using computations to reconstruct, analyze and redirect metabolism
DESCRIPTION
Maranas discussed how to speed up up the process of building and correcting organism-specific metabolic models using the recently developmed MetRxn knowledgebase of standardized metabolite and reaction information.TRANSCRIPT
Using Computations to Reconstruct, Analyze and Redirect Metabolism
E-mail: [email protected] Web page: http://maranas.che.psu.edu
Penn State University University Park, PA 16802
Costas D. Maranas
Chemical factories on the µm scale
Escherichia coli Chemical Process Plant
Outline q Reconstruct: Organism-specific genome-scale models
Automated assembly and curation of genome-scale models of metabolism (Suthers et al., PLoS Comp Biol, 2009; M. AbuOun et al., J Biol Chem, 2009; Satish Kumar et al., BMS Sys. Biol., 2011; Saha et al., PLoS ONE, 2011)
q Redesign: Computational strain design Pathway prospecting and identification engineering strategies leading to targeted overproductions (Ranganathan et al., PLoS Comp Biol, 2010; Ranganathan and Maranas, Biotech. J., 2010; Xu et al., Metab. Engr., 2011)
Compile and standardize genome-scale models and databases with consistent naming and balanced reactions (Kumar et al. submitted)
q Standardize: MetRxn: standardized knowledgebase of metabolites and reactions
Nature Reviews Genetics, 7, 130-141, 2006
Genome-Scale Metabolic Models Linking reactions proteins genes (GPRs)
Genome-scale metabolic models vs. fully sequenced genomes
sequenced genomes (genomesonline.org)
genome-scale metabolic models
year
# co
mpl
eted
Genome-scale metabolic models § Mycoplasma genitalium 274 met., 262 rxn’s
(Suthers et al., PloS Comp Biol, 2009) (Collaboration with J.C. Venter Inst.) § Salmonella enterica 945 met., 1,964 rxn’s
(M. AbuOun et al., J Biol Chem, 2009) (Collaboration with M.F. Anjum and M.J. Woodward @ Veterinary Laboratories Agency-Weybridge (UK))
§ Methanosarcina acetivorans 779 met., 776 rxn’s (Satish Kumar et al., BMC Sys. Biol., 2011) (Collaboration with J. Ferry @ PSU)
§ Zeo mays (maize) 1,825 met., 1,983 rxn’s (Saha et al., PLoS ONE, 2011)
Tools for reconstruction § GapFind/GapFill Network connectivity analysis and
restoration (Satish Kumar, et al., BMC Bioinformatics, 2008)
§ GrowMatch Reconcile consistency with growth/no growth experiments upon genetic and/or environmental perturbations
(Satish Kumar and Maranas, PLoS Comp Biol, 2009;Zomorrodi and Maranas, BMC Sys. Biol., 2010)
?
G/G
NG/G
G/NG
NG/NG
Zea mays
• Major food crop 31% of the world production of cereals (Sanchez and Cardona, Bioresource Technology 2008)
• Important source of biofuels 3.4 billion gallons of ethanol in 2004 accounting 99% of all biofuels in the USA (Farrell et al., Science 2006)
Model plant species (vintageprintable.com) Zea mays genome
(Schnable et al., Science 2009)
• Zea mays genome 2.3 gigabase pairs – 14 χ A. thaliana
genome (Schnable et al., Science 2009)
• Filtered Gene Set (FGS) 32,540 genes and 53,764 transcripts (Schnable et al., Science 2009)
Important to study Zea mays as a food crop, biofuels production platform and a model for studying plant genetics
• Functional annotation 54% of total genes associated with specific metabolic functions (Schnable et al., Science 2009)
Model reconstruction workflow
STEP 3: Assemble into genome-scale metabolic model
STEP 1: Transfer GPR from AraGEM via orthologs in Zea mays
(Saha et al., PLoS ONE 2011)
STEP 4: Network connectivity analysis and restoration
STEP 2: Identify additional biotransformations using homology searches
Zea mays genome (Schnable et al., Science 2009)
AraGEM (Dal’molin et al., Plant Physiology 2010)
Step 1: Transfer GPR from AraGEM via orthologs in Zea mays
(Saha et al., PLoS ONE 2011)
Tpi
AT2G21170
R01015
ACF85433
AUTOGRAPH Method (Notebaart et al., BMC Bioinformatics 2006)
Tpi
R01015
Gene
Protein
Rxn
A. thaliana Z. mays
BLASTp (E value = 10 -30 )
Auto model: 1186 rxns
Step 2: Identify additional biotransformations using homology searches
(Saha et al., PLoS ONE 2011)
Functionally annotated genome
Zmxxxx
Osxxxx
Forward BLASTp
(E value = 10 -30 )
Reverse BLASTp
(E value = 10 -30 )
Z. Mays genome
A. thaliana
O. sativa
S. bicolor
T. aestivum
G. max
Species origin of newly added reactions in the
draft model
Draft model: 1667 rxns
Step 3: Assemble into genome-scale metabolic model
(Saha et al., PLoS ONE 2011)
Generate biomass equation from biomass composition of young and vegetative Maize plants. (Penningd et al., Journal of Theoretical Biology 1974)
Generate stoichiometric matrix (Sij).
