and DXS2, encoded by ZMO1243 and ZMO1598, respectively) and IspD and IspF are fused and expressed as the single bifunctional enzyme IspDF.
Intracellular levels of all detected intermediates of the MEP pathway were lower during N 2 fixation. Under continuous N 2 -fixing conditions, MEP pathway intermediates were between 2-fold and 4-fold lower than NH 4 1 -replete controls (Fig. 10B). During the dynamic shift to N 2 fixation, an even more pronounced depletion was observed. All detected MEP pathway intermediates decreased by between 8-fold and 32-fold during the 6-h NH 4 1 downshift time course (Fig. 10B). The intermediates 4-diphosphocytidyl-2-C-methyl-D-erythritol 2-phosphate (CDP-MEP) and 4-hydroxy-3-methyl-butenyl 1diphosphate (HMBDP) displayed the largest decreases in abundance during both continuous N 2 fixation and the shift to N 2 -fixing conditions. NH 4 1 upshift did not induce any increase in intracellular levels of MEP pathway intermediates, which all remained within 2-fold of the nitrogen-fixing control (Fig. 10B).
Of the seven enzymes of the MEP pathway, only the two DXS enzymes demonstrated significant changes in abundance in response to changes in NH 4 1 availability (Fig. 10C). All other MEP pathway enzymes remained within 1.5-fold of the control condition during all three nitrogen regimes. During continuous N 2 fixation, DXS2 protein levels were significantly lower than NH 4 1 -replete controls. During the NH 4 1 downshift experiment, both DXS1 and DXS2 decreased in abundance by over 3-fold within 3 h. However, following NH 4 1 upshift, neither DXS1 nor DXS2 changed in abundance during the 2-h time course.
Taken together, the trends in metabolite abundance of MEP pathway intermediates and protein abundance of MEP pathway enzymes suggest that MEP pathway activity is regulated during conditions of nitrogen limitation via abundance of the DXS enzyme. This is an interesting observation considering that activity of the MEP pathway does not directly consume nitrogen. The fitness benefit of regulating the MEP pathway in response to nitrogen availability may be related to the pathway's consumption of ATP, CTP, or reduced cofactors. Additionally, decreased growth rates likely correspond to decreased demand for isoprenoid metabolites downstream of the MEP pathway.
Others have reported experimental findings suggesting that the Clp protease complex may regulate MEP pathway activity in Arabidopsis plastids via degradation of DXS (84). More research is required to confirm targeted proteolytic degradation of DXS in Z. mobilis and to identify the mechanism of such degradation as well as the precise physiological cue, either nitrogen limitation itself or some secondary signal (for example depletion of NTP levels), that initiates the response. Depletion of ED pathway intermediates and increased abundance of ADHI during N 2 fixation. Z. mobilis utilizes the ED pathway exclusively for glucose catabolism (85,86). Intracellular concentrations of all quantified ED glycolytic intermediates were significantly depleted (FC . 1.5, FDR , 0.05) during continuous N 2 fixation (Fig. 11A). Additionally, 2-keto-3-deoxy-6-phosphogluconate (KDPG), 1,3-bisphosphoglycerate (BPG), and phosphoenolpyruvate (PEP) levels decreased significantly during the shift to N 2 -fixing conditions (Fig. 11A). During the NH 4 1 upshift, no significant changes were observed for any intermediates of the ED pathway. Relative protein abundances of ED pathway enzymes were not significantly different during any of the three conditions of nitrogen availability, except for 6-phosphogluconolactonase (Pgl), which was 1.7-fold more abundant during continuous N 2 fixation (Fig. 11B).
During N 2 fixation, Z. mobilis exhibits increased specific rates of glucose consumption and ethanol production (17,19,20). However, the physiological factors that drive this response are not well understood. The Z. mobilis genome encodes two alcohol dehydrogenases, ADHI and ADHII. ADHI (encoded by adhA) is a zinc-dependent alcohol dehydrogenase, while ADHII (encoded by adhB) is iron dependent. Previous studies of purified ADHI and ADHII, and analysis of DadhB strains, indicate that both ADHI and ADHII contribute to ethanol production (87)(88)(89)(90)(91). We found that protein levels of ADHI increased by 10-fold during the shift to N 2 fixation, placing it among the top 50 differentially abundant proteins observed in this study (Fig. 11C). No changes were observed in ADHI levels during continuous N 2 fixation or during the NH 4 1 upshift. ADHII did not significantly change in abundance during changes in NH 4 1 availability (Fig. 11C). The difference in metal cofactors used by the two ADH isozymes is especially relevant considering that expression of nitrogenase holoenzyme significantly increases the cellular demand for iron. Increasing expression of zinc-dependent ADHI rather than iron-dependent ADHII may be important for maintaining iron homeostasis. Increased ADHI expression may help explain the previously reported increase in specific ethanol production during N 2 fixation (17,19,20). However, the fact that no difference in ADHI abundance was detected during continuous N 2 -fixing conditions implies a possible difference in short-term and long-term mechanisms for increasing ethanol production.
