Download - Novel approach to engineer strains for simultaneous sugar utilization

Transcript
Page 1: Novel approach to engineer strains for simultaneous sugar utilization

Author's Accepted Manuscript

Novel Approach to Engineer Strains for Si-multaneous Sugar Utilization

Pratish Gawand, Patrick Hyland, Andrew Ekins,Vincent J.J. Martin, Radhakrishnan Mahadevan

PII: S1096-7176(13)00073-6DOI: http://dx.doi.org/10.1016/j.ymben.2013.08.003Reference: YMBEN813

To appear in: Metabolic Engineering

Cite this article as: Pratish Gawand, Patrick Hyland, Andrew Ekins, Vincent J.J.Martin, Radhakrishnan Mahadevan, Novel Approach to Engineer Strains forSimultaneous Sugar Utilization, Metabolic Engineering, http://dx.doi.org/10.1016/j.ymben.2013.08.003

This is a PDF file of an unedited manuscript that has been accepted forpublication. As a service to our customers we are providing this early version ofthe manuscript. The manuscript will undergo copyediting, typesetting, andreview of the resulting galley proof before it is published in its final citable form.Please note that during the production process errors may be discovered whichcould affect the content, and all legal disclaimers that apply to the journalpertain.

www.elsevier.com/locate/ymben

Page 2: Novel approach to engineer strains for simultaneous sugar utilization

1  

Novel Approach to Engineer Strains for Simultaneous Sugar Utilization 

Pratish Gawand1, Patrick Hyland1, Andrew Ekins2, Vincent J. J. Martin2, Radhakrishnan 

Mahadevan1, 3* 

1Department of Chemical Engineering and Applied Chemistry, University of Toronto, 200 

College Street, Toronto, ON, Canada M5S 3E5  

2Department of Biology, Centre for Structural and Functional Genomics, Concordia 

University, 7141 Sherbrooke Street West, Montreal, QC, Canada H4B 1R6 

3Institute of Biomaterials and Biomedical Engineering, University of Toronto, 164 College 

Street, Toronto, ON, Canada M5S 3G9  

Corresponding author 

Prof. Radhakrishnan Mahadevan 

Department of Chemical Engineering and Applied Chemistry,  University of Toronto, 200 College Street, Toronto, ON, Canada M5S 3E5  email: [email protected]  Ph: 1‐416‐946‐0996   Fax: 1‐416‐978‐8605 

Abstract 

Use of lignocellulosic biomass as a second generation feedstock in the biofuels industry 

is a pressing challenge. Among other difficulties in using lignocellulosic biomass, one 

major challenge is the optimal utilization of both 6‐carbon (glucose) and 5‐carbon 

(xylose) sugars by industrial microorganisms. Most industrial microorganisms 

sequentially utilize glucose over xylose owing to the regulatory phenomenon of carbon 

catabolite repression (CCR). Microorganisms that can co‐utilize glucose and xylose are of 

Page 3: Novel approach to engineer strains for simultaneous sugar utilization

2  

considerable interest to the biofuels industry due to their ability to simplify the 

fermentation processes. However, elimination of CCR in microorganisms is challenging 

due to the multiple coordinating mechanisms involved. We report a novel algorithm, 

SIMUP, which finds metabolic engineering strategies to force co‐utilization of two 

sugars, without targeting the regulatory pathways of CCR. Mutants of Escherichia coli 

based on SIMUP algorithm showed predicted growth phenotypes and co‐utilized 

glucose and xylose; however, consumed the sugars slower than the wild‐type. Some 

solutions identified by the algorithm were based on stoichiometric imbalance and were 

not obvious from the metabolic network topology. Furthermore, sequencing studies on 

the genes involved in CCR showed that the mechanism for co‐utilization of the sugars 

could be different from previously known mechanisms. 

Keywords 

Strain design, metabolic modeling, glucose‐xylose co‐utilization, bilevel optimization, 

carbon catabolite repression, Escherichia coli 

Abbreviations 

CCR – carbon catabolite repression, PTS – phosphotransferase system, PPP – pentose 

phosphate pathway, G6P – glucose‐6‐phosphate, F6P – fructose‐6‐phosphate, GA3P – 

glyceraldehydes‐3‐phosphate, PEP – phosphoenolpyruvate, PYR – pyruvate, R5P – 

ribose‐5‐phosphate, X5P – xylulose‐5‐phosphate, E4P – erythrose‐4‐phosphate. 

   

Page 4: Novel approach to engineer strains for simultaneous sugar utilization

3  

1. Introduction   

Lignocellulosic biomass is being increasingly sought after as an alternative feedstock for 

biofuels production due  to  the associated benefits  in sustainability  (Jang et al., 2011). 

Hydrolysis of lignocellulosic biomass yields two major sugars, glucose and xylose (Hahn‐

Hagerdal et al., 2001), efficient utilization of which is a major technical challenge (Kim et 

al., 2010).  Industrial microorganisms that can simultaneously utilize glucose and xylose 

are valuable, as  they can shorten and simplify  the  fermentation processes  (Kim et al., 

2010, Kim et al., 2012, Slininger et al., 2011). However, most  industrial organisms such 

as  E.  coli  (Görke  and  Stülke,  2008)  and  Saccharomyces  cerevisiae  (Ho  et  al.,  1998), 

(Kuyper et al., 2005) show a clear preference for glucose over xylose. Similar sequential 

utilization  of  the  sugars  also  exists  in  other  industrial  organisms  such  as  Zymomonas 

mobilis  (Mohagheghi et al., 2002), Scheffersomyces stipitis  (Slininger et al., 2011), and 

Pseudomonas putida (Meijnen et al., 2008). 

Sequential utilization of sugars results in a characteristic growth pattern with an 

intermediate lag‐phase. Such a growth is known as diauxic growth and is controlled by 

the regulatory phenomenon called carbon catabolite repression (CCR). CCR causes the 

genes encoding catabolic proteins for non‐preferred sugars to be repressed during the 

growth on a preferred sugar (Görke and Stülke, 2008, Deutscher, 2008, Rolland et al., 

2002, Bettenbrock et al., 2006).  Mechanistic details of CCR have been rigorously studied 

for many model organisms including E. coli and S. cerevisiae. In E. coli, CCR is mediated 

by the interplay between the transcription activator CRP (cyclic‐AMP receptor protein), 

Page 5: Novel approach to engineer strains for simultaneous sugar utilization

4  

the signaling metabolite cAMP, adenylatecyclase, and the component EIIAGlc of the 

glucose specific phoshphoenolpyruvate carbohydrate phosphotransferase system (PTS) 

(Deutscher et al., 2006). In S. cerevisiae, CCR entails the main glucose repression 

pathway, involving MIG1 transcriptional repressor complex, SNF1 protein kinase 

complex, and protein phosphatase 1 (Rolland et al., 2002, Gancedo, 1998). 

Due to the extensive information available on CCR in E. coli and S. cerevisiae, 

engineering of the regulatory process has been previously attempted to generate 

mutants for glucose‐xylose co‐utilization. For instance, a mutant devoid of PTS (with 

deletion of ptsH, ptsI and crr, encoding HPr, EI and EIIAGlc, respectively) was constructed 

to force non‐PTS mediated glucose uptake. This mutant showed diminished glucose 

repression on xylose (Hernandéz‐Montalvo et al., 2001). E. coli carrying a mutation in 

ptsG (encoding the glucose transporter EIIBCGlc) was found to co‐utilize glucose and 

xylose at low sugar concentrations (Nichols et al., 2001). Allosteric alteration of the 

global regulatory protein crp (designated as crp* or crp+) is known to confer relaxed CCR 

in E. coli (Khankal et al., 2009). Recently, a crp* mutant (with an amino acid substitution 

from Gly122 to Ser122) was shown to co‐utilize glucose and xylose (Yao et al., 2011). 

