Fine and dynamic tuning the glycolytic ux ratio of an articial carbon saving pathway for high yield of mevalonate in Escherichia coli

In natural Escherichia coli, glucose is mainly metabolized via the Embden-Meyerhoff-Parnas (EMP) pathway. However, in the metabolic process of conversion of pyruvate to acetyl-CoA, one-third of the carbon is lost at CO 2 . To decrease the loss of glucose in the metabolic process and enhance the carbon conversion eciency production of desired products by E. coli, we constructed a carbon saving pathway, EP-bido pathway. As the balance of energy and reducing power was not optimal, we use synthetic biology methods to precisely and dynamically adjust the EMP pathway and pentose phosphate pathway (PPP) ux to improve the production of mevalonate (MVA) via the EP-bido pathway.

BP10BF accumulated 11.2 g/L of MVA after 72 h of fermentation and the molar conversion rate from glucose reached 62.2%. Second, the expression of pfkA was suppressed at a certain time by the clustered regularly interspaced short palindromic repeats interference (CRISPRi) system to avoid the growth defect caused by pfkA direct knock-out. The resulting MVA yield of strain BiB1F was 8.53 g/L, and the conversion rate from glucose reached 68.7%.

Conclusion
This is the highest MVA conversion rate reported in shaken ask fermentation. The CRISPRi and promoter ne-tuning provided an effective strategy for metabolic ux redistribution in many metabolic pathways and promotes the chemicals production.

Background
It is an important challenge in metabolic engineering to reasonably allocate metabolic ux to achieve high yields of target products [1]. Traditional metabolic engineering methods modify and optimize an organism for production of chemicals by decreasing ow through competing pathways and introducing heterogeneous production pathways. As such, metabolic rewiring designs are necessary to increase ux towards essential metabolites, for example, overexpressing native pathways [2,3], inhibition of competing pathways [4], increasing Coenzyme A (CoA) availability [5], and construction of pyruvate dehydrogenase bypass. Many strategies have been applied to improve production of chemicals [6].
A new strategy is to decrease the generation of harmful byproducts such as CO 2 or increase the reuse of byproducts by constructing arti cial synthetic pathways. With the rapid development of synthetic biology and molecular biotechnology, scientists have made great efforts to maximize microbial chemical yields focusing on enhancing the e ciency of CO 2 xation and decreasing CO 2 emission. Many unnatural metabolic pathways have been constructed, such as CETCH [7], MCG [8], NOG [9], MOG [10], and so on.
These pathways provide a variety of new ideas to use CO 2 or one-carbon chemicals as carbon sources, and rewire metabolic pathways [11,12].
In natural glycolysis, a variety of carbon sources are metabolized through the Embden-Meyerhoff-Parnas (EMP) pathway, which synthesizes C3 (pyruvate) and C2 (acetyl-CoA) metabolites. Acetyl-CoA is a precursor of almost all biosynthesis and energy metabolism pathways. It is normally produced via pyruvate decarboxylation, in which one-third of the carbon is lost as CO 2 [14]. Therefore, the arti cial pathway must be optimized to be robust and catabolize sugar at a reasonable rate.
The design of arti cial synthetic pathways is usually static, blocking a competing pathway or introducing heterogeneous pathways permanently and continuously [15][16][17]. Sometimes this has a detrimental effect, for example in the early growth period when resources would ideally be dedicated to building biomass.
Implementing exible regulation would be valuable in engineering projects by rebalancing synthetic pathways to respond to the growth phase or the buildup of precursor metabolites [18][19][20].
The PPP is an important energy metabolism pathway in all organisms. The PPP, tricarboxylic acid (TCA) cycle, and hydrogen transfer provide 35%-45%, 20%-25%, and 30%-45% of the NADPH required during aerobic growth. The production/consumption balance of NADPH suggested that the oxidative PPP provided inadequate NADPH for higher MVA production [21]. The ux ratio between the EMP and oxPP pathways affects the ux to MVA and acetate from acetyl-CoA [22]. Increase of the dehydrogenase reactions of the PPP is effective in increasing the yield of NADPH-dependent products [23][24][25][26][27].
To overcome the above challenges, here, we design and construct an Escherichia coli strain that relies on the EP-bi do pathway for carbon catabolism to support growth. The metabolic ow through the PPP was enhanced to supply su cient NADPH by introducing different strength promoters of zwf. And then to avoid the growth inhibition caused by deletion of pfkA, we used the clustered regularly interspaced short palindromic repeats interference (CRISPRi) system to suppress the expression of pfkA by designing three sgRNAs for suppression. The CRISPRi gene regulation system requires only two components, dCas9 protein and a gRNA, to achieve regulation of the transcription level of any gene in the genome. The degree of suppression of gene expression can be controlled by adjusting the binding position and expression amount of the gRNA. The system also has the advantages of strong applicability and no obvious offtarget phenomenon. These characteristics mean CRISPRi is widely used in the eld of metabolic engineering [28,29]. After CRISPRi was applied, we obtained a titer of 8.53 g/L MVA and a yield of 68.7% (mol/mol). This is the highest MVA yield reported in shaken ask fermentation.