Increased thermodynamic favorability of the ED pathway during N 2 fixation. Thermodynamics constitutes a key determinant of flux and enzyme efficiency in metabolic networks. A pathway with a strong thermodynamic driving force (i.e., with an overall large negative change in Gibb's free energy or DG) will achieve a higher net flux given a fixed amount of enzyme activity than one closer to equilibrium (86,92,93). Within a pathway, steps closer to equilibrium will be the least enzyme efficient. A reaction's DG is related to its reversibility or reverse-to-forward flux ratio (J 2 /J 1 ) by the equation DG = RT ln(J 2 /J 1 ), where R is the gas constant and T is the temperature in Kelvin. In vivo reaction reversibility, and thus thermodynamics, can be examined using isotope tracers.
To examine the impact of NH 4 1 availability on ED pathway thermodynamics, we grew cells on glucose positionally labeled with either 13 C or 2 H under continuous N 2fixing conditions and NH 4 1 -replete conditions. These experiments revealed that several reactions in the ED pathway were more reversible under NH 4 1 -replete conditions, implying increased thermodynamic favorability of the ED pathway during N 2 fixation (Fig. S5). In particular, labeling patterns of KDPG in cells fed either 100% [1-13 C]glucose (98 to 99% isotopic purity) or 100% [6-13 C]glucose (98 to 99% isotopic purity) indicated decreased reversibility of the KDPG aldolase reaction during N 2 fixation (Fig. 12A)  (Continued on next page) of GAP dehydrogenase during N 2 fixation (Fig. 12A). Labeling patterns of 3-phosphoglycerate (3PG) from cells fed 100% [5-2 H]glucose (98% isotopic purity) were indicative of decreased reversibility of the two-step conversion of 3PG to PEP (Fig. 12A). Overall, the glycolytic reactions we observed were less reversible, and therefore more thermodynamically favorable, during N 2 fixation.
To provide a quantitative estimate of the effect of N 2 fixation on the thermodynamics of ED glycolysis, we fit glucose uptake rates, ethanol production rates, and labeling data from four 13 (Table S4). The equation DG = RT ln(J 2 /J 1 ) was used to calculate reaction free energies for glycolytic reactions using forward-to-reverse flux ratios derived from metabolic flux analysis (MFA) (85,94). For highly thermodynamically favorable reactions in upper ED glycolysis, i.e., glucose-6-phosphate 1-dehydrogenase (ZWF), 6-phosphogluconolactonase (PGL), and 6-phosphogluconate dehydratase (EDD), and the pyruvate kinase (PYK) reaction, forward-toreverse flux ratios were not well resolved by MFA. Therefore, intracellular metabolite concentrations were used to provide tighter bounds on reaction free energies for these reactions, following the equation DG = DG°9 RT ln(Q), (where Q is the ratio of products to reactants) as previously described (85) (Table S4).
For highly favorable reactions (e.g., upper ED glycolysis), a small increase in DG has a minimal impact on net forward flux, whereas for reactions that are closer to equilibrium (e.g., lower ED glycolysis), a modest increase in DG can result in a substantial increase in net flux (92,95). The MFA-derived free energies of lower ED pathway reactions are displayed in Fig. 12B and support the qualitative interpretation of the labeling data: glycolytic reactions tend to be more thermodynamically favorable during N 2 fixation than when NH 4 1 is supplied in the medium. The glyceraldehyde-3-phosphate dehydrogenase (GAPDH:GAP 1 NAD 1 1 P i ! BPG 1 NADH 1 H 1 ) reaction in particular was significantly more thermodynamically favorable during N 2 fixation (P , 0.05). The optimal solution DG value of the GAPDH reaction was 21.617 kJ/mol during N 2 fixation and 20.897 kJ/mol when NH 4 1 was provided. The increased thermodynamic favorability of GAPDH may be due to increased consumption of reducing power by nitrogenase, thereby depleting one of the products (i.e., NADH) of the GAPDH reaction and driving the reaction forward. Other ED pathway reactions did not display significantly different free energies (P . 0.05) but did trend toward greater thermodynamic favorability under N 2 -fixing conditions than under NH 4 1 -replete conditions ( Fig. 12B; Table S4).