Establish GPR associations.
Protein Carbohydrates Lipids Ions L-alanine Ribose Glycerotripalmitate Potassium L-arginine Glucose Glycerotristearate Chloride L-aspartic acid Fructose Glycerotrioleate L-Cystine Mannose Glycerotrilinolate RNA L-glutamic acid Galactose Glycerotrilinoleate ATP L-glycine Sucrose GTP L-histidine Cellulose Lignin CTP L-Isoleucine Pectin 4-coumaryl alcohol UTP L-leucine Coniferyl alcohol L-lysine Hemicellulose Sinapyl alcohol DNA
L-phenylalanine Arabinose dATP
L-methionine Xylose Organic acids dGTP L-proline Mannose Oxalic acid dCTP L-serine Galactose Glyoxalic acid dUTP L-threonine Glucose Oxalo-acetic acid L-tryptophan Uronic acids Malic acid L-tyrosine Citric acid L-valine Aconitic acid
Functional model: 1821 rxns
Step 4: Network connectivity analysis and restoration
(Saha et al., PLoS ONE 2011)
Enforce network connectivity by finding & filling gaps in model (GapFind & GapFill)
?
(Satish Kumar, et al., BMC Bioinformatics, 2008)
Example of gap filling
(E value = 1χ10-24 )
Final model: 1985 rxns
Maize iRS1563 Included genes 1,563 Proteins 876 Single functional proteins 463 Multifunctional proteins 170 Protein complexes 4 Isozymes 36 Multimeric proteins 148 Others 55 Reactions 1,985 Metabolic reactions 1,900 Transport reactions 70 Exchange reactions 15 GPR associations Gene associated 1,668 Nongene associated 175 Nonenzyme associated 86 Spontaneous 41 Metabolites 1,825 Cytoplasmic 1,744 Plastidic 115 Peroxisomic 93 Mitochondrial 86 Vacuolic 5 Extracellular 15
Distribution of metabolites in cytoplasm and organelles
• All reactions are elementally and charged balanced • 42% of reaction entries have direct literature evidence • 448 reactions and 369 metabolites are unique to iRS1563 compared to A. thaliana • 674 reactions and 893 metabolites are unique to maize iRS1563 compared to C4GEM
C4 Photosynthesis
C4 photosynthesis
Secondary metabolism
• Phenylpropanoid metabolism H, G, S lignins • Flavonoid biosynthesis Fungal defense • Others include steroid biosynthesis, caffeine metabolism, streptomycin biosynthesis, etc.
• CO2 fixation is carried out in mesophyll cell • The Calvin cycle (RuBisCO) works in bundle sheath cell • Photo respiration is impeded due to separation • It requires more energy (ATP) to power additional steps
Unique features of C4 PS
Ribulose 1,5-bisphosphate + CO2 3 phosphoglycerate RuBisCO
First reaction of Calvin cycle
Photosynthesis/respiration
CO2
O2
CO2 transport Uptake Sucrose transport Disabled Photon transport Uptake H2O transport Uptake Inorganic nutrient transport Uptake O2 transport RUBISCO: EC 4.1.1.39
Photosynthesis (PS)
O2
Biomass
CO2
Biomass CO2 transport Release Sucrose transport Uptake Photon transport Disabled H2O transport Uptake Inorganic nutrient transport Uptake O2 transport Uptake RUBISCO: EC 4.1.1.39 Both disabled
Respiration (R)
Photorespiration (PR)
Unconstrained Release Carboxylation:
Oxygenation = 3:1 Carboxylation
(Wise et al., 2007)
Models (maize iRS1563 & A. thaliana iRS1597) available at http://maranas.che.psu.edu/models.htm
Cyanothece 51142 and Synechocystis 6803 Collaboration with H. Pakrasi lab (Wash. U.)
