Expanding the Computable Reactome in Pseudomonas Putida Reveals Metabolic Cycles Providing Robustness

Expanding the computable reactome in Pseudomonas putida reveals metabolic cycles providing robustness.

Juan Nogales,a,b#* Steinn Gudmundsson,c Estrella Duque,d Juan Luis Ramos,d and Bernhard O. Palssona

Running Head: Metabolic Robustness Cycles in P. putida.

aDepartment of Bioengineering, University of California, San Diego, La Jolla, CA, USA bDepartment of Environmental Biology, Centro de Investigaciones Biológicas-CSIC, Madrid, Spain cCenter for Systems Biology, University of Iceland, Iceland dBiotechnology Research Area, Abengoa Research, Sevilla, Spain

# Address correspondence to: Juan Nogales, [email protected]

* Present address: Systems and Synthetic Biology Program, Centro Nacional de Biotecnología (CNB-CSIC), Madrid, Spain

1. Reconstruction process

Figure. S1. Overall workflow followed for reconstructing the metabolic network of P. putida KT2440. The reconstruction was performed through an iterative tri-dimensional expansion based on genome annotation (red lane), biochemical (blue line) and phenotypic (green lane) information. Along this approach up to 336 genes included in previous reconstruction were discarded due insufficient information available while 246 ORFs from the genome of P. putida were re-annotated (Table S1).

Figure. S2. Venn diagrams showing the gene comparison between the current P. putida reconstructions. Our previous reconstruction iJN746 (left) and iJN1411 (right) are shown as reference.

2. Reconstruction Validation.

Figure S3. Reconstruction of the sarcosine catabolom. Based on legacy data and manual search in the genome of P. putida KT2440 it was possible to reconstruct the catabolic pathway for choline sulfate, choline, glycine betaine, dimethylglycine, creatine, carnitine, carnosine and creatinine which converge in the key metabolite sarcosine. The reconstruction was validated by growth experiments (Table S2). Furthermore, knockouts available at PRCC were used to validate at molecular level the reconstruction. Hence, while the wild type strain grew on glycine betaine and creatine, P. putida strains knockouts in the genes PP_0310 and PP_0116 were unable to growth in glycine betaine and a knockout strain in the gene PP_3637 was unable to growth using creatine as only carbon source. The abbreviations for metabolites and reactions can be found in Table S1.

2 Figure S4. Reconstruction of the 2,5 dioxopentanoate catabolom. Based on legacy data and 3 manual search in the genome of P. putida KT2440 it was possible to reconstruct the catabolic 4 pathway for D-galacturate, D-galactarate, D-glucuronate, D-glucarate, trans-4-Hydroxy-L- 5 proline and cis-4-Hydroxy-L-proline which converge in the key metabolite 2,5 6 dioxopentanoate. The reconstruction was validated by growth experiments (Table S2). 7 Furthermore, knockouts available at PRCC were used to validate at molecular level the 8 reconstruction. Hence, P. putida strains knockouts in the genes PP_1168 and PP_1169 were 9 unable to growth D-Galacturonate and D-glucoronate, a knockout strain in the gene PP_4757 10 failed in to growth in Glucarate. Finally, a knockout strain in the gen PP_1259 was unable to 11 growth using cis-4-Hydroxy-L-proline as only carbon source. The abbreviations for metabolites 12 and reactions can be found in Table S1. 13 14 15 16 17 18 19 20 21 22 23 24

1 2 3 Figure S5. Reconstruction of polyamine metabolism. Based on legacy data and manual search 4 in the genome of P. putida KT2440 it was possible to reconstruct the catabolic pathway for 5 polyamines. The reconstruction was validated by growth experiments (Table S2). Furthermore, 6 knockouts available at PRCC were used to validate at molecular level the reconstruction. 7 Hence, P. putida strain knockout in the key gene PP_5278 was unable to growth using 8 Spermidine or 1,3-Diaminopropane as carbon sources. The abbreviations for metabolites and 9 reactions can be found in Table S1. 10 11 12 13 14 15 16 17 18 19 20 21

