an nmr-based pharmacometabonomic study of cyp3a4 activity

1
An NMR-based pharmacometabonomic study of CYP3A4 activity Variability in Drug Response - “One size does not fit all” 1 Institute of Food Research, Norwich Research Park, Colney, Norwich, UK 2 Department of Twin Research and Genetic Epidemiology, King's College London, London, UK 3 Micro Separations Group, Pharmaceutical Science Division, King's College London, UK Discrimination before/after St John’s Wort intervention We thank the British Biotechnological Science Research Council (BBSRC), Prof Tim Spector the director of the Department of Twin Research and Genetic Epidemiology at St Thomas Hospital in London, UK and the Twin Participants (TwinsUK) Gwénaëlle Le Gall 1 , Nilufer Rahmioglu 2 , James Heaton 3 , Norman Smith 3 , Ian Colquhoun 1 , Kourosh R Ahmadi 2 and Kate Kemsley 1 • Response to medication is highly variable, unpredictable, and at times fatal • “Personalised” treatment has the potential to increase efficacy and decrease toxicity if “response” can be predicted accurately [email protected] • A clear discrimination between pre and post samples is observed for both urine and plasma. Markers (not shown) include exo and endogenous compounds (quinine and derivatives, tyrosine, N- acetylated metabolites, pyruvate, acetate, glycine, etc.) The aim of the study - • Assemble a large cohort phenotyped for induced CYP3A4 activity with St. John’s Wort, a mild, herbal antidepressant - potent inducer of CYP3A4 • Obtain metabolite profiles and identify biomarkers for predicting CYP3A4 induction Intervention study, quinine as probe drug - Recruitment Goal: 400 healthy individuals (100MZ:300DZ) Start taking SJW Take Quinine Visit St. Thomas’ Hospital 1 st Day 14 th Day 15 th Day 14 days pre- urines prior to day 1 post- urines on day 15 1 H NMR spectra • 415 pre-urines • 412 post- urines • 315 pre-plasma • 272 post- plasma Figure 1. Cross-validated PLS-DA models based on urine (A) and plasma samples (B) before and after chronic St John’s Wort exposure for two weeks and acute intake of quinine 100 200 300 400 500 600 -3 -2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 Sample Y CV Predicted 1 (Class 1) cross validated predictions 6LV PLS-DA model 50 100 150 200 250 300 350 400 450 500 550 -1 -0.5 0 0.5 1 1.5 2 Sample Y CV Predicted 1 (Class 1) cross validated predictions 4LV PLS-DA model pre-dose samples post-dose samples Cross Validation: Random subset; 10 data splits 20 iterations Sensitivity 92% Specificity 89% Cross Validation: Random subset; 10 data splits 20 iterations. • Sensitivity 94% • Specificity 95% • Genetic and environmental factors affect variability of the response of Drug Metabolizing Enzymes (DME) in particular the the cytochrome P450 (CYP) Quinine response in post-urines Figure 3. Validated MLR models predicting the quinine ratio response of NMR (A) and UPLC data (B) based on respectively 9 buckets (buckets 101, 28, 144, 30, 124, 93, 14, 64 and 98) and 8 buckets (buckets 80, 93, 107, 153, 111,102,144 and 87); bucket 93: glycine and bucket 144: N-acetylated metabolites at 1.974 ppm; note that although r 2 is lower for the UPLC model, the permutation test shows that the model is more robust giving a p value < 0.005 0.5 1 1.5 2 2.5 3 3.5 0 0.5 1 1.5 2 2.5 3 3.5 P redicted Y Actual Y R 2 =0.27 p<0.05 Independent Test set Training set -0.5 0 0.5 1 1.5 -0.5 0 0.5 1 1.5 Log(actual quinine ratio) P redicted y Predicted versus actualplot,from 8-variate m odel(blue triangles = independenttestset) R 2 =0.21 p<0.005 A B Predicted Y Actual Y Predicted Y Log (actual quinine ratio) Independent test set Training set Both NMR and UPLC based quinine ratios can be predicted modelling profiles from pre-urines. The Pearson correlation coefficients (r 2 ) are high for this type of data and the p values are highly significant especially for the UPLC data MLR predictive models based on NMR (A) and UPLC (B) quinine ratio MLR: Multi Liner Regression NMR: Nuclear Magnetic Resonance; PLS-DA: Partial Least Square-Discriminant Analysis; UPLC: Ultra Performance Liquid Chromatography Main outcomes: It is possible to detect urinary and plasmatic responses to St John’s Wort and quinine by 1 H NMR More importantly good prediction of the CYP3A4 induction response can be obtained using the healthy individual’s metabolite levels from pre-urine spectra • High resolution signals of quinine at 8.74 ppm and 3-OH quinine at 8.72 ppm were used to calculate 3OHQ/Q • UPLC measurements of quinine and 3-OH quinine were performed on 367 samples Run on 600 MHz NMR spectrometer with cryoprobe 8.74 ppm ppm Figure 2. Example of 6 post-urine NMR spectra 8.72 ppm 3-Hydroxy-Quinine Quinine A B