Mutations in various metabolic genes are also known to confer co‐utilization phenotype 

to E. coli. Deletion of mgsA, for example, has been implicated in improved co‐utilization 

of glucose and xylose (Yomano et al., 2009). A minimal E. coli mutant that was 

engineered to produce ethanol, co‐utilized glucose and xylose using an unknown 

mechanism (Trinh et al., 2008). An alternative strategy to force co‐utilization involves 

cultivation of a co‐culture of substrate‐selective E. coli mutants (Eiteman et al., 2009). 

Page 6: Novel approach to engineer strains for simultaneous sugar utilization

5  

Though these mutants of E. coli show improved co‐utilization of glucose and xylose, 

none of the mutants have shown complete elimination of CCR. Furthermore, similar 

approaches for altering the regulatory genes in S. cerevisiae have not been particularly 

successful. For instance, deletion of the regulatory gene MIG1 did not alleviate glucose 

repression on xylose utilization (Roca et al., 2004). Evolutionary engineering of S. 

cerevisiae also did not yield a strain capable of co‐utilizing glucose and xylose (Wouter‐

Wisselink et al., 2009). These examples suggest that the glucose‐xylose co‐utilization 

problem has not been entirely solved, even in the most intensively studied organisms. 

Due to the natural tendency of the organisms to use the most efficient substrate first, 

which is coordinated by multiple interacting regulatory pathways, the problem of 

engineering a co‐utilization phenotype is especially challenging. This problem is 

compounded further by the lack of knowledge of CCR in many organisms of interest. 

Thus, a method that can generate sugar co‐utilizing mutants of industrial organisms, 

without any knowledge of their regulatory pathways, is of great interest.  

In this study, we present a rational bilevel optimization algorithm, SIMUP, to engineer 

the metabolic network of organisms in such a way that the predicted mutants are forced 

to co‐utilize two sugars for growth. As a major advantage, the algorithm obviates the 

knowledge of underlying regulatory mechanisms of CCR of the target organism. Many 

bilevel algorithms have been previously developed to design strains with improved 

production of metabolites. Notable among these are: OptKnock (Burgard et al., 2003), 

OptGene (Patil et al., 2005), GDLS (Lun et al., 2009), OptORF (Kim and Reed, 2010), 

optFORCE (Ranganathan et al., 2010), EMILiO (Yang et al., 2011). In addition, bilevel 

Page 7: Novel approach to engineer strains for simultaneous sugar utilization

6  

algorithms have also been used for finding enzyme capacity constraints (Yang et al., 

2008), for selection of the best objective function that explains observed experimental 

fluxes (Gianchandani et al., 2008), and for the design of auxotrophy dependent 

microbial sensors (Tepper and Shlomi, 2011). Though similar in framework to these 

algorithms, SIMUP’s distinguishing feature is the use of multiple external media 

conditions in a single optimization problem. 

Solutions predicted by SIMUP were verified using E. coli as a test organism. The mutants 

constructed based on the model predictions were characterized using phenotypic tests 

and batch cultivation studies.   

2. Materials and methods 

2.1. SIMUP algorithm  

SIMUP algorithm identifies strategies for deletion of metabolic reactions to force co‐

utilization of glucose and xylose. The algorithm is based on the fact that lethality of 

metabolic gene deletions depends on the external nutrient conditions. For instance, 

gene deletions that are lethal with glucose as a carbon source may not be lethal with 

xylose as a carbon source. The algorithm uses the metabolic network of an organism to 

find deletion combinations (of reactions) that are lethal when either glucose or xylose is 

provided, but are not lethal only when both glucose and xylose are provided. As these 

reaction deletions do not allow the organism to grow on either glucose or xylose alone, 

but allow the organism to grow on a mixture of glucose and xylose, the organism is 

forced to consume both the sugars simultaneously. As the algorithm uses only the 

Page 8: Novel approach to engineer strains for simultaneous sugar utilization

7  

metabolic network as an input, no knowledge of the underlying regulatory network is 

required. The algorithm was formulated as a bilevel optimization problem (Fig. 1) shown 

in equation (1).  

        2 2                                                   

                     ,             

                                               

                                                                      , 1,2,3

                                                              ,                                                , 1,2,3  

                                                              ,                                                , 1,2,3

                                                              ,                                                 , 1,2,3                                                              0                                                                                                                                                                                                                                                                                                                                                      0,1                                                        

                     1

                                        

          1  

 

Where,   is the growth of the mutant and    is the growth of the wild‐type. 

Superscripts correspond to different growth conditions: 1) with both glucose and xylose, 

2) with glucose alone, and 3) with xylose alone. The outer function is subjected to the 

inner constraints of the three independent flux balance analysis (FBA) (Orth et al., 2010) 

problems.   is the set of metabolites,   and   are the sets of reversible and 

Page 9: Novel approach to engineer strains for simultaneous sugar utilization

8  

irreversible reactions respectively. Variables   and   are the flux variables;   

and   are the lower and upper bounds on the reversible reaction fluxes;   and   

are the lower and upper bounds on the irreversible reaction fluxes;   and   are the 

stoichiometric matrices for reversible and irreversible reactions; and   and   are 

the coefficient vectors of the objective function. The parameter   is the limit on the 

number of reactions that can be deleted. Availability of substrates was altered by 

changing the bounds on the exchange reactions.  

The decision variables for the outer objective function were the binary variables  , 

which could delete or retain a reaction by assuming a value of 0 or 1, respectively. This 

was achieved by multiplying the binary variables to the corresponding upper and lower 

bounds of each reaction as shown in equation (2).  

  . .        ,                                                    2  

1,      0,          

The problem was solved using the strong duality theorem of LP theory by converting the 

bilevel problem into a single level mixed integer linear program (MILP). A detailed, step‐

by‐step description of the solution to the problem can be found in the SI Text online.   

The MILP was implemented in AMPL (ILOG) and was solved using CPLEX 11.2 (ILOG). To 

solve for the bilevel optimization problem, we used a central metabolism model of E. 

coli reported in a previous study with additional reactions for xylose metabolism 

Page 10: Novel approach to engineer strains for simultaneous sugar utilization

9  

(Schilling et al., 2000). To verify the solutions obtained from SIMUP on a genome‐scale 

model, we used iAF1260 model of E. coli metabolism (Feist et al., 2007).   

2.2. Feasibility analysis of metabolites 

To understand the growth phenotypes observed for some mutants, analysis of feasibility 

of formation of each biomass precursor in the metabolic network was performed. The 

problem was formulated using an FBA with an objective function to maximize one 

biomass precursor at a time. The stoichiometric matrix was modified by adding an 

exchange flux for the biomass precursor under consideration. The maximization 

problem shown in equation 3 was iteratively used to find the feasibility of formation of 

each biomass precursor in the background of reaction deletions of the mutants LMSE2 

and LMSE5. Similar method was used in a previous study to find network gaps in 

metabolic network of E. coli (Feist et al., 2007).  

                              

                    . 0                                   

                  

                                                               3  

Where,   is additional exchange reaction for the biomass precursor under 

consideration,   is the stoichiometric matrix,   is the flux vector,   and   are the 

upper and lower bounds on the flux variable, and   is the number of reactions. All the 

constraints on the co‐factors were removed by allowing free exchange and supply.  

   

Page 11: Novel approach to engineer strains for simultaneous sugar utilization

10  

2.3.  Strains, media and plasmids 

All the strains and plasmids were procured from Coli Genetic Stock Center (CGSC), Yale 

University. E. coli K‐12 MG1655 was the wild type control strain and was used to make 

the gene knockout mutants ‐ LMSE1 (Δeda,Δpgi, Δgnd::kanR), LMSE2 

(Δeda, Δpgi, Δrpe::kanR) and LMSE5 (Δeda, Δpgi, Δfbp, ΔpfkB, ΔpfkA::kanR). The 

mutants were constructed using sequential generalized P1 transduction method and the 

gene deletions were confirmed by PCR (sequences of primers used are provided in the 

Supplementary Table 1 online). For isolation of the final knockouts for the mutants 

LMSE2 and LMSE5, in addition to the antibiotic, LB plates were supplemented with 

glucose (0.5%) and xylose (0.5%). 