1 2 3 Figure S6. Reconstruction of isovaleryl-Coa metabolism. Based on legacy data and manual 4 search in the genome of P. putida KT2440 it was possible to reconstruct the catabolic pathway 5 for isovaleryl-CoA. The reconstruction was validated by growth experiments (Table S2). 6 Furthermore, knockouts available at PRCC were used to validate at molecular level the 7 reconstruction. Hence, P. putida strain knockouts in the genes PP_4063, PP_4065 and PP_4067 8 were unable to growth using 3-Methylbutanoic acid as carbon sources. The abbreviations for 9 metabolites and reactions can be found in Table S1. 10 11 12 13

MCA factor map

0.5

0.0

2 2 (7.248%)

Dim

0.5

- 1.0 - -0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0 Dim 1 (9.758%)

1.0 0.9

0.8 0.7 0.6

0.5 0.4 0.3

0.2 0.1

0.0

C6:0 C8:0

C6:0 C4:0

C4:0

C8:0 C9:0 C7:0 C5:0

C6:0

C9:0 C5:0 C9:0

- -

- - -

C16:0

C12:0 C14:1 C14:0

C10:0

C10:0 C14:2

C12:1 C10:1

C14:1d5 C12:1d6

Thio Thio

Phenyl

Phenyl Phenyl Phenyl 1 Phenyl

2 Figure. S7. Relative production of PHA monomers by iJN1411. The 223 different carbon 3 sources supporting growth of iJN1411 (vertical axis) were used to maximize the production of 4 the 24 different PHA monomers included in iJN1411 (horizontal axis). The relative production 5 was computed as the fraction of carbon being transformed in the target monomer. The 6 analysis shows that P. putida is well suited for producing aliphatic PHA monomers, irrespective 7 of the carbon source, while the aromatic, thiol, and polyunsaturated PHA monomers are highly 8 carbon source dependent. R-Hydroxyacid PHA monomers: CX:Y where the X indicates the 9 chain length and Y the number of unsaturations. The Phenyl and Thio prefixes indicate 10 monomers including phenyl and thiol groups in their structures, respectively.

11 C 12 13 14

1 3. Biomass Comparison

Metabolite present in BOF Metabolite absent in BOF 2 Precursor and/or no specie-specific

iJN1411 Belda's PpuQY1140 iJN1411 Belda's PpuQY1140 Protein ala-L Soluble Pool k arg-L nh4 asn-L mg2 asp-L ca2 cys-L fe2 gln-L fe3 glu-L cu2 gly mn2 his-L mobd ile-L cobalt2 leu-L zn2 lys-L cl met-L ni2 phe-L na1 pro-L so4 ser-L pi** thr-L ptrc trp-L accoa tyr-L coa val-L succoa DNA datp malcoa dctp nad dgtp nadh dttp nadp RNA ctp nadph gtp fad utp thf atp** mlthf Peptidoglycan murein4p4p 5mthf murein3p3p thmpp murein4px4p q8h2 murein3px4p pydx5p murein4px4px4p hemeO LPS lpspput pheme LIPIDS pe140 sheme pe160 pyovd-kt pe161 gthrd pe181 adocbl pe180 udcpdp pg140 10fthf pg160 chor pg161 amet pg181 ribflv pg180 2fe2s clpn140 4fe4s clpn160 mocogdp clpn161 btamp clpn181 lipopb clpn180 pqqh2 cpe160 bmocogdp 3 cpg160 thmnp

4 Figure S8. Qualitative comparison of Biomass components present in iJN1411, iEB1050 and 5 PpuQY1140

1 4. Nutrients source Validation.