Cyanothece 51142 (Image courtesy of The Pakrasi Lab)
Synechocystis 6803 (Image courtesy of The Pakrasi Lab)
• Efficient nitrogen fixation Highest fixation rate than many filamentous cyanobacteria
(Zehr et al., 2005; Montoya et al., Nature 2004) • Biofuel producer
Fermentative pathways for the production of butanol and other organic acids (Stal and Moelzaar, FEMS Microbiol Rev 1997)
Proposed synthetic Biology experiments
Validated experimental findings
• “Chassis” for synthetic biology Useful for performing gene manipulations and building synthetic pathways (Ng et al. Arcg Microbiol 2000)
• Source of valuable bioproducts Existence of pathways leading to the production of ethanol and alkane (Schirmer et al., J Bacteriol 1997)
• Cyanothece 51142 genome 5.46 Mbp and 5304 ORFs (Welsh et al., PNAS 2008)
• Synechocystis 6803 genome 3.57 Mbp and 3168 ORFs (Kaneko et al., DNA Res 1996)
Elemental and charge balancing
• Impact of elemental and charge balancing
Accurate prediction • Growth • Product yield
Balanced genome-scale model
• Test: Prediction of biomass yield (M/M CO2) with and without elemental and charge balancing under high light intensity and phototrophic condition
368 and 454 rxns were rebalanced for iRS706 and iRS764, respectively. (Fu, Journal of Chemical Technology and Biotechnology 2009; Knoop et al., Plant Physiology 2010; Montagud et al., Bmc Systems Biology 2010)
Cyanobacterium Model With balancing
Without balancing
Exp. observation
Synechocystis 6803
iRS706 0.098 0.0007 0.120
Fu’s model - 0.0024 0.120
Knoop’s model - 0.0012 0.120
Montagud’s model - 0.0037 0.120
Cyanothece 51142 iRS764 0.316 0.0002 0.540
Cyanothece iRS764 model (light/dark)
Cyanothece central metabolism (Stockel et al., PNAS 2008)
Coexpression network of strongly cycling genes (Stockel et al., PNAS 2008)
+
Cyanothece 51142 (light) model Cyanothece 51142 (dark) model
+ Distinct biomass equations for
lightand dark phases
• Photosynthesis • Calvin cycle • Reductive PPP • Glycogen synthesis
• Glycogen degradation • Glycolysis • Nitrogen fixation • Oxidative PPP • TCA cycle • AA biosynthesis
Upregulated pathways
Upregulated pathways
Comparison between iRS764 and iRS706
Cyanothece 51152 iRS764 vs
Synechocystis 6803 iRS706
Genes Reactions Metabolites
• 282 unique reactions in Cyanothece 51142 iRS764 compared to Synechocystis 6803 iRS706 q Primary metabolism (i.e., central metabolism, nitrogen metabolism, amino acid biostnthesis, etc)
q Secondary metabolism (i.e., biosynthesis of terpenoid, glucosinolate, porphyrin, etc.)
