Dynamic Metabolic Rewiring Enables Efficient Acetyl Coenzyme A Assimilation in Paracoccus denitrificans

Central carbon metabolism provides organisms with energy and cellular building blocks during growth and is considered the invariable “operating system” of the cell. Here, we describe a new phenomenon in bacterial central carbon metabolism. In contrast to many other bacteria that employ only one pathway for the conversion of the central metabolite acetyl-CoA, Paracoccus denitrificans possesses two different acetyl-CoA assimilation pathways. These two pathways are dynamically recruited during different stages of growth, which allows P. denitrificans to achieve both high biomass yield and high growth rates under changing environmental conditions. Overall, this dynamic rewiring of central carbon metabolism in P. denitrificans represents a new strategy compared to those of other organisms employing only one acetyl-CoA assimilation pathway.

D uring growth, bacteria have to distribute metabolic flux between catabolism and anabolism. An important central metabolic pathway is the tricarboxylic acid (TCA) cycle, which interfaces anabolism and catabolism. In the tricarboxylic acid (TCA) cycle, the C 2 -unit acetyl coenzyme A (acetyl-CoA) is catabolized into CO 2 , generating reducing equivalents and ATP for the cell. At the same time, the TCA cycle produces several intermediates that are committed to biosynthesis.
Note that the TCA cycle poses a special challenge for many compounds that are exclusively metabolized via acetyl-CoA, such as acetate, alcohols, short-and long-chain fatty acids, esters, and waxes. Because all carbon of acetyl-CoA is lost to CO 2 , this allows energy conservation but not carbon assimilation through the TCA cycle. Consequently, dedicated pathways for the assimilation of acetyl-CoA are required to allow growth on these ubiquitous substrates.
Two different acetyl-CoA assimilation pathways in bacteria have been described: the glyoxylate cycle (GC) (1, 2) and the ethylmalonyl CoA pathway (EMCP) (3,4). The GC uses two enzymes in addition to the enzymes of the TCA cycle. The first enzyme of the GC, isocitrate lyase (Icl), cleaves isocitrate into succinate and glyoxylate. This step is followed by the condensation of glyoxylate with a second molecule of acetyl-CoA to form malate and free CoA in a reaction catalyzed by the second enzyme of the GC, malate synthase (1, 2).
The EMCP also forms malate and succinate. However, unlike the GC, the EMCP is a linear pathway that employs 13 different enzymes that collectively convert a total of three acetyl-CoA and two CO 2 molecules into the TCA cycle intermediates malate and succinate. The key enzyme of the EMCP is crotonyl-CoA carboxylase/reductase (Ccr), which catalyzes the reductive carboxylation of crotonyl-CoA to ethylmalonyl-CoA (3)(4)(5).
Overall, the GC and the EMCP differ substantially in terms of cofactor and carbon requirements. The GC requires four ATPs and generates reducing equivalents at the level of two nicotinamides as well as two quinols per four acetates converted into malate (estimated change in reaction Gibbs free energy [Δ r G=°] ϭ Ϫ129 kJ/mol). To generate two malates from acetate in the EMCP, three molecules of acetate and two CO 2 molecules are converted at the expense of three ATPs and two reduced nicotinamides, while two quinols are generated (estimated Δ r G=°ϭ Ϫ142 kJ/mol). When acetate is used to generate the CO 2 and fuel the cycle, this changes the stoichiometry to three consumed ATPs and one reduced nicotinamide, as well as three quinols, generated (estimated Δ r G=°ϭ Ϫ190 kJ/mol). While the EMCP requires more enzymatic steps (and likely more cellular resources), it enables the fixation of inorganic carbon into biomass, as well as the coassimilation of different C-5 and C-3 carbonic acids compared to in the GC, indicating that the two pathways provide different physiological advantages. In most bacteria, only one of the two acetyl-CoA assimilation pathways is present. Importantly, the alphaproteobacterium Paracoccus denitrificans encodes the enzymes for both acetyl-CoA assimilation pathways (Fig. 1), raising the question of which role each individual pathway plays in the cell.