2 We validated the growth performance accuracy of iJN1411 taken advantage of the 3 legacy data. For those nutrients with contradictory reports or never before tested in P. putida, 4 we proceeded to its experimental validation (72 and 46 carbon and nitrogen sources, 5 respectively). The carbon and nitrogen sources supporting growth in iJN1411 together with the 6 validation sources is provided in Table S2. Excluding those metabolites for which the validation 7 was not possible due technical issues, the overall accuracy of the updated model was very high 8 and it predicted correctly the 77.6 % and 83.2% (two-sided p-values of Fisher‘s exact test were 9 less than 10-3) of the phenotypes observed for carbon and nitrogen sources, respectively 10 (Figure 2). Just a few false negatives were found and being confined to metabolites which 11 catabolic pathways are unknown in Pseudomonas e.g., 3,4-dihydroxyphenylacetate, methyl 12 pyruvate, and Tween 20 as carbon source and L-cysteine as nitrogen source. However, 13 significant higher number of false positives (nutrients supporting growth in silico but not in 14 vivo were detected. Many of they supported growth as nitrogen source but not as carbon 15 source e.g., nitrogenous bases such as xanthine and hypoxanthine and polyamines like 16 agmatine, suggesting regulatory and/or adaptive limitations. In fact, KT2440 grew on xanthine 17 when this metabolite was used as only carbon and nitrogen source (data non-shown), 18 suggesting that the metabolism of xanthine is repressed in presence of ammonium. In addition 19 when P. putida was cultured in xanthine it growth after 96 h. On the other hands, it is become 20 evident the presence of silent metabolic pathways without apparent function encoded in 21 bacterial genomes as result of insufficient adaptive and/or regulatory evolution (1-3). Since 22 GENREs ignore these limitations, often the in silico metabolic capabilities predicted overpass 23 those experimental observed phenotypes. This fact explains, to some extent, the presence of 24 false positives in growth predictions. Interestingly, adaptive laboratory experiments (ALE) have 25 shown that in silico predicted phenotypes can be achieved in vivo through short-term 26 evolution (1, 3). In order to investigate if some of the false positives found here respond to lack 27 of evolutionary adaptation we growth to KT2440 on ethylene glycol since it has been described 28 that several P. putida are able to use this compound as only carbon and energy source but no 29 KT2440 (4). We observe that KT2440 was unable to grow on EG after 48 h of incubation as 30 previously has been reported, however it grew efficiently after 100 hours. This result shows 31 that the EG catabolic pathway is indeed a latent pathway in KT2440, and it suggests that the 32 rest of false positives detected could be suitable targets for the catabolic expansion of this 33 biotechnological interesting strain. 34

1 5. Definition of a CORE biomass objective function.

2 We proceed to formulate the core biomass objective function based on the minimal 3 biosynthetic precursors supporting growth according the knockouts present at PRCC (5). For 4 instance, we included in the core biomass function the metabolite phosphor-6-heptosyl-1,3- 5 ethanolaminephosphate-2-phospho-4-heptosyl-1,5-kdo2_lipidA (phethapphlipa) instead of the 6 complete lipopolysaccharide (lpspput) since knockouts strains upstream of this metabolite in 7 the biosynthetic pathway were not present at PRCC. Thereby we assume that phethapphlipa is 8 the minimal lipopolysaccharide supporting growth in P. putida.

10 Figure S9. Reconstruction of the lipopolysaccharide biosynthetic pathway in P. putida 11 KT2440. Precursor and intermediate metabolites in the biosynthesis of the lipopolysaccharide 12 of KT2440 are shown as green and grey balls, respectively. The complete lipopolysaccharide 13 included in the wild type biomass is shown in blue while the metabolite phethapphlipa present 14 in the core biomass is shown in red. Dashed lines indicate lumped reactions in the biosynthetic 15 pathway. P. putida gene knockouts strains present at PRCC are indicated in green lines. Note 16 that no knockouts strains upstream of phethapphlipa are present at PRCC. The abbreviations 17 for metabolites and reactions can be found in Table S1. 18

1 6. Carbon flux predictions.

3 Figure S10. Carbon flux predictions from previous models of P. putida KT2440.

4 7. Gene-essentiality analysis validation and contextualization.

