Lecture 7 Protein Folding

James Chou

BCMP201 Spring 2008

 Lecture Outline

The problem of protein folding

How proteins fold?

How to predict protein folds?

 An observation by Anfinsen and colleagues

Control experiment

Bovine pancreatic RNase

Urea Urea denatures protein mainly by making the

polar and non-polar residues have similar
solubility.
Refolding experiment

Haber and Anfinsen. JBC 1961; 236(2):422-4

 Anfinsens Dogma

Anfinsen et al., PNAS 1961; 47(9):1309

 The problem of protein folding

?
Amino acid Tertiary
sequence structure
 Levinthal paradox

Assume each amino acid backbone can be in 3 conformational states,

for 101 residues, there are 3100 = 5 x 1047 conformations.

If the protein can sample a new conformation at a rate of 1013 s-1, it will
take 1027 years to try them all. Longer than the age of the universe!

Therefore, proteins must fold in pre-arranged pathways and in a

cooperative manner.

Levinthal C. Extrait du Journal de Chimie Physique 1968; 65:44

Zwanzig et al., PNAS 1992; 89:20-22

Cooperativity in protein folding : How a globally optimal state can
be found without a global search?

Origin of cooperativity -- The probability of forming contact C2

is much higher if C1 is formed than in the absence of C1.

Dill et al., PNAS 1993; 90:1942-6

Coil-Helix transition -- the paradigm for cooperativity in biopolymers

Nucleation of
alpha helix

The first H-bond

Initiation of a helical turn is much harder than appending another
residue to a helical segment, due to higher entropic penalty

! " #60 ! " #40

!
! " 180, # 60, + 60 favored rotamer
 !k
!!!
1st helical turn ...cccccccc...#!!
f
"
! ...ccchcccc... ! << 1
kb

k
!!!
2nd helical turn ...ccchcccc...#!!
f
"
! ...ccchhccc...
kb

k
!!!
3rd helical turn ...ccchhccc...#!!
f
"
! ...ccchhhcc...
kb

kf
equilibrium const for each reaction s =
kb

!!!
A#
k
f
" dA
= !k f [ A ] + kb [ B ] = 0 K eq =
[ B] k f
=
!! !B
kb
dt [ A ] kb
The Zimm-Bragg-Lifson-Roig theory of coil-helix transition

!k k k k k
!!!
C#!!
f
"
! H 1
!
#!!f
!! "
! H 2
!
#!!f
!! "
! $ !
#!!f
!! "
! H n "1
!
#!!f
!! "
! Hn
kb kb kb kb kb

!!
C#!"
! Hn Kn =
[ H ]
n
= !s n

[C]
N"3
! # ( k + 3) ( N " k " 2 ) s k
Fraction of residues in helix = k =1

$ N "3
k'
,
N &1 + ! # ( N " k " 2 ) s )
% k =1 (

where N is the number of residues in the polypeptide chain.

 Cooperative transition has a sigmoidal profile

N "3 Melting curve of a coiled-coil

! # ( k + 3) ( N " k " 2 ) s k
protein in water
f = k =1

$ N "3
'
N &1 + ! # ( N " k " 2 ) s k )
% k =1 (

The sharpness of the transition

depends both on and N, being
sharper at smaller values of and
at larger N. The temperature
dependence of s is given by
ln s G o k BT .
Cooperativity results roughly in a two-state system

Luminescence decay kinetics of

Cyt c measured at various
times after initial denaturation.

The process of unfolding

shows a bimodal distribution,
one for largely unfolded state
and other includes nearly
folded states.
 The free energy of protein folding

~50 kJ/mol for

10 kDa protein

From Alan Fersht. Structure and Mechanism in Protein Science

 Hydrophobic interaction (entropic)

Pure H2O H2O around a hydrophobic molecule

Water molecules have less

Water molecules have more
degrees of freedom as the H-
bonds are in the true tetrahedral
arrangement.
degrees of freedom in the
clathrate cage arrangements
because some H-bonds
cannot point inside toward the
hydrophobic sphere
 Formation of protein hydrophobic core in water

S > 0

Unfolded Folded

More Hydrocarbon-Water Less Hydrocarbon-Water

Interfacial Area, Interfacial Area,
More Water Ordered Less Water Ordered
 Hydrogen bonds (enthalpic)

Hydrogen bonds within protein Hydrogen bonds with H2O (hydration)

!H of H-bond = " ( 4 to 20 ) kJ mol"1

 Other Enthalpic Interactions

Electrostatic or Coulomb interactions between ions

van der Waals interactions

Metal coordination

Disulfide bonds
 Dissecting the free energy of protein folding

!G
Unfolded #!!!
!! "
! Folded
!G = !H " T !S < 0, !G =~ "50 kJ/mol

Hydrophobic effect ! T "S ! 0 !G ~ "50 kJ/mol

~ !200 kJ/mol

H-bonds !H ! 0
chain conformational
" "500 kJ/mol
entropy ! T "S ! 0
~ 750 kJ/mol

