Biophysical methods to guide protein crystallizationand inhibitor binding studies

Paul Erbel, Frederic Villard, Allan d’Arcy

RAMC: September 2011

Introductionoverview

N

N

O

N

NO

N N

N

N

O

N

NO

N N

N NH

FF

F Cl

OH

N

O

N+

O

OO -

N

O

O

Cl

F

FF

NH

NH

O

OH

Cl

N

O

N+

O

OO -

N

O

O

Cl

F

FF

NH

NH

O

OH

Cl

N

S

N

N

NH

N

N

C

N

O

Cl

Biophysical methods to characterize protein quality and guide crystallization

• Limited proteolysis

• Protein purification

• Construct design

• Special focus on NMR spectroscopy

Biophysical methods to characterize compound binding

• Support crystallization: ligand binding alters crystallization properties

• Select compounds for cocrystallizationBiophysical methods: expensive toys or powerful tools?

Proteins that are perfect but difficult to crystallize

Introduction: Classes of proteinsan oversimplified view

Proteins that cannot crystallize

Aggregation/wrongly folded

Proteins that are easy to crystallize

Introduction: critical parameters for protein crystallization?Can we measure those parameters? Can we affect those parameters?

Purity

Aggregation state

Stability

Structural order

Crystal contacts

Nucleation

Molecular biology - Protein construct

Introduction: critical parameters for protein crystallization?Can we measure those parameters? Can we affect those parameters?

Purity

Aggregation state

Stability

Structural order

Crystal contacts

Nucleation

Molecular biology - Protein construct

No Crystals

SDS Page

Size Exclusion Chromatography

Dynamic Light Scattering

Does the protein concentrate?

Thermostability

Differential Scanning Fluorimetry (DSF)

-> Change buffer conditions

Specific Enzymatic activity

Limited proteolysis

Nuclear Magnetic Resonance

-> Ligand binding

-> Binding partners (additional domains)

-> Natural variants

-> Engineering/surface mutations

-> Binding partners

-> Seeding

-> More screens / narrow screens

Some background: Protein folding as seen by NMRNMR = No Meaningful Results (quote by famous colleague)

N

N

O

N

NO

N N

N

N

O

N

NO

N N

N NH

FF

F Cl

OH

N

O

N+

O

OO -

N

O

O

Cl

F

FF

NH

NH

O

OH

Cl

N

O

N+

O

OO -

N

O

O

Cl

F

FF

NH

NH

O

OH

Cl

N

S

N

N

NH

N

N

C

N

O

Cl

R-CH3

HN-R

AromaticFolded

MMP9; Mw: 18 kDa (no fibronectin)

- poor chemical dispersion in methyl region

- poor chemical dispersion in amide region

- broad resonance (=multiply conformations)

MMP12; Mw: 18 kDa

+ chemical dispersion in methyl region

+ chemical dispersion in amide region

+ sharp resonances (=unique conformation)

Unfolded

Simple experiment: requirements ~0.2mg of protein, Mw < 30kDa

• Protein can be reused after NMR (concentrate for crystallization?)

Folding assessment by NMR: situation often not black or white

Refolded protein: Tolloid like protease BMP1 and TLL12D HSQC spectrum: folding seen by NMR

N

O

N+

O

OO -

N

O

O

Cl

F

FF

NH

NH

O

OH

Cl

N

O

N+

O

OO -

N

O

O

Cl

F

FF

NH

NH

O

OH

Cl

N

S

N

N

NH

N

N

C

N

O

Cl

Requirements: ~0.5 mg of 15N labeled protein (E.Coli expression)

• Mw < 40 kDa

• Higher resolution (each peak corresponds to backbone amide)

• BMP1 has 200 amino acids

Bone Morphogenetic Protease 1

15N

1H

Refolded protein: Tolloid like protease BMP1 and TLL1Literature: Mac Sweeney et al., J. Mol. Biol. 384 (2008)

N

O

N+

O

OO -

N

O

O

Cl

F

FF

NH

NH

O

OH

Cl

N

O

N+

O

OO -

N

O

O

Cl

F

FF

NH

NH

O

OH

Cl

N

S

N

N

NH

N

N

C

N

O

Cl

Refolded proteins:

• Sequence identity: 87%

• Interpretation of TLL spectrum: two species (only minor fraction well folded)

Tolloid like MetalloproteaseBone Morphogenetic Protease 1

15N

1H

15N

1H

Refolded protein: Tolloid like protease BMP1 and TLL1two species: quantification

N

N

O

N

NO

N N

N

N

O

N

NO

N N

N NH

FF

F Cl

OH

N

O

N+

O

OO -

N

O

O

Cl

F

FF

NH

NH

O

OH

Cl

N

O

N+

O

OO -

N

O

O

Cl

F

FF

NH

NH

O

OH

Cl

N

S

N

N

NH

N

N

C

N

O

Cl

Quantification of the protease preparation by biochemical assay

• Specific activity in U per mg (1umol substrate turned over per minute per mg of enzyme)

• Active-site titration (molarity by active site titration / molarity as protein)

- In our experience this is not straightforward: potent inhibitor required

Active site titration by NMR is very simple

• Spy molecule with Kd < 1uM (19F or 13C nice to have)

Spy + proteinfree spy

IC50=0.2uM

13C-HSQC

Tris

BRZ357

O

O

O NH

N

O

NH

C31

HO

Rest

50uM Spy

Refolded protein: Tolloid like protease BMP1 and TLL1ONLY 15% OF TLL1 IS FOLDED

N

N

O

N

NO

N N

N

N

O

N

NO

N N

N NH

FF

F Cl

OH

N

O

N+

O

OO -

N

O

O

Cl

F

FF

NH

NH

O

OH

Cl

N

O

N+

O

OO -

N

O

O

Cl

F

FF

NH

NH

O

OH

Cl

N

S

N

N

NH

N

N

C

N

O

Cl

100% of BMP1 protein

preparation binds to Spy

Xtal grow spontaneous

• In Tris buffer at 4 °C

• Resolution 1.3 Å

Xtal obtained in PEG screen

• Screening conditions 11mg/ml at RT (1.4 Å)

Folded protein crystallizes out?

Seen this more often for refolded proteins

- presence of detergent, glycerol, arginine?

- mix of disulphide bonds?

50uM Spy

Only 15% of TLL protein

preparation binds to Spy

50uM Spy

Rather difficult case: Dengue Virus ProteaseD’Arcy et al., Acta Cryst. F62 (2006) and Erbel et al., Nat. Struct. Biol. 13 (2006)

Purity

Aggregation state

Stability

Structural order

Crystal contacts

Nucleation

Molecular biology - Protein construct

No Crystals

DLS=OK

SDS page = OK

SEC = OK

No degradation over time even at elevated

temperature (crystallization at 80mg/ml)

(Specific) Enzymatic activity

Cpd binding confirmed

~ NMR spectrum: Ugly, suggesting disordered regions

?

Dengue virus proteaseFirst assessment of structural order

N

N

O

N

NO

N N

N

N

O

N

NO

N N

N NH

FF

F Cl

OH

N

O

N+

O

OO -

N

O

O

Cl

F

FF

NH

NH

O

OH

Cl

N

O

N+

O

OO -

N

O

O

Cl

F

FF

NH

NH

O

OH

Cl

N

S

N

N

NH

N

N

C

N

O

Cl

R-CH3

Unfolded

HN-R

AromaticFolded

MMP9; Mw: 18 kDa (no fibronectin)

- poor chemical dispersion in methyl region

- poor chemical dispersion in amide region

- broad resonance (=multiply conformations)

OK, but uglyDengue2; Mw: 28 kDa

+ chemical dispersion in methyl region

- poor chemical dispersion in amide region

+ acceptable line shapes (e.q. Trp imidazole amide)

MMP12; Mw: 18 kDa

+ chemical dispersion in methyl region

+ chemical dispersion in amide region

+ sharp resonances (=unique conformation)

Construct optimized (further reduction makes construct unstable / inactive)

• Still NMR spectrum not great

• Ligand binds and affects protein dynamics / structure

=> No crystallization conditions found with the short construct

49 NS2B 95 EVKKQR↓AG 17 NS3pro 170

(33 amino acids removed)

49 NS2B 95 GGGGSGGGG 1 NS3pro 185

Reducing structural disorder of Dengue proteaseimprove construct, ligand binding, ...