Increased thermodynamic favorability of glycolysis during N 2 fixation is a striking observation considering that the ED pathway in Z. mobilis is already highly thermodynamically favorable under NH 4 1 -replete conditions (85). High thermodynamic favorability of glycolysis has been proposed to contribute to high rates of glucose consumption (85,86). A previous study found that in E. coli, thermodynamic favorability of the Embden-Meyerhof-Parnas (EMP) glycolytic pathway increased during an upshift in nitrogen or phosphorus availability, coinciding with an increased rate of glucose uptake (95). In Z. mobilis, increased thermodynamic favorability of the ED pathway may thus contribute to increased rates of glucose consumption and ethanol production. Conclusions. (i) Regulation of nitrogenase activity in Z. mobilis. The results of our metabolomics and proteomics analysis indicate that acute NH 4 1 upshift induces decreased production of nitrogenase cofactors (e.g., FeS clusters, homocitrate) but not immediate, pronounced degradation of nitrogenase itself (Fig. 3 and 4). The decreased abundance of proteins involved in cofactor biogenesis (e.g., NifZ, NifB, NifU) and the decline in intracellular levels of homocitrate during NH 4 1 upshift may play a role in downregulating nitrogenase activity. Proteolytic degradation of NifB, NifE, and NifN has been observed in Azotobacter vinelandii under N 2 -fixing conditions and was found to be important for iron conservation (96). In Z. mobilis, rapid degradation of proteins involved in nitrogenase cofactor biogenesis in response to NH 4 1 upshift may help to optimize iron utilization as soon as nitrogenase activity is no longer required.
Decreased biogenesis of nitrogenase cofactors may inhibit maturation or repair of new or damaged nitrogenase complexes but is not expected to inactivate holoenzyme. Indeed, tracer analysis using 15 NH 4 1 indicated that residual nitrogenase activity may persist for at least 2 h following NH 4 1 upshift (Fig. 6). However, based on labeling patterns, only a small fraction of assimilated nitrogen is derived from non-15 NH 4 1 sources following NH 4 1 upshift, suggesting that if residual nitrogenase activity persists, it is inhibited compared to its fully active state (Fig. 6). It is possible that posttranslational modification plays a role in reversibly inactivating nitrogenase during NH 4 1 upshift, as has been demonstrated in other N 2 -fixing bacteria (68)(69)(70). Differential phosphorylation of NifH and NifD during N 2 fixation has been observed in Z. mobilis (60). However, draT and draG, the genes responsible for reversible inactivation of NifH via ADP-ribosylation in other organisms, have not been annotated in the Z. mobilis genome (30).
We previously reported a multi-omics analysis of oxygen exposure in Z. mobilis (48). Interestingly, unlike during the NH 4 1 upshift, pronounced degradation of nitrogenase proteins did occur when cells were rapidly transferred from anaerobic to aerobic conditions during exponential growth. NifH, NifD, and NifK levels dropped to between 5-fold and 45fold below anaerobic baseline within 2 h after oxygen exposure (48). The oxygen exposure time course was performed under NH 4 1 -replete conditions, so the initial abundance of nitrogenase was lower than it was at the beginning of the NH 4 1 upshift time course. However, the relative decrease in nitrogenase protein was much more pronounced during oxygen exposure than it was during NH 4 1 upshift. Z. mobilis cultures immediately stop growing following a shift to aerobic conditions when NH 4 1 is not supplied in the medium, suggesting that Z. mobilis lacks the ability to protect nitrogenase from oxidative damage (Fig. S6). A regulatory mechanism that stimulates high rates of nitrogenase protein degradation during oxygen exposure but delays complete degradation during NH 4 1 upshift would therefore prevent detrimental accumulation of damaged nitrogenase while potentially allowing for rapid adaptation to fluctuating NH 4 1 availability. (ii) Increased glycolytic activity may help meet increased energy demands associated with N 2 fixation. Regulation of carbon metabolism during changes in nitrogen availability is required to maintain the optimal ratio of carbon to nitrogen for biomass production. For most model microorganisms (e.g., Escherichia coli, Bacillus subtilis, Saccharomyces cerevisiae), decreased nitrogen availability results in a decreased rate of carbon uptake (97)(98)(99). However, for Z. mobilis, the opposite is true; nitrogen limitation results in increased specific glucose consumption despite decreased growth yield (17,19,100). In this study, we provide insight into the potential mechanisms driving this metabolic response, including decreased activity of biosynthetic pathways, increased protein abundance of ADHI, and increased thermodynamic favorability of the ED pathway during N 2 fixation. The metabolic strategy for balancing carbon and nitrogen levels in Z. mobilis appears to leverage redirecting glucose toward ethanol production and away from biosynthetic reactions rather than decreasing glucose uptake. For Z. mobilis, glucose catabolism via the ED pathway is the only means of producing ATP, which is required to power N 2 fixation by nitrogenase. Increased ED pathway activity may therefore be required to meet the high energy demands of N 2 fixation. Additionally, the resulting increase in biomass-specific glucose consumption and ethanol production may serve a competitive advantage in the glucose-rich ecological niches to which Z. mobilis is adapted (101).
