Quantitative properties of protein-protein interactions Ed Evans, T-cell biology group

Why this lecture? Protein/protein interactions are fundamental to biology and therefore to medicine! In the past much of the focus has been on qualitative information. a)What proteins interact? b)What is the function of the interaction? Now quantitative information is also considered increasingly important

An example: T-cell co-stimulation CD80 CD28 CTLA-4 CD28 CTLA-4 CD86 Co-stimulationInhibition 20M4M4M3M3M0.2M Affinity Valency Stoichiometry ~10,000 times stronger

Why this lecture? Protein/protein interactions are fundamental to biology and therefore to medicine! In the past much of the focus has been on qualitative information. a)What proteins interact? b)What is the function of the interaction? Now quantitative information is also considered increasingly important a)Helps understand molecular mechanisms b)Essential for modelling complex processes c)Important for drug discovery

What we’ll cover (hopefully) 1.What binding properties are important? 2.How might we measure them? (introduction – more tomorrow!) 3.Comparison of interactions ‘in solution’ vs. at the cell surface

1. What binding properties are important? One protein soluble: Affinity Valency (Kinetics) (Thermodynamics) Cell-cell interaction: Mechanical Stress (unbinding force) Real membranes: Lateral diffusion rate Inter-membrane distance (size) Abundance => Avidity A T O M I C S T R U C T U R E

Molecules involved … T-cell surface CS1 CD11a* ‡ TCR  1* CD6*CD50* CD48* CD164 CD96*CD99 CD7* CD244 CD97* NKp30* CD147 CD46 CD162/R* CD5* CDw210* ESL-1 CD69* ?? CD71 CD103 ‡ CD222 CD223 CD43 NTBA CD100 CD107a ? ? CD95 CD8   1* Toso CD120a CD150 CD178 CD49a ‡ CD224 ? CD26 ? CD70 CD86RAGE CD101 CD160 CD168 ? CD201 CD226 CD229 ? CD38 CD44 CD44R CD49c ‡ CD51 ‡ CD56 CD58CD72MAFA-L ST2L SRCLPorimin CD146 CD45* P P ? ? ? ? MDC-L CD230 ? CD2* CD52* Galectin-1 CD225* HM1.24 ? CD39CD184* TIRC7* CD53* CD122* CD37* Galectin-3 CD195*CD63 EDG4 CD47R CD183 CD132* CD25 Galectin-9 ? apoB48R CCR1 CCR2 ? CD132* CD127 CXCR5 TRAILR2 CCR4 CRTH2 Galectin-11 Galectin-13 IL-11R  OAP-1 CD59 TR2 ? ? GPR68*FLJ23270* CD247(  )* CD3  * TCR  2* CD236R  2* NKG2E/H* CD94 CD68 ? NKG2-F

Molecules involved … T-cell surface Including less well characterised / housekeeping molecules...

Background: binding models 1.The majority of protein:protein interactions are simple 1:1 associations: A + B  AB This is what we will focus on in this lecture. 2.The most common variation of this scheme is where one protein (e.g. B) has additional binding sites. e.g. AB + A  A 2 B or AB + C  ABC 3.If these binding sites are independent then one simply treats each interaction as a 1:1 association and adds them together. 4.If the binding sites are not independent then one has positive or negative cooperativity (i.e. allosteric effects) and more complex modeling is required.

What binding properties are important? Affinity –K A (affinity constant) or K D (dissociation constant) –K D = 1/K A Kinetics –k ass or k on (association rate constant or on rate) –k diss or k off (dissociation rate constant or off rate) Thermodynamic properties –H (enthalpy change on binding) –S (entropy change on binding) –C (heat capacity change on binding)

Affinity 1.Measures how favourable an interaction is 2.Best expressed as affinity constant: K A 3.For A + B  AB –Best thought of as the ratio of [products] vs. [reactants] at equilibrium –Note the units (M -1 ) –Higher affinity = higher K A

Affinity 4.Also expressed as dissociation constant: K D –The inverse of K A –Usually thought of as concentration of A at which half of B is bound ([B]=[AB]) at equilibrium –Units are M –Higher affinity = lower K D

Measuring the affinity constant 1.One could simply measure [A], [B] and [AB] at equilibrium and calculate K D 2.In practice this is difficult and the following approach is used. 3.Increasing fixed concentrations of one molecules A ([A]) are added to a fixed small amount of its ligand B and you measure the amount of bound A (Bound) 4.Plot the results and fit the 1:1 Langmuir equation to the data to determine K D and Bound max