5 The essentiality analysis and validation experiments provided a unique opportunity 6 to deciphering new insights in the metabolism of KT2440 while increasing the accuracy of 7 iJN1411. For instance, several cob genes involved in the biosynthesis of vitamin B12 were 8 predicted as essential in the first version of iJN1411 contrary to the phenotype displayed by 9 these gene knockouts in vivo (6). Even more, based on the lack of B12 requirement of these 10 mutants together the presence of several cbi genes (responsible of synthetizing B12 under 11 anaerobic conditions) it was suggested a putative alternative B12 biosynthetic pathway thus, 12 uncertainties remain on the biosynthesis of this vitamin in KT2440 (6). In order to resolve this 13 gap of knowledge, we perform a comprehensive search looking for B12-dependent enzymes 14 encoded in the genome of KT2440 finding just three, two of which are included in iJN1411 e.g., 15 the methionine synthase MetH (PP_2375) and the ethanolamine ammonia Lyase EutCB 16 (PP_0542 and PP_0543). Since a B12-independent methionine synthase (MetE) exist in KT2440 17 becoming in non-essential to MetH in any case, we decided validate the role of the cob genes 18 in B12 biosynthesis by growing the cob knockouts in ethanolamine as only nitrogen source. If 19 such mutant strains were unable to growth under these conditions, the presence of an 20 alternative pathway for synthesizing B12 would be discarded, while the role of cob genes in 21 this process would unequivocally demonstrated. We found that this was the case (Fig. S9), thus 22 we proceeded to remove the B12 from the wild-type BOF solving the initial discrepancy 23 between in vivo and in silico essentiality data.

Nitrogen Source NH4 Ethanolamine KT2440 + + PP_1680 (cobC ) + - PP_1678 (cobU ) + - PP_1673 (cobB) + - 1 PP_3410 (cobM) + +

2 Figure S11. Role of cob genes in B12 biosynthesis. P. putida KT2440 and isogenic knockouts 3 strains lacking cob genes were grown in glucose minimal medium using ammonia or 4 ethanolamine as nitrogen source. Growth or not growth are indicated as (+) or (-), respectively. 5

6 8. Definition of In silico Luria Broth (LB) medium.

7 In order to define an accurate in silico LB medium (iLB) we used the basic composition based 8 on previous reports (7) and conditional essential gene analysis from KT2440 (6).

9 The default Exchange reactions were constrained as follow.

10 Carbohydrates 11 About ~ 29% weight; glucose was assumed as a representative carbohydrate source. 12 13 model=changeRxnBounds(model,'EX_glc(e)',-5,'l'); 14 15 Amino acids 16 model=changeRxnBounds(model,'EX_ala_L(e)',-5,'l'); 17 model=changeRxnBounds(model,'EX_asp_L(e)',-5,'l'); 18 model=changeRxnBounds(model,'EX_glu_L(e)',-5,'l'); 19 model=changeRxnBounds(model,'EX_his_L(e)',-5,'l'); 20 model=changeRxnBounds(model,'EX_leu_L(e)',-5,'l'); 21 model=changeRxnBounds(model,'EX_met_L(e)',-5,'l'); 22 model=changeRxnBounds(model,'EX_pro_L(e)',-5,'l'); 23 model=changeRxnBounds(model,'EX_thr_L(e)',-5,'l'); 24 model=changeRxnBounds(model,'EX_tyr_L(e)',-5,'l'); 25 model=changeRxnBounds(model,'EX_arg_L(e)',-5,'l'); 26 model=changeRxnBounds(model,'EX_cys_L(e)',-5,'l'); 27 model=changeRxnBounds(model,'EX_gly(e)',-5,'l'); 28 model=changeRxnBounds(model,'EX_ile_L(e)',-5,'l'); 29 model=changeRxnBounds(model,'EX_lys_L(e)',-5,'l'); 30 model=changeRxnBounds(model,'EX_phe_L(e)',-5,'l'); 31 model=changeRxnBounds(model,'EX_ser_L(e)',-5,'l'); 32 model=changeRxnBounds(model,'EX_trp_L(e)',-5,'l'); 33 model=changeRxnBounds(model,'EX_val_L(e)',-5,'l'); 34 35 Vitamins 36 37 Cobalamine (B12). Conditional essential genes found in Molina-Henares et al, 2010: 38 PMID:20158506. 39 40 model=changeRxnBounds(model,'EX_cbl1(e)',-0.1,'l'); 41