• 216 unique reactions in Synechocystis 6803 iRS706 with no counterpart in Cyanothece 51142 iRS764 q 202 from a wide range of primary metabolism pathways such as central metabolism, benjoate degradtion, starch, sucrose and lipid metabolism, amino acid and fatty acid biosynthesis
q 14 from secondary metabolism such as brassinosteroid metabolism and fluorene degradation
Comparison between iRS764 and iRS706 • Fermentative butanol pathway
• Citramalate pathway
Preliminary models testing
MEP pathway in Synechocystis 6803
• Isoprene is a precursor chemical and biofuel candidate • Upon inclusion of lspS to model the maximum isoprene theoretical yield is found to be (1.2 χ 10-5 mM/gDW-24hr) • This value is in the same order of magnitude of the experimentally achieved# (3.0 χ 10-5 mM/gDW-24hr)
• Isoprene synthesis in Synechocystis 6803
• H2 production in Cyanothece 51142 and Synechocystis 6803
Cyanobacterium Reported production Max Theoretical rate (mM/gDW)* Yield (mM/gDW)
Cyanothece 51142 0.193 0.082 Synechocystis 6803 6.99×10-3 5.32 × 10-4
* Bandyopadhay et al., Nature Comm, 2011
# Lindberg et al., Met Eng 2011
Outline q Reconstruct: Organism-specific genome-scale models
Automated assembly and curation of genome-scale models of metabolism (Suthers et al., PLoS Comp Biol, 2009; M. AbuOun et al., J Biol Chem, 2009; Satish Kumar et al., BMS Sys. Biol., 2011; Saha et al., PLoS ONE, 2011)
q Redesign: Computational strain design Pathway prospecting and identification engineering strategies leading to targeted overproductions (Ranganathan et al., PLoS Comp Biol, 2010; Ranganathan and Maranas, Biotech. J., 2010; Xu et al., Metab. Engr., 2011)
Compile and standardize genome-scale models and databases with consistent naming and balanced reactions (Kumar et al. submitted)
q Standardize: MetRxn: standardized knowledgebase of metabolites and reactions
MetRxn primary metabolite and reaction data sources
“Raw” dataset in MetRxn
KEGG
MetaCyc BRENDA RHEA
ChEBI
Reactome
HMDB
44 Metabolic
models
322,936 Metabolite 121,236 Reactions
Metabolites : 73659 Reactions: 50416
Metabolites : 16145 Reactions: 8123
Metabolites : 10477 Reactions: 8711
Reactions : 2907
Metabolites: 63344
Reactions : 1686
Metabolites: 7900
BKM Metabolites : 22367
Reactions: 18172
Incongruence across databases and models
Example: 2-Oxoglutarate + L-Alanine <=> Pyruvate + L-Glutamate
1. Naming inconsistencies
KEGG C00026 + C00041 <=> C00022 + C00025 BRENDA alpha-ketoglutarate + L-alanine <=> L-glutamate + pyruvate E. coli (iAf1260) [c] : akg + ala-L --> glu-L + pyr Acinetobacter baylyi 1 GLT + 1 PYRUVATE <-> 1 2-KETOGLUTARATE + 1 L-ALPHA-ALANINE Leishmania major [m] : akg + ala-L -> glu-L + pyr Mannheimia succiniciproducens PYR + GLU --> AKG + ALA
AMP: adenosine 5-monophosphate or ampicillin? Example:
Balanced: (R)-Lactate + NAD+ <=> Pyruvate + NADH + H+ KEGG [c] : lac-D + nad --> h + nadh + pyr iAF1260 E.coli (Feist et al. Mol Sys Biol, 2007)
Unbalanced: 1 D-LACTATE + 1 NAD <==> 1 NADH + 1 PYRUVATE Acinetobacter baylyi (Durot et al. BMC Systems Biology, 2008 )
2. Elemental and charge imbalances
Non-specific structural information Multiple structures associated with the same metabolite name
3. Incompleteness, degeneracy, and errors in information
R
# of
met
abol
ites
# of structures
Workflow Metabolite & reaction
information extraction
Download / identify metabolite bond
connectivity information
Metabolite identity analysis
Reaction identity analysis
Disambiguation of metabolites using
structure & phonetic comparisons
Canonical / isomeric SMILES at pH 7.2
Elemental & charge balancing
MetRxn (as of October 2011)
322,936 Metabolite 121,236 Reactions
entities
71,089 Metabolite, 63,243 Reactions
entities
31,177 Metabolite, 7,180 Reactions
entities
Initial repository
Non-resolved (no atomistic detail;
sometimes no chemical formula)
Resolved repository
42,540 Metabolite, 35,473 Reactions
entities