Enhancement of PPP ux by increasing zwf expression
One of the essential purposes of the EP-bi do pathway is to lessen EMP pathway ux and improve PPP ux to enhance the NADPH supply. Suitable ux distribution between the EMP pathway and PPP has a positive effect on the production of target compounds. In the EP-bi do pathway, the theoretical optimal carbon split to the EMP pathway and PPP in the EP-bi do route for MVA production is 1:6 [13]; therefore, the maximum carbon theoretical conversion rate of the EP-bi do pathway is 86% (mol/mol). In our previously constructed EP-bi do pathway, the carbon split ratio was 0.43:0.57. To further increase the shunt to the PPP, we aimed to enhance the expression of the rst key gene of the PPP, zwf, by replacing its promoter with a high strength promoter.
We selected ve constitutive promoters with different strengths from the Anderson promoter library. The theoretical strength of each promoter is shown in Table 3. We compared the actual expression strength of these ve synthetic promoters with the original zwf promoter by placing a green uorescent protein gene downstream of the promoters and detecting uorescence intensity. The uorescence/OD 600 at 16 h is shown in Fig. 1A. The strength of the promoters was relatively consistent with that stated by the Anderson promoter library. The strength of promoters BBa-J23100 and BBa-J23104 was relatively high, and BBa-J23100 was the strongest. The strength of the original (native) zwf promoter is between that of BBa-J23108 and BBa-J23114, and is relatively weak. Strain BW-P (pfkA deletion in BW25113) grew slightly poorer than that of the parental strain BW25113.
To detect the effect of PPP enhancement on MVA production, plasmids pBSA (expressing three enzymes catalysis aceytl-CoA to mevalonate) and pFF (carrying fbp and fxpk gene) were transformed into the ve zwf-enhanced strains and cultivated with the control strain BW-P/pBSA pFF (abbreviated to BW-P BF). Strain BW-P10 BF showed almost the same growth and glucose consumption as the others, while the conversion rate of MVA was far higher than that in the control strain due to production of less byproducts. Promoters BBa-J23100 and BBa-J23108 resulted in the highest yield of MVA, 64.3% (mol/mol) and 62.3% (mol/mol) respectively, although the strength of the promoters did not show an obvious positive correlation with the MVA yield. This proved that enhancing expression of gene zwf was effective for increasing the PPP ux. 13 C-Metabolic ux analysis of changes in central carbon metabolism ux and energy metabolism Strains BW-P10 BF and BW-P08 BF and control strain BW-P BF were chosen for metabolic ux analysis.
With the zwf promoter replaced, the normalized data showed that the carbon ux through the oxidative part of the PPP was signi cantly increased, and the carbon ux through the TCA cycle was decreased, which was consistent with our expectations. More carbon ux moved towards the EP-bi do pathway. The two zwf-expression-enhanced strains showed a large difference in TCA cycle ux, which may explain the growth difference between these strains (Fig. 1C).
In addition, the ATP, NADPH, and NADH synthesis capacity and glucose consumption of the three strains were compared based on the 13 C-MFA data. After the EP-bi do pathway and the MVA synthetic pathway were introduced, the NADPH content and yield of the strain were signi cantly improved. pfkA de ciency shunted carbon ux to the PPP and the expression level of zwf was increased. Comparison of the zwfexpression-enhanced strains showed that overexpression of zwf enhanced NADPH synthesis, and the NADPH level was positively correlated with the promoter strength. Taking strains BW25113, BW-P BF, and BW-P10 BF as examples, the introduction of the EP-bi do pathway and overexpression of zwf changed the main source of NADPH: The main NADPH generating pathway shifted from isocitrate dehydrogenase in the TCA cycle to glucose-6-phosphate dehydrogenase in the PPP. This further proved that we have redirected part of the carbon ux of the EMP pathway to the PPP.
