When cells are shifted from one metabolic mode to another, significant remodeling of metabolic networks occurs. With the advent of global profiling capabilities, it is now possible to study metabolic response using a systems biology approach. In this paper, we used whole genome microarrays in conjunction with targeted enzyme assays and metabolite and flux measurements to assess the dynamic response that M. extorquens AM1 undergoes when switching from multi-carbon growth to single-carbon growth. Integrating datasets targeting multiple layers in the metabolic hierarchy has generated a much more complete picture of how the cells respond and adapt to changes in their environment than any of the datasets alone, and has provided insights into the mechanisms of this response.
To switch from succinate to methanol growth, energy metabolism shifts from NADH oxidation via NADH dehydrogenase to a cytochrome-based electron transport system that accepts electrons from methanol dehydrogenase during methanol oxidation 
. Carbon metabolism shifts from the TCA cycle and its associated pathways to a methylene H4
F- and CO2
-based assimilatory scheme involving the serine cycle and the EMC pathway 
, while NADH production shifts from the TCA cycle to the methylene H4
MPT dehydrogenase and formate dehydrogenase enzymes (
; ,). To accomplish this shift, over 100 genes must be up-regulated in M extorquens
AM1, producing over 50 enzymes and a number of regulatory proteins 
, and the cell must accommodate high flux through toxic metabolites, such as formaldehyde, glyoxylate, and glycine.
Our time course experiments demonstrated that when the cells were transitioned from a state of succinate growth to methanol growth, they underwent a period of transient carbon starvation followed by periods of recovery and growth. During the starvation phase, the cells were thrown out of metabolic balance and growth ceased, and during the recovery period, all of the methylotrophic metabolic pathways rapidly changed to allow the cells to shift metabolic modes.
During the starvation period (0–30 min), gene expression patterns largely reflected those for a starvation control. In both the starvation control and the methanol switch experiment, expression of genes predicted to be involved in many core cellular processes such as cell structure synthesis, ribosome synthesis and protein production, flagellar synthesis, and iron uptake were down-regulated, conserving energy. Stress response genes were up-regulated along with several protease genes, which correlated with the increase in free amino acid pools detected in cell extracts. Expression of cytochrome c oxidase genes decreased by 30 min and then rose by 2 h mirroring expression of methanol dehydrogenase genes and cytochrome c (encoded by mxaG) used during methanol metabolism suggesting the possibility of a coordinated response. Immediately after the transition, ATP dropped, suggesting that energy metabolism was curtailed, while NADPH rose transiently, in keeping with a drop in biosynthetic consumption, and then dropped, consistent with a drop in production.
Cells growing on succinate maintain significant capacity to oxidize methanol to formaldehyde and then to formate. For the first h after the switch from succinate to methanol, the cells excreted about 1/3 of the total production of these compounds into the medium, mainly as formate. This result suggests that the capacity to consume formate was initially less than the formate production capacity, but by converting most of the formaldehyde to formate and excreting the excess, the cells could survive the transition without buildup of formaldehyde to toxic levels. It is likely that excretion serves as a buffer for flux capacity to these intermediates, allowing the cell to adjust formate consumption capacity according to need, rather than maintaining a high constitutive capacity. During this time period, the formate dehydrogenases rapidly induced and the flux of methanol to CO2 reached nearly full capacity by the time the culture started growing, at about 2 h. By maintaining significant methanol oxidation capacity in the absence of substrate, the cells can take advantage of transient methanol sources to obtain energy, without committing to the considerable effort of inducing the entire system of methylotrophic metabolism.
After the first 30 min, virtually all of the genes involved in methylotrophy began to induce, as expected. However, a number of examples were identified in which enzyme activities did not reflect transcript changes, suggesting post-transcriptional regulation. These included: 1) methanol dehydrogenase, for which enzyme activity increased rapidly during the time that transcripts dropped, suggesting a mechanism of enzyme activation; 2) methylene H4F dehydrogenase, for which transcripts increased 14-fold but enzyme activity stayed constant for the first hour, suggesting a mechanism of post-transcriptional inhibition, either of translation or of enzyme activity, and 3) TCA cycle enzymes (Icd, Sdh, and Mdh), for which the activities stayed relatively constant while transcripts dropped, even after cell division, suggesting either a mechanism of enzyme stabilization or activation.
Surprisingly, during the time period from 30 min to 2 h while the flux capacity from methanol to CO2 was increasing, net flux through methylotrophy assimilatory pathways to biomass did not occur, even though both transcripts and enzyme activities increased significantly. Since intermediates of the EMC pathway did not accumulate, the assimilatory block must have occurred before that pathway. A candidate step for this block was identified, the conversion of methenyl H4F to methylene H4F by methylene H4F dehydrogenase, since as noted above, this enzyme did not increase in activity during this time period. Once this activity increased, flux to biomass occurred.
These results suggest a metabolic strategy in which the assimilatory pathways are primed to receive assimilatory carbon, but carbon flux occurs only after they are poised and ready, and only after the dissimilatory flux has increased substantially. Such a strategy might decrease the likelihood of accumulation of intermediates produced in the biomass cycles that could inhibit growth, such as glyoxylate and glycine 
, ensuring a smooth transition between these two metabolic modes. Our measurements showed that glyoxylate and glycine did not show major changes during this transition.