VDW !H ~ " 50 kJ/mol

Electrostatic !H ! "50 kJ/mol
 Denaturation by Heat -- break H-bonds and other enthalpically
favorable interactions

!G = !H " T !S < 0

Hydrophobic effect ! T "S ! 0 !G ~ "50 kJ/mol

~ !200 kJ/mol

H-bonds chain conformational

Heating makes H entropy ! T "S ! 0
less negative ~ 750 kJ/mol

VDW
Electrostatic
 Denaturation by Cold -- reduce the contribution from
hydrophobic effect

!G = !H " T !S < 0

!G ~ "50 kJ/mol
Hydrophobic effect

Near freezing T, entropy

of H2O around non-polar
residues is less different H-bonds chain conformational
from those around polar entropy ! T "S ! 0
residues
~ 750 kJ/mol

VDW
Electrostatic
 Pathways of protein folding

From Alan Fersht. Structure and Mechanism in Protein Science

 Folding pathway of Barnase

Barnase is a bacterial protein that consists of 110 amino acids and has
ribonuclease activity.

Bond et al., PNAS 1997; 94:13409-13

NMR studies of denatured Barnase reveal residual 2nd structures

1H (ppm)

Arcus et al., JMB 1995; 254(2):305-21

 Chaperone assisted protein folding

Two most important types of chaperones, Hsp60 and Hsp70

Hsp60 - In bacteria, provide a folding chamber

Hsp70 - In all living organisms, mainly to block aggregation

Hsp60 - GroEL-GroES chaperone
 Hsp60 - GroEL-GroES chaperone

Fenton and Horwich, Quarterly Rev of Biophysics 2003; 36(2):229-56

DnaK binds unfolded proteins by recognizing an extended region
of the polypeptide chain that is rich in hydrophobic residues

substrate
 free unfolded protein

E J

Bukau and Horwich, Cell 1998; 92:351-66

 Computational Protein Folding

Amino acid Tertiary

sequence structure
 The free energy landscape of protein folding

Ideal funnel Reality funnel

Dill and Chan, Nature Struct Biol 1997; 4:10-19

 A brief note on molecular dynamics simulation

The basic idea is to solve Newtons equation of motion for every atom in the system

d 2 ri
!" qi V = mi 2 ,
dt
ri position of the ith atom

mi mass of the ith atom

V total potential energy of the system

The total potential energy is a function of the atomic positions (3N) of all the atoms in the
system. Due to the complicated nature of this function, there is no analytical solution to
the equations of motion above; they must be solved numerically.
Example of computer simulation using the Verlet algorithm

All the integration algorithms assume the positions, velocities and accelerations
can be approximated by a Taylor series expansion:

1
r( t + !t ) = r( t ) + v ( t )!t + a( t )!t 2
2
1
r( t " !t ) = r( t ) " v ( t )!t + a( t )!t 2
2
d 2 ri
!"qi V = mi 2
dt
Summing the above two equations, we get

r( t + !t ) = 2r( t ) " r( t " !t ) + a( t )!t 2

The Verlet algorithm uses positions and accelerations at time t and the positions
from time t-dt to calculate new positions at time t+dt.
MD simulation of membrane, water and water channel
http://www.stanford.edu/group/pandegroup/folding/villin/

Game - Folding@home by the Pande group at Stanford

Jumping out of the false minima

Temperature annealing
Melt the system at 2000 C and
slowly cool down to 20 C.
Perform MD simulation at each T
step.

Metropolis Monte Carlo

1. Make a move (alter the conf.)

2. Calculate G for the move

3. Generate a random number,

x, between 0 and 1
# !"G &
accept the move if x < exp %
$ kBT ('
and reject otherwise.
Statistical methods for predicting protein structures

On Feb 12, 2008, there are 48891

Structures in PDB. Pure statistical or
machine-learning methods are becoming
more and more powerful due to the rapidly
expanding data base.
Early predictions of protein structural properties use amino
acid composition

The Chou-Fasman Method

v = [ p1 , p2 ,, p20 ], pi = % residues of AA type i

Generate similar vectors for known classes of

folds, e.g., , , /, etc, from the PDB.
coil
V
/
Chou PY, Fasman GD. (1974). Prediction of
protein conformation. Biochemistry.
13(2):222-45.

Prediction of Protein Cellular Attributes

Using Pseudo- Amino Acid Composition.
Proteins 2001; 43:246-55.
Knowledge-based methods for predicting protein structures

Homology Modeling - Multiple AA sequence alignment

PSI-BLAST (take into account
sequence evolution information)

Artificial Intelligence - Neural network

Pattern Recognition - Evidence theory, Support

Vector Machine

Combining All Youve Got - Rosetta from the Baker Lab