In addition, the production of NADH also changed signi cantly, as shown in Figs. 3C and Figs. 3D. Through zwf enhancement, the total amount and the yield of NADH decreased signi cantly. The NADH was produced distinctly in strain BW-P10 BF compared with wild-type strain BW25113: In strain BW25113, ve dehydrogenases were the main source of NADH [glyceraldehyde-3-phosphate dehydrogenase (GAPDH), pyruvate dehydrogenase (PDH), malate dehydrogenase, α-ketoglutarate dehydrogenase, and succinate dehydrogenase]; in strain BW-P10 BF, NADH was mainly formed via GAPDH and PDH.
The EP-bi do pathway and MVA synthetic pathway expression increased the ATP yield from glucose, but the total amount of ATP decreased (Fig. 3E). The pfkA de ciency greatly impaired the EMP pathway and thereby blocked the three steps of substrate level phosphorylation absorption (glycerate-1, 3-diphosphate, phosphoenolpyruvate kinase, and succinyl CoA synthetase). This can also be veri ed from the origin ratio of ATP (Fig. 3F). Enhancing zwf expression resulted in increased ATP production and yield in strain BW-P08 BF compared with BW-P BF.
Down regulation of EMP pathway ux by targeting pfkA using CRISPRi system To further rationally use the carbon source, we tried to suppress pfkA in a time-controlled way through exogenous addition of inducers or self-induction to ensure su cient growth and building block production before activation of MVA synthesis. The CRISPRi system has shown a relatively good inhibition effect in single gene suppression. To further explore the optimal ux through the EMP pathway and improve the conversion rate of highly-NADPH-dependent products, we introduced the CRISPRi system to our EP-bi do pathway (Fig. 4).
To avoid the growth inhibition that may be caused by dCas9 from the CRISPRi gene regulation system, we selected a relatively low strength promoter, BBa-J23134, to promote dcas9. In order to obtain a different repression range, three different sgRNAs targeting the promoter or coding region of pfkA were designed. sgRNA1 were designed on the promoter region of pfkA, sgRNA2 and sgRNA3 targeted the coding chain of pfkA, at the region of 100 bp and 200 bp downstream of the initial codon, which may cause different repression effect [30]. After dcas9 and the sgRNAs were incorporated into pFF and pBSA respectively, six CRISPRi-regulated strains were generated. The fermentation results showed that the introduction of CRISPRi signi cantly inhibited the growth of cells and the glucose consumption was also reduced compared with that of the control strain BW25113 zwf-23100 pFF pBSA (abbreviated to BBF). This may be caused by the toxicity or leaky expression of dCas9. The CRISPRi-regulated strains were induced at 12 h by adding IPTG. From the fermentation results, we can see that the three inhibitory sites we selected had different inhibitory effects in Fig. 5. sgRNA1 showed a better inhibition effect. Although strain BW25113 pFF-dCas9 pBSA-sgRNA1 produced only 8.53 g/L MVA, its MVA conversion rate reached 68.7%, which exceeded the previous best MVA conversion rate. Four control strains were also constructed to con rm the effect of the CRISPRi system on cell growth (strains BW25113 zwf-23100 pBSA-sgRNA1 pFF, BW25113 zwf-23100 pBSA pFF-dCas9, BW25113 zwf-23100 pBSA-sgRNA1-dCas9 pFF, and BW25113 zwf-23100 pBSA pFF, abbreviated to BB1F, BBFd, BB1dF, and BBF, respectively). The results proved our speculation that dCas9 had negative effective on cell growth in our engineered strains ( Figure  S1). The fermentation result showed that inhibitory effect of sgRNA1 on_pfkA is suitable for enhancing MVA fermentation in the EP-bi do system (Fig. 5). In the CRISPRi-regulated strains, we hardly detected any acetic acid, ethanol, or other byproducts during the fermentation process, which was in line with our expectations. The timely inhibition of pfkA reduces the ux of glycolysis, so that the carbon source cannot over ow excessively.

