Notes 20100929 CHEM 731 Meiering Protein Engineering and Design

From SnOwy - Ed's Wiki Notebook

Note to self: do I want to audit this course?

Administrative

• a book for this course is on reserve in the library

Display systems, protein engineering

• phage display
• cell surface display
  • binding proteins
  • present structure on membrane
• cell-free display

Cell free displays

• directed evolution

In vitro compartmentalization

Directed evolution by in vitro compartmentalization (July 2006)

• constructing artificial cells
• emulsion compartments
• lipid emulsion surrounds each environment
• on average, one DNA sequence per droplet -- incl. RNA manufacturing machinery in each droplet
• 10¹⁰ genes per droplet
  • (other techniques go to 12 to 15 magnitudes)
• ligate
• the droplets look like monolayers
  • droplets are nanometer size
• particularly useful for enzymes
• enzymes must be complementary to a transition state
• enzymes are then trapped in each droplet
• examples
  • DNA altering enzymes -- DNA methyl transferases

Directed evolution methods

in vitro evolution

• faulty DNA polymerase
• primers with amino acid substitutions
• product is bound and ligated
• substrate is covalently linked to the DNA
• retrieve droplets with fluorescent screening
• i.e. find the changed substrate → correct enzyme

Near future

• microfluidics -- engineering attempt to sort the drops better and better

Water-Oil-Water (w/o/w) emulsions

• proteins may not translate well in vitro
• capture each E. coli cell in its own membrane

Back to general directed evolution

• enzyme modifications include
  1. increase equilibrium affinity
  2. slow dissociation rates
  3. protein stability and folding
  4. evolve enzymes -- the above compartmentalization methods
• the evolution tasks require different approaches

Rational design

force fields, computation

• protein fitness landscapes -- directed evolution

Library construction -- protein engineering

see text -- follows display methods

• error-prone PCR
  • low return
  • proteins won't tolerate a high density of changes
• recombination
  • generate fragments of your protein and recombine them
  • orthologous genes
  • protein modularity
  • site-directed diversity (single nucleotide polymorphism (SNP) / point mutation)
  • bioinformatics powered?
  • computation --
• scanning mutagenesis
  • mutate each amino acid in turn -- perhaps into an alanine -- determines which positions are mandatory
  • natural source diversity -- antibody libraries from germline sequences
• things that aren't well discuss
  • the extended nature of the diversification -- sufficient diversity to solve the problem / sufficient number of amino acid types
  • sampling size problem (degrees of freedom!)
• library quantity -- library must represent intended design
  • design versus randomization
• drive to use random approaches since there is insufficient mathematical background still -- reminds me of machine learning

Protein fitness landscapes

Exploring protein fitness landscapes by directed evolution (2009)

• mutations far from active site matter -- not well understood
• proteins in general are not very stable
• directed evolution has allowed customization of fluorescence proteins
• single-site modifications allow us to see the mutations that won't survive in nature
  • we are suffering from an 'extreme' observer's bias in bioinformatics -- we're only seeing the stuff that survives!
• fitness space
• epistasis: an amino acid's behaviour depends on its neighbours
  • general -- a gene's behaviour depends on other genes.
  • some proteins cannot be improved -- we're assuming that a starting candidate is evolvable
    • extra evolvable: TIM barrels (catalytic apparatus), antibodies (VDJ madness!) ...

Steps

• parent selection (template, candidate sequence)
• ...

Misc

• screening is difficult -- fluorescence / binding based screens
  • incorrect screening → poor results
• screen for iterative improvement
  • gradually increase selection pressure
• deleterious mutations occur often -- of random mutations,
  • ... 30-50% are deleterious
  • ... 50-90% are neutral
  • ... 0.01-1% are beneficial
• note -- search is biased due to the candidate

Recombination

• splicing modules of different proteins
  • spliced orthologous proteins
  • may produce more stable proteins with poorer function
  • creates new candidate parents for further point mutagenesis
• example: cellobiohydrase II

Findings for recombined enzymes

• optimization requires trading stability against function
• example enzyme -- cytochrome P450 mutant gained a long-chain fatty acid catabolysis into propane function
• proteins are only as stable as they need to be
• most mutagenesis must cause destabilization -- 80% of diseases involving proteins: the proteins are destabilized
• possible to use more stable mutant before attempting to enhance function

Computational approaches

see book

• computational approaches, homology modelling and bioinformatics
• consensus motifs

Repeat proteins

A recurring theme in protein engineering: the design, stability and folding of repeat proteins (2005)

• fibrillar (non-globular)
• widely used in nature -- relatively simple
• repeating units are 20-40 aa
• different folding properties in terms of hydrophobicity
  • structural capping at the end of the long chain of repeats
• know your fibrillars ...
  • ANK Ankyrin
  • TPR Tetratrico peptide repeat
  • LRR Leucine rich ...
• coupling networks -- direct correlation between two specific positions -- e.g. salt bridge Lysine---Glutamate
• novel binding specificity
• unlikely to fold in the way globulars do
  • possible to have multiple folding pathways (needs more research)
• repeats in a fibrillar repeat protein might not be perfect lego blocks -- loss of stability if insertion of repeat
  • however, consensus sequences are OK
• ANK proteins are used very often in protein engineering