Upon addition of Bz-Nleu-Lys-Arg-Arg-H

232 amino acids

Rather difficult case: Dengue Virus ProteaseWell diffracting crystals obtained with Lysine mutants

One crystallization condition found out of screens

• Diffraction is good (~1.7 Å)

• Novel protease structure solved

Blue-red: NS3 protease

Yellow: NS2B cofactor

Cofactor contributes b-strand to N-terminal b-barrel

49

82

18

167

49 NS2B 95 GGGGSGGGG 1 NS3pro 185Construct

Lys-> Arg

49 NS2B 95 EVKKQR↓AG 17 NS3pro 170Construct

optimization

49 NS2B 80 18 NS3pro 167Visible in

structure

Rather difficult case: Dengue Virus Proteasewe feel really smart ...

N

N

O

N

NO

N N

N

N

O

N

NO

N N

N NH

FF

F Cl

OH

N

O

N+

O

OO -

N

O

O

Cl

F

FF

NH

NH

O

OH

Cl

N

O

N+

O

OO -

N

O

O

Cl

F

FF

NH

NH

O

OH

Cl

N

S

N

N

NH

N

N

C

N

O

Cl

Protein construct: indeed very dynamic

• Quickly observed by NMR

• Optimal truncated construct made (see crystal structure)

• However, not the critical factor for crystallization of Dengue Protease

Introducing crystal contacts (7x Lys->Arg) did the trick !?

• Repetitive purification of Dengue protease Lysine mutants allow to optimize purification protocol

affinity – thrombine digestion at 4 °C - anion exchange - size exclusion chromatography

Better stop the story now, but for once we took the time to look back

• Can we obtain crystals with original construct using the improved purification protocol?

Dengue protease purificationoriginal construct - two different purifications

No Crystal Crystal

Rather difficult case: Dengue Virus Proteaseour learning, take home message

N

N

O

N

NO

N N

N

N

O

N

NO

N N

N NH

FF

F Cl

OH

N

O

N+

O

OO -

N

O

O

Cl

F

FF

NH

NH

O

OH

Cl

N

O

N+

O

OO -

N

O

O

Cl

F

FF

NH

NH

O

OH

Cl

N

S

N

N

NH

N

N

C

N

O

Cl

Critical for crystallization of Dengue protease: protein purification

• How clean needs a protein preparation be?

• SDS page does not show DNA / RNA / carbohydrates ....

Standard Operating Procedure: 3 purification steps

• Affinity purification (typically Ni-NTA)

• Ionic exchange (or HIC) (=removal of DNA/RNA)

• Size exclusion: only polishing step (=analytics, buffer exchange)

Rather difficult case: Cysteine proteaseno structural information: 56 kDa multidomain protein required for biochemical activity

N

N

O

N

NO

N N

N

N

O

N

NO

N N

N NH

FF

F Cl

OH

N

O

N+

O

OO -

N

O

O

Cl

F

FF

NH

NH

O

OH

Cl

N

O

N+

O

OO -

N

O

O

Cl

F

FF

NH

NH

O

OH

Cl

N

S

N

N

NH

N

N

C

N

O

Cl

DLS=OK

SDS page = OK

SEC = OK

No degradation over time even at elevated

temperature (protein concentration up to 20mg/ml)

(Specific) Enzymatic activity

? Too big to study by NMR

? Sequence alignment indicate ‘unknown regions’

Purity

Aggregation State

Stability

Structural order

Crystal contacts

Nucleation

Molecular biology - Protein construct

No Crystals

?

Rather difficult case: Cysteine proteaselimited proteolysis: an efficient way to probe the structural order of protein

N

N

O

N

NO

N N

N

N

O

N

NO

N N

N NH

FF

F Cl

OH

N

O

N+

O

OO -

N

O

O

Cl

F

FF

NH

NH

O

OH

Cl

N

O

N+

O

OO -

N

O

O

Cl

F

FF

NH

NH

O

OH

Cl

N

S

N

N

NH

N

N

C

N

O

Cl

kDa

250

150

100

75

50

37

25

20

15

10

1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8

Room temperature 4deg 1. Marker

2. Control

3. O/N with a-chymotrypsin

4. O/N with trypsin

5. O/N with elastase

6. O/N with papain

7. O/N with subtilisin

8. O/N with EndoGlu-C

28 kDa

46 kDa

56 kDa

Limited proteolysis:

• Requirement ~ 1mg of protein

• Proteolysis profile of up to 18 different proteases (Hampton kit, see F. Villard for details)

• Digestion O/N at room temperature and 4 °C, typically at 1:1000 dilution

• Analysis cleavage product by SDS page and LC-MS

Rather difficult case: Cysteine proteasetruncated construct: only the first step.

