Post on 13-Feb-2017
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