In this work, we show that P. denitrificans uses both the GC and the EMCP. The EMCP serves as the default acetyl-CoA assimilation pathway, which is always present with low activity in the cell during growth on various substrates, including carbon sources that do not require acetyl-CoA assimilation. In contrast, the GC is specifically induced after an organism switches to carbon sources that depend on acetyl-CoA assimilation, such as acetate or crotonate. We further demonstrate that the two acetyl-CoA assimilation pathways confer distinct physiological advantages. Growth with the EMCP results in increased biomass yield in P. denitrificans, while the GC allows for higher growth rates, suggestive for a rate yield tradeoff between the two pathways. Phylogenetic analysis indicates that P. denitrificans and several other alphaproteobacteria acquired the GC through lateral gene transfer, consisting with a specific adaptation of the central carbon metabolism of these organisms during evolution. Collectively, our experiments show a surprising flexibility in central carbon metabolism of alphaproteobacteria that allows the complete rewiring of metabolic flux depending on the type and availability of the carbon and free-energy source according to the physiological needs of the cell.

RESULTS
P. denitrificans uses the EMCP and the GC for acetyl-CoA assimilation. Earlier studies suggested that Paracoccus might use the EMCP as well as the GC for acetyl-CoA assimilation (4,6,7). We therefore searched the genome of the fully sequenced strain P. denitrificans Pd1222 for homologs of genes involved in the EMCP and the GC from Rhodobacter sphaeroides 2.4.1 and Escherichia coli K-12, respectively. Our analysis verified that P. denitrificans has the genetic potential for both the GC and the EMCP ( Fig. 1; Table 1) (4).
To test whether both the EMCP and the GC are functional and thus operate in P. denitrificans, we grew bacterial cells in batch culture on minimal medium supplemented with different carbon sources and quantified the activity of Ccr, the key enzyme of the EMCP, as well as Icl, the key enzyme of the GC, in cell extracts obtained from the cultures at mid-log phase (Fig. 2).
The activity of Ccr was almost undetectable in cells grown on succinate and at a very low basal level (Ͻ100 mU/mg) in extracts of cells grown on glucose and glycolate. In contrast, the activity of Ccr was significantly higher in extracts of cells grown on crotonate, acetate, and ethanol. In these extracts, Icl activity was also present, while no Icl activity was detected in extracts of cells grown on glucose, succinate, or glycolate. Together, our data suggest that both the EMCP and the GC are active in P. denitrificans. However, while the EMCP is always active at low levels and upregulated in response to growth on substrates requiring acetyl-CoA assimilation, the GC is specifically activated only during the growth of P. denitrificans on these substrates.  Table 1.
Metabolic Rewiring in P. denitrificans ® The EMCP is sufficient to sustain the growth of P. denitrificans on acetate. Next, we tested whether both pathways are essential for acetyl-CoA assimilation in P. denitrificans. To that end, we aimed to selectively block the GC or EMCP using external inhibitors. Several inhibitors of Icl are known (8). We screened three of them to identify 3-nitropropionate (3-NPA [9]) as a potential candidate for a selective GC inhibitor of P. denitrificans in vivo.
3-NPA inhibited purified P. denitrificans Icl in vitro with an apparent 50% inhibitory concentration (IC 50 ) of 34 M at concentrations of 100 M D,L-isocitrate (Fig. 3A). Moreover, 3-NPA did not affect the growth of P. denitrificans on carbon substrates that do not require Icl activity, such as succinate ( Fig. 2B; see also Fig. S1E in the supplemental material). Notably, 3-NPA also did not affect the growth of Methylobacterium extorquens AM1, an EMCP-positive organism that is closely related to P. denitrificans, on acetate ( Fig. S1B and E). This suggested to us that 3-NPA acts as a GC inhibitor that shows negligible off-target effects in the cell, although the possibility of additional effects of 3-NPA could not be completely excluded.
a Key enzymes of the two pathways are in bold. Missing enzyme homologues are indicated by dashes. When we added increasing concentrations of 3-NPA to P. denitrificans growing on acetate, the growth rate was successively decreased from 0.26 Ϯ 0.043 h Ϫ1 (0 M 3-NPA) to 0.13 Ϯ 0.005 h Ϫ1 (600 M 3-NPA) but not completely abolished ( Fig. 3B; Fig. S1D). This indicated that the GC is not essential in P. denitrificans and that the EMCP is sufficient for acetyl-CoA assimilation.