1 Riboflavin (Vitamin B2). No auxotrophs were found in Molina-Henares et al, 2010: 2 PMID:20158506. No gene encoding riboflavin transport. Assumed that P. putida is 3 unable to transport this vitamin. 4 5 Biotin (B8). Conditional essential genes found in Molina-Henares et al, 2010: 6 PMID:20158506. 7 8 model=changeRxnBounds(model,'EX_btn(e)',-0.1,'l'); 9 10 Niacin (PP). Conditional essential genes found in Molina-Henares et al, 2010: 11 PMID:20158506. NadA knockout required nicotinic acid for growth. 12 13 model=changeRxnBounds(model,'EX_nac(e)',-0.1,'l'); 14 15 Thiamine (B1). No auxotrophs were found in Molina-Henares et al, 2010: 16 PMID:20158506. No gene encoding thiamine transport. Assumed that P. putida is 17 unable to transport this vitamin. 18 19 Pyridoxine (B6). No auxotrophs were found in Molina-Henares et al, 2010: 20 PMID:20158506. No gene encoding pyridoxine transport. Assumed that P. putida is 21 unable to transport this vitamin. 22 23 Pantotenate. Synthesis of pnto-R is not essential in LB (This study) 24 25 model=changeRxnBounds(model,'EX_pnto_R(e)',-0.1,'l'); 26 27 Folic acid (B9). No auxotrophs were found in Molina-Henares et al, 2010: 28 PMID:20158506. No gene encoding folic acid transport. Assumed that P. putida is 29 unable to transport this vitamin. 30 31 Chorismate. No auxotrophs found, P. putida is able to transport aromatic aminoacid as 32 well. Assumed present in LB medium based on Oh et al, 2007 (PMID:17573341). 33 34 model=changeRxnBounds(model,'EX_chor(e)',-0.1,'l'); 35 36 Diaminopimelic acid. An dapA (PP_1237) knockout required diaminopimelic acid for 37 grow in minimal media with glucose as carbon source, since dapA is not essential LB 38 medium, was assumed that 26dap_M is present in LB. 39 40 model=changeRxnBounds(model,'EX_26dap_M(e)',-0.1,'l'); 41 42 Siroheme. CysG, (PP_3999) the key enzyme in sheme biosynthesis is not essential in LB 43 medium. It was assumed that sheme is present in LB. 44 45 model=changeRxnBounds(model,'EX_sheme(e)',-0.1,'l'); 46 47 Other metabolites included based on Oh et al, 2007. 48 49 model=changeRxnBounds(model,'EX_na1(e)',-100,'l');

1 model=changeRxnBounds(model,'EX_cl(e)',-100,'l'); 2 model=changeRxnBounds(model,'EX_so4(e)',-100,'l'); 3 model=changeRxnBounds(model,'EX_k(e)',-100,'l'); 4 model=changeRxnBounds(model,'EX_pi(e)',-100,'l'); 5 model=changeRxnBounds(model,'EX_ca2(e)',-100,'l'); 6 model=changeRxnBounds(model,'EX_mg2(e)',-100,'l'); 7 model=changeRxnBounds(model,'EX_sel(e)',-100,'l'); 8 model=changeRxnBounds(model,'EX_zn2(e)',-100,'l'); 9 model=changeRxnBounds(model,'EX_aso4(e)',-100,'l'); 10 model=changeRxnBounds(model,'EX_cd2(e)',-100,'l'); 11 model=changeRxnBounds(model,'EX_hg2(e)',-100,'l'); 12 model=changeRxnBounds(model,'EX_pb2(e)',-100,'l'); 13 14 Other chemicals (included based on P.putida BOF) 15 model=changeRxnBounds(model,'EX_ni2(e)',-100,'l'); 16 model=changeRxnBounds(model,'EX_cu2(e)',-100,'l'); 17 model=changeRxnBounds(model,'EX_fe2(e)',-100,'l'); 18 model=changeRxnBounds(model,'EX_fe3(e)',-100,'l'); 19 model=changeRxnBounds(model,'EX_mn2(e)',-100,'l'); 20 model=changeRxnBounds(model,'EX_mobd(e)',-100,'l'); 21 model=changeRxnBounds(model,'EX_cobalt2(e)',-100,'l'); 22 23 24 Nucleotides/nucleosides 25 model=changeRxnBounds(model,'EX_ins(e)',-5,'l'); 26 model=changeRxnBounds(model,'EX_hxan(e)',-5,'l'); 27 model=changeRxnBounds(model,'EX_ura(e)',-5,'l'); 28 model=changeRxnBounds(model,'EX_uri(e)',-5,'l'); 29 model=changeRxnBounds(model,'EX_adn(e)',-5,'l'); 30 31 Additional constrains. 32 model=changeRxnBounds(model,'EX_o2(e)',-21.5,'l'); 33 model=changeRxnBounds(model,'EX_nh4(e)',0,'l'); 34

