Eugene Shakhnovich Chemistry And Chemical Biology Harvard .

Climbing the scale ladder in biologyLecture 3: Biophysical walks on fitnesslandscapes of viruses and BacteriaEugene ShakhnovichChemistry and Chemical BiologyHarvard UniversityENS mini seriesMay 27, 2015

Length scales in biology and evolutionProbability of Fixation:PopulationPfix 1- exp(2s)1- exp(N e s)Part 2:NorvirusescapeFitness Effect:OrganismPart 1:E. GCTAAfitnessmutant - fitnesswildtypes otein-protein interactionProtein-DNA interaction

Exploring Fitness Landscape of E.ColiBottom UpA Problem:Analysis of existing sequences gives a biased view offitness landscape because mutations that we observehave been selected.Can be interpreted only using an evolutionary modelA Solution:Introduce controllable genetic variation and evaluateits fitness consequences without interveningselection.

Biophysics is a stepping stone to linksequences to phenotype

The experimental setupin vitroin silico12Identify mutations withdesired effect on stabilityand activityGenerate constructsfor recombinantexpressionDG, Tm, kcat/KmMeasure molecularproperties of thepurified proteinsin vivoIntroduce mutation(s) on the chromosome bysite-directed mutagenesisendogenous PcmR5folAkanRMeasure DFEcompetition with wt strainGenotype-phenotype linkage43pETBershtein S .&Shakhnovich EI. PNAS 2012

Determination of the bacterial growthparameters: ‘lag’ l and growth rate m:𝑦 𝐴 exp exp0.8This is a Gompertz’s equation usuallyused for fitting bacterial growth curvesin ln(OD/OD0) scale.Here,A height of the plotmm the growth rate (slope at themidpoint of the transition)l lag time (time where the tangent atthe inflexion point has y 0)OD6000.60.40.20.0lag0𝜇𝑚 𝑒𝐴 𝜆 𝑡 1120240360Time (min)480

Biophysics is a stepping stone to linksequences to phenotype

Theory: Evolution of Stability as a Random Walk(Biased Diffusional Motion) on the ‘’Fermi-Dirac’’Fitness LandscapeASSUMPTION: Fitness # of folded proteinsMutational wind - stronger at higher mFolding Free EnergyDeath cliffselection stronger with N

DHFR Is An Essential Core Metabolism Enzyme;Expressed in Low copy number of 100/cell

Adenylate kinase- a high copy # stand-alone enzyme.Expressed at 10,000 copies per cellNMP 214 amino acids inE. coli variant Three structuraldomains:LIDLID (ATP binding): 118-160NMP (AMP binding): 30-67CORE: 1-29; 68-117; 161-214 Large conformationaldynamics of LID andNMP domainCORE𝐴𝑇𝑃 𝐴𝑀𝑃𝑀𝑔 2𝐴𝐷𝑃

Selected mutations based on sequence analysisG080A, G080NL082F, L082I, L082VL083A, L083E, L083F, L083I, L083TMuts: did not occur in MSAMuts: with higher conservationMuts: with varying degrees ofconservationA093F, A093I, A093L, A093V, A093YM096L, M096NV106A, V106H, V106L, V106N, V106WY182A, Y182F, Y182VL209A,L209F, L209I, L209SL213A, L213FA total of 31 mutants at 9positions were selected

Location of the selected mutationsAll mutations are in the CORE domain and are buried

Thermal denaturation using DSCBlack trace is for WT

Chemical denaturation by CD using ureaBlack trace is for WTCD signal (mdeg)0-5-10L83E-15-20-2501234Urea (M)567

2.453.2-4-8-12-16-20-2.0-1.5-1.0-0.50.0Cm(M)Almost all the mutants aredestabilizing, as expectedCm and Tm correlate very well,indicating similar unfoldingmechanism (2 state unfolding)* Fit not reliable0.5

2Ifluorescence (au)ANS fluorescence (au)Aggregation andmolten globuleWTL83EV106H400440480520560600640wavelength (nm)ProteoStatANSIncreasing stability

16300122258150475KM ( M)kcat (min -1)Activities of all mutants have been measuredA positive correlation betweenstability and activity00-16-12-8-4Tm (oC)0-16-12-8-40Tm (oC)17

Fitness landscape for AdK projected onthe stability axis: a perfect cliff:Growth 0444852564044485256Tm (ºC)18

Fitness landscape for Adk fits roughly the‘’Fermi-Dirac’’ 68 0246G (kcal/mol)8 02468