The use of the GC increases the growth rate of P. denitrificans on acetate. To estimate the contribution of each pathway to acetyl-CoA assimilation, we individually deleted the genes coding for Ccr and Icl from the genome of P. denitrificans, yielding Δccr and Δicl strains, respectively ( Fig. 4; Fig. S2). During growth on succinate, both deletion strains grew similarly to the wild type. However, a shift from succinate to acetate caused pronounced and different strain-specific growth defects.
The Δccr strain grew with wild type-like growth rates but displayed an extended lag phase (Fig. 4). Notably, this lag phase ranged from 15 to 50 h between different experiments (Fig. S3). However, when the Δccr strain was subcultured on acetate, the variation in the lag phase disappeared (Fig. 4B, solid blue line), and the growth rate corresponded to that of the wild type ( Fig. 4B and C), indicating that the mutant adapted to acetate.
The Δicl strain, in contrast, did not exhibit a prolonged lag phase but showed a decreased growth rate on acetate, which is consistent with the results of the 3-NPA inhibition experiments. Together, these results suggest that the GC enables fast growth of P. denitrificans on acetate, but unlike with the EMCP, some time is required for the strain's growth to be induced after a switch to acetate.
The EMCP and GC show a complex expression pattern during switches from succinate to acetate. To follow the expression of the EMCP and the GC in vivo, the native ccr and icl genes of P. denitrificans were replaced with constructs encoding translational fusions of Ccr to the red fluorescent protein mCherry and of Icl to the cyan fluorescent protein Cerulean, respectively. The resulting strain (P. denitrificans ccr::ccr-mCherry icl::icl-Cerulean) was continuously monitored by determining the total fluorescence of the population during cultivation in 96-well plates (Fig. 5).
During growth on succinate, cells showed very low levels of mCherry fluorescence and even lower levels of Cerulean fluorescence, which is consistent with our earlier finding that basal Ccr, but no Icl, activity could be detected in cell extracts of P. denitrificans grown on succinate (Fig. 2). When shifted from succinate to acetate, the reporter strain showed biphasic ccr-mCherry and icl-Cerulean expression patterns. Ccr-mCherry fluorescence increased within the first 20 h and then decreased gradually at the onset of exponential phase until it again reached a low basal level. In contrast, the production of Icl-Cerulean started only after 20 h but continued to increase until Metabolic Rewiring in P. denitrificans ® shortly before the cells entered stationary phase. Subsequently, Icl-Cerulean fluorescence dropped transiently before continuing to rise again in the late stationary phase. Very similar growth-linked expression patterns were observed in reporter strains of P. denitrificans carrying only a single Ccr-mCherry or Icl-mCherry fusion (Fig. S5). In summary, our results suggest that acetyl-CoA assimilation in P. denitrificans follows a complex regulatory pattern in which the GC is used for fast and efficient acetyl-CoA assimilation during exponential phase, while the EMCP is used to bypass the time needed for activation of the GC.
Population-wide switch responses monitored by single-cell microscopy. To understand whether the transition from EMCP-to GC-driven acetyl-CoA assimilation during the switch to acetate occurred in the whole population or in only a subset of Asterisks mark the level of significance of the growth rate or maximal optical density between the wild type (wt) and the individual mutants as determined by t test, with the number of asterisks indicating the decimal place of P (e.g., ***, P Ͻ 5 ϫ 10 Ϫ3 ; n.s., not significant). AU, arbitrary units; suc, succinate; ac, acetate.