Discussion
Global warming is mainly due to excess CO 2 emission; it is urgently necessary to nd sustainable solutions to address this issue. Moreover, this wasted carbon may have a major impact on the overall economy of biobased products derived from fermentable carbon sources. Scientists have explored the possibility of using microbial systems to optimize carbon conservation during metabolic processes. The pyruvate decarboxylation step of glycolysis releases CO 2 into the environment, resulting in 33% loss of carbon yield; this carbon loss has now been challenged by many scientists. We previously constructed an EP-bi d pathway in E. coli to reduce CO 2 emissions and successfully applied it to produce a series of acetyl-CoA-derived compounds such as PHA, PHB, and MVA. This provides a new approach for the e cient production of acetyl-CoA as a precursor in E. coli. However, in the EP-bi do pathway, the EMP pathway was blocked by knocking out pfkA to attenuate the carbon ow from pyruvate to acetyl-CoA, which limits the growth of the engineered strains. Based on theoretical calculation and metabolic ux analysis, this EP-bi do pathway has great potential for further optimization.
It is possible to control the ux ratio between the EMP and oxPP pathways by ne-tuning and dynamic control of the expression of pfkA and the rst key gene in the PPP, zwf, whose expression acts as a gateway into the EMP and PP pathways. Here, our ne-tuning strategy to improve NADPH availability was to enhance the expression of zwf by replacing its promoter; we used ve promoters with different strengths. The engineered strains BW-P08 BF and BW-P10 BF produced higher MVA titers than the control strain BW-P BF, 9.12 and 11.2 g/L, respectively. The MVA production by these strains did not show an obvious positive correlation with the zwf promoter strength. Since the MVA yield did not represent the ux distribution between the EMP pathway and the PPP, 13 C-MFA was performed to detected the metabolic ux distribution in strains BW-P08 BF and BW-P10 BF (which had high MVA titers) and the control strain BW-P BF. The ux ratio between the PPP and the EMP pathway in strains BW-P08 BF and BW-P10 BF was much higher than that in strain BW-P BF (Fig. 2 ), indicating an improved shunt to the PPP. Also, enhanced zwf expression increased the total amount and molar yield of NADPH (Fig. 3). The NADPHgenerating pathway shifted from isocitrate dehydrogenase in the TCA cycle to glucose-6-phosphate dehydrogenase in the PPP, further proving that carbon ux was redirected from the EMP pathway to the PPP. The NADH level also showed a decreased TCA cycle activity. In terms of the ATP level, enhancement of zwf expression increased the ratio of substrate level phosphorylation (Fig. 3) and reduced the energy supply ratio of pyruvate kinase. Thus, we identi ed the metabolic ux distribution following the netuning of central metabolic nodes, which helps us to understand the impact on metabolism.
In dynamic regulation of metabolic pathways, the CRISPRi system has recently been used to improve ux through different pathways [31,32]. One bene t of using the CRISPRi system over promoter replacement methods is that it does not require genome editing of the target gene, which remains a challenge. The introduction of the CRISPRi system is achieved by adding an inducer at a certain time to start the CRISPRi system. Here, we used it to adjust the inhibition level of pfkA, so as to achieve timely adjustment of EMP pathway/PPP ux. To reduce the growth inhibition caused by pfkA knockout, considering that the strain itself already harbors two plasmids and the CRISPRi system, we integrated dcas9 and the sgRNA onto the two plasmids respectively. After introduction of the CRISPRi system, cell growth and sugar consumption of the engineered strains were signi cantly decreased. The introduction of dCas9 may also have an inhibitory effect on bacterial growth. The relevant limitation of the CRISPRi system is therefore the toxicity of dCas9 expression in certain hosts [33, 34] that would affect the growth of pathway-expressing cells that typically already suffer from growth defects. The three targeting sites played a role in netuning of pfkA and sgRNA1 showed the best inhibition effect. The MVA yield of strain BW25113 pFF-dCas9 pBSA-sgRNA1 was only 8.53 g/L, its yield reached 68.7%, exceeding the previous best conversion rate.