2.4. Phenotypic test and batch cultivation 

The mutants LMSE1, LMSE2 and LMSE5 were characterized for growth on minimal 

medium and complex medium with different carbon sources (2% glucose, 2%xylose, and 

1% each of glucose and xylose).  

Each strain was then cultivated in 5 L fermenters with 3 L working volume to find the 

growth and sugar consumption profiles. The minimal medium used was described in a 

previous study (Causey et al., 2003). Glucose and xylose, 5 g/L each were used as carbon 

sources. The pH was maintained at 7.0 using 6 M KOH and dissolved oxygen levels were 

maintained above 50 % by varying airflow rate and impeller speed. The samples were 

analyzed for OD550, and glucose and xylose concentrations. The sugars were separated 

using a Bio‐Rad HPX‐87H cation‐exchange column (5 mM H2SO4 mobile phase, 0.5 

Page 12: Novel approach to engineer strains for simultaneous sugar utilization

11  

mL/min flow rate, 42o C column temperature, 20 μL injection volume). The study was 

carried out in duplicates. Seed cultures were prepared by inoculating a fresh colony 

overnight in 3 mL minimal medium. The overnight grown culture was then transferred 

to 500 mL Erlenmeyer flasks with 150 mL minimal medium containing glucose and 

xylose (5 g/L each). The fermenters were inoculated with appropriate amount of 

harvested biomass in exponential growth phase to get an initial OD550 of 0.1. 

Additional characterization of the mutant LMSE2 was carried out for sugar utilization for 

media with different ratios of glucose and xylose: 1) 3.5 g/L glucose and 7.0 g/L xylose, 

and 2) 3.5 g/L glucose and 7.0 g/L xylose.   

2.5. Gene sequencing  

Genes that play a role in carbon catabolite repression and xylose uptake were 

sequenced using Sanger sequencing. The list of genes sequenced, the regions 

sequenced, and the primers used for sequencing are provided in the Supplementary 

Table 2 online. The genes were PCR amplified from the wild‐type E. coli and the mutant 

LMSE2. The PCR products were purified using Fermetas PCR purification kit and were 

sent for sequencing at the The Sanger Sequencing Facility at The Center for Applied 

Genomics (TCAG), Hospital for Sick Children, Toronto. Alignment of the sequences was 

carried out using Geneious 5.5.6 (Biomatters Ltd.). 

2.6. Expression of ptsG* 

The mutated ptsG gene (ptsG*) was amplified from the mutant LMSE2 using primers 

with restriction sites for EcoR1 (Forward: 5’‐

Page 13: Novel approach to engineer strains for simultaneous sugar utilization

12  

AGTCATTGAATTCCGTTGAATGAGTTTTTTTAAAGC‐3’) and HindIII (Reverse:5’‐

GGATCGTAAGCTTTTAGTGGTTACGGATGTACTCA‐3’), and ligated into pBT4 (Lynch and 

Gill, 2006)  (Addgene plasmid number 22831). The mutation was confirmed using Sanger 

sequencing before ligation into the vector. The constructed vector was transformed into 

an E. coli ΔptsG mutant and the strain was designated as E. coli ptsG*. Both E. coli ΔptsG 

and E. coli ptsG* were characterized in fermenters for glucose‐xylose consumption 

profiles using media composition (5 g/L glucose and 5 g/L xylose) and fermentation 

conditions mentioned above. 

3. Results 

3.1.  Solution to SIMUP algorithm   

Lethality in the mutants with metabolic gene deletions can be explained by the inability 

of these mutants to synthesize one or more biomass precursors essential for growth. 

Such auxotrophic lethal mutants can be rescued by providing the missing biomass 

precursor(s) or metabolite(s) that can be converted to the missing precursor(s) (Suthers 

et al., 2009). The solutions predicted by SIMUP can be considered as auxotrophies that 

can be rescued only with a combination of sugars. These solutions allow only a subset of 

the growth precursors to be synthesized from one sugar. The complete set of biomass 

precursors can be synthesized only if both the sugars are consumed simultaneously. 

Using this framework, the solutions found by the algorithm were grouped together 

based on the unique set of biomass precursors that could be synthesized from each 

sugar. A group of equivalent solutions was called a strategy.   

Page 14: Novel approach to engineer strains for simultaneous sugar utilization

13  

By setting the maximum gene deletion limit to three and four, a total of eleven solutions 

were found, which were grouped into three strategies, A, B, and C. Three representative 

mutants were constructed based on one solution from each strategy for experimental 

validation. Though the strategies A and B could be explained from the connectivity of 

the metabolic network, the strategy C could not be explained solely from the 

connectivity. Additional analysis, involving assessment of the feasibility of formation of 

each biomass precursor from the given substrates (section 2.2), was carried out to 

explain the strategy C.  

3.1.1. Strategy A and mutant LMSE1 

Strategy A consisted of three solutions which were grouped together based on the 

unique set of biomass precursors that could be synthesized from glucose and xylose. All 

the solutions allowed the synthesis of only glucose‐6‐phosphate (G6P) from glucose, and 

all the remaining precursors were synthesized from xylose. One of the solutions, with 

three reaction deletions, pgi, gnd and eda, was used to construct the mutant LMSE1, 

with three corresponding gene deletions, pgi, gnd and eda. The solutions belonging to 

strategy A and the metabolic network for the mutant LMSE1 are shown in Figure 2a. 

Deletion of pgi, encoding phosphoglucose isomerase, restricted the entry of glucose into 

glycolysis and conversion of xylose to G6P using reductive pentose phosphate pathway 

(PPP) (conversion to fructose‐6‐phosphate) and reverse activity of phosphoglucose 

isomerase(Fig. 2a). Deletion of gnd, encoding 6‐phosphogluconate dehydrogenase, 

prevented glucose from entering pentose phosphate pathway and no intermediate from 

Page 15: Novel approach to engineer strains for simultaneous sugar utilization

14  

PPP could be used to synthesize glucose. Finally, deletion of eda, encoding the second 

enzyme of Entner‐Doudoroff pathway, ensured that glucose could not be used to 

synthesize glyceraldehyde‐3‐phosphate (G3P), which could otherwise be channeled into 

glycolysis. By deleting these three genes, glucose was confined to synthesis of G6P, and 

could not be used in the downstream metabolic reactions. All the other biomass 

precursors were provided by xylose alone, which entered PPP as xylulose‐5‐phosphate 

(X5P) (Fig. 2a). The mutant was thus predicted to grow only when both glucose and 

xylose were provided and not on glucose or xylose individually.  

The gene deletion combination of LMSE1 was also tested on the genome scale model of 

E. coli iAF1260 (Feist et al., 2007), which predicted non‐zero growth for the mutant 

LMSE1 on xylose alone. This discrepancy was attributed to the fact that glycogen was 

not included as a biomass precursor in the model iAF1260, thereby rendering G6P 

synthesis non‐essential for growth. However, since G6P has been customarily 

considered a biomass precursor and is one of the hub metabolites (Ma and Zeng, 2003), 

we decided to verify the model predictions by constructing the mutant LMSE1.  

The mutant LMSE1 was first phenotypically tested in the complex and minimal medium, 

followed by batch cultivation to determine its sugar utilization characteristics. The sugar 

consumption and growth profiles of the wild type and the mutant LMSE1 are shown in 

Figure 3a and3b, respectively. The phenotypic characterization of the mutant showed 

that it could grow on minimal medium with xylose as the sole carbon source (in 

agreement with iAF1260 predictions). Also, the mutant did not co‐utilize glucose and 

Page 16: Novel approach to engineer strains for simultaneous sugar utilization

15  

xylose, but consumed xylose alone (Fig. 3b). These results indicated that either G6P is 

not a biomass precursor or the mutant could synthesize G6P from xylose.  

In the model iAF1260, G6P is not channeled into any important component except 

glycogen, and since glycogen is not a biomass precursor, it was likely that the mutant 

LMSE1 did not synthesize G6P. Based on these results, the central metabolism model 

was updated by removing G6P as a biomass precursor. This new biomass objective 

function was used for all further simulations to solve SIMUP and to identify strategy B 

and C mutants. 