Measuring affinity constant Bound (arbitrary units) [A], M K D = 19 M Bound max =200 Data are circles Line is non-linear fit of the equation performed by a computer (e.g. Origin, R) Gives the indicated values for K D and Bound max If the fit is good it indicates that binding follows the simple 1:1 model Difficult to see if fit is poor in this plot

Scatchard plot A plot of Linear for a 1:1 interaction If curved it indicates wrong model and possible problem with the experiment Most commonly concave up Usually caused by experimental error (often heterogeneity) Sometimes due to negative cooperativity Far less common is to see concave down Usually caused by positive cooperativity Bound/[A] Bound max Slope = -1/K d

****** T cell bilayer boundfree Dustin et al JCB 132, 465 Experimental set-upData collection Measuring 2D Kds

Comparing 2D and 3D affinities D K d (  M) hCD2 TCR CD28 CTLA-4 KIR CD8 SLAM CD4 2D K d (mols/  m 2 ) hCD2 rCD2 K d > 2  M “barely adequate”strong interaction

Thermodynamics of binding 1.Binding is favoured if it leads to a net increase in disorder or entropy. 2.This includes entropy of…. a)the system (interacting molecules and solvent) represented as change in entropy or S b)the environment (everything else) as the system releases or absorbs heat it changes the entropy of the surroundings heat release is measure as change in enthalpy or H

Gibbs free energy change 1.The change in Gibbs free energy (G) is a measure of the net change in universal entropy - i.e. the extent to which binding is favoured. G = H -T S If G < 0 then binding is favoured. 2.G depends on concentration. At equilibrium G = 0 3.G o is the standard state G which assumes all components are at the standard state concentration of 1 M (mol.L -1 ) 4.It can be calculated from the affinity constant G o = RTlnK D R = Gas Constant (2 cal.mol -1.K -1 ) T = absolute temp. in Kelvin ( o C ) and K D is expressed in units M

Origins of enthalpy and entropy changes G o = H -TS o 1.Change in enthalpy (H) a)Release of heat (H <0) favours binding b)This happens when bonds are formed e.g. hydrogen bonds, salt bridges, van der Waals contacts c)However bonds are also broken upon binding displacement of water and ions (always) conformational change (sometimes) 2.Change in entropy (TS) a)Increase in entropy (S >0) favours binding b)Protein/protein interactions leads to decrease in entropy Stabilise conformation at the binding interface Decreased rotation/translation of proteins c)However displacement of water from the binding interface leads to an increase in entropy (the hydrophobic effect)

The key role of water 1.Water is present at very high concentrations (55 M) and interacts with protein surfaces 2.Thus, many water bonds need to be broken, which has an unfavourable enthalpic effect 3.Water can also act as glue filling in gaps between surfaces that lack surface shape complementarity 4.Water is believed to form an organised shell over hydrophobic surfaces. Ejection of water from these surfaces into free solution has favourable entropic effect. This is the ‘hydrophobic effect’. 5.Note that there is a weak unfavourable enthalpic effect as well since the water molecules in the shell interact weakly Hydrophilic patch in binding site binding Hydrophobic patch

TCR and antibody binding have distinct thermodynamic properties (Data from Willcox et al 1999 and Stites 1997) Favourable Unfavourable

Changes in conformation at a T cell receptor/peptide-MHC interface TCR pMHC Garcia et al (1998)

Heat capacity change (C) 1. H and TS usually vary with temperature 2.The extent of this variation is given by C 3.This is a consequence of changes in water with temperature binding Low temp – binding disrupts water ‘shell’ with unfavourable effects on H and favorable effects on S Hydophobic patch binding High temp – water shell already ‘melted’ so both effects are lost Hydophobic patch

Why measure heat capacity change? S includes contributions from changes in solvent entropy (hydrophobic effect) and protein entropy 2.The heat capacity change can be used to estimate solvent entropy change, enabling estimation of the protein entropy change. DETAILED EXPLANATION ( Spolar and Record (1994) Science 263:777) C correlates with non-polar surface area that is buried by binding (A np ) => C can be used to estimate the contribution of the hydrophobic effect (S he ) to total entropy change (S Total ) The change in rotational and translational entropy (S rt ) can be calculated, and is same for all protein- protein interactions. Thus S other can be calculated since S Total = S he + S rot/trans + S other Main contribution to S other is thought to be reductions in conformational flexibility accompanying binding i.e. it’s a measure of amount of conf. change