A recent publication examined the effect of type 2 quorum sensing autoinducer (AI-2) on Z. mobilis physiology. Because Z. mobilis does not synthesize AI-2, its presence can be interpreted as an indication of the proliferation of competing microorganisms. The study found that AI-2 induced increased nitrogenase activity together with increased biomass-specific glucose consumption and ethanol production by Z. mobilis (20). Interestingly, this study also found that ED pathway genes were not consistently differentially transcribed in the presence of AI-2, but several carbohydrate transporters were. We did not observe any consistent increases in protein abundance of carbohydrate transporters during N 2 fixation in this study (Table S2). The glucose-facilitated diffusion protein (Glf) encoded directly upstream of the operon containing glucose-6phosphate 1-dehydrogenase (zwf), phosphogluconate dehydratase (edd), and glucokinase (glk) genes was not differentially abundant at the protein level during any conditions of NH 4 1 availability (Fig. 11B) (102). Increased thermodynamic favorability of the ED pathway helps to explain the increase in biomass-specific glucose consumption rates during N 2 fixation despite no increases in the abundance of glucose transporters or glycolytic enzymes. Nitrogenase activity may directly contribute to the increased thermodynamic favorability of the ED pathway by consuming both energy (ATP) and reducing equivalents [NAD(P)H], two of the products of glycolysis. We found that intracellular levels of all NTPs were significantly lower under continuous N 2 -fixing conditions (Fig. 9). It is therefore possible that N 2 -fixing conditions result in increased specific rates of glucose consumption and ethanol production by reducing intracellular ATP and NAD(P)H concentrations, thereby driving increased flux through the ED pathway via thermodynamic control.
(iii) Implications for biofuel production. The results presented in this study provide valuable insight into the native regulation of metabolic pathways required for biofuel production in Z. mobilis. In particular, the MEP pathway is a metabolic engineering target for microbial production of isoprenoid bioproducts, including transport fuels, polymers, pharmaceuticals, fragrances, and flavor additives (103). Decreased concentrations of MEP pathway intermediates coincided with decreased abundance of DXS protein. The rate of DXS protein depletion was higher than the growth rate, implying active protein degradation rather than dilution by cell division. This suggests that DXS degradation natively inhibits MEP pathway activity in Z. mobilis, as has been proposed for Arabidopsis plastids (84). This finding has significant ramifications for metabolic engineering efforts to develop Z. mobilis as a platform organism for isoprenoid production. Metabolic engineering strategies often rely on transcription-based control of gene expression with the assumption that increased transcription will result in increased protein abundance. It will therefore be important to further investigate the mechanism of DXS protein degradation in order to avoid degradation-driven repression of MEP pathway activity in engineered strains.
The observation that the abundance of motility proteins decreased during N 2 fixation provides an interesting insight into native mechanisms for energy conservation in Z. mobilis. The impact of flagellar motility on the electrochemical gradient across the inner membrane is an important factor to consider for metabolic engineering. It is possible that deletion of motility genes may help conserve cellular energy, which could be directed to energy-intensive biofuel-producing pathways such as the MEP pathway.
Aside from isoprenoids, many bioproducts such as acetone, isobutanol, and butanediol are derived from the products of glycolysis and compete with ethanol production for substrates. Understanding the native regulation of glycolysis and ethanol production is therefore critical in optimizing the production of these biofuels in a microbial system. Previous studies showed that specific ethanol production and specific glucose uptake both increased during N 2 fixation in Z. mobilis (17,19,20). In this study, we showed that the thermodynamic favorability of the ED glycolytic pathway increased during N 2 fixation and that the protein abundance of zinc-dependent alcohol dehydrogenase ADHI increased during NH 4 1 downshift. These findings help elucidate the underlying physiological mechanisms that drive changes in carbon utilization in Z. mobilis and will help inform metabolic engineering for increased conversion of sugars to bioproducts.
Finally, optimized performance of microbial biofuel producers requires not only maximal product generation but also robust and consistent growth despite the presence of diverse chemical inhibitors. Polyamines have long been appreciated to participate in the bacterial response to environmental stress, but the metabolic pathway for polyamine production in Z. mobilis is still unknown. Accumulation of arginine despite depletion of arginine biosynthetic intermediates suggests that polyamine production from arginine may be repressed during the transition to N 2 -fixing conditions. Understanding the mechanism behind this metabolic response, and the physiological role of sym-homospermidine biosynthesis in stress tolerance, will therefore likely be informative in developing a robust biofuel-producing strain of Z. mobilis.

MATERIALS AND METHODS
Medium preparation. To prepare the liquid minimal medium used in this study, a 10Â glucose solution (200 g/liter), 10Â base solution [10 g/liter KH 2 PO 4 , 10 g/liter K 2 HPO 4 , 5 g/liter NaCl, and either 10 g/ liter (NH 4 ) 2 SO 4 or 2 g/liter MgSO 4 ], and 1,000Â solutions of MgSO 4 Á7 H 2 O (200 g/liter), Na 2 MoO 4 Á2 H 2 O (25 g/liter), and CaCl 2 (10 g/liter) were prepared and autoclaved separately. One thousand-fold solutions of FeSO 4 (2.5 g/liter) and calcium pantothenate (1 g/liter) were also prepared separately and filter sterilized using a 0.2-mm-pore-size filter. Autoclaved deionized water was then aseptically combined with the separately sterilized solutions to produce 1Â concentrations of the added components. Finally, pH was measured to ensure that the medium was within the pH 6 to 6.5 range.