28,549 Metabolite, 27,770 Reactions
entities
Full atomistic detail
Partial atomistic detail generic side chains unspecified repeats
“known unknowns”
lipids generics (e.g., “electron donor”)
macromolecules (3,490 structural proteins and enzymes)
MetRxn content (Oct 2011)
71,089 Metabolite, 63,243 Reactions
entities
31,177 Metabolite, 7,180 Reactions
entities
Non-resolved (no atomistic
detail)
Resolved repository
42,540 Metabolite, 35,473 Reactions
entities
28,549 Metabolite, 27,770 Reactions
entities
Full atomistic detail
Partial atomistic detail
# of metabolites
mod
els
data
base
s
MetRxn Home (http://metrxn.che.psu.edu)
1. Model selection, viewing, exporting
1. GSM Re-balancing
Charge unbalanced: D-LACTATE + NAD <==> NADH + PYRUVATE Balanced by MetRnx: D-LACTATE + NAD <==> NADH + PYRUVATE + PROTON
q iAF1260 E.coli (Feist et al. Mol Sys Biol, 2007)
q Acinetobacter baylyi (Durot et al. BMC Systems Biology, 2008 )
1,039 metab, 2,077 rxn
703 metab, 853 rxn
arbtn-fe3 Aerobactin C22H33FeN4O13 C05554 iAF1260
189 rxn balanced by MetRxn
Elemental and charge unbalanced: GTP + 2 H2O <-> FORMATE + DIHYDRONEOPTERIN-P3 Balanced by MetRnx: GTP + 1 H2O <-> FORMATE + DIHYDRONEOPTERIN-P3 + PROTON
MetRxn fixed link to incorrect structure
Incorrect:
Corrected:
C05554 Aerobactin C22H36N4O13 KEGG ferric-aerobactin C22H33FeN4O13 PubChem
2. Model comparisons
2. Model comparisons (clostridia)
C. acetobutylicum C. thermocellum
solventogenesis, CoB12 pathway
cellulosome rxns
charged/uncharged tRNA
58
173
224 66
37
1181 79
29
79
57 90
61
642 266
B. subtilis
amino acids biosynthesis pathways carbohydrate metabolism nucleoside metabolism
Overlaps occur in
C. thermocellum (Roberts, et al. 2010)
C. acetobutylicum (Lee, et al. 2008)
reactions
295
140
147
137
210
290
Differences occur in
Reactions
Metabolites
Outline q Reconstruct: Organism-specific genome-scale models
Automated assembly and curation of genome-scale models of metabolism (Suthers et al., PLoS Comp Biol, 2009; M. AbuOun et al., J Biol Chem, 2009; Satish Kumar et al., BMS Sys. Biol., 2011; Saha et al., PLoS ONE, 2011)
q Redesign: Computational strain design Pathway prospecting and identification engineering strategies leading to targeted overproductions (Ranganathan et al., PLoS Comp Biol, 2010; Ranganathan and Maranas, Biotech. J., 2010; Xu et al., Metab. Engr., 2011)
Compile and standardize genome-scale models and databases with consistent naming and balanced reactions (Kumar et al. submitted)
q Standardize: MetRxn: standardized knowledgebase of metabolites and reactions
Computational strain design: OptForce (Ranganathan et al., PLoS Comput. Biol., 2010)
Limitations: 1. Generate one “redesign” at a time 2. Use of surrogate objective functions (e.g., max
biomass or min MOMA) 3. No direct use of MFA or other flux data
Existing Strategies:
Wild-type flux ranges (with MFA data)
Wild-type flux ranges (without MFA data)
Flux ranges required for overproduction
Min / Max vj s.t. MFA data Stoichiometry Uptake
Min / Max vj s.t. Stoichiometry Uptake
Min / Max vj s.t. Stoichiometry Uptake Vproduct > target
MFA data Vproduct > target
OptKnock (Burgard
et al. 2003)
OptStrain (Pharkya
et al. 2004)
FSEOF (Choi et al.
2010)
OptGene (Patil
et al. 2005)
OptORF (Kim and Reed
2010)
RobustKnock (Tepper and Schlomi
2010)
Identify all individual reactions and combinations thereof whose total flux value MUST increase, decrease or be knocked out to meet a newly imposed production target
Key Idea:
Flux range classifications (MUST sets)
Wild-type phenotype
must increase must decrease must knockout
can increase
Sum of two fluxes
v1 or v2 must increase
v1 or v2 must decrease
v1, v2, or v3 must increase
Sum of three fluxes
: :
can decrease
Desired phenotype
Flux range classifications (MUST sets)
Singles Pairs Triples Higher order
v1 v2
MUSTU
MUSTL
v3 v4
MUSTUU
MUSTUL
MUSTLL
v5 v6 v7
MUSTUUU
MUSTULL
MUSTUUL
MUSTLLL
. . . .
