Using retrosynthesis in synthetic biology and metabolic engineering

Jean-LoupFaulon20161

Jean-LoupFaulon

jfaulon.com [email protected]

UsingRetroSynthesisinSynthe/cBiologyandMetabolic

Engineering

and Synthetic Biologyi n s t i t u t e o f S y s t e m s

iSSB

mSSB iGEM


CHASSIS

TARGET

EFFECTOR

CHASSIS

TARGET(signal)

Metabolicengineering Biosensorengineering

Findingmetabolicpathwaysbetweentwopoolsofcompounds


CHASSIS

TARGET

Findingpathwaysfromchassistotarget

Forward-synthesis

Findingpathwaysbetweentwopoolsofcompounds

SINK

SOURCE

Retro-synthesis

SOURCE

SINK

metabolicreac8on

Reversedmetabolicreac8on


Findingpathwaysbetweentwopoolsofcompounds:state-of-the-art

BNICE SimPheny/GEM-Path RetroPath


Findingpathwaysbetweentwopoolsofcompounds:theworkflow


UsingtheworkflowtofindretrosyntheMcpathwaystoproduce3-methylphenolinE.coli

SINK

SOURCE-SOURCE:trimethylphenol-SINK:E.colimetabolites-RULES:reversedMetaNetXreacEons


SINK

SOURCE

37productsgeneratedbyFIRE FILTER15productstobecomenewsubstratesfornextiteraEon

1SOURCEsubstrate



SINK

SOURCE

Iter2:15substrates1281productsgeneratedbyFIRE214newsubstratesIter3:214substrates25016productsgeneratedbyFIRE2789newsubstrates



SINK

SOURCE

FinalretrosyntheEcnetwork:•  1compoundinSOURCE•  163compoundsinSINK•  26312compoundstotal•  4soluEonpathways


L-tyrosineC3-methyltransferase(Streptomyceslavendulae)


L-Glutamate Oxaloacetate 2-Oxaloglutarate L-Aspartate

Transaminase(2.6.1)

BNICERULE SIMPHENYRULE

[H][#7]([H])-[#6](-[#6]-[#6]-[#6](-[#8])=O)-[#6](-[#8])=O.[#6,#1]-[#6](-[#6,#1])=O>>[#8]-[#6](=O)-[#6]-[#6]-[#6](=O)-[#6](-[#8])=O.[H][#7]([H])-[#6](-[#6,#1])-[#6,#1]

[#6,#1]-[#6](-[#6,#1])=O>>[H][#7]([H])-[#6](-[#6,#1])-[#6,#1]

Findingpathwaysbetweentwopoolsofcompounds:allthatmaOersisthesetofreacMonrules

•  86mulE-substratesrulesmanuallycuratedatthethirdEClevelclass(HenryCS,etal.BiotechnolBioeng.2010;106(3):462-73)

•  Rulescover30%ofallknowmetabolicreacEons(havinganECnumber)

•  50rulesmanuallycreated,allmono-substrate(Yim,H.etal.NatChemBiol,20117(7):p.445-5)

•  Rulescover20%ofallknowmetabolicreacEons(havinganECnumber)


-1(4−10)-1(15−17)+1(4−17)+1(15−10)

2.6.1

L-Glutamate Oxaloacetate 2-Oxaloglutarate L-Aspartate

RetroPathrulecomputaMonworkflow

[#6:1]-[#6:2]-[#6:4](-[#7:10])-[#6:5](-[#8:6])=[O:7]>>[#6:1]-[#6:2]-[#6:4](=O)-[#6:5](-[#8:6])=[O:7].[#7:10]-[#6](-[#6]-[#6](-[#8])=O)-[#6](-[#8])=O

Substrate+Oxaloacetate=Product+L-Aspartate

D=4

[#6:4]-[#7:10]>>[#6:4]=O.[#7:10]-[#6](-[#6]-[#6](-[#8])=O)-[#6](-[#8])=O

D=0Substrate+Oxaloacetate=Product+L-Aspartate

•  From12000(forD=∞)to6000(forD=0)rules•  Rulescover100%ofallknowmetabolicreacEons(havinganECnumber)


SOURCE:E.colimodeliJO1366(752compounds)

RULES:BNICE,SimPheny,RetroPathD=8onMetaNetXE.colireac8ons

Numberofproductsgenerated

inModel inBRENDA TotalSimPheny 376/752 2874 9782BNICE 546/752 3526 150228RetroPathD=8 633/752 6298 14178

SOURCE

SINK

HowwellcanthedifferentrulesystemsbuildanExtendedMetabolicSpaceforE.coli?

•  Carbonell,etal.ACSSynth.Biol2014+addeddistribuEonsforSimPhenyandBNICE

SINK:BRENDA

SimPheny


?