Culture conditions. Zymomonas mobilis subsp. mobilis strain ZM4 (ATCC 31821) was struck onto rich medium-glucose plates (20 g/liter glucose, 2 g/liter KH 2 PO 4 , 10 g/liter yeast extract, 18 g/liter agar) from a frozen 25% glycerol stock and incubated at 30°C in an anaerobic chamber for 3 to 6 days. For each biological replicate, a single colony was used to inoculate a flask of liquid minimal medium with replete ammonia [20 g/liter glucose, 1 g/liter KH 2 PO 4 , 1 g/liter K 2 HPO 4 , 0.5 g/liter NaCl, 1 g/liter (NH 4 4 1 or no NH 4 1 . In the case of no-NH 4 1 minimal medium, the 1 g/liter (NH 4 ) 2 SO 4 was replaced with 0.2 g/liter MgSO 4 to achieve similar molarity. All cultures were grown in foil-covered Erlenmeyer flasks with an approximately 1:5 ratio of liquid volume-to-flask capacity. The total liquid volume ranged from 25 ml (in 125-ml flasks) to 100 ml (in 500-ml flasks). Cultures were stirred with a magnetic stir bar set to 120 rpm. All medium was kept anaerobic for at least 16 h prior to inoculation. The atmosphere in the anaerobic chamber was composed of 2 to 4% H 2 , 5% CO 2 , and the remaining percentage N 2 . Oxygen levels were kept below 100 ppm.
Comparison of N 2 and NH 4 + . To compare continuous N 2 fixation to NH 4 1 -replete conditions, each of three separate NH 4 1 -replete starting cultures were used to inoculate minimal medium with 15 mM NH 4 1 (replete) and minimal medium with no NH 4 1 (N 2 -fixing conditions), resulting in three biological replicates per condition. These cultures were grown anaerobically for up to 16 h. Before reaching stationary phase, cultures were used to inoculate fresh medium, maintaining the same NH 4 1 availability. Again, before reaching stationary phase, the second set of cultures was used to inoculate experimental flasks with a starting optical density at 600 nm (OD 600 ) of approximately 0.05. This passaging was performed both to ensure continuous NH 4 1 availability and to dilute any nutrients provided by the rich medium plate. The result was that experimental cultures for the N 2 -fixing condition were inoculated with cells that had been growing under N 2 -fixing conditions for at least 6 doublings. Following inoculation of experimental cultures, extractions were performed for intracellular metabolite analysis during early, mid-, and late exponential phase (OD 600 of 0.3, 0.5, and 0.7, respectively) for both NH 4 1 -replete and N 2 -fixing cultures. For proteomics, a separate experiment was conducted with the same culture inoculation protocol except that there were five replicates instead of three. Extractions for proteomic analysis were performed at mid-exponential phase (OD 600 of 0.5) only. NH 4 + downshift. To examine the transition to N 2 -fixing conditions, two rounds of passaging were performed as described for the N 2 versus NH 4 1 experiment, except that all passaging was done in medium containing 15 mM NH 4 1 . Flasks containing medium with either no NH 4 1 or 15 mM NH 4 1 were then inoculated using the cultures that had been passaged in replete NH 4 1 . The inoculation volume was such that NH 4 1 carryover was sufficient to sustain a 2-h doubling time for 6 h after inoculation into medium containing no additional NH 4 1 . Based on inoculation volume, the carryover from inoculation resulted in an initial NH 4 1 concentration of less than 2 mM. During the NH 4 1 downshift experiment, metabolomics and proteomics analyses were performed at the same time, sampling from the same cultures. The NH 4 1 downshift condition had biological triplicates and the NH 4 1 -replete controls had biological duplicates. The first extraction for metabolomics and proteomics analysis was performed for both conditions 5.5 h after inoculation (OD 600 of 0.35), when growth was still exponential for both conditions. For the NH 4 1 downshift condition, subsequent samples were taken at 1, 1.5, 2, 3, and 6 h after the first sample. For the NH 4 1 repletion condition, samples were taken at 1 and 2 h after the first sample.
NH 4 + upshift. For the NH 4 1 upshift experiment, two rounds of passaging were performed as described for the N 2 versus NH 4 1 experiment, except that all passaging was done in medium containing no NH 4 1 . Flasks containing medium with no NH 4 1 were then inoculated using the cultures that had been passaged under N 2 -fixing conditions. Approximately 6 h after inoculation (OD 600 of approximately 0.35), the first extraction for metabolomics and proteomics analysis was performed for all cultures. Immediately following this extraction, 1.5 ml of a 1 M NH 4 Cl solution (15 mM final concentration) was added to three experimental cultures, leaving two cultures as N 2 -fixing controls. Subsequently, extractions were taken at 5, 15, 30, 60, and 120 min after addition of NH 4 Cl for the NH 4 1 upshift condition. For the N 2 -fixing controls, extractions were taken at 60 and 120 min after the first extraction. This experiment was also performed using 15 NH 4 Cl, without proteomics sampling. For the 15 N-labeled iteration, metabolomics time point samples were collected at 1,5,15,30,60, and 120 min after addition of NH 4 1 for the NH 4 1 upshift condition and at 15, 30, 60, and 120 min after the first extraction for the N 2 -fixing controls. An additional replicate was also included for the control condition in this iteration, resulting in biological triplicates for both conditions.