Kremer et al. ® cells, we followed the production of Ccr-mCherry and Icl-Cerulean in P. denitrificans ccr::ccr-mCherry icl::icl-Cerulean at the single-cell level using time-lapse fluorescence microscopy. Here, we imaged cells growing on succinate ( Fig. 6A and B) and acetate ( Fig. 6C and D) at 2-h intervals and subjected the images to automated analysis (10) to determine the average fluorescence per cell ( Fig. 6A and C). While repeated handling of the cultures for sample collection led to slightly decreased growth rates, the overall expression patterns of ccr-mCherry and icl-Cerulean matched those measured during continuous cultivation in the 96-well plates (Fig. 5).
While no fluorescence was detected for the wild type (negative control) (Fig. S4), the reporter strain showed different production patterns for Ccr-mCherry and Icl-Cerulean. On succinate, Ccr was produced at low levels during exponential phase, with increased expression in the stationary phase. In contrast, Icl was detected only at the end of the exponential phase and only in a small fraction of cells, which appeared strongly fluorescent. This implies that Icl expression is heterogeneous in the late exponential phase on succinate and thus may suggest a bet-hedging-like behavior under these conditions ( Fig. 6A and B; Fig. S6).
Upon the switch to acetate, Ccr was produced first, whereas Icl was induced only at the onset of the early exponential phase. While some cells switched earlier to Icl production than others, all cells produced Icl in the exponential phase, indicating that the whole population and not only a subset of cells shifted from using the EMCP to using the GC (Fig. 6C and D; Fig. S6).
To follow the expression dynamics of ccr and icl continuously in individual cells, we trapped succinate-grown P. denitrificans ccr::ccr-mCherry icl::icl-Cerulean in a microfluidic device. We then provided acetate as the sole source of carbon and tracked cell growth and division as well as reporter production by microscopy over the course of time (Movie S1). This analysis confirmed that essentially all cells switched from the EMCP to the GC, demonstrating that faster acetyl-CoA assimilation is switched on in the whole population and not by individual cells that outgrow the population (Fig. 7).

DISCUSSION
Here, we describe an unprecedented (and unexpected) metabolic redundancy in bacterial central carbon metabolism. The chromosome of P. denitrificans carries genes for two fundamentally different acetyl-CoA assimilation pathways, the EMCP and the GC. Both pathways are functional and regulated dynamically depending on the growth stage of the cells. The EMCP serves a default function and is expressed at basal levels at all times and on all carbon and free-energy sources tested. Upon a switch to acetate, the EMCP is strongly induced at the early stage of growth but then decreases in activity again. The GC, in contrast, is exclusively induced on acetate and reaches peak activity at the late stage of growth, indicating a surprising rewiring of central carbon metabolism in P. denitrificans at the onset of the exponential phase.
What might be the reasons for the rewiring of central carbon metabolism in P. denitrificans? Individual knockout strains of the EMCP and GC show that the two pathways confer distinct advantages. The EMCP increases the growth yield, while the GC allows faster growth on acetate. How can these differences be explained, and what are their consequences? Importantly, the EMCP is able to fix inorganic carbon to increase biomass yield according to the following equation: 3 acetate ϩ 2 CO 2 ϩ 2 NADPH ϩ 2 quinones ¡ 2 malate ϩ 2 NADP ϩ ϩ 2 quinols. This stoichiometry allows organisms using the EMCP to gain extra carbon from the environment if additional reducing equivalents are available to the cell, e.g., through internal storage compounds, such as polyhydroxybutyrate (PHB), or through growth substrates that are comparatively more reduced than average cellular carbon. This is supported by recent calculations that predict an approximately 3%-higher yield for M. extorquens on methanol when the EMCP is used than when the GC is used (11). Another important feature of the EMCP is that it not only enables acetyl-CoA assimilation but also is directly linked to PHB metabolism and can also function in the assimilation of propionate, in addition to several dicarboxylic acids. This versatility makes the EMCP an all-purpose pathway that might allow P. denitrificans to assimilate several different carbon and free-energy sources in parallel, which may explain the constant expression of the EMCP on diverse growth substrates.