3.1.2. Strategy B and mutant LMSE2 

The strategy B was a group of five solutions which allowed synthesis of only ribose‐5‐

phosphate (R5P) from glucose and only xylulose‐5‐phosphate (X5P) from xylose. The 

mutant LMSE2, with three gene deletions, pgi, rpe and eda, each corresponding to a 

single reaction, was constructed to verify the strategy B. The solutions belonging to the 

strategy B and the metabolic network for the mutant LMSE2 are shown in Figure 2b. 

The genes pgi and eda were common to the mutant LMSE1 and served the same 

purpose of restricting the entry of glucose into glycolysis and xylose to G6P (Fig. 2b). 

Glucose could not be used to synthesize downstream metabolites beyond R5P in PPP. 

Deletion of rpe prevented xylose from entering PPP, and the only metabolite that could 

be synthesized from xylose was X5P. In presence of both glucose and xylose, R5P (from 

glucose) and X5P (from xylose) could be used to synthesize all the remaining PPP 

Page 17: Novel approach to engineer strains for simultaneous sugar utilization

16  

metabolites, which then could be used to synthesize the remaining biomass precursors. 

The solution was verified on the model iAF1260. The equivalent strategy needed 

deletion of six reactions (The strategies based on genome scale models are provided in 

Supplementary Table 3 online). 

For the mutant LMSE2, deletion of only three genes was required to confer the 

expected phenotype. Additional deletions, based on iAF1260, were not needed, 

indicating that the additional reactions could not support growth, most likely due to a 

limited flux–carrying capacity. The characterization results of the mutant LMSE2 are 

shown in Figure3c and Table 1. The mutant showed model predicted phenotype and did 

not grow on either glucose or xylose minimal medium. However, it grew well on 

minimal medium containing both glucose and xylose. When cultivated in a batch, it 

showed simultaneous utilization of glucose and xylose, thereby confirming SIMUP 

prediction. 

The trend of the sugar uptake rates in the mutant LMSE2 showed that xylose was 

consumed at a higher rate than glucose (Table 1). Even though the stoichiometry 

suggests that glucose and xylose are required in almost equal amounts (for synthesis of 

R5P and X5P, respectively) in the mutant LMSE2, the excess xylose utilization rate can be 

explained by the fact that additional xylose could be utilized using the reaction tkt2 (Fig 

2b). Using the reaction tkt2, X5P and erythrose‐4‐phosphate (E4P) could react to give 

fructose‐6‐phosphate (F6P) and glyceraldehydes‐3‐phosphate (G3P) which were 

channeled into glycolysis, thereby, resulting in a higher xylose to glucose ratio. 

Page 18: Novel approach to engineer strains for simultaneous sugar utilization

17  

It is also clear from the simulation results that the deletions of the genes in the mutants 

LMSE2 slightly decreased the biomass yields per mmol C when compared to the wild 

type (Table 1). Accordingly, experimentally observed values for biomass yields were low 

for the mutant LMSE2. Thus, the co‐utilization phenotype in the mutant LMSE2, which 

required rerouting of the fluxes in the metabolic network, came at the cost of decreased 

efficiency of the metabolic network.  

3.1.3. Strategy C and mutant LMSE5 

The strategy C consisted of three solutions, two with three reaction deletions and one 

with four reaction deletions. The solutions in the strategy C could not be explained from 

the network connectivity like the strategies A and B. The mutant LMSE5 was constructed 

based on the solution with four reaction deletions, pgi, gnd, fbp, and pfk. Due to the 

presence of two isozymes, pfkA and pfkB, the mutant LMSE5 had five gene deletions 

(pgi, gnd, fbp, pfkA, and pfkB). The solutions in the strategy C and the related metabolic 

network for the mutant LMSE5 are shown in Figure 2c. 

With the deletion of the four reactions, glucose could not be used to synthesize 

fructose‐6‐phosphate (F6P).  Furthermore, gluconeogenic flux and entry of glucose into 

PPP were stopped. The gene deletions hence were lethal in the glucose minimal 

medium. Xylose, however, could enter through PPP as X5P, followed by conversion to 

R5P (using the reactions rpi and rpe) (Fig. 2c). With X5P and R5P synthesized from 

xylose, all the remaining metabolites of PPP could possibly be synthesized and 

channeled into glycolysis. Thus, considering the connectivity of the metabolic network, 

Page 19: Novel approach to engineer strains for simultaneous sugar utilization

18  

it appears that the strain would grow on xylose minimal medium. However, SIMUP 

predicted that the deletion of the four reactions would be lethal on xylose. An 

equivalent strategy on the genome scale model iAF1260 needed deletion of seven 

reactions (Supplementary Table 3 online). As the metabolic network was fully 

connected with respect to xylose, lethality on xylose in the strategy C was most likely 

due to stoichiometric constraints. 

On construction of the mutant LMSE5, it was found that the mutant showed no growth 

on glucose, limited growth on xylose (Supplementary Fig. 1), and significant growth only 

when both glucose and xylose were provided (Fig. 3d). The mutant co‐utilized glucose 

and xylose in batch cultivation, in agreement with the SIMUP predictions.   

Similar to the mutant LMSE2, the predicted growth yield of the mutant LMSE5 was 

lower than the wild type, and reduced growth yield was observed even in experimental 

results (Table 1). Additionally, the sugar uptake rates of the mutant LMSE5 were 

different from the mutant LMSE2 due to different stoichiometric requirements. As 

glucose supported the synthesis of most of the biomass precursors in the mutant 

LMSE5, and xylose was required mainly for synthesis of F6P (Fig 2c), glucose was 

consumed at much a higher rate than xylose. 

To investigate the stoichiometric constraints of the mutant LMSE5, further analysis of 

the metabolic network was carried out.  

   

Page 20: Novel approach to engineer strains for simultaneous sugar utilization

19  

 

3.2.  Feasibility analysis of biomass precursors  

Lethality caused due to metabolic gene deletions can be explained by the mutants’ 

inability to synthesize the complete set of biomass precursors (Suthers et al., 2009). To 

identify the cause of lethality of the mutant LMSE5 (on xylose), the feasibility of 

formation of its biomass precursors was analyzed under three different substrate 

conditions: glucose alone, xylose alone, and glucose and xylose. Similar analysis was also 

carried out for the mutant LMSE2.  

Figure 4 shows the results for the feasibility analysis for the mutants LMSE2 and LMSE5. 

As expected, the mutant LMSE2 could not synthesize any precursor except R5P from 

glucose and X5P (not a biomass precursor) from xylose (Fig. 4a). However, in presence 

of both glucose and xylose, all the precursors could be synthesized and hence growth of 

the mutant was possible. For the mutant LMSE5, growth on glucose was not possible 

because R5P and erythrose‐4‐phosphate (E4P) could not be synthesized from glucose 

(as deletion of gnd prevented glucose from entering PPP) (Fig. 4b). The only metabolites 

that were feasible from xylose were R5P and F6P. No other metabolites, such as G3P or 

E4P, could be synthesized, despite the existing reactions that connected these 

metabolites to xylose. To identify whether any stoichiometric imbalance was causing the 

lethality on xylose, we iteratively added secretion reaction for each biomass precursor 

to the metabolic network of the mutant LMSE5 and observed its effect on the growth on 

xylose. 

Page 21: Novel approach to engineer strains for simultaneous sugar utilization

20  

Figure 4c shows the feasibilities of biomass precursors in the mutant LMSE5 after 

addition of a secretion reaction for F6P. Allowing secretion of F6P enabled the mutant 

LMSE5 to grow on xylose alone, suggesting that the cause of reduced growth of the 

mutant LMSE5 on xylose was intracellular accumulation of F6P. The mutant was most 

likely forced to synthesize excess F6P to meet the imposed stoichiometric constraints 

due to gene deletions. As sugar phosphates cannot be secreted out of the cell, a likely 

accumulation of F6P resulted in the limited growth phenotype of the mutant LMSE5 on 

xylose. 

