PhysicochemicalProfiling for Early Drug Discovery at

UCB

Barbara MasonUCB, Slough, UK

ACDLabs 2nd AnnualPhysChem Symposium

Obernai, 20th October 2005

Target ID Target Validation HTS Hit 2 Lead Lead

Optimisation

Later stage research is supported by recognised, industry standard methods which are capable of generating gold standard data but on a much reduced number of compounds, to assist in candidate selection

Physical Chemistry work in collaboration with DMPK colleagues to provide a full data package for every compound using systems which are directly relevant to the assay system used to generate potency and selectivity data.

Discovery PhysChem, UCB Slough

• The assays we routinely carry out

• Some that we are developing

• Data

GI PAMPA(all NCEs)

BBB PAMPA(all NCEs)

Caco-2(request)

Permeability

SnapSol(request)

BioSol(all NCEs)

AKAS(all NCEs)

QSol(request)

TSol(request)

Solubility

HT Log D7.4(all NCEs)

BIO-PAMPA(all NCEs)

HT pKa(all NCEs)

In Progress

10 mM DMSO Stock

PPB(request)

Log P(request)

pKa(request)

Discovery PhysChem (Slh)

PhysChem (Brne + Cam)

DMPK (Slh)

Solubility1. Intrinsic solubility – solubility of the neutral species regardless

of physiological relevance of pH (similar to Log P)

2. Unbuffered solubility – solubility of a saturated solution at whatever pH the solution ends up at (self – buffering)

3. Buffered solubility – solubility at a specific pH, 5 or 7.4 for example (similar to Log D – takes pKa into account)

4. Kinetic solubility – solubility of the fastest dissolving or fastest precipitating species

5. Thermodynamic solubility – equilibrium solubility of all species

Solubility – A Cascade of Assays Needed

• Greater Confidence in in-vitro and in-vivo assay results to avoid false negatives

• Poorly soluble compounds may undergo non-specific binding with proteins leading to false positives

• Measurement at multiple pH levels is more useful than single point in some circumstances

• Solubility – pH profile provides a better understanding of absorption through the pH gradient of the GI tract

• Kinetics, thermodynamics and DMSO content

GI PAMPA(all NCEs)

BBB PAMPA(all NCEs)

Caco-2(request)

Permeability

SnapSol(request)

BioSol(all NCEs)

AKAS(all NCEs)

QSol(request)

TSol(request)

Solubility

HT Log D7.4(all NCEs)

BIO-PAMPA(all NCEs)

HT pKa(all NCEs)

In Progress

10 mM DMSO Stock

PPB(request)

Log P(request)

pKa(request)

Discovery PhysChem (Slh)

PhysChem (Brne + Cam)

DMPK (Slh)

AKAS – Automated Kinetic Aqueous Solubility• All novel NCEs pass through this assay (~100/week)

• 10mM DMSO stock solutions are diluted in buffer at 4 pHs– 5% final DMSO concentration

• Samples are shaken for 90 minutes then filtered

• Concentration of sample in the filtrate is determined by UV plate reader against a calibration curve- Spectra measured from 240-400nm− λmax / isosbestic point

• Measuring at 4 pHs gives us a number of advantages– Any pH instability is flagged early in the discovery process– Approximate pKa values can be estimated by effects on solubility

ND 0 - 30 uM 30 - 150 uM 150 - 350 uM >350 uMpoor modest good

BioSol – Solubility under in-vitro conditions

• In a high throughput drug discovery screen, primary assay data are generated in enzyme or protein based assays

• Solubility of the compounds may be compromised by the presence of these proteins and incubation media

• Maximum quantity of sample present is very low• For cellular systems DMSO content has to be kept to a minimum

• This assay gives a handle on sample solubility at relevant DMSO concentration and also flags potential protein binding issues in a high throughput method which can then be studied using more traditional PPB methods by the DMPK group.