Stable isotope labeling. NH 4 Cl (NLM-467-PK) (98 to 99% isotopic purity) were purchased from Cambridge Isotope Laboratories, Inc. For all labeling data in this study, masses were adjusted to account for the natural abundance of 15 N, 13 C, or 2 H using ElemCor (104). For thermodynamics analysis, 10Â glucose stock solutions were prepared using autoclaved deionized water and sterilized by passage through a 0.2-mm filter. Both NH 4 1 and N 2 conditions were grown with 15 g/liter glucose to reduce costs. Growth rates were the same as in 20 g/liter glucose. Cultures were passaged as described for the N 2 versus NH 4 1 experiment, except that the final passage before inoculation into experimental flasks was performed in 4-ml volumes in culture tubes containing medium with labeled glucose matching the labeled glucose present in experimental flasks. This was done to minimize unlabeled carryover from inoculation. Metabolite extractions were performed from experimental flasks 5 to 12 h after inoculation, when cultures reached an OD 600 of 0.35.
Metabolic flux analysis and goodness of fit. A Z. mobilis metabolic model was adapted from a report by Jacobson et al. in 2019 and simplified to include only reactions for glucose uptake, ED glycolysis, and ethanol production (85). Reversible reactions were modeled as separate forward and backward reactions. Within the model, cellular H 1 and CO 2 were allowed to freely exchange with naturally labeled equivalents. Metabolic flux analysis was performed using the INCA software suite (105), which is implemented in MATLAB and uses the elementary metabolite unit (EMU) framework to simulate isotopic distributions (106). We combined labeling data from our 13 (17) to create a single, statistically acceptable flux map using the COMPLETE-MFA technique (107). Glucose tracer inputs were defined by label type ( 13 C or 2 H) and position, but the proportion of unlabeled glucose was allowed to vary by modeling glucose uptake as two glucose inputs, one labeled and one unlabeled, whose relative contribution was controlled by the flux fit optimization process. Final model solutions estimate the unlabeled fraction of glucose at approximately 1%, consistent with nominal tracer purity. Metabolite mass isotopomer distribution (MID) precision was estimated by combining the variance of each measurement across biological replicates and the maximum error observed from naturally labeled metabolites compared to the theoretical MID calculated from natural isotope abundances with a minimum allowable error of 0.3% for each MID, as previously described (85). The combined 2 H and 13 C best-fit flux solutions are contained in Table S4 in the supplemental material. Labeling data from 13 C and 2 H tracer experiments were used in INCA without prior correction for naturally abundant heavy isotopes. Intracellular fluxes were estimated by solving a nonlinear least-squares regression problem that minimizes the variance-weighted sum of square residuals (SSR) between simulated and measured isotopic distributions of intracellular and extracellular metabolites. Because the solver does not guarantee a global SSR optimum, we used a random multistart approach until SSR improvement ceased. Using the optimal solution, we calculated 95% confidence intervals for all estimated fluxes by individually varying each flux and testing the sensitivity of the optimal SSR to changes in that flux. Upper and lower bounds were assigned by varying each flux until the SSR was perturbed beyond a critical point corresponding to a chi-square distribution with a single degree of freedom.
A x 2 test was used to determine whether estimated fluxes adequately describe the measured labeling data. A correct model and data set have an optimized SSR that falls within a x 2 distribution with degrees of freedom equal to the fitted measurements (i.e., non-zero MIDs and measured fluxes, such as uptake and excretion rates) minus the number of independent parameters (i.e., all fluxes estimated by the analysis). We set the critical threshold of our x 2 test at 0.05 (95% confidence) and required that optimized SSRs fell within this distribution for acceptance. We attempted to fit labeling data from Z. mobilis grown under NH 4 1 -replete conditions together with data from N 2 -fixing conditions to a single flux map and were unable to find a statistically acceptable fit, suggesting that these two conditions are distinct metabolic states for Z. mobilis.