35 9. Definition of in silico M9 glucose minimal medium.

36 In order to define an in silico M9 (iM9) glucose minimal medium, the lower and upper bounds 37 of several exchange reactions were constrained as follows.

Rxn name Rxn description Lower Bound Upper Bound (mmol.gDW-1.h-1) (mmol.gDW-1.h-1) EX_co2(e) CO2 exchange -100 1000 EX_h2o(e) H2O exchange -100 1000 EX_h(e) H exchange -100 1000 EX_o2(e) O2 exchange -30 1000 EX_ca2(e) Calcium exchange -10 1000 EX_cl(e) Chloride exchange -10 1000 EX_cobalt2(e) Co2 exchange -10 1000 EX_cu2(e) Cu2 exchange -10 1000 EX_fe2(e) Fe2 exchange -10 1000 EX_hco3(e) Bicarbonate exchange -10 1000 EX_k(e) K exchange -10 1000

EX_mg2(e) Mg exchange -10 1000 EX_mn2(e) Mn2 exchange -10 1000 EX_mobd(e) Molybdate exchange -10 1000 EX_na1(e) Sodium exchange -10 1000 EX_nh4(e) Ammonia exchange -10 1000 EX_ni2(e) Ni2 exchange -10 1000 EX_pi(e) Phosphate exchange -10 1000 EX_sel(e) Selenate exchange -10 1000 EX_so4(e) Sulfate exchange -10 1000 EX_tungs(e) tungstate exchange -10 1000 EX_zn2(e) Zinc exchange -10 1000 EX_glc(e) D Glucose exchange -6,3 1000 1

2 10. Multi-strain P. putida modeling.

3 For the modeling of the sequenced P. putida strains we follow a similar approach than 4 used previously by Orth and colleagues (8). The list of orthologous genes between the 5 different strains was downloaded from Pseudomonas database 6 (http://www.pseudomonas.com/) taking the genome of P. putida KT2440 as reference (Table 7 S6). After that, we removed from iJN1411 those genes missing and their reactions associated 8 in each P. putida strain being reconstructed by using the function “deleteModelGenes” 9 implemented in the Cobra Toolbox. Finally we tested the ability of the new models to produce 10 biomass in iLB using the core biomass objective function as BOF. We noted that any of the 11 models was able to produce biomass under these conditions. In order to construct functional 12 models, we cross the list of deleted genes in each model with the list of essential genes of 13 iJN1411 in iLB. Those genes predicted to be essential in iJN1411 under these conditions and 14 deleted in the rest of the models were added back to the corresponding model assuming that 15 alternative(s) gene(s) should be present in the target P. putida strain. The genes added back to 16 each model are shown in figure S12.

17 In addition and because the gene PP_2519 is annoted as pseudogene in the genome of 18 P. putida KT2440 but it is essential for gallic acid metabolism, we included the PP_2519 and it 19 reaction associated in those models corresponding to P. putida strains which have the rest of 20 the gallic acid degradation pathway encoded in their genomes.