BioSol• All novel NCEs pass through this assay ~100/week

• 10mM DMSO stock solutions diluted in D-PBS– 0.2% final DMSO concentration– 10% bovine serum albumin– 10% glucose– MgCl2, CaCl2 (1µM)

• Control plate without protein

• Shaken for 90 minutes at 30oC then filtered

• Filtrate centrifuged through 10KDa cut off plates– 20 mins– 2000 rpm

• Analysis by HPLC

• Solubility data is returned for 0.2% DMSO along with an indication of potential protein binding issues

QuickSol – Pseudothermodynamic solubility

• 5% DMSO in AKAS is a rather high quantity of co-solvent.• True thermodynamic assay is expensive with respect to sample

quantity and slow.• QuickSol starts from solid material – 0% DMSO

– Solid material not necessarily present in excess (0.5mg / pH)– Single time point, not equilibrium

• Measured on selected compounds by request achieving >150µM at either pH 5 or pH 7.4 in AKAS

• Buffer added to solid material– gentle shaking for 90 mins– filtered

• Filtrate analysed by HPLC against a calibration curve

SnapSol – fragment screening single point solubility

• This has been developed as an aid to fragment screening to ensure that compounds meet minimum solubility criteria

• Can be used to screen a selection under conditions which mimic closely those of the screening deck

• 10mM DMSO stock solutions are diluted with buffer system used inscreening deck to give final DMSO concentration at a relevant level

• Shaken for 90 minutes then filtered

• Analysis of the filtrate by UV plate reader at fixed wavelength against a calibration curve

GI PAMPA(all NCEs)

BBB PAMPA(all NCEs)

Caco-2(request)

Permeability

SnapSol(request)

BioSol(all NCEs)

AKAS(all NCEs)

QSol(request)

TSol(request)

Solubility

HT Log D7.4(all NCEs)

BIO-PAMPA(all NCEs)

HT pKa(all NCEs)

In Progress

10 mM DMSO Stock

PPB(request)

Log P(request)

pKa(request)

Discovery PhysChem (Slh)

PhysChem (Brne + Cam)

DMPK (Slh)

Permeability• Industry wide, in-vitro cell based assays still extensively used to

predict permeability across phospholipid membranes• Advantages and disadvantages are widely known and debated

+ closer mimic to in-vivo systems+ transport mechanisms can be monitored+ varying cell lines can be used for more applicable assays– relatively low throughput (despite new 96 well format)– experimental errors are high– time consuming and expensive

• We have moved to artificial membranes as a means to measure passive diffusion using the PAMPA methodology+ fast and cheap– no information about active transport or efflux mechanisms

Permeability – GI PAMPA• All novel NCEs pass through this assay (~100/week)

• Phosphatidyl choline artificial membrane on solid filter support• 10mM DMSO stock solutions are diluted in buffer to 5%DMSO • Donor / acceptor sandwich incubated

– for 4 hours at 20oC with gentle shaking

• Ratio of sample in the donor and acceptor wells is measured by UV plate reader

• Membrane retention is calculated as a % from mass balance equations and the permeability is returned at 10-7cms-1

ND 0 - 50 50 - 150 >150poor modest good

PAMPA – Blood Brain Barrier• The Blood Brain Barrier is an important membrane – it has very

tight intercellular boundaries

• CNS active drugs must cross the BBB while to avoid undesirable side effects non-CNS active drugs must not– Particularly important to UCB targeting both CNS+ and CNS- disease

states

• We have therefore developed a PAMPA-BBB assay to indicate whether our compounds are CNS+ or CNS-.– Validated against known CNS active compounds

• Methodology is as for the traditional PAMPA assay, with the phosphatidyl choline membrane being replaced with porcine polar brain lipid

* L.Di, E.Kerns et al Eur. J. Med. Chem 38 (2003)223-232

GI PAMPA(all NCEs)

BBB PAMPA(all NCEs)

Caco-2(request)

Permeability

SnapSol(request)

BioSol(all NCEs)

AKAS(all NCEs)

QSol(request)

TSol(request)

Solubility

HT Log D7.4(all NCEs)

BIO-PAMPA(all NCEs)

HT pKa(all NCEs)

In Progress

10 mM DMSO Stock

PPB(request)

Log P(request)

pKa(request)

Discovery PhysChem (Slh)

PhysChem (Brne + Cam)

DMPK (Slh)

Thermodynamic solubility, pKa and Lipophilicity• No physical chemistry profiling would be complete

without thermodynamic solubility, pKa and Log P

– These are performed by colleagues in Braine (Belgium) and Cambridge (UK) on a request basis using their Gold Standard or traditional shake flask methods as part of the development characterisation

• For discovery screening these methods do not lend themselves easily to high throughput.