Thus, the co‐utilization phenotype in the mutant LMSE5 can be attributed to 

stoichiometric imbalance in addition to the topological constraints of the metabolic 

network observed in the mutant LMSE2.  

3.3. Characterization of the mutant LMSE2 on different ratios of glucose and xylose 

The mutant LMSE2 showed co‐utilization of glucose and xylose when the two sugars 

were provided in equal concentrations. However, it was important to characterize the 

mutant on different concentrations of glucose and xylose as lignocellulosic hydrolysate 

contains unequal concentrations of glucose and xylose (Hahn‐Hagerdal et al., 2001). We 

characterized the mutant LMSE2 on two different sugar ratios, 1:2 glucose:xylose and 

2:1 glucose:xylose. The growth characteristics obtained from the characterization 

studies are summarized in Figure 5 and Table 1. The mutant LMSE2 grew slower in both 

1:2 glucose:xylose and 2:1 glucose:xylose media as compared to the medium containing 

equal concentrations of glucose and xylose. In addition, glucose and xylose uptake rates 

Page 22: Novel approach to engineer strains for simultaneous sugar utilization

21  

were also found to be lower in media containing unequal concentrations of glucose and 

xylose. From the values of glucose and xylose utilization rates, it is apparent that growth 

of the mutant LMSE2 was controlled by xylose metabolism and excess glucose did not 

improve the growth rate of the mutant (Table 1). The growth yield, however, was found 

to be better when the mutant LMSE2 grew on different sugar mixtures, most likely due 

to slower metabolism and no production of acetate. In addition, excess glucose and 

xylose were found to accumulate in both batches with unequal glucose and xylose 

concentrations. These results indicate that further engineering of the mutant LMSE2 is 

required to adjust to different sugar ratios. Interestingly, with different concentrations 

of sugars, the mutant LMSE2 changed the ratios of sugar uptake rates suggesting that 

the mutant does not need to utilize sugars in strictly equal proportion. This result 

indicates a possibility of further engineering to channel the excess sugar consumed into 

useful products.  

 

4. Investigation of the mechanism of CCR‐elimination in the mutant LMSE2 

To identify whether the molecular mechanism of CCR‐elimination in our co‐utilizing 

strain LMSE2 was similar to the previously reported mutants (Hernandéz‐Montalvo et 

al., 2001, Nichols et al., 2001, Khankal et al., 2009, Yao et al., 2011, Yomano et al., 2009, 

Trinh et al., 2008), where the key genes implicated in CCR were mutated, we sequenced 

eleven genes (and their relevant upstream and downstream regions) involved in CCR in 

E. coli (Supplementary Table 2). The only gene that was found to carry a mutation was 

ptsG, which encodes the glucose transporter EIIBCGlc. A base‐pair substitution from T to 

Page 23: Novel approach to engineer strains for simultaneous sugar utilization

22  

G at position 89 within the gene was identified, which resulted in the amino acid 

substitution from Lys30 to Arg30.  A sequence comparison of the wild‐type and LMSE2 

ptsG genes is provided in the Supplementary Data online. The protein EIIBCGlc consists 

of two domains: a hydrophobic transmembrane EIIC domain and a hydrophilic EIIB 

domain, which are connected by a conserved linker (Jeckelmann et al., 2011). The 

identified mutation was present in the transmembrane EIIC domain of the protein. 

Mutations in this region have been associated with non‐specific utilization of ribose, 

mannitol and fructose (Siebold et al., 2001). Mutations in ptsG are also known to induce 

co‐utilization of glucose and xylose (Nichols et al., 2001). 

To characterize the identified mutation in ptsG, the mutated gene (ptsG*) was expressed 

in the wild‐type E. coli devoid of ptsG (E. coli ΔptsG). Characterization of the two 

mutants (E. coli ΔptsG and E. coli ptsG*) indicated that the mutation in ptsG* was not 

responsible for the elimination of CCR in the mutant LMSE2. This conclusion could be 

drawn as glucose still repressed xylose‐utilization in the E. coli ptsG* mutant (Fig. 6). 

Thus, no CCR gene was found to carry any significant mutation that would eliminate CCR 

in the mutant LMSE2. This result is in contrast to the earlier attempts to eliminate CCR, 

which involved gene deletions of critical CCR genes or employed mutated versions of 

CCR genes. These results suggest involvement of novel genes or mechanisms in the 

observed co‐utilization phenotype of the mutant LMSE2. 

   

Page 24: Novel approach to engineer strains for simultaneous sugar utilization

23  

 

5. Discussion 

Essentiality of metabolic genes is known to be dependent on the external nutrient 

conditions (Harrison et al., 2007). SIMUP algorithm extends this notion to find the gene 

deletion combinations that force co‐utilization of glucose and xylose. Two of the three 

mutants constructed in this study showed the desired co‐utilization phenotype. 

Whereas the mutant LMSE2 showed strict co‐utilization of the two sugars, the mutant 

LMSE5 exemplified a metabolic engineering strategy that was not obvious from 

metabolic network topology. Under different concentrations of sugars the mutant 

LMSE2 consumed sugar at different rates suggesting a possibility of channeling the 

excess sugar flux towards products. Supplementary Table 4 online compares the 

important properties of the mutants constructed in this study with the previously 

reported co‐utilizing mutants.   

One of the limitations of the mutants constructed in the study was that the mutants 

consumed glucose and xylose at a lower rate than the wild‐type, thereby increasing the 

batch times of the fermentation process.  Thus, the co‐utilization phenotype observed in 

the mutants came at the cost of decreased productivity of any potential products. This 

limitation of the mutants constructed can be attributed to the steady‐state metabolic 

modeling approach that was used to design the mutants.  The steady‐state metabolic 

models cannot simulate the dynamic nature of the metabolism, and thus cannot be 

used to predict or optimize the dynamic growth of the metabolic mutants (Song and 

Page 25: Novel approach to engineer strains for simultaneous sugar utilization

24  

Ramkrishna, 2012). To improve the fermentation productivities along with sugar co‐

utilization phenotypes, more detailed dynamic models of metabolism need to be 

implemented (Song and Ramkrishna, 2012, Anesiadis et al., 2008). 

The most important feature of the mutants constructed in this study was that the co‐

utilization phenotype in these mutants was achieved without any prior knowledge of 

the regulatory mechanisms involved in CCR. As the gene deletions identified in the study 

force the mutants to co‐utilize glucose and xylose, the mutants are anticipated to 

accommodate the change by accumulating mutations in their regulatory genes. Thus, 

this study can be useful to find novel gene targets by which mutants can co‐utilize 

glucose and xylose. This approach is especially important when the targeted mutations 

in the regulatory pathways are unsuccessful in forcing the co‐utilization phenotype.  

 To identify the molecular mechanism used by the mutant LMSE2, the genes involved in 

CCR in E. coli were sequenced. Two major mechanisms are implicated in CCR in E. coli: a 

global regulatory mechanism (involving cAMP‐CRP complex), and operon‐specific 

inducer exclusion mechanisms (Görke and Stülke, 2008). A key participant in CCR of E. 

coli is the protein EIIAGlc encoded by crr. For an unrepressed uptake of secondary sugars, 

it is important that the protein EIIAGlc remains in its phosphorylated state 

(Supplementary Fig. 2), which can be sustained by deactivating the subsequent 

phosphate acceptor: EIIBCGlc (encoded by ptsG). Mutations in ptsG have been previously 

reported to induce co‐utilization of glucose and other secondary sugars such as ribose, 

fructose and xylose (Siebold et al., 2001). Characterization of the mutated ptsG in the 

Page 26: Novel approach to engineer strains for simultaneous sugar utilization

25  

mutant LMSE2 showed that the identified mutation was silent and was not responsible 

for the co‐utilization phenotype. In addition, no other mutation in the known CCR genes 

was identified. The mutant LMSE2 thus co‐utilized glucose and xylose most likely by 

accumulating mutations in non‐CCR genes exerting a pleotropic effect on sugar 

utilization, or by varying its gene expression levels without any permanent mutations. 