Limited proteolysis – option 1:

• Size exclusion chromatography to polish digested protein

• Crystallization trials with digested protein

fastest way to change protein construct

• Crystals only obtained in presence of substrate analogue

- substrate induces dimers

no use for drug discovery

Limited proteolysis – option 2:

• Generate new constructs based on results of limited proteolysis

• Issue: protein only expresses as monomer (expression system Baculo virus expression)

Truncated construct (46kDa) revived interest in biophysical studies

• Protein expression in E.coli explored

• Soluble expressed protein in high yield obtained for truncated construct

Rather difficult case: Cysteine proteasepurification of truncated construct

SOP purification: interesting IEX profile

Observations:

• No crystals obtained with monomeric protein (used for binding studies by NMR/SPR)

• Intensity of dimeric peak C depends on protein yield

- E.coli > Baculovirus expression system (dimeric signal overlooked)

A

tag

B

C

IEX on Source 15Q

A B

C

10000.00 20000.00 30000.00 40000.00 50000.00 60000.00 70000.00 80000.00m/z0

100

%

1:TOF MS ES+

9.8e+0051:

45639.3

22819.7

15213.111409.8

15227.1 22839.9 43237.230426.1

45681.5

45718.368459.048957.4 60852.3 76064.9

79868.0

45639 Da

(Analytical) SEC

Monomer

dimer

Dimeric form: crystals obtained in absence of substrate analogue

Rather difficult case: Cysteine proteasestructural of novel protease solved

N

N

O

N

NO

N N

N

N

O

N

NO

N N

N NH

FF

F Cl

OH

N

O

N+

O

OO -

N

O

O

Cl

F

FF

NH

NH

O

OH

Cl

N

O

N+

O

OO -

N

O

O

Cl

F

FF

NH

NH

O

OH

Cl

N

S

N

N

NH

N

N

C

N

O

Cl

Key steps:

Improving structural order:

• Limited proteolysis: C-terminal truncation of ~100 amino acids

• Focus on dimeric form

Expression and Purification:

• Changing the expression system

• Ion Exchange chromatography

Which parameters are critical for protein crystallization?Can we measure those parameters? Can we affect those parameters?

Monodisperse

Purity

Protein

stabilityConstruct design

Structural order

crystal

contacts

Nucleation

Analytics

Biophysics

Rational approach

Quality assessment

Change weak points

Increase chance of crystallization

- experience

- surprises

No Crystals

what next?

Predict crystallization = NO

SnowCrystals.com

Biophysical characterization for compound bindingsupport crystallization and drug discovery

Compound binding to support crystallization

• Compound binding can alter protein conformation and dynamics.

• This can affect crystallization behavior.

Drug discovery: protein crystallization in industry requires complex structures

• IC50 sufficient for compounds selection?

• Early HTS hits, Fragment based screening hits

Critical factors to select compounds for crystallization:

• Solubility - NMR application

• Binding (yes/no) and Binding pocket - NMR application

• Affinity (Kd), Specific or Unspecific binding - SPR application

N

N

O

N

NO

N N

N

N

O

N

NO

N N

N NH

FF

F Cl

OH

N

O

N+

O

OO -

N

O

O

Cl

F

FF

NH

NH

O

OH

Cl

N

O

N+

O

OO -

N

O

O

Cl

F

FF

NH

NH

O

OH

Cl

N

S

N

N

NH

N

N

C

N

O

Cl

Biophysical characterization for compound bindingsolubility determination in buffer

Soluble compound: ~55uM

Compound is micelle at nominal 200uM

Poorly soluble compound: <5uM

Compound solubility by NMR

• reliable method in our experience

• Acceptable throughput (30min / sample)

• Solubility can be determined >5uM

• Good range of buffers (including detergents)

Impact of solubility determination

• ratio IC50 vs solubility

- problematic if solubility < IC50

- IC50 valid? What is the assay measuring?