RetroPathapplicaMonforflavanones

SINK:Flavanones

RULES:FromD=∞toD=4onMetaNetXreac8ons

SOURCE

SINK

SOURCE:E.colimodeliJO1366


AllowingpromiscuousreacEonrules(D=4)

SpecificreacEonrulesonly(D=∞)

UsingdiametertocontrolreacMonspecificityanddesignnovelmetabolicpathways

11pathwaysenumeratedfollowingCarbonelletal.BMCSysBio2012

1pathway


SearchingforenzymesequencescatalyzingreacMons

Tensorproduct(MachinelearningSVMandGP)

Enzymesequence?

www.jfaulon.com/category/tools/

(µM)

PredicMngKMvalues

• Protein-protein:MarEn,Roe,FaulonBioinforma/cs2005• Drug-target:Oprea,Trospha,Faulon,Rintoul,Nat.Chem.Bio2007• Enzyme-reacEon:Faulonetal.Bioinforma/cs2008

Melloretal.ACSSynth.Bio.2016

SOURCE

SINK


EvaluaMngreacMonthermodynamicfeasibilityandproducttoxicity

Kanamycin

Screen-168compounds

Q2=0.71

StructureToxicityRelaMonship

Model

MeasureE.colibacterialgrowth(IC50)toxicity?

•  PlansonAGetal.Biotechnology&Bioengineering2011

RejectreacEonsthermodynamicallyunfavorable(Jankowskietal.BiophysJ,2008)andscorefavorablereacEonsthoughproducttoxicitySOURCE

SINK

www.jfaulon.com/category/tools/


DESIGN:Rankingpathways

•  Benchmark:rankingthepathwaysforaminoacidbiosynthesisinE.coli

Allna8vepathwaysfoundinthetop10pathwaysreturnedbytherankingfunc8on

11.1510.9710.4610.4410.0810.039.939.579.527.757.74

scoregenesscorereacMonsscorepathways

Scoreandselectbestpathway(s)

λpathλgene

λtox

S

•  CarbonellP,etal.BMCSystemsBiology2011•  CarbonellP,etal.ACSSynthBiol.2014•  CarbonellPetal.NucleicAcidRes.2014

GPandSVMtensorproduct

ToxicityTheoreEcalflux

xtms.issb.genopole.fr


Emzyme6.2.1.12(4CL)

L:Streptomycesmari8mus

M:Arabidopsisthaliana

H:Streptomycescoelocolor

Emzyme2.3.1.74(CHS)

L:Bacillussub8lis

H:Arabidopsisthaliana

Emzyme5.5.1.6(CHI)

L:Arabidopsisthaliana

H:Arabidopsisthaliana

EnzymeproducMvityvs.predictedscore

L-Phenylalanine

Cinnamoyl-CoA

Pinocembrin-chalcone

Pinocembrin

5.5.1.6

Trans-Cinnamate

CoA

ATP

Malonyl-CoA

2.3.1.74

6.2.1.12

4.3.1.25

BUILDANDTEST:ToppathwayconstrucMonandvalidaMonoftherankingfuncMon

[Pinocembrin-H]-

• FeherTetal.BiotechJ.2014

L/M/H:Low/Medium/HighpredictedacMvity


LEARNwithfluxanalysis

E.colimodel:FeistAetal.MolSystBiol2007+heterologouspathway2038reacEons(includingtransport)1043metabolites

@24haserinducEon:Growthrate:0.4847h-1Trans-cinnamate:49.0mg/LPinocembrin:<1.0mg/L

Fumarate

cinnamicacid-dihydrodiol

TCACycle

Trans-2,3-dihydroxycinnamate

Growth

2-Hydroxy-6-oxonona-trienedioate

1.14.12.19

1.3.1.87

1.13.11.16

3.7.1.-

L-Phenylalanine

Cinnamoyl-CoA


Pinocembrin

5.5.1.6

Trans-Cinnamate

CoA

ATP

2.3.1.74

6.2.1.12

4.3.1.25

Malonyl-ACP

FaOyAcids

1.1695

2.3.1.39

Malonyl-CoA

6.4.1.2

Acetyl-CoA

9.24e-70.4847

0.0026

0.2909

0.2935

1.1695

9.24e-7

Weneedtoboostmalonyl-CoA

FluxesinmmolgDW-1h-1

Jean-LoupFaulon201620FluxesinmmolgDW-1h-1

4.

Coxiella

1.

Rhizobium 3.

E. coli

2.

Geobacter

Fumarate


TCACycle


Growth


1.14.12.19

1.3.1.87

1.13.11.16

3.7.1.-

L-Phenylalanine

Cinnamoyl-CoA


Pinocembrin

5.5.1.6

Trans-Cinnamate

CoA

ATP

2.3.1.74

6.2.1.12

4.3.1.25

Malonyl-ACP

FaOyAcids

1.1695

2.3.1.39

Malonyl-CoA

6.4.1.2

Acetyl-CoA

9.24e-70.4847

0.0026

0.2909

0.2935

1.1695

9.24e-7

LEARNàBacktoDESIGN

• FeherTetal.BiotechJ.2014

• Melloretal.ACSSynth.Bio.2016


Fumarate


TCACycle


Growth


1.14.12.19

1.3.1.87

1.13.11.16

3.7.1.-

L-Phenylalanine

Cinnamoyl-CoA


Pinocembrin

5.5.1.6

Trans-Cinnamate

CoA

ATP

2.3.1.74

6.2.1.12

4.3.1.25

Malonyl-ACP

FaOyAcids

1.1695

FluxesinmmolgDW-1h-1

2.3.1.39

Malonyl-CoA

0.3412

6.4.1.2

Acetyl-CoA Malonate

6.2.1.-

0.00120.06139.24e-70.4847

0.0050

0.6932

0.0026

0.2909

0.69950.2935

0.33761.1695

0.00129.24e-7

@24haserinducEon:Growthrate:0.0613h-1Trans-cinnamate:50.44mg/LPinocembrin:20.92mg/L