Another hypothesis generated from these data was that M. extorquens AM1 may fine-tune the levels of these toxic metabolites through the Mcd step (mesaconyl-CoA hydratase) in the EMC pathway, which produces/regenerates the glyoxylate required to drive the serine cycle. The expression pattern of mcd was notable, since it decreased significantly initially, and only slowly returned to a level slightly above the succinate level. Once intermediates of the EMC pathway began to accumulate, those generated before the Mcd step accumulated to a much higher level than those generated after the Mcd step, consistent with a flux bottleneck at this step. Because glyoxylate is toxic at high levels, it would be advantageous for the cell to regulate the production and consumption of this metabolite. Since glyoxylate is the precursor to glycine during methylotrophic growth, by controlling glyoxylate levels, both of these potentially toxic metabolites are managed. This potential flux bottleneck at the Mcd step would allow the cells to control the balance between need and toxicity for glyoxylate and is an interesting candidate for further study.
Another possible metabolic strategy for carbon partitioning during this transition suggested by these results involved 3-hydroxybutyryl-CoA. This metabolite is an intermediate for both PHB synthesis (carbon storage) and the EMC pathway (carbon assimilation), and is thus a key branchpoint. Succinate-grown cells contain higher levels of PHB than methanol-grown cells 
, and so during the transition, a shift must occur that redirects flux from PHB synthesis to the EMC pathway. Our results showed that 3-hydroxybutyryl-CoA increased almost 6-fold in the first 30 min, before significant increases were observed for the EMC pathway intermediates, at the time when PHB synthesis gene expression was beginning to decrease. In other organisms, the partitioning of acetyl-CoA flux between PHB synthesis and assimilation is controlled by levels of NADPH or the ratio of NADPH/NADP 
. The accumulation of 3-hydroxybutyryl-CoA correlated with the increase in NADPH levels and the NADPH/NADP ratio seen during the first 30 min after methanol addition, suggesting that this mechanism of control may also be present in M. extorquens
AM1. Later, when flux began to increase through the EMC pathway, both 3-hydroxybutyryl-CoA and NADPH levels dropped. By decreasing flux into the PHB cycle first, then increasing the enzymes of the EMC pathway, the cells could effectively shunt 3-hydroxybutyryl-CoA into primary assimilation and away from synthesis of the storage product, PHB.
The results presented here showed that in general, each pathway responded as a separate module, with gene expression changes and enzymatic activities responding as a group. One notable exception was the EMC pathway, in which the majority of the genes increased during the transition yet the patterns of this increase differed ().
Another example was the H4
transfer pathway in which expression of mtdB
stood out as different from the other genes involved in the oxidation of formaldehyde to formate. Fae is present in the cell at very high levels 
and likely is involved in keeping intracellular levels of free formaldehyde low. The rapid increase of fae
transcription may reflect the need to respond to a sudden increase in flux from methanol to formaldehyde. Likewise, MtdB is the first NAD(P)H-generating enzyme in the oxidative pathway, and it responded more similarly to the other NADH-generating step, the formate oxidation pathway. It may be advantageous to the cell to control reduced pyridine nucleotide-generating capacity separately from the other oxidative steps, and act to increase that capacity immediately upon downshift.
Metabolic set points
These datasets also revealed another major feature, which is the prevalence of metabolic set points. During the transition from succinate to methanol growth, central metabolic pathways change significantly, with correspondingly large changes in flux through these pathways, on the order of 10-fold 
. Such changes might be expected to reflect transient swings in metabolite concentrations as the enzyme requirements and activities change. However, the concentration of those metabolites that are involved in both succinate and methanol growth such as fumarate, malate, α-ketoglutarate, citrate, pyruvate and glycerate stayed remarkably constant, with less than 2-fold changes during the transition. A similar phenomenon has been reported for E. coli
wild-type and mutant strains in steady-state cultures under different growth conditions, and these authors suggested that metabolic setpoints may be a mechanism to maintain metabolic robustness in a changing environment 
. Metabolites that were required for growth on both succinate and methanol but did not appear as metabolic setpoints in this study include succinate and serine. Succinate decreased from 12.2 mM to 3.7 mM within 10 min of stopping medium flow, consistent with the termination of succinate as the carbon source. Serine responded similarly to other amino acids which correlated with an increase in protease gene expression (, ).
In summary, the multi-level datasets we have generated in this study coupled to our pathway-level analysis have resulted in new insights into the metabolic strategies carried out to effect this shift between two dramatically different modes of growth, one energy-limited (succinate), and the other reducing-power limited (methanol). Analysis of the dynamic response of modular sections of the central metabolic network is an important approach to determine how cells respond and adapt to carbon substrate perturbation. Our work showed that the cells use a “down-stream priming” strategy, inducing biomass cycles before formate production to carry out this transition in a way that does not accumulate toxic intermediates to inhibitory levels. In addition, this work has uncovered key candidates for regulatory control points, as a further step towards understanding metabolic adaptation and the cellular strategies employed to maintain metabolic setpoints.