• ‘absolute’ solubility

- Prioritization of compounds

- Design co-crystallization experiments

Compound solubility: low tech, high impact

Biophysical characterization for compound bindingbinding (yes / no) and where?

Bias toward NMR spectroscopy for determining compound binding

NMR ligand based methods like STD, Waterlogsy, relaxation filtered exp. (also as 19F NMR)

• Fine screening technologies, but not good enough to (de)validate hits for crystallization

• Neat applications as reporter set-up (provides Kd information)

Only way: NMR protein observed

• Very low false positives (experimental mistakes like changes in pH and DMSO concentration)

• Very low false negatives (low solubility in combination with weak affinity)

• Protein Mw <50 kDa

Big trick box to simplify NMR spectra – amino acid selective labeling

• Provides binding site information

N

N

O

N

NO

N N

N

N

O

N

NO

N N

N NH

FF

F Cl

OH

N

O

N+

O

OO -

N

O

O

Cl

F

FF

NH

NH

O

OH

Cl

N

O

N+

O

OO -

N

O

O

Cl

F

FF

NH

NH

O

OH

Cl

N

S

N

N

NH

N

N

C

N

O

Cl

Biophysical characterization for compound bindingbinding (yes / no) and where?

Amino acid selective labeling

• BEV and E.coli (in minimal growth medium supplemented with amino acids)

• Reduces signal overlap, increased resolution

All backbone 15NH-labeled Only Trp 15N labeled

O

N15

H

R2 N15

C R1

H

Biophysical characterization for compound bindingbinding (yes / no) and where?

Chemical shift perturbation:

• Binding confirmed for both compounds

Different chemical shift pattern:

• Class 1 and class 2 compounds bind in different pocket

Assignment of amino acid / pocket:

• Mutation of labeled amino acid

class 1

class 2

Methionine 13C labeled Cysteine 15N labeled

13C Methionine

Met007-> Ile

Met007

wild type

Category 2:

• NMR binding (yes)

• SPR: response

- slow kinetics, super stoichiometric

- Binding to surface?

Low priority for crystallization

Biophysical characterization for compound bindingAffinity (Kd), Specific or Unspecific binding

N

N

O

N

NO

N N

N

N

O

N

NO

N N

N NH

FF

F Cl

OH

N

O

N+

O

OO -

N

O

O

Cl

F

FF

NH

NH

O

OH

Cl

N

O

N+

O

OO -

N

O

O

Cl

F

FF

NH

NH

O

OH

Cl

N

S

N

N

NH

N

N

C

N

O

Cl

SPR = Slightly Plausible Results

Category 1:

• NMR binding (yes)

• SPR: 1-1 binding with KD

NMR and SPR aligned

First priority for crystallization

[uM]1000 6704403002001309060

Cat IKD = 275uM

0 200 400 600 800 1000

0

5

10

15

20

25

30

35

40

45

50

R^2 = 0.99

KD = 275uM

RU

[PKF054-108]

Xray: NoXray: Yes

Lessons learnt:

• SPR: strict filter for crystallization (solubility, cpd aggregation, non specific binding, affinity)?

• Combination of NMR and SPR links robustness, sensitivity and hit characterization (reversed strategy?)

>60%

~20%

Structural Science Unit of Protease Platform

Allan d’Arcy

Frederic Villard

Martin Renatus

Arnaud Decock

Aengus MacSweeney

Nicola Hughes

Daniela Vinzenz

Simon Ruedisser

Nikolaus Schiering

Christian Wiesmann

Biology Chemistry

Structural Sciences

and biophysics

sPoC’s

Integrated target knowledge

Lead optimization

Assays development & compound profiling

Crystal structures & crystallization panels

Protein production

Target validation

Hit finding

Focused libraries

Hit validation

DAs

Expertise Platform Proteases (EPP)an integrated approach

Some theorysolubility vs IC50 – specific vs unspecific

potent

IC50 weak

IC50 or Kd

0 50 500 (nM)

Compound soluble / insoluble

[cmp]1 100.1 100 R

esp

on

se

Specific

binding

Unspecific

interaction