• FernandezAetal.J.Biotechnology2014

@48haserinducEon:Pinocembrin:27mg/L

1.

Rhizobium

Pathway ranking:

3.

E. coli

4.

Coxiella

2.

Geobacter

LEARNàBacktoDESIGN


Biosensorengineering

•  DelepineBetal.NAR2016

SOURCEDrugBankHMDB(biomarkers)Tox21

SINKRegulonDBRegTransBaseRegPreciseBioNemo

RULESD=∞onRhea,MetaCycandBRENDAreac8ons

EFFECTOR

CHASSIS

TARGET(signal)

www.sensipath.micalis.fr


Usingmetabolicengineeringtoengineeringbiosensors

• LibisV,etal.ACSSynthBio2016


10 100 10000

5

10

15

20

Concentration (µM)

Rela

tive

fluor

esce

nce

(A.U

.)

1 10 100 10001

2

3

4

5

6

7

8

Concentration (µM)

Rela

tive

fluor

esce

nce

(A.U

.)

10 100 10000

20

40

60

Concentration (µM)

Rela

tive

fluor

esce

nce

(A.U

.)

0 50 100

1.0

1.2

1.4

1.6

1.8

Concentration (µM)

Rela

tive

fluor

esce

nce

(A.U

.)

10 100 10000

10

20

10 100 10000

10

20

0 50 100

1.0

1.9

2.8

Without metabolic module Without metabolic module

Without metabolic moduleWithout metabolic module1 10 100 1000

0

4

8

12

E.colibiosensorsinE.colimanymoleculesofinterest

• LibisV,etal.ACSSynthBio2016


Usingbiosensorsformetabolicengineering

TyrR

FapR

FdeR

Pathwayscanbedynamicallyregulatedandvariantscanbescreenedviafluorescentbiosensor

• FeherTetal.Fron/erinBioengineering&Biotechnology,2015

>100mg/L

2.5mg/L


LACTOSE GALACTOSE+GLUCOSE

ALLOLACTOSELacI LacZYA

+

TheundergroundacEvityof𝛽-GalactosidaseispresentonlywhenLacIispresentThisnetworkmoEfisconservedbyevoluEon(RWheatleyetal.,2013JBiolChem)

WhensyntheMcbiologybringsnewsystemsbiologyquesMons

𝛽-Gal

host Sensed compound intermediates regulator

E. coli lactose allolactose LacI E. coli nitroglycerin nitrite NarL

Thauera aromatica

toluene benzylsuccinate TutBC

Mycobacterium tuberculosis

cholesterol

cholest-4-en-3-one; 3-oxocholest-4-en-26-oyl-CoA

KstR

Mycobacterium smegmatis

cholesterol

cholest-4-en-3-one ; 3-oxo-4-cholestenoic acid

KstR

Paracoccus sp. L-gluconate L-5-ketogluconate ; D-idonate

LgnR

Azoarcus sp. 3-methylbenzoate 3-methylbenzoyl-CoA MbdR Sphingobium sp.

ferulate feruloyl-CoA FerR

Rhodopseudomonas palustris

p-Coumarate p-Coumaroyl-CoA CouR

Comamonas testosteroni

benzoic acid benzoyl-CoA GenR

Thermus thermophilus

Phenylacetic acid phenylacetyl-CoA PaaR

•  Libis,Delepine,FaulonCurrentOpinionMicrobiology,2016


Acknowledgements

ThomasDuigou,IGR,MicalisINRA (reacMonruleworkflow)DhaferLhabib,Post-doc,MicalisINRA (retrosynthesisalgorithm)BaudoinDelepine,PhDstudent,MicalisINRA (biosensordesign)MathildeKoch,PhDstudent,MicalisINRA (biosensordesign&engineering)PabloCarbonell,SeniorResearcher,U.Manchester (metabolicpathwayscoring)JoeMellor,Post-doc,U.Manchester (MachineLearning)VincentLibis,PhDstudent,MicalisINRA (biosensorengineering)HeykelTrabelsi,Post-doc,MicalisINRA (metabolicengineering)AmirPandi,PhDstudent,MicalisINRA (biosensorengineering)CecileJacry,Tech.iSSB,Evry (metabolicengineering)IoanaPopescu,MCF,iSSB,Evry (metabolicengineering)CyrillePauthenier,Abolis,Evry (metabolicengineering)

Using retrosynthesis in synthetic biology and metabolic engineering

Science