Metabolite extraction. At the time of extraction, 5 to 10 ml of liquid culture was extracted using a serological pipette. The culture was then rapidly filtered through a 0.45-mm nylon filter (Millipore catalog no. HNWP04700) using a vacuum flask fitted with a sintered glass funnel, separating cells from the growth medium. Immediately after the medium passed through the filter, the cells captured on the filter were plunged into cold extraction solvent, simultaneously quenching metabolism, lysing cells, and dissolving intracellular metabolites. This was done by placing the filter facedown in a small (5.5-cm-diameter) plastic petri dish containing 1.5 ml extraction solvent (40:40:20 methanol-acetonitrile-water; all high-performance liquid chromatography [HPLC] grade). The dish containing extraction solvent was kept on dry ice or an aluminum block that had been stored at 280°C. The entire process of extraction was done in 30 s or less. The filter was then rinsed in the extraction solvent within the dish using a pipette to dislodge remaining cell debris and metabolites. The 1.5 ml of extract was then transferred to a microcentrifuge tube and centrifuged at 16,000 Â g for 3 min to remove debris. The supernatant was stored at 280°C until analysis by LC-MS. For analysis, 300 to 200 ml of extract was dried down under N 2 gas. Samples were concentrated three times by resuspension in one-third the dry-down volume of solvent A (see "Metabolomics LC-MS analysis"), vortexed for 10 s, and centrifuged at 16,000 Â g for 3 min to remove any remaining cell debris. Fifty microliters of the supernatant was then transferred to an HPLC vial for LC-MS analysis.
Metabolomics LC-MS analysis. Metabolomics analysis by LC-MS was performed using a Vanquish ultra-high-performance LC (UHPLC) system (Thermo Scientific) coupled to a hybrid quadrupole orbitrap mass spectrometer (Q Exactive, Thermo Scientific), as previously described (23,48,108). The chromatography was done using a reverse-phase C 18 column (1.7-mm particle size, 2.1-by 100-mm column; Acquity UPLC BEH). Solvent A was 97% H 2 O and 3% methanol with 10 mM tributylamine (TBA) and ;10 mM acetic acid for a pH of 8.2. Solvent B was 100% methanol. The total run time was 25 min. The flow rate was held constant at 0.2 ml/min. The chromatography gradient was as follows: 5% solvent B for 2.5 min, linear increase to 95% B over 14.5 min, maintenance of 95% B for 2.5 min, linear decrease back to 5% B over 0.5 min, maintenance of 5% B for 5 min. Eluent from the column was analyzed by mass spectrometry from the start of the run until 19 min, at which time flow was directed to waste for the remainder of the run. Compounds separated by HPLC were ionized by electrospray ionization (negative polarity) and analyzed by full MS-selected ion monitoring (MS-SIM) with a scanning range of 70 to 1,000 m/z, an automatic gain control (AGC) target value of 1 Â 10 6 , maximum injection time (IT) of 40 ms, and resolution of 70,000.
Metabolomics computational analysis. LC-MS raw files were converted to mzXML format and visualized using MAVEN (109). Peaks were chosen by comparison with retention times obtained using analytical standards. To account for slight signal variation from injection to injection, samples were either run twice (technical duplicate) and averaged or mixed 1:1 with a universally 13 C-labeled intracellular metabolite sample harvested from E. coli grown in [U-13 C]D-glucose and normalized by U-13 C signal, as previously described (48). For each metabolite, signal intensity was divided by OD 600 to account for variation in culture density between samples. These values were then divided by the average of the three replicates in the control sample to generate fold change values. For N 2 versus NH 4 1 , the control sample was NH 4 1replete conditions at early exponential phase. For NH 4 1 downshift, the control sample was NH 4 1 -replete conditions at time zero. For NH 4 1 upshift, the control sample was N 2 -fixing conditions at time zero. The log 2 of the fold change values was then averaged to obtain the data displayed in this study.
Protein extraction and preparation. At the time of extraction, 4 ml of culture was collected and cells were pelleted by centrifugation for 3 min at 16,000 Â g. Supernatant was discarded, and pellets were frozen and stored at 280°C until further analysis. Samples were prepared for LC-tandem MS (LC-MS/MS) analysis by thawing and then lysing cells in 900 ml of methanol, resulting in a final concentration of over 90% methanol. Samples were kept cold at 4°C for 30 min and then centrifuged for 20 min at 15,000 Â g. Supernatant was removed, and protein extract was air dried at room temperature. The protein pellet was resuspended in 50 ml 8 M urea, 100 mM Tris (pH 8.0)-10 mM TCEP [Tris(2-carboxyethyl)phosphine hydrochloride], and 40 mM chloroacetamide to denature, reduce, and alkylate proteins. Sonication for 10 min ensured that all protein was in solution. The protein concentration was determined with NanoDrop using the A280 method. The protein resuspension was diluted to 1.5 M urea in 100 mM Tris (pH 8.0) and sonicated for 10 min. Trypsin was added at an estimated 50:1 ratio, and samples were incubated overnight (12 h) at ambient temperature. After incubation with digestion enzyme, each sample was prepared for desalting using a 96-well Strata polymeric reversed-phase 10-mg SPE (styrene divinylbenzene) cartridge. Preparation included priming the cartridge wells with 1 ml of ACN (acetonitrile), followed by 1 ml of 0.1% trifluoroacetic acid (TFA). Each sample was acidified with TFA to a final pH of 2.0 or less and then centrifuged for 15 min at 2,000 Â g to remove all nonprotein material. Acidified sample was then loaded onto the cartridge, washed with 1 ml of 0.1% TFA, and then eluted with 600 ml of 80% ACN-0.1% TFA into a clean 96-well plate to be dried. Samples were resuspended in 0.2% formic acid, and peptide mass was assayed with NanoDrop A280 for a final concentration close to 1 mg/ml.