Quite surprisingly,evolution optimizes lag time as well!:Wt stays just above the cliff:-20Lagapp02040601001200.00 0.04 0.080.200.400.60Catalytic ContingencyCatalytic contingency Protein Abundance*kcat/KM20

DHFR: (100 molecules per cell)26 Point mutations in DHFR (Bershtein et al PNAS’12),% conservationE.coli’ DHFR

Thermodynamic, structural, and catalytic propertiesof the in vitro purified DHFR rshtein S .&Shakhnovich EI. PNAS 2012

30oC1.510.5000.511.52oFitness (predicted), 30 C237oCFitness (measured), 42oC2Fitness (measured), 37oCFitness (measured), 30oCFitness is at the plateau at low T butinversely correlated with protein stability at elevated T1.510.5000.511.52oFitness (predicted), 37 C2.542oC21.510.5000.511.52oFitness (predicted), 42 CBershtein S .&Shakhnovich EI. PNAS 2012

Why?Shimon on odd daysShimon on even daysme

Fitness prediction based on protein thermodynamicsassumes:1. Equilibrium2. No flux of matter or energy

Free EnergyDestabilization can promoteoligomerization and/or aggregationDGmut DG Destabilization(mutation, heat)DGdimerizationmonomers(Partially) unfoldedmonomersoligomerAssociation pathwayLiu Y & Eisenberg D. Protein Science 2002

Cross-linking experiment shows in vitrooligomerization of high fitness and aggregation of low fitnessmutant DHFRs at high temperature(W133V)(I155A)DHFR mutants can be roughly dividedinto two groups based on their solubilityat 42C

Soluble oligomerization of the DHFR mutantscontributes to improved fitness at high temperatureBershtein S .&Shakhnovich EI. PNAS 2012

Why are these 5 mutant strains lethal? probably they fail to fold properly.(and fail to oligomerize as well)Finkelstein,Shakhnovich,Biopolymers,1989A fluorescent hydrophobic dye ANS is an excellent probe for MG state - it fluorsescesProteins like materials can be in three states: crystal-like (Native), liquid-like (Moltenupon penetration into MG but does not interact with N (too rigid) or RC (noGlobule) and gas-like(Randomhydrophobicclustersto bind) Coil)29

DHFR of low fitness mutants are Molten Globules:Evidence from ANS fluorescence in vitroLethal Mutants

Mutants which exhibit high Molten Globule Content do not growwell at 42C but are rescued by GroEL overexpression or by theknockout of an MG-sensitive protease Lon.Lon and GroEL compete for Molten Globule DHFR in thecytoplasm hence their symmetric action.Mol Cell 201331

GroEL overexpression affects fitness of each mutant inthe same way as Lon knockoutMol Cell in press32

How about Adk?Chaperone over expression does not rescue AdKmutants, unlike DHFR

Why does GroEL act onfitness same as Lon?Image credit:Ruah Edelstein34

Active cytoplasm environment vs protein solution:key differencesProtein solution:mutations affectthe distributionbetween foldedand unfoldedStates accordingtoBoltzmann law35Active environment:Mutations affect thebalance betweenproduction anddegradation, i.e.of proteinsin the cell

The ‘’Corral Reef’’ model:A simplest kinetic model of active cellular milieu,no aggregationF – folded proteinIB – intermediate bound to GroELIU – free (unbound) intermediateC – production of protein molecules at steady-statek0 – degradation of F (constant and unrelated to Lon protease)

Soluble abundance of DHFR mutants isproportional to signature ANS fluorescence,as predicted by the modelB1.5Soluble abundance at 30 oCSoluble abundance at 30 oCA10.500510152025o30ANS Fluorescence, 30 CFig. 5Mol Cell 2013, v.49, p1331.510.5000.51/AN

At steady-state:Key prediction of the dynamic theory:Lon knockout is an equalizer:Dlon growth rates do not depend on stability!CG Gmaxas k Lon 0C F0 k0Mol Cell 2013, v.49, p.133

The genetic proteostatic network of folA withGroEL/ES and Lon:Full stochastic simulationdegradedkr0[DNA]d F0kP0[mRNA][FolA F]kF0 - Idr0degraded[FolA I]kIeff- F kI0- F kIGroEL/ES [GroEL / ES]- Fdegradedd eff dI0 dILon [Lon]Gillespie algorithm, 3000 trajectories simulatedMuyoung Heo