The GC, on the other hand, is a very specialized route that requires only two additional enzymes. The GC requires fewer protein resources, and the thermodynamics of the cycle (i.e., the free energy of the overall process) might become more favorable with the upper limit of the free Gibbs energy dissipation rate in P. denitrificans (12). Rerouting acetyl-CoA assimilation to the GC thus might allow a higher metabolic flux and consequently faster growth, providing a potential explanation as to why P. denitrificans switches to the GC when high concentrations of acetate are available.
Overall, a picture emerges in which metabolic rewiring is a highly coordinated strategy in P. denitrificans that allows optimal growth of this species in a changing environment. When the growth substrate changes from succinate to acetate, it is conceivable that the constant expression of the EMCP facilitates the immediate assimilation of acetate. This might help P. denitrificans during the lag phase until the GC is fully induced and thus provide an advantage over other microorganisms that rely on the GC only. Once induced, the GC then may enable increased acetate assimilation rates, which might enable P. denitrificans to outcompete microorganisms that possess solely the EMCP.
Out of 48 Paracoccus genomes analyzed, 34 possess only a ccr homolog, while 9 additionally contain a homolog of icl ( Fig. 8; see also Fig. S7 and S8 in the supplemental material). Only five strains seem to exclusively possess an icl homolog. Notably, ccr phylogeny correspond largely to overall strain phylogeny, suggesting that the EMCP is the ancient acetyl-CoA assimilation strategy in the genus Paracoccus. This hypothesis is in line with our experimental observation that the EMCP serves a default function in P. denitrificans and is further supported by the fact that other, closely related alphaproteobacteria possess only the EMCP (4).
Notably, the phenomenon of metabolic rewiring is presumably not restricted to the Paracoccus clade alone. Several alphaproteobacteria show the genetic potential to express multiple acetyl-CoA assimilation pathways (4), suggesting that dynamic rewiring of the central carbon metabolism is a more widespread strategy in nature. This discovery extends recent findings that metabolic degeneracy plays an important role in alphaproteobacterial metabolism (13). Future experiments will focus on understanding the molecular mechanisms that underlie the coordination of the two acetyl-CoA pathways and their regulation through transcriptional, posttranslational, or allosteric regulatory mechanisms (14)(15)(16)(17) and on clarifying the evolutionary and ecological significance of dynamic metabolic rewiring as a new principle in microbial central carbon metabolism.

MATERIALS AND METHODS
Strains, media, and growth conditions. All experiments were performed with Paracoccus denitrificans strains DSM413 and Pd1222 ( Table 2). The strain preferentially used for all experiments was DSM413, which is the original wild-type strain isolated by Beijerinck and Minkman in 1908 (18). In cases where genetic manipulation of this strain was not successful, its genetically more tractable derivative Pd1222 (19) was used alternatively. P. denitrificans was grown at 30°C in mineral salt medium (trace elements, TE3-Zn) (20) supplemented with defined carbon sources and adjusted to a total carbon concentration of 120 mM. The density of cultures was determined photometrically at a wavelength of 600 nm. Bacterial growth was monitored over time using Tecan Infinite 200 Pro plate reader systems (Tecan Trading AG, Switzerland) with Nunclon Delta Surface 96-well plates (Thermo Scientific, USA). In silico analyses of growth data were performed using the software Prism 7 (GraphPad Software, USA). Escherichia coli was grown at 37°C in Luria-Bertani broth or M9 minimal medium (337 mM NaH 2 PO 4 , 220 mM KH 2 PO 4 , 85.5 mM NaCl, 18.7 mM NH 4 Cl, 1 mM MgSO 4 , 0.3 mM CaCl 2 , 0.13 mM Na 2 EDTA, 0.03 mM FeSO 4 , 1 g/ml biotin, 1 g/ml thiamine) supplemented with 60 mM acetate. Methylobacterium extorquens AM1 was grown as described previously (21) at 30°C in mineral medium supplemented with 10 mM acetate. Antibiotic concentrations were used as follows: kanamycin at 25 or 50 g ml Ϫ1 , spectinomycin at 50 g ml Ϫ1 , and tetracycline at 10 g ml Ϫ1 .