21 Finally we noted that the strain HB3267 is unable to synthetize chorismic acid, thus it 22 was predicted to be auxotrophic for aromatic amino acids. For analyzing the metabolic 23 potential of this specific strain we allow the uptake of chorismic acid.

Gene PP_0037 PP_3959 PP_4976 PP_5013 PP_5418 PP_4016 Rxn PItex CLt3_2pp AHC OPHHX ATPS4rpp ADSL1r BIRD-1 DOT-T1E F1

strain GB1 H83267 HB3267 NBCR ND6 S16

Pseudomonas W619 1 L48

2 Figure S12. Genes add back to the P. putida models in order to construct functional models.

3 Comparisons between experimentally reported flux values in the central metabolism of P. 4 putida growing on glucose (9) and predicted flux values obtained with iJN746, iEB1050 and 5 PpuQY1140.

7 11. Contextualization of gene expression of metabolic robustness cycles.

8 9 Figure S13. The absolute expression levels of the genes included in iJN1411 in glucose, 10 fructose, glycerol and succinate minimal media (red dots). The relative growth rates after 11 single gene knockouts under these conditions are shown as well (blue dots), enabling the

1 grouping of the genes as essential (EG) or non-essential (NEG). Genes encoding for metabolic 2 robustness cycles (MC) are shown as green dots while genes known to be unexpressed (UE) 3 under these conditions are shown as purple dots.

4 12. Supplementary references.

6 1. Conrad TM, Lewis NE, Palsson BØ. 2011. Microbial laboratory evolution in the era of 7 genome‐scale science, vol 7. 8 2. Cornelius SP, Lee JS, Motter AE. 2011. Dispensability of Escherichia coli’s latent 9 pathways. Proceedings of the National Academy of Sciences 108:3124-3129. 10 3. Fong SS, Nanchen A, Palsson BO, Sauer U. 2006. Latent Pathway Activation and 11 Increased Pathway Capacity Enable Escherichia coli Adaptation to Loss of Key 12 Metabolic Enzymes. Journal of Biological Chemistry 281:8024-8033. 13 4. Mückschel B, Simon O, Klebensberger J, Graf N, Rosche B, Altenbuchner J, Pfannstiel 14 J, Huber A, Hauer B. 2012. Ethylene Glycol Metabolism by Pseudomonas putida. 15 Applied and Environmental Microbiology 78:8531-8539. 16 5. Duque E, Molina-Henares A, Torre J, Molina-Henares M, Castillo T, Lam J, Ramos J. 17 2007. Towards a Genome-Wide Mutant Library of Pseudomonas putida Strain KT2440, 18 p 227-251. In Ramos J-L, Filloux A (ed), Pseudomonas. Springer Netherlands. 19 6. Molina-Henares MA, De La Torre J, García-Salamanca A, Molina-Henares AJ, Herrera 20 MC, Ramos JL, Duque E. 2010. Identification of conditionally essential genes for 21 growth of Pseudomonas putida KT2440 on minimal medium through the screening of a 22 genome-wide mutant library. Environmental Microbiology 12:1468-1485. 23 7. Oh Y-K, Palsson BO, Park SM, Schilling CH, Mahadevan R. 2007. Genome-scale 24 Reconstruction of Metabolic Network in Bacillus subtilis Based on High-throughput 25 Phenotyping and Gene Essentiality Data. Journal of Biological Chemistry 282:28791- 26 28799. 27 8. Orth JD, Conrad TM, Na J, Lerman JA, Nam H, Feist AM, Palsson BO. 2011. A 28 comprehensive genome-scale reconstruction of Escherichia coli metabolism--2011. 29 Mol Syst Biol 7:535. 30 9. Blank LM, Ionidis G, Ebert BE, Bühler B, Schmid A. 2008. Metabolic response of 31 Pseudomonas putida during redox biocatalysis in the presence of a second octanol 32 phase. FEBS Journal 275:5173-5190.