Measuring thermodynamic parameters 1.S can’t be measured directly 2.G and H are measured and G = H -TS 3.H can be measured in 2 ways a)calorimetry (see later) or b)van’t Hoff analysis Van’t Hoff analysis 1.G is measured over a range of temperature and plotted 2.The non-linear van’t Hoff equation* is fitted to the data to determine H, S and C 3.The slope represents H 4.This plot is curved for macromolecular interactions as H varies with temperature 5.The curvature represents C

Kinetics Since biological systems are not at equilibrium, the rate of binding and dissociation is critical For a simple 1:1 interaction ( A + B  AB )… 1.Rate of dissociation a)d[AB]/dt = k diss [AB] b)where k diss is the dissociation rate constant (k off ) 2.Rate of association a) d[AB]/dt = k ass [A][B] b)where k ass is the association rate constant (k on ) 3.At equilibrium the rate of association must equal the rate of dissociation k diss [AB] = k ass [A][B]=> k diss /k ass = [A][B]/[AB] = K D

Dissociation Any reaction of the form d[AB]/dt ∞ [AB] will be exponential so a)i.e. [AB] t = [AB] o e -k diss t b)k diss determined directly by curve fitting The half life (t 1/2 ) can be calculate as follows: Since at t = t 1/2 [AB] t /[AB] o =0.5=e -k diss t 1/2 It follows that -k diss t 1/2 = ln(0.5) = Thus t 1/2 = 0.693/k off Dissociation of A from B Symbols are data, lines are fitted curves t 1/2

Association In most experimental system it is impossible to follow association alone in the absence of simultaneous dissociation For the simple interaction A + B  AB d[AB]/dt = k ass [A][B] – k diss [AB] It follows that [AB] t =[AB] final (1-e -k obs t ) where k obs = k ass [A]+k off Thus one needs to know k off and [A] as well as measuring [AB] to calculate the k on

Determination of binding kinetics k obs = [A]k ass + k diss k ass M -1.s -1 k diss 0.5 s -1 K D 1.2 x M Residuals plot (difference between data and fitted curve) Association phase (k obs ) Dissociation phase (k diss )

Factors affecting kinetics 1.The association rate constant does not vary that much a)Association requires two proteins to collide in the correct orientation and in the correct conformation b)Depends on diffusion so will be similar for most proteins c)The basic rate is about 10 5 M -1.s -1 d)Can be accelerated by long range electrostatic forces Increased rate of collision Steer binding sites into correct orientation E.g. barnase/barnstar interaction 2.The dissociation rate constant varies considerably and is responsible for most variation in affinity constants a)It is determined by the number and strength of bonds in the contact interface b)Depends on size of interface and the degree of surface- shape and electrostatic complementarity

Summary of average affinity and kinetic constants for biological interactions Interaction k on (M -1 s -1 )k off (s -1 )K D (M) Cell-cell recognition molecules to Antibody/antigen Cytokine/receptor Enzyme/inhibitor (eg barnase/barnstar)

Transition state theory (or activation complex theory) 1.Proposes that when two molecules interact they traverse a mountain-like energy ‘landscape’. 2.Highest energy point is the putative transition state (AB ‡ ). 3.The height of this point determines rate constant 4.Transition state then ‘relaxes’ into the final complex (AB).  ‡G‡G diss A+B AB ‡ Binding coordinate Potential energy (kcal.Mol -1 ) AB  ‡ G ass GG  ‡ G diss

Assume reactants are in equilibrium with transition state => it is possible to apply thermodynamic principles A + B ↔ AB ‡ =>  ‡ G =  ‡ H-T ‡ S  ‡ G : calculate from k ass  ‡ H : determine by measured temperature dependence of k ass => T ‡ S can be calculated Analyse in the same way as described previously Application: to study the structure of the transition state complex (i.e. how binding occurs) e.g. by examining the effect of mutations on these values one can determine which residues in the binding site interact in the transition state complex The same approach can be used for dissociation AB ↔ AB ‡ Transition state theory

Breaking a bond analogous to pulling a cart up a hill height of hill = work required = bond energy slope of hill = force required = mechanical strength Force = work/distance Mechanical strength = bond energy/bond length bond energy bond length long bond length short bond length slope = mechanical strength The importance of bond length