Page 8/21
This study showed glycolytic ux ratio ne-tuning strategies applying in an arti cial carbon saving pathway for e cient MVA production. The strategies presented in this work serve as a guide to metabolic engineering projects requiring acetyl-CoA as a metabolic precursor.

Media and Culture
For plasmid preparation, E. coli strains were cultured at 37°C on a rotary shaker (220rpm) in test tubes containing 5mL Luria-Bertani (LB) medium. For MVA production, 50-mL shake ask cultures were started by 2% inoculation from the 5-mL LB culture. The 50-mL cultures contained M9 minimal medium with 0.2% yeast extract containing 20 g/Lglucose and shaken at 37°C in a rotary shaker (120rpm) for 48h. Overnight cultures were shaken at 37°C in a rotary shaker (220rpm). Antibiotics were added as follows: ampicillin (Amp) 100 mg/mL, spectinomycin (Spc) 50 mg/mL and chloromycetin (Cm) 25mg/mL. For promoter integration and replacement procedure, strains were cultivated in SOB medium.

Strains and plasmids
All E. coli strains and plasmids used are listed in Table 1. Strain BW-P was used as the starting strain for further genetic manipulation [13]. All primers used for molecular manipulations are listed in Table S1. All promoter used for genetic manipulation are listed in Table 2.
Plasmid construction for MVA production To replace the original tac promoter of pBSA plasmid, ve promoters BBa-J 23119, BBa-J 23100, BBa-J 23102, BBa-J 23104, BBa-J 23118 was designed into primers to construct plasmids pBSA-23119, pBSA-23100, pBSA-23102, pBSA-23104, pBSA-23118. Two primers were designed in the opposite direction. Five PCR amplicons were obtained using the original plasmid as the template with primier pcr-23119-F/R, pcr-23100-F/R, pcr-23102-F/R, pcr-23104-F/R, pcr-23118-F/R. DpnI was added to the PCR system in 37 o C for 1 hour to remove the methylated template. A mixture of 50 μL was transformed into competent cells using chemical transformation. Colony PCR was performed by picking monoclonal from resistance plate to eliminate false positives and template interference. Finally, the ve plasmids were transformed into the BW-P 23100 strain with the plasmid pFF.