Another explanation for the co‐utilization phenotype could be changes in the metabolite 

concentrations that could allosterically regulate the proteins involved in CCR. For 

example, phosphorylation state of the protein EIIAGlc depends on the concentration 

ratio of pyruvate and phopshpoenolpyruvate (PEP) (Görke and Stülke, 2008).  The 

possibility of co‐utilization of glucose and xylose, either by varying gene expression 

levels or by changing concentrations of metabolites, is especially likely since the mutant 

was not adaptively evolved. Both these phenomena provide interesting possibilities to 

investigate the mechanism involved in co‐utilization of glucose and xylose. In the future, 

we expect to perform genome re‐sequencing, transcriptomic and metabolomic studies, 

that can potentially distinguish the mechanism underlying the co‐utilization phenotype. 

Finally, as the gene deletion strategies identified by SIMUP strongly couple sugar co‐

utilization to growth, by not allowing growth on individual sugars, it is ensured that the 

mutants do not revert back to their original CCR phenotype. To the best of our 

knowledge the mutants LMSE2 and LMSE5 are the only mutants that have growth‐

coupled co‐utilization of glucose and xylose, thereby making them the best candidates 

for adaptive evolution. 

Page 27: Novel approach to engineer strains for simultaneous sugar utilization

26  

 

6. Conclusions 

We developed a rational algorithm that uses conditional synthetic lethality to force the 

co‐utilization of two sugars. We demonstrated a successful use of the algorithm by 

constructing E. coli mutants capable of co‐utilizing glucose and xylose; however, the 

mutants consumed sugars at lower rate than the wild‐type. No previous knowledge of 

the regulatory network was needed to construct these mutants. We believe that this 

novel approach is not limited to E. coli and can be conveniently extended to other 

industrially relevant organisms with genome‐scale metabolic models and other sugar 

combinations. 

   

Page 28: Novel approach to engineer strains for simultaneous sugar utilization

27  

Table 1. Growth characteristics of the mutants LMSE2 and LMSE5. 

Mutant  Sugar ratio 

μexp 

(h‐1) 

qglc(mmol/ 

g dcw/h) 

qxyl (mmol/ g dcw/h) 

Yx/s (g/mmol C) 

μFBA (h‐1) 

YX/S (predicted)(g/mmol C)

Wild type  1G:1X  0.60 (glc)0.28 (xyl) 

9.3  6.53  0.011 (glc)0.008  (xyl) 

0.86 0.49 

0.0150.015 

LMSE2  1G:1X  0.38  3.89  4.91  0.008  0.67  0.014 

LMSE2  2G:1X  0.26  2.17  2.48  0.010  0.32  0.012 

LMSE2  1G:2X  0.34  2.76  3.78  0.010  0.51  0.014 

LMSE5  1G:1X  0.20  3.86  1.12  0.007  0.36  0.012 

Growth rates and biomass yields were calculated using genome‐scale metabolic model iAF12610 of E. coli. Experimentally obtained substrate uptake rates were used for growth rate calculations. 

   

Page 29: Novel approach to engineer strains for simultaneous sugar utilization

28  

Figure legends 

Figure 1. The schematic shows the logic behinds the SIMUP algorithm. The algorithm 

finds an optimum strategy that allows growth only when both glucose and xylose are 

provided together and does not allow growth on either glucose or xylose alone. This 

strategy causes the organism to co‐utilize glucose and xylose. The problem was 

formulated as a bilevel optimization problem. The outer objective function maximized 

the growth on both the substrates and penalized the growth on either one substrate, by 

deleting the reactions in the metabolic network. The inner optimization problems find 

the maximum possible growth under each substrate condition while conforming to the 

reaction deletions by the outer optimization problem.  

 

Figure 2. The solutions predicted by the SIMUP algorithm. Based on how each solution 

splits the metabolic network, they have been divided into three strategies: a) Strategy A 

in which glucose is used to synthesize glucose‐6‐phosphate and all other metabolites are 

synthesized from xylose. Among the three gene combinations shown, the highlighted 

(red) solution was used to construct mutant LMSE1. The metabolic map of the mutant 

LMSE1 is shown with equivalent solutions. b) Strategy B has five solutions. The 

highlighted solution was used to construct the mutant LMSE2.  Glucose was used to 

synthesize ribose‐5‐phosphate and xylose was used to synthesize xylulose‐5‐phosphate, 

and using these two metabolites all the remaining biomass precursors were synthesized. 

c) Strategy C contains three solutions. The four gene deletion strategy highlighted was 

Page 30: Novel approach to engineer strains for simultaneous sugar utilization

29  

used to construct mutant LMSE5. The strategy has non‐intuitive solutions as xylose was 

connected to all the metabolites. However, the strain was capable of growth only when 

glucose and xylose both were provided.  

Figure 3. Results for batch cultivation studies and phenotypic results for the wild type 

and the mutants. The phenotypic test results are shown as insets in the plots (  ‐ no‐

growth/ lethal phenotype;  ‐ growth/non‐lethal phenotype;  ‐ limited 

growth/synthetic‐sick phenotype). The batch cultivation studies show time profiles of 

dry cell weight (•), glucose (♦), and xylose (▲). The abbreviations used are LB – Luria‐

Bertani broth; MM – minimal medium; G – glucose; X – xylose; GX – glucose + xylose; NS 

– no sugars. a) Batch cultivation results for wild type E. coli. b) Characterization results 

for LMSE1. c) Characterization results for LMSE2. d) Characterization results for LMSE5. 

Figure 4. Stoichiometric feasibilities for the mutant strains LMSE2 and LMSE5.  G – 

glucose as carbon source; X – xylose as carbon source; GX – glucose and xylose as 

carbon source. a) Feasibilities of biomass precursors for the mutant LMSE2. b) 

Feasibilities of biomass precursors for the mutant LMSE5. c) Feasibilities for all the 

biomass precursors with additional secretion reaction for F6P in the metabolic network 

of the mutant LMSE5. F6P – fructose‐6‐phosphate; GA3P – glyceraldehydes‐3‐

phosphate; 3PG – 3‐phosphoglycerate; PEP – phosphoenolpyruvate; PYR – pyruvate; 

R5P – ribose‐5‐phosphate; E4P – erythrose‐4‐phosphate; OA – oxaloacetate; AKG – α‐

ketoglutarate.  

Page 31: Novel approach to engineer strains for simultaneous sugar utilization

30  

Figure 5. Results for batch cultivation studies for the LMSE2 mutant on unequal 

concentration of glucose and xylose. The batch cultivation plots show time profiles of 

dry cell weight (•), glucose (♦), and xylose (▲). a) batch cultivation results on 7 g/L 

xylose and 3.5 g/L glucose, b) batch cultivation results on 3.5 g/L xylose and 7 g/L 

glucose.  

Figure 6. Results for batch cultivation studies for E. coli ΔptsG and E. coli ptsG* mutants. 

The batch cultivation studies show time profiles of dry cell weight (•), glucose (♦), and 

xylose (▲).  a) characterization results for E. coli ΔptsG, b) characterization results for E. 

coli ptsG*. 

Acknowledgements 

We thank Laurence Yang (University of Toronto) for the insightful discussions on the 

computational work, and Echo Da Zhang and Saba Khan (University of Toronto) for their 

help in experiments. This research was funded by ABIP Canada and NSERC 

Bioconversion Network. 

   

Page 32: Novel approach to engineer strains for simultaneous sugar utilization

31  

References 

Anesiadis, N., Cluett, W.R., Mahadevan, R., 2008. Dynamic metabolic engineering for increasing bioprocess productivity. Metab. Eng. 10, 255‐266.  

Bettenbrock, K., Fischer, S., Kremling, A., Jahreis, K., Sauter, T., Gilles, E.D., 2006. A quantitative approach to catabolite repression in Escherichia coli. J. Biol. Chem. 281, 2578‐2584.  