Chemicals. All chemicals were obtained from Sigma-Aldrich (Steinheim, Germany) and Carl Roth (Karlsruhe, Germany) and were of the highest purity available.
Construction of plasmids. Oligonucleotides and synthetic genes were purchased from Eurofins Deutschland (Hamburg, Germany). All standard cloning techniques were carried out according to the instructions in reference 22.
Markerless genomic deletions and integrations of reporter genes were generated using the pK18mobsacB sucrose suicide plasmid system (23).
For the deletion of ccr (Pden_3873) and icl (Pden_1363), respectively, fusions of the respective downstream and upstream flanking regions of the genes, each approximately 700 bp in length, were purchased as synthetic genes cloned into the pEX-K4 backbone (pSYNccrflanks [pTE1602] pSYNiclflanks [pTE1604]) ( Table 3). Restriction of the plasmids with PstI and EcoRI, gel purification of the flanking For the introduction of fluorescent reporter fusions, ccr and icl were ordered as synthetic genes fused to evoglowPp1 (24) and mCherry (25), respectively, each preceded by a 30-bp linker and followed by the downstream region of the respective genes, yielding plasmids pEX4_ccrevoglowPp1ccrds (pTE1601) and pEX4_iclmCherryiclds (pTE1603).
pTE1603 was digested with SalI and BamHI. Subsequent ligation into SalI/BamHI-cut pK18mobsacB yielded pK18mobsacB_iclmCherryiclds (pTE1616). pTE1601 was amplified with targeted omission of the evoglowPp1 gene via inverted PCR using primers NdeI_pKK21-f and pKK21-r. The mCherry gene was amplified from pTE1603 using primers mCherry-f and NdeI_mCherry-r. Restriction of both fragments with AvrII and NdeI and subsequent ligation of the individual fragments to each other resulted in the plasmid pEX4_ccrmCherryccrds. pEX4-ccrmCherryccrds was digested with SalI and BamHI. Subsequent ligation into SalI/BamHI-cut pK18mobsacB yielded pK18mobsacB_ccrmCherryccrds (pTE1624).
pTE1603 was amplified with targeted omission of the mCherry gene via inverted PCR using primers NdeI_pSYNiclXiclds-f and HindIII_pSYNccrXccrds-r. The Cerulean gene (26) was amplified from plasmid pVCERC-6 (27) using primers HindIII_cerulean-f and NdeI_cerulean-r. Restriction of both fragments with HindIII and NdeI and subsequent ligation of the individual fragments to each other resulted in the plasmid pEX4_iclceruleaniclds. Restriction of pEX4_iclceruleaniclds with SalI and BamHI and ligation into SalI/BamHI-cut pK18mobsacB yielded pK18mobsacB_iclceruleaniclds (pTE1625).
For the heterologous expression of icl, icl was amplified from pTE1603 using primers NdeI_icl-for and HindIII_icl. The resulting product was digested with NdeI and HindIII and ligated into NdeI/HindIIIdigested pET16b, yielding pET16b_icl (pTE1614).
Genetic manipulation of P. denitrificans. The transfer of plasmids into P. denitrificans was achieved by biparental mating with the donor strain E. coli ST18 (28). Selection of the first integration was performed on LB or methanol mineral medium plates supplemented with 25 g/ml kanamycin. Colonies were picked and restreaked on LB with kanamycin and LB with 3% sucrose in parallel. Colonies that were kanamycin resistant and sucrose sensitive were grown in plain LB for 2 days. Subsequently, cells were plated on methanol mineral medium plates supplemented with 6% sucrose in serial dilution. Colonies were restreaked on LB with kanamycin and LB with 3% sucrose in parallel. Kanamycin-sensitive, sucrose-resistant clones were screened for the successful deletion or integration of genes by colony PCR.
Preparation of cell extracts. P. denitrificans cultures were grown at 30°C in mineral medium supplemented with various carbon sources. Cells were harvested at mid-exponential phase, resuspended  We declare no competing interests.