Proteomics LC-MS/MS analysis. Proteomics analysis was performed as previously described (22,25,26). For each analysis, 1 mg of peptides was loaded onto a 75-mm-inside-diameter (i.d.), 30-cm-long capillary with an embedded electrospray emitter and packed in a 1.7-mm-particle-size C 18 BEH column (stationary phase). The mobile phases used were as follows: phase A, 0.2% formic acid; phase B, 0.2% formic acid-70% acetonitrile. The peptides were eluted with a gradient of acetonitrile increasing from 0% to 50% B over 74 min followed by a 1-min increase to 100% B, 5 min sustained at 100% B, and a final 10 min of equilibration in 100% A. The eluting peptides were analyzed with an Orbitrap Eclipse (Thermo Fischer Scientific) mass spectrometer. Survey scans were performed at a resolution of 240,000 with an isolation analysis at m/z 300 to 1,350 and 250% normalized automatic gain control (AGC) target. Datadependent top-speed (1-s) MS/MS sampling of peptide precursors was enabled with dynamic exclusion set to 10 s on precursors with charge states 2 to 5. Data-dependent MS/MS sampling was performed with 0.5-Da quadrupole isolation, with fragmentation by higher-energy collisional dissociation (HCD) with a normal collisional energy (NCE) value of 300%. The mass analysis was performed in the ion trap using the "turbo" scan speed for a mass range of 150 to 1,350 m/z with a maximum inject time of 14 ms, and the normalized AGC target set to 300%.
Proteomics computational analysis. Raw files were analyzed using MaxQuant 1.5.8.3 (110). Spectra were searched using the Andromeda search engine against a decoy target list. Label-free quantitation and match between runs were toggled on, MS/MS tolerance was set to 0.4 Da, and the number of measurements for each protein was set to 1. Default values were used for all other analysis parameters. The peptides were grouped into subsumable protein groups and filtered to reach 1% FDR, based on the target decoy approach. The fasta database "Zm4.CDS.AA.fasta" was used to generate the protein list utilizing the ZM4 tag for protein names. Using RStudio, the label-free quantitation (LFQ) values and protein intensities in each sample were log 2 transformed and filtered to contain proteins that fulfill a .50% cutoff of measurements across the samples. The average and standard deviation were calculated across the replicates for each protein, and fold changes are relative to control time zero for each experiment.
Statistical analysis. Statistical analysis for metabolomics and proteomics data sets was performed in R. For metabolomics, raw signal intensity was normalized by OD 600 and then log 2 transformed before statistical analysis. For proteomics, the log 2 of LFQ values was used. For both metabolomics and proteomics time courses, a repeated-measures analysis of variance (ANOVA) test was performed for each metabolite or protein, comparing time points within the treatment group. This tests the null hypothesis that the mean value for metabolite or protein abundance was not different at any time during the time course. The P values from the repeated-measures ANOVA tests were then adjusted for multiple hypothesis testing using the Benjamini-Hochberg method to control for false discovery rate (FDR) (111). In this case, an FDR-adjusted P value below 0.05 indicates that less than 5% of proteins or metabolites identified as changing during changes in NH 4 1 availability are false positives. For metabolomics during continuous N 2 -fixing conditions compared to continuous NH 4 1 -replete conditions, a two-way repeated-measures ANOVA test was performed to test the null hypothesis that the average metabolite abundance was not different between N 2 -fixing and NH 4 1 -replete conditions, independent of growth stage. The P values from this test were adjusted for multiple hypothesis testing using the Benjamini-Hochberg method. For proteomics during continuous NH 4 1 availability, an unpaired t test was performed to test the null hypothesis that the average protein abundance was not different between N 2 -fixing and NH 4 1 -replete conditions. The P values from this test were adjusted for multiple hypothesis testing using the Benjamini-Hochberg method. For 15 N isotope tracers, a repeated-measures ANOVA was performed to test the null hypothesis that the average fraction of M 1 2 glutamine was not different at any time point between 15 min and 2 h. For thermodynamic analysis using isotopic labeling, statistical analysis was performed in GraphPad Prism. For this data set, an unpaired t test was performed for each metabolite shown in Fig. 12, testing the null hypothesis that the average fraction of reverse-flux-associated labeled form was not different between N 2 -fixing and NH 4 1 -replete conditions. Data availability. The mass spectrometry proteomics data have been deposited in the ProteomeXchange Consortium via the PRIDE (112) partner repository with the data set identifier PXD028526. Metabolomics data have been deposited in the Open Science Framework (https://osf.io) (113) under the project "N2 fixation in Zymomonas mobilis" at https://doi.org/10.17605/OSF.IO/GJVYW.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.