Burgard, A.P., Pharkya, P., Maranas, C.D., 2003. Optknock: a bilevel programming framework for identifying gene knockout strategies for microbial strain optimization. Biotechnol. Bioeng. 84, 647‐657.  

Causey, T.B., Zhou, S., Shanmugam, K.T., Ingram, L.O., 2003. Engineering the metabolism of Escherichia coli W3110 for the conversion of sugar to redox‐neutral and oxidized products: Homoacetate production. Proc. Natl. Acad. Sci. U. S. A. 100, 825‐832.  

Deutscher, J., 2008. The mechanisms of carbon catabolite repression in bacteria. Curr. Opin. Microbiol. 11, 87‐93.  

Deutscher, J., Francke, C., Postma, P.W., 2006. How phosphotransferase system‐related protein phosphorylation regulates carbohydrate metabolism in bacteria. Microbiol. Mol. Biol. Rev. 70, 555‐1031.  

Eiteman, M.A., Lee, S.A., Altman, R., Altman, E., 2009. A substrate‐selective co‐fermentation strategy with Escherichia coli produces lactate by simultaneously consuming xylose and glucose. Biotechnol. Bioeng. 102, 822‐827.  

Feist, A.M., Henry, C.S., Reed, J.L., Krummenacker, M., Joyce, A.R., Karp, P.D., Broadbelt, L.J., Hatzimanikatis, V., Palsson, B.O., 2007. A genome‐scale metabolic reconstruction for Escherichia coli K‐12 MG1655 that accounts for 1260 ORFs and thermodynamic information. Mol. Syst. Biol. 3, 121.  

Gancedo, J.M., 1998. Yeast carbon catabolite repression. Microbiol. Mol. Biol. Rev. 62, 334‐361.  

Gianchandani, E.P., Oberhardt, M.A., Burgard, A.P., Maranas, C.D., Papin, J.A., 2008. Predicting biological system objectives de novo from internal state measurements. BMC Bioinformatics 9.  

Görke, B., Stülke, J., 2008. Carbon catabolite repression in bacteria: Many ways to make the most out of nutrients. Nat. Rev. Microbiol. 6, 613‐624.  

Page 33: Novel approach to engineer strains for simultaneous sugar utilization

32  

Hahn‐Hagerdal, B., Wahlbom, C.F., Gardonyi, M., van Zyl, W.H., Cordero Otero, R.R., Jonsson, L.J., 2001. Metabolic engineering of Saccharomyces cerevisiae for xylose utilization. Adv. Biochem. Eng. Biotechnol. 73, 53‐84.  

Harrison, R., Papp, B., Pál, C., Oliver, S.G., Delneri, D., 2007. Plasticity of genetic interactions in metabolic networks of yeast. Proc. Natl. Acad. Sci. U. S. A. 104, 2307‐2312.  

Hernandéz‐Montalvo, V., Valle, F., Bolivar, F., Gosset, G., 2001. Characterization of sugar mixtures utilization by an Escherichia coli mutant devoid of the phosphotransferase system. Appl. Microbiol. Biotechnol. 57, 186‐191.  

Ho, N., Chen, Z., Brainard, A., 1998. Genetically engineered Saccharomyces yeast capable of effective cofermentation of glucose and xylose. Appl. Environ. Microbiol. 64, 1852‐1859.  

Jang, Y.S., Park, J.M., Choi, S., Choi, Y.J., Seung, D.Y., Cho, J.H., Lee, S.Y., 2011. Engineering of microorganisms for the production of biofuels and perspectives based on systems metabolic engineering approaches. Biotechnol. Adv. .  

Jeckelmann, J.M., Harder, D., Mari, S.A., Meury, M., Ucurum, Z., Müller, D.J., Erni, B., Fotiadis, D., 2011. Structure and function of the glucose PTS transporter from Escherichia coli. J. Struct. Biol. 176, 395‐403.  

Khankal, R., Chin, J.W., Ghosh, D., Cirino, P.C., 2009. Transcriptional effects of CRP* expression in Escherichia coli. Journal of Biological Engineering 3, 13.  

Kim, J.H., Block, D.E., Mills, D.A., 2010. Simultaneous consumption of pentose and hexose sugars: An optimal microbial phenotype for efficient fermentation of lignocellulosic biomass. Appl. Microbiol. Biotechnol. 88, 1077‐1085.  

Kim, J., Reed, J.L., 2010. OptORF: Optimal metabolic and regulatory perturbations for metabolic engineering of microbial strains. BMC Syst Biol 4, 53‐53.  

Kim, S.R., Ha, S.J., Wei, N., Oh, E.J., Jin, Y.‐., 2012. Simultaneous co‐fermentation of mixed sugars: a promising strategy for producing cellulosic ethanol. Trends Biotechnol. .  

Kuyper, M., Toirkens, M.J., Diderich, J.A., Winkler, A.A., Van Dijken, J.P., Pronk, J.T., 2005. Evolutionary engineering of mixed‐sugar utilization by a xylose‐fermenting Saccharomyces cerevisiae strain. FEMS Yeast Res. 5, 925‐934.  

Lun, D.S., Rockwell, G., Guido, N.J., Baym, M., Kelner, J.A., Berger, B., Galagan, J.E., Church, G.M., 2009. Large‐scale identification of genetic design strategies using local search. Mol. Syst. Biol. 5, 296.  

Page 34: Novel approach to engineer strains for simultaneous sugar utilization

33  

Lynch, M.D., Gill, R.T., 2006. Broad host range vectors for stable genomic library construction. Biotechnol. Bioeng. 94, 151‐158.  

Ma, H., Zeng, A.P., 2003. Reconstruction of metabolic networks from genome data and analysis of their global structure for various organisms. Bioinformatics 19, 270‐277.  

Meijnen, J.P., De Winde, J.H., Ruijssenaars, H.J., 2008. Engineering Pseudomonas putida S12 for efficient utilization of D‐xylose and L‐arabinose. Appl. Environ. Microbiol. 74, 5031‐5037.  

Mohagheghi, A., Evans, K., Chou, Y.C., Zhang, M., 2002. Cofermentation of glucose, xylose, and arabinose by genomic dna‐integrated xylose/arabinose fermenting strain of Zymomonas mobilis AX101. Applied Biochemistry and Biotechnology ‐ Part A Enzyme Engineering and Biotechnology 98‐100, 885‐898.  

Nichols, N.N., Dien, B.S., Bothast, R.J., 2001. Use of catabolite repression mutants for fermentation of sugar mixtures to ethanol. Appl. Microbiol. Biotechnol. 56, 120‐125.  

Orth, J.D., Thiele, I., Palsson, B.O., 2010. What is flux balance analysis? Nat. Biotechnol. 28, 245‐248.  

Patil, K.R., Rocha, I., Forster, J., Nielsen, J., 2005. Evolutionary programming as a platform for in silico metabolic engineering. BMC Bioinformatics 6, 308.  

Ranganathan, S., Suthers, P.F., Maranas, C.D., 2010. OptForce: An optimization procedure for identifying all genetic manipulations leading to targeted overproductions. PLoS Computational Biology 6.  

Roca, C., Haack, M.B., Olsson, L., 2004. Engineering of carbon catabolite repression in recombinant xylose fermenting Saccharomyces cerevisiae. Appl. Microbiol. Biotechnol. 63, 578‐583.  

Rolland, F., Winderickx, J., Thevelein, J.M., 2002. Glucose‐sensing and signalling mechanisms in yeast. FEMS Yeast Res. 2, 183‐201.  

Schilling, C.H., Edwards, J.S., Letscher, D., Palsson, B.O., 2000. Combining pathway analysis with flux balance analysis for the comprehensive study of metabolic systems. Biotechnol. Bioeng. 71, 286‐306.  

Siebold, C., Flükiger, K., Beutler, R., Erni, B., 2001. Carbohydrate transporters of the bacterial phosphoenolpyruvate: Sugar phosphotransferase system (PTS). FEBS Lett. 504, 104‐111.  