Lon knockout in simulations indeedequalized DHFR abundance for allmutants

Lon knockout indeed makes all the ‘’phylogeny’’DHFR strains grow with (approximately) same rates.41Mol Cell, 2013, v.49, p.133

Why?RsProtein synthesisDegradation by LonDegradation by other proteasesWhen Lon is knocked out the F-I interconversion becomesThe amount of F in the cytoplasm is determined by the balancebetween production of F and its degradation42

Biophysics is a stepping stone to linksequences to phenotype

Projecting bacterial fitness to biophysicalcoordinatesE. coliDihydrofolate reductaseDHFR (folA gene)Kacser & Burns Genetics 1981Flux dynamics theory:æ kcat öæ 1 öAç çbDG è K m øè 1 e øfitness º flux µæ kcat öæ 1 öconstant A ç ç bDG øè K m øè 1 e “Michaelis-Menten-like’’dependence on abundance Aand activity Boltzmann-like dependence onstability

Orthologous DHFRs from diverse mesophiles20Frequency151050102030 40 50 60 70 80 90 100SequenceSequenceidentityIdentitywith respect toE. coli DHFR, (%)(with Shimon Bershtein)

Orthologous replacements on E. coli chromosomeallow to explore broadly the biophysics-fitness mapOrthologous replacement in the chromosomewhile preserving the endogenous promoter6cmRMeasure fitnessfolA OrthologkanRln(OD/OD0)5endogenous P43210050 100 150 200 250 300 350 400Time (min)E. coliGrowth rate:(slope of exponential phase)42 orthologous DHFRs have been purified and theirbiophysical properties in solution/cell(stability, activity, abundance in cytoplasm etc) have been determined.*That allows us to map biophysics to fitness in a broad range of properties:Large steps on FL.*6 strains with unsuccessful othologous replacements.

Distribution of molecular properties and fitness effectsamong orthologous DHFR showthe major ‘’fitness barrier’’ to interspecies replacementsStability:Activity:Fitness (Growth Rates)

The cause of HGT fitness barrier is partial loss of DHFRfunction rather than toxicity of foreign DHFRNo toxicity: Adding orthologous DHFR from pBAD plasmid does not affect fitness of wt EcLack of DHFR function in the HGT strains:Adding wt DHFR from pBAD plasmidrestores fitness of HGT strains48

Propagating the bacteria under neutral evolution36 strains with DHFRchromosomal replacementsBefore evolution:E. coliLab evolution of each strain at 8 replicates: M9 glucose thiamine 37 Celsius Passage 2x per dayAfter evolution:3 weeks ( 500 generations)

Fitness effect of HGT before and after evolution does notdepend on the evolutionary distance between Ec andorthologous DHFRBlue symbols – after propagationOrange symbols – just after HGT50

Fitness Landscape for HGT:Projecting fitness into molecular properties51

Fitness Landscape for HGT:Projecting fitness into molecular properties52

Fitness Landscape for HGT:Projecting fitness into molecular properties53

Fitness Landscape for HGT:Projecting fitness into molecular propertiesParabola-like dependence oncharge of transferred DHFR-upon HGTLess so after propagation54

Next Generation Sequencing of the Evolved StrainsQ: What did evolution do to remove fitness barrier(s) to HGT?55

Next Generation Sequencing of the Evolved StrainsQ: What did evolution do to remove fitness barrier(s) to HGT?A: It inserted the IS186 186 bp mobile element upstream of lon protease and inactivated it!56

Inactivation of lon protease in the evolved strains:Direct Evidence from Western Blot with anti-lon AbApparently lon distinguishes between ‘’self’’ and ‘’non-self’’ proteins.The mechanism of recognition is unclear57

Evolution dramatically increased DHFRabundance in cytoplasm58

Part II:Predicting evolution:Biophysical Fitness Landscape ofViral Escape from Antibodies59

Norovirus( )ss-RNA, 7.5 kbpDonaldson et al. Nat Rev Microbiol 2010.The Guardian“Most common cause of acute gastroenteritis in the United States.“Most common cause of foodborne-disease outbreaks in the United States.“Causes 20 million illnesses and contributes to 70,000 hospitalizations”-Center for Disease Control and Prevention, US

Viral evolution in the labSerial passaging:Total volume: 2 mLNumber of cells:8 x106 cells/mLNumber of viral particles: 106 pfu/mL 108 viral particles/mLBurst size: 104 virions/infected cellMutation Rate: 1 mutation/genome/generationNext-gen sequencing (Illumina):59186293250 bp250 bp