Can we address these issues with our current technology?

• Thermodynamic Solubility– While the QuickSol assay does not give a true thermodynamic

solubility, it does give a clear indication of DMSO co-solvent effects

• pKa– AKAS solubility gives a good handle on pKa (allowing for the DMSO

effect) since it is run across a pH range• In-house algorithms have demonstrated that it is possible to take the UV

data and use it to produce a better approximation of the pKa (not fully validated yet)

• ACDLabs pKa prediction software

• Lipophilicity– Higher throughput Log D 7.4 robotics based method is being developed

(with knowledge of pKa we can then determine a lipophilicity profile)

Data Package

• List of standard compounds– All well known drugs commonly used in the literature and well

validated

• Data we have measured in our assays

Impact of our data package on discovery projects

• No single piece of assay data should make or break a project– informed interpretation of physical chemistry data alongside

biological data

• Very fine line– But, must not cloud the issue with too much data– Although, much can be learned from large amounts of historical

data to use in future projects

• So, to demonstrate what information can we get from our cascade of assays…..

QuickSol BBB PAMPApH3 (uM) pH5 (uM) pH7.4 (uM) pH9 (uM) pH7.4 (uM) pH7.4 (uM) PPB? Pa (10-7 cm/sec) %R Pa (10-7 cm/sec)

Alprazolam ND >350 >350 >350 820 18 37 30Alprenolol >350 >350 >350 >350 ND 20 122 26Amiloride >350 >350 >350 >350 >1500 0 0

Astemizole >350 >350 68 13 0 14 125 63 19Atenolol >350 >350 >350 >350 >1500 7 0

Caffeine >350 >350 >350 >350 >1500 10 5Chlorpromazine >350 >350 >350 52 >1500 15 94 12Corticosterone >350 >350 >350 >350 526 4 ND

Desipramine >350 >350 >350 >350 >1500 16 222 28Dexamethasone >350 >350 >350 >350 275 18 0 16

Diazepam ND >350 >350 >350 172 19 175 16 25Diltiazem >350 >350 >350 >350 >1500 12 223 38Enoxacin >350 >350 >350 >350 >1500 1 18

Famotidine >350 >350 >350 244 >1500 0 0Furosemide 283 >350 >350 >350 >1500 13 Yes 0 0

Hydrocortisone >350 >350 >350 >350 1237 0 0Imipramine >350 >350 >350 >350 >1500 16 116 40 17Isoxicam 2 >350 >350 >350 >1500 12 Yes 0 0

Ketoconazole >350 >350 252 236 7 10 63 32 8Ketoprofen >350 >350 >350 >350 >1500 7 Yes 0 1

Labetolol >350 >350 >350 >350 >1500 18 0 78Metolazone >350 >350 >350 >350 109 0 3

Naproxen 164 >350 >350 >350 >1500 9 Yes 0 0Nifedipine >350 >350 >350 >350 20 20 133 12 13

Norfloxacin >350 >350 >350 >350 >1500 0 0Ofloxacin >350 >350 >350 >350 >1500 18 0 0

Oxazepam >350 >350 >350 >350 76 8 14 42Piroxicam >350 >350 >350 >350 >1500 11 Yes 10 6

Promazine >350 >350 >350 >350 >1500 17 193 19 12Propranolol >350 >350 >350 >350 >1500 17 109 32

Quinidine >350 >350 >350 >350 >1500 9 28Ranitidine >350 >350 >350 >350 >1500 0 0

Sulfasalazine 144 >350 >350 >350 >1500 0 0Sulpiride >350 >350 >350 >350 >1500 0 0