Page 35: Novel approach to engineer strains for simultaneous sugar utilization

34  

Slininger, P.J., Thompson, S.R., Weber, S., Liu, Z.L., Moon, J., 2011. Repression of xylose‐specific enzymes by ethanol in Scheffersomyces (Pichia) stipitis and utility of repitching xylose‐grown populations to eliminate diauxic lag. Biotechnol. Bioeng. 108, 1801‐1815.  

Song, H.S., Ramkrishna, D., 2012. Prediction of dynamic behavior of mutant strains from limited wild‐type data. Metab. Eng. 14, 69‐80.  

Suthers, P.F., Zomorrodi, A., Maranas, C.D., 2009. Genome‐scale gene/reaction essentiality and synthetic lethality analysis. Molecular Systems Biology 5.  

Tepper, N., Shlomi, T., 2011. Computational design of auxotrophy‐dependent microbial biosensors for combinatorial metabolic engineering experiments. PLoS ONE 6, e16274.  

Trinh, C.T., Unrean, P., Srienc, F., 2008. Minimal Eschenchia coli cell for the most efficient production of ethanol from hexoses and pentoses. Appl. Environ. Microbiol. 74, 3634‐3643.  

Wouter Wisselink, H., Toirkens, M.J., Wu, Q., Pronk, J.T., Van Maris, A.J.A., 2009. Novel evolutionary engineering approach for accelerated utilization of glucose, xylose, and arabinose mixtures by engineered Saccharomyces cerevisiae strains. Appl. Environ. Microbiol. 75, 907‐914.  

Yang, L., Cluett, W.R., Mahadevan, R., 2011. EMILiO: A fast algorithm for genome‐scale strain design. Metab. Eng. 13, 272‐281.  

Yang, L., Mahadevan, R., Cluett, W.R., 2008. A bilevel optimization algorithm to identify enzymatic capacity constraints in metabolic networks. Computers and Chemical Engineering 32, 2072‐2085.  

Yao, R., Hirose, Y., Sarkar, D., Nakahigashi, K., Ye, Q., Shimizu, K., 2011. Catabolic regulation analysis of Escherichia coli and its crp, mlc, mgsA, pgi and ptsG mutants. Microb Cell Fact. 10, 78.  

Yomano, L., York, S., Shanmugam, K., Ingram, L., 2009. Deletion of methylglyoxal synthase gene (mgsA) increased sugar co‐metabolism in ethanol‐producing Escherichia coli. Biotechnol. Lett. 31, 1389‐1398.  

   

Page 36: Novel approach to engineer strains for simultaneous sugar utilization

35  

Highlights 

• We developed an algorithm, SIMUP, to engineer organisms for sugar co‐utilization. 

• The algorithm predicted multiple strategies including some non‐obvious strategies. 

• Two model‐based Escherichia coli mutants showed co‐utilization of glucose and 

xylose. 

• No CCR gene had significant mutations, suggesting novel mechanisms of co‐

utilization. 

 

Page 37: Novel approach to engineer strains for simultaneous sugar utilization

Maximize Growth on glucose and xylose

Subject to Network stoichiometry

Reaction bounds Gene knockouts identified by outer problem

Maximize Growth on glucoseSubject to

Network stoichiometryReaction bounds

Gene knockouts identified by outer problem

Maximize Growth on xylose

Subject to Network stoichiometry

Reaction bounds

Gene knockouts identified by outer problem

Maximize (Growth on glucose and xylose) –

(Growth on glucose) – (Growth on xylose) by deleting genes

Maximum gene knockouts ≤ Limit

Subject to

Figure

Page 38: Novel approach to engineer strains for simultaneous sugar utilization

PGI

RPE

FBP EDD EDA TPI FBA

eda

pts

zwf pgl

gnd

pgi

GLC

G6P

F6P

FDP

GA3PDHAP

6PGA 6PG Ru5P

GA3P

rpe

XuL

2DDG6Pedd

rpi

xkspfkA

fbp

fba

tpi

To other biomass precursors

tktA2

X5PR5P

S7P

F6P E4P

tktA1

talA

XYL

xylA

eda

pts

zwf pgl

gnd

pgi

GLC

G6P

F6P

FDP

GA3PDHAP

6PGA 6PG Ru5P

GA3P

rpe

XuL

2DDG6Pedd

rpi

xkspfkA

fbp

fba

tpi

To other biomass precursors

tktA2

X5PR5P

S7P

F6P E4P

tktA1

talA

XYL

xylA

Deleted reactions

Glucose flux

Glucose and xylose flux

Xylose flux

F6P Biomass precursor

rpe Deleted gene

PGI

GND

EDD EDA FBP

eda

pts

zwf pgl

gnd

pgi

GLC

G6P

F6P

FDP

GA3PDHAP

6PGA 6PG Ru5P

GA3P

rpe

XuL

2DDG6Pedd

rpi

xkspfkA

fbp

fba

tpi

To other biomass precursors

tktA2

X5PR5P

S7P

F6P E4P

tktA1

talA

XYL

xylA

PGI

GND

FBA FBP TPI

PFK

a b

c

Figure

Page 39: Novel approach to engineer strains for simultaneous sugar utilization

0.00

0.50

1.00

1.50

2.00

2.50

0.00

1.00

2.00

3.00

4.00

5.00

6.00

0 4 8 12 16 20 24

bio

mass (g/L

)

glu

cose ,

xylo

se (g/L

)

Time h

0.00

0.50

1.00

1.50

2.00

2.50

3.00

0.00

1.00

2.00

3.00

4.00

5.00

6.00

0 5 10 15 20

Dry

cell w

eig

ht

(g/L

)

glu

cose ,

xylo

se (g/L

)

Time h

a

0.00

0.50

1.00

1.50

2.00

2.50

3.00

0.00

1.00

2.00

3.00

4.00

5.00

6.00

0 5 10 15 20

bio

mass (g/L

)

glu

cose ,

xylo

se (g/L

)

Time h

c

b

dMM

LB

G X GX

NS

MM

LB

G X GX

NS

MM

LB

G X GX

NS

0.00

0.50

1.00

1.50

2.00

2.50

0.00

1.00

2.00

3.00

4.00

5.00

6.00

7.00

0 5 10 15 20

bio

mass (g/L

)

glu

cose ,

xylo

se (g/L

)

Time h

Figure

Page 40: Novel approach to engineer strains for simultaneous sugar utilization

G X G+X

F6P - - +

GA3P - - +

3PG - - +

PEP - - +

PYR - - +

R5P + - +

E4P - - +

OA - - +

AKG - - +

G X G+X

F6P - + +

GA3P + - +

3PG + - +

PEP + - +

PYR + - +

R5P - + +

E4P - - +

OA + - +

AKG + - +

G X G+X

F6P - + +

GA3P + + +

3PG + + +

PEP + + +

PYR + + +

R5P - + +

E4P - + +

OA + + +

AKG + + +

a b c

Not feasibleFeasible+ -

Figure

Page 41: Novel approach to engineer strains for simultaneous sugar utilization

0.00

0.50

1.00

1.50

2.00

0.00

1.00

2.00

3.00

4.00

5.00

6.00

7.00

8.00

0 5 10 15 20

bio

mass (g/L

)

glu

cose, xylo

se (g/L

)

Time h

0.00

0.50

1.00

1.50

2.00

0.00

1.00

2.00

3.00

4.00

5.00

6.00

7.00

8.00

0 5 10 15 20

bio

mass (g/L

)

glu

cose, xylo

se (g/L

)

Time h

a b

Figure

Page 42: Novel approach to engineer strains for simultaneous sugar utilization

0.00

0.50

1.00

1.50

2.00

2.50

3.00

0.00

1.00

2.00

3.00

4.00

5.00

6.00

0 5 10

bio

mass (g/L

)

glu

cose, x

ylo

se (g/L

)

Time h

0.00

0.50

1.00

1.50

2.00

2.50

0.00

1.00

2.00

3.00

4.00

5.00

6.00

0 5 10

bio

mass (g/L

)

glu

cose, x

ylo

se (g/L

)

Time h

a b

Figure