Define
Logic Relations
(V1 AND V2 ) (V3 OR V4 ) AND AND (V5 OR V6 OR V7 )
Encode changes that must happen in the metabolic network MUST sets:
è Identify set of required direct genetic interventions
Identify the minimal set of genetic interventions that guarantee the imposed yield by satisfying all the MUST relations
Max-min problem:
FORCE set
Maximize (over MUST sets)
s.t. Minimize (over fluxes) s.t. Stoichiometry
Environmental conditions
MUST set constraints
vproduct
• Prioritization of genetic interventions
• Mostly additive contribution of interventions
• Alternate minimal FORCE sets
vproduct
vproduct
Number of interventions (k)
4 6 8
Target yield
2
∑ # of direct manipulations < k
Alternate interventions
New reactions added: 4CL: 4-coumaric acid lyase CHS: chalcone synthase CHI: chalcone isomerase
Flavanone synthesis in E. coli (Xu et al., Metab. Engr., 2011)
Fowler Z.L. and M.A.G. Koffas, Applied Microbiology and Biotechnology, 2009, 83 (5)
(Collaboration with Prof. Mattheos Koffas group)
(van de Walle and Shiloach J, 1998; Noronha et. al, 2000)
q Metabolic flux data for wild-type strain BL21*
è Use OptForce to identify minimal interventions (FORCE set) for malonyl-CoA availability
Results for , and set of reactions Flavanone Biosynthesis
Phosphoenolpyruvate Carboxylase (PPC)
PDH
ENO
GAPD
ACCOAC
FBA
PGK
HSK
PGI
SERD
PPC
PFL
ACONTa
ACONTb
ICDHyr
CD
AKGDH
MDH
PYK
FUM
ASPTA
ACLS
TKT1
TALA
RPE
CHORS
DHAD1
MTHFD
Pyruvate:Formate Lyase (PFL)
Aconitase (ACONT)
Glyceraldehyde-3-phosphate dehydrogenase (GAPD)
Enolase (ENO)
Pyruvate Dehydrogenase (PDH)
HSK
HSK
HSK
THRS
THRS
THRS
3HAD181
3OAS181
3OAR181
3HAD181
3OAS181
3OAR181
SUCOAS
RPI
PPCSCT
PPCK
Succinyl-CoA Synthase (SUCOAS)
Propanoyl-CoA:Succinyl-CoA transferase (PPCSCT)
FORCE set for flavanone synthesis in E. coli
fumB
Δ sucC
acnA
GPR Associations
accABD mdh
gapA
pgk pdh
Δ scpC
and and and and
or
or or or
mdh
fum
Succinyl-CoA
ppcsct/sucoas
Experimental Results
Glyceraldehyde-3-phosphate dehydrogenase (GAPD)
Phosphoglycerate Kinase (PGK)
Pyruvate Dehydrogenase (PDH)
Acetyl-CoA Carboxylase (ACCOAC)
Malate Dehydrogenase (MDH)
Fumarase (FUM)
Aconitase (ACONT)
Δ Propanoyl-CoA:Succinyl-CoA Carboxylase (PPCSCT)
Δ Succinyl-CoA Synthetase (SUCOAS)
fumC or
or Δ sucD
Experimental Results (Koffas lab, RPI) (Xu et al., Metab. Engr., 2011)
BL21*
Naringenin yield
(mg / gr glucose)
accABD gapA pgk ΔfumB ΔfumC ΔsucC
57
112 113
157 153
199
55
Δmdh
BL21* ↑ gapA
BL21* ↑ pgk
↑pgk
• Up-regulation of pgk and/or gapA increases yield by about 98%
• Knock-outs of mdh or acnA decreases yield
• Knock-outs of fumB or fumC and sucC further increases yield by about 76%
• Overexpression of pdh boosts yield by 8% resulting in a final yield of 504 mg/L
fumB
Δ sucC
acnA
accABD mdh
gapA
pgk pdh
Δ scpC
and and and and
or
or or or fumC
or
or Δ sucD
53
ΔacnA
52
BL21* Δ mdh
52
155 150
↑gapA ↑pgk
BL21* Δ acnA
↑gapA ↑pgk
BL21* Δ fumB
↑gapA ↑pgk
BL21* Δ fumC
↑gapA ↑pgk
BL21* Δ sucC
↑gapA
196
203
198
↑pgk ↑gapA
Δ fumB Δ fumC
BL21* Δ sucC ↑pdh
↑pgk ↑gapA
Δ fumC
219 213
Δ fumB
pdh
Summary & Acknowledgements
Funding Source: DOE DE-FG02-05ER25684
Patrick Suthers (GSM)
Rajib Saha (Maize, cyano)
Akhil Kumar (MetRxn)
Sridhar Ranganathan (OptForce)
Vinay Satish Kumar (GapFill,
GrowMatch)