Biophysical characterization

• melting point
• ΔG

Consensus Analysis

Evolutionarily Conserved Pathways of Energetic Connectivity in Protein Families

• energetically coupled residues that are structurally distant
• enzymatic and substrate binding is beyond binding site analysis
  • coupled networks
  • can be calculated
  • statistical free energy for sites i, j
  • probability of positions used in Boltzmann distribution
• free energy G
• G = -RTln(k)
• k = [a]/[b]
• background probability given by genome
• the networks -- allosteric regulation networks
  • can go right through the core of the protein from the binding site to the opposite side.

Pairwise Interactions

Surface Sites for Engineering Allosteric Control in Proteins

• coupling DHFR with PAS to make PAS-DHFR complex -- tuning activity of DHFR with light-sensing PAS
• continuation of previous paper ...
• PAS-DHFR does not normally exist
• PDZ binding site of DHFR connected to core which will talk to PAS
• chromophore absorbs light on surface of PAS
• change on back-side in C-terminus -- coupled using a allosteric regulation network
• helices carry allosteric signal
• DHFR binds folate -- there is a corresponding network there on the back site, βF-βG loop
  • βF-βG loop must be dynamic to enable enzymatic action
• input PAS light
• output DHFR
• two sets of constructs created
  • things expected to work
  • negative controls -- with intentional sabotage of allostery by removal of important loops (communication cutting)
• conclusion -- allosteric regulatory networks needed during protein design

Graphical Analysis

• homology modelling
• stability, affinity through steric complementarity
• affinity through electrostatic complementarity

Fast methods for mutation analysis

Action-at-a-distance interactions enhance protein binding affinity

• folding problem and inverse folding problem (threading problem)
  • folding: sequence → structure → function
  • dead-end elimination
  • self-consistent meanfield theory
  • simulated annealing
• calculation of effects of individual mutations
• effect on binding
• calculation capable of determining binding energy
• hydrogen bonds, van der Waals forces -- these ended up not being needed in this case
• charge mutation library possible
• residues aren't directly in binding interface

Math and software used

• CHARMM etc.,
• Poisson-Boltzmann equation with DELPHI / PARSE
• simplified force fields

Molecular force fields

• vastly simplified here (thanks!)
• physics-based force fields used for
  • protein stability
  • denatured state -- required for a reference state
• computing free energy -- explicit / implict representation
  • commonly -- implicit water represented as an all-purpose dielectric medium
• statistical potentials
  • potential of mean force (energies based on statistics of protein structures) -- information-based -- ROSETTA
• reference states ← ?
• fundamental interactions between atoms, molecules

Internal energy function

• Hooke's law for springs!
  • E_total = Σ_bondsK_r(r-r_eq)² + Σ_anglesK_θ(θ-θ_eq)² + Σ_dihedrals (V_n / 2)(1 + cos(nφ - γ)] + Σ_i<j[(A_ij / R_ij¹²) - (B_ij / R_ij⁶) - (q_iq_j / εR_ij)]
    • A is repulsion of very close molecules
    • B is attraction of close molecules
• Newton's laws solved for very small time intervals
  • no quantum mechanical terms here yet-- but they can be included!
• we can try to minimize the energies of the system
  • these energy formulas relate to the enthalpy of the system
  • the entropy for proteins is not solved -- we can infer some useful entropic properties with molecular dynamics
• Monte Carlo simulation useful for evaluating a random sampling of conformations -- step by step approaching a minimal energy
• Rosetta -- uses conformations observed in real proteins

Molecular Dynamics / Mechanical -- Force Fields

BIOMOLECULAR SIMULATIONS: Recent Developments in Force Fields, Simulations of Enzyme Catalysis, Protein-Ligand, Protein-Protein, and Protein-Nucleic Acid Noncovalent Interactions

• explicitly including water is computationally tough
  • implicit is easier -- use force fields
• see the alphabet soup of energy functions
  • AMBER, CHARMm, . . .
• using various force fields
  • keep in mind which force fields work in reality, and which don't -- must draw from literature based on success
• hybridizing molecular dynamics with quantum mechanical methods
  • requires Hamiltonian matrix

Rosetta Overview

Practically Useful: What the ROSETTA Protein Modeling Suite Can Do for You

• conformational sampling space
• sequence sampling space

Intrigue

• consensus motifs bother the heck out of me because they drop all notion of column to column correlation
  • would like to figure out if we can improve similarity scoring with that knowledge
  • sounds like Bayes
• see "Consensus Analysis" above -- this has been touched!
• Hamiltonian missing assembly -- undergraduate chemistry :(