Tenoxicam >350 >350 >350 >350 >1500 14 Yes 6 2Terfenadine >350 >350 204 0 0 0 0 0Testosterone >350 >350 >350 >350 199 201 21

Theophylline >350 >350 >350 >350 >1500 0 0Timolol >350 >350 >350 >350 >1500 0 13

Trimethoprim >350 >350 >350 >350 1159 0 1Verapamil >350 >350 >350 >350 >1500 16 188 13 25

Warfarin >350 >350 >350 >350 >1500 6 Yes 2 1

GI PAMPAAKAS BioSol

Ketoconazole:

• Solubility assays show clear effect of DMSO

• Distinction between permeability assays and membrane retention also noted

AKASpH7.4 (uM)

252

N

N

O

OO

ON N

Cl

Cl

BioSol QuickSolpH7.4 (uM) pH7.4 (uM)

10 7

BBB PAMPAPa (10-7 cm/sec) %R Pa (10-7 cm/sec)

63 32 8

GI PAMPA

Warfarin:

• Solubility assays show no effect by DMSO (AKAS vs QuickSol)

• BioSol shows drop despite 0.2% DMSO and evidence of protein binding – consistent with literature

O

OH O

AKAS BioSol PPB? QuickSolpH7.4 (uM) pH7.4 (uM) pH7.4 (uM)

>350 6 yes >1500

Astemizole:

• Relatively low solubility while the PAMPA data is relatively high

• Can be explained by membrane retention– Lipophilic compound

N

N NH

N

O

F

AKASpH7.4 (uM) Pa (10-7 cm/sec) %R

68 125 63

GI PAMPA

pKa

pH3 (uM) pH5 (uM) pH7.4 (uM) pH9 (uM)2 >350 >350 >350

AKASIsoxicam

pH3 (uM) pH5 (uM) pH7.4 (uM) pH9 (uM)144 >350 >350 >350

AKASSulfasalazine

pH3 (uM) pH5 (uM) pH7.4 (uM) pH9 (uM)>350 >350 240 0

AKASTerfenadine

NS

O

NH

OH

OO

ON

N NH

SO O

NN

OH

O

OH

NOH

OH

pH Stability issues – from AKAS UV spectra

Alprazolam

pH3pH5

pH7.4pH9

N

NN

N

Cl

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

2

240 260 280 300 320 340 360 380 400

Wavelength / nm

Abs

orba

nce

0.5uM0.2uM

0.05uM0.01uM

pH3pH9

Abs

orba

nce

Wavelength / nm

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

2

240 260 280 300 320 340 360 380 400

Calibration Spectra

Wavelength / nm

Sample Spectra

And finally a word of Caution..

N

N N

N

O

O

GI PAMPA PAMPA-BBBPa 10-7cm/sec Pa 10-7cm/sec

10 5

Caffeine

O

O

N O

O

N

GI PAMPA Caco-2Pa 10-7cm/sec Pa 10-7cm/sec

188 263

Verapamil

“high” PAMPA, but “low” Caco-2 due to efflux pump

N

N N

NH

O

O

GI PAMPA Caco-2Pa 10-7cm/sec Pa 10-7cm/sec

0 447

Theophylline

“low” PAMPA, but “high” Caco-2 due to active transport

“low” GI-PAMPA and “low” PAMPA-BBB but known to be CNS+ due to active transport across BBB

Acknowledgements

• PhysChem, Slough, UK– Richard Taylor– Christine Prosser– Emily Freeman

• PhysChem, Cambridge, UK– John Cooper– Benedicte Fau– Dave Sherwood

• PhysChem, Braine, Belgium– Luc Quere– Liliane Ellens– Geraldine Longfils

• DMPK, Slough, UK– Ted Parton– Lloyd King– Hanna Hailu– Mark Baker– Sarah Bartlett– Simon Carter– Judith van Asperen

Bit of a Cheek…….

• Physical Chemistry Symposium

• PerkinElmer, Seer Green, Buckinghamshire • Wednesday 30 November 2005.• No registration fee

– Register at: info@physchem.org.uk• Bring a poster!!

• Who Should Attend?– Scientists who have been actively involved in

driving forward this area of the industry within their own institutions.