The mechanical strength of a bond is only indirectly related to affinity Mechanical strength = bond energy/bond length => 1.The bond energy is likely to be more closely related to the enthalpy change (ΔH) than the affinity (ΔG) since ΔH measures net number of bonds broken 2.Since some bonds reform during dissociation mechanical strength is related to the number of bonds broken to reach the transition state of dissociation 3.This is given by the activation enthalpy of dissociation or Δ ‡ H diss

IMPLICATIONS: Mechanical strength should be measured directly if possible. This can be done using techniques such atomic force microscopy The mechanical strength of a bond is only indirectly related to affinity

2. How can we measure them? SPR (BIAcore) AUCITC (microcalorimetry) Surface Plasmon Resonance Analytical Ultracentrifugation Isothermal Calorimetry

Measuring key properties of protein-protein interactions PropertyAUCBIAcoreCalorimetry Affinity++++ Enthalpyno+++ Entropyno+++ Heat capacityno+++ Kineticsno++no Stochiometry++++ Size & Shape+no

3. Comparison of interactions ‘in solution’ vs. at the cell surface For most interactions of soluble proteins, want strong, specific interactions But what about on the cell surface? How can you get TRANSIENT adhesion? (especially given number of molecules on a cell!) PROBLEM: how to reduce affinity without reducing specificity? Reducing area of interaction will reduce both

3D K d (  M) hCD2 TCR CD28 CTLA-4KIR CD8 SLAM CD4rCD2Ab:Ag Inactive LFA-1fully active LFA-1 fully active Mac-1 Selectins The range of affinities seen for transient interactions at the cell surface

Best studied example: CD2:ligand interactions CD48LFA-3 CD2 CD2: cell adhesion molecule enhances antigen recognition by T-cells human: K d = 15  Mmurine: K d ~ 65  M 2B4CD2

Clues from the Rat sCD2 structure T86 R87 K43 E41 E33 structureelec. potentialmutations

Charged residues & binding specificity CD48: R31 to ? E44 to ? CD2: E41A K43A

A new protein recognition paradigm? CD2:LFA-3 K d = 15  m J Bloggs mAb:lysozyme K d = 1 nM

The value of electrostatic interactions Specificity is generated by electrostatic rather than surface/shape complementarity BUT this does not result in high affinity because salt bridges are approximately energy neutral. This is because binding energy from these interactions is counteracted by the need to disrupt the interactions between the charged residues and the solvent (i.e. water) before binding.

hydration of the charged residues in the unliganded receptor exclusion of water from the interface The value of electrostatic interactions

Electrostatic complementarity predicts CD2:LFA-3 complex topology Harvard complex Wang et al. Oxford prediction Ikemizu et al.

But it’s not all like that... B7-1:CTLA-4 ~ High surface complementarity Complementarity (S) = (antibodies = ; CD2 = 0.58; TCR = 0.45)

Thermodynamics of sB7-1/CTLA-4: some compensation? kcal/mole of injectant molar ratio  H =  G = -8.9 T  S = -2.7 kcal/mol -1

Interactions are NOT uniform! All transient cell surface interactions are relatively weak but mechanism varies. Range of affinities/avidities is still large Precise affinity/avidity and structural mechanisms used to determine it (and specificity) depend on FUNCTION.

Willcox et al. (1999) Immunity 10:357 TCR recognition Another reminder: TCR entropy barrier – linked to function/nature of TCR?

Interactions are NOT uniform! All transient cell surface interactions are relatively weak but mechanism varies. Range of affinities/avidities is still large Precise affinity/avidity and structural mechanisms used to determine it (and specificity) depend on FUNCTION. Oligomerisation state and valency are key factors in determining avidity (remember full e.g. at start)

Interactions between two cell surface proteins are generally weak but remain highly specific Hierarchical affinities may determine the sequence of events in key processes such as T-cell activation Co-operative, avidity-driven interactions canprofoundly alter the strength of signalling In general, it is more important that these interactions are weak than how this is achieved – but there are other functional constraints. A variety of structural mechanisms underlie this e.g. Clustered charged residues allow weak specific recognition by CD2 and its ligands Summary of the nature of recognition at the cell surface

What we’ve covered 1.What binding properties are important? a)Affinity b)Thermodynamics c)Kinetics d)Stoichiometry, avidity etc. e)Mechanical binding strength 2.How might we measure them? a)SPR (BIAcore) b)ITC / microcalorimetry c)AUC 3.Comparison of interactions ‘in solution’ vs. at the cell surface