Construction of CRISPRi suppression system
To select the CRISPRi inhibition site, three different sgRNAs was designed by targeting pfkA promoter sequence, 100bp downstream of the promoter sequence, and 200bp downstream of the initiation codon. dcas9 and sgRNAs were assembled into pFF and pBSA plasmids respectively downstream of an IPTG-induced promoter. Primer dCas9FF-F/R, dCas9-F/R were used to amplify dCas9 sequence. Primer sgRNA-F1, sgRNA1-R and sgRNA-F2, sgRNA1-R were used to amplify sgRNA1 sequence by PCR. The two ampli ed sequences were overlapped from homology arms. Sequences sgRNA2 and sgRNA3 were ampli ed as above. All constructed plasmids were electro-transformed into E. coli strains. To accomplished the timely control of pfkA by CRISPRi system, cells was induced by 200 μM IPTG after 12h of fermentation.
Promoter replacement on E. coli genome Promoter for replacing the promoter of zwf gene were selected from the Anderson promoter library (http://parts.igem.org/Promoters/Catalog/Anderson). Promoter replacement primers homoarm-F and homoarm-cm-R were designed using homology arms at about 300-500 bp at both ends of the target gene promoter, and plasmid pKD3 or pKD4 was used as a template to obtain recombinant fragments with kan or Cm resistance by using PKD3-cm-F/R or PKD4-cm-F/R. Three ampli ed sequences were overlapped from homology arms and resistance tag. All ve replacing sequences were ampli ed as above.
The Red homologous recombination method was employed for gene integration. The pTKRED complementary plasmid was transformed into the target strain. The electrotransfection were performed by growing BW-P in 5 mL LB medium at 30°C and shaking at 220rpm for 12h. 5-mL shake ask cultures using SOB broth were started with a 1% inoculation from the overnight culture. Isopropyl-β-Dthiogalactopyranoside (IPTG) was added to a nal concentration of 0.5 mM to induce λ-prophage (bet, gam,and exo) gene expression. Cells were then incubated at 30°C and shaking at 220rpm until reaching an OD 600 of 0.5 to 0.6. Cell were collected (2 mL), pelleted, and washed three times with cold sterile water to make them electrocompetent. ssDNA mixture (1μM) was added to electrocompetent cells and electroporated at 2.5kV. Add 1 mL LB liquid medium and cultivate for 1 hour. After centrifuged, the collected bacteria were plated onto plates containing 25ug/ml kan or 18ug/ml spc for overnight incubation at 37°C. Transformed strains were selected by their kanR phenotype and were veri ed by PCR.

Measurement of extracellular metabolites
A spectrophotometer was used to measure the optical density at 600nm (OD 600 ) of the bacterial culture.
For extracellular metabolite analysis, 1 mL of culture was centrifuged at12,000 ×g for 2 min. The supernatant was ltered through a 0.22-μm syringe lter for high-performance liquid chromatography analysis. Glucose, MVA, acetate, and pyruvate were measured on an ion exchange column (HPX-87H; Bio-Rad Labs) with a differential refractive index detector (Shimadzu RID-10A). A 0.5-mL/min mobile phase using a 5 mM H 2 SO 4 solution was applied to the column. The column was operated at 65°C.

C-MFA
To investigate if the carbon ux was really redistributed to the newly constructed EP-bi do pathway, 13C-MFA was performed using 100% 1-13 C 1 glucose as the feeding substrate was added to a concentration of Page 10/21°C . The cell pellet was then washed twice with chemical de ned medium and hydrolyzed in 6 M HCl for 24 h at 120 °C (Schwender et al., 2006). The resulting proteinogenic acids were derivatized with N-(tertbutyldimethylsilyl)-N-methyl-tri uoroacetamide containing tert-butyldimethylchlorosilane in acetonitrile at 105 °C for 1 h, and then analyzed by a GC-MS [Agilent 7890 A GC and 5975 C Mass Selective Detector (Agilent Technologies, Santa Clara, USA)] equipped with a DB-1column (Agilent Technologies). The data obtained from GC-MS were corrected by reduction of the natural abundance ratio of C, H, O, N, and Si isotopes [30]. Metabolic uxes were estimated by minimizing the residual sum of squares between experimentally measured and model predicted 13 C-enrichment using 13  Tables Table 1 Strains

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.