Upload: bradley-barrera

Post on 02-Jan-2016

36 views

Category:

Documents


3 download

DESCRIPTION

2. 1.5. 1. Y CV Predicted 1 (Class 1). 0.5. 0. -0.5. -1. 50. 100. 150. 200. 250. 300. 350. 400. 450. 500. 550. Sample. pre-dose samples. post-dose samples. Quinine. 8.74 ppm. 3-Hydroxy-Quinine. 8.72 ppm. ppm. Figure 2. Example of 6 post-urine NMR spectra. - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: An NMR-based pharmacometabonomic study of CYP3A4 activity

An NMR-based pharmacometabonomic study of

CYP3A4 activity

Variability in Drug Response - “One size does not fit all”

1Institute of Food Research, Norwich Research Park, Colney, Norwich, UK 2Department of Twin Research and Genetic Epidemiology, King's College London, London, UK

3Micro Separations Group, Pharmaceutical Science Division, King's College London, UK

Discrimination before/after St John’s Wort intervention

We thank the British Biotechnological Science Research Council (BBSRC), Prof Tim Spector the director of the Department of Twin Research and Genetic Epidemiology at St Thomas Hospital in London, UK and the Twin Participants (TwinsUK)

Gwénaëlle Le Gall1, Nilufer Rahmioglu2, James Heaton3, Norman Smith3, Ian Colquhoun1, Kourosh R Ahmadi2 and Kate Kemsley1

• Response to medication is highly variable, unpredictable, and at times fatal

• “Personalised” treatment has the potential to increase efficacy and decrease toxicity if “response” can be predicted accurately

[email protected]

• A clear discrimination between pre and post samples is observed for both urine and plasma. Markers (not shown) include exo and endogenous compounds (quinine and derivatives, tyrosine, N-acetylated metabolites, pyruvate, acetate, glycine, etc.)

The aim of the study -• Assemble a large cohort phenotyped for induced CYP3A4 activity with St. John’s Wort, a mild, herbal antidepressant - potent inducer of CYP3A4

• Obtain metabolite profiles and identify biomarkers for predicting CYP3A4 induction

Intervention study, quinine as probe drug -

Recruitment Goal: 400 healthy individuals (100MZ:300DZ)

Start taking SJW

Take Quinine

Visit St. Thomas’ Hospital

1st Day 14th Day 15th Day

14 days

pre-urines prior to day 1

post-urines on day 15

1H NMR spectra• 415 pre-urines

• 412 post-urines • 315 pre-plasma

• 272 post-plasma

Figure 1. Cross-validated PLS-DA models based on urine (A) and plasma samples (B) before and after chronic St John’s Wort exposure for two weeks and acute intake of quinine

100 200 300 400 500 600-3

-2.5

-2

-1.5

-1

-0.5

0

0.5

1

1.5

Sample

Y C

V P

red

icte

d 1

(C

lass

1)

cross validated predictions 6LV PLS-DA model

50 100 150 200 250 300 350 400 450 500 550-1

-0.5

0

0.5

1

1.5

2

Sample

Y C

V P

red

icte

d 1

(C

lass

1)

cross validated predictions 4LV PLS-DA model

pre-dose samples

post-dose samples

Cross Validation: Random subset; 10 data splits 20 iterations

Sensitivity 92% Specificity 89%

Cross Validation: Random subset; 10 data splits 20 iterations.

• Sensitivity 94%• Specificity 95%• Genetic and environmental factors affect variability of the response of

Drug Metabolizing Enzymes (DME) in particular the the cytochrome P450 (CYP)

Quinine response in post-urines

Figure 3. Validated MLR models predicting the quinine ratio response of NMR (A) and UPLC data (B) based on respectively 9 buckets (buckets 101, 28, 144, 30, 124, 93, 14, 64 and 98) and 8 buckets (buckets 80, 93, 107, 153, 111,102,144 and 87); bucket 93: glycine and bucket 144: N-acetylated metabolites at 1.974 ppm; note that although r2 is lower for the UPLC model, the permutation test shows that the model is more robust giving a p value < 0.005

0.5 1 1.5 2 2.5 3 3.50

0.5

1

1.5

2

2.5

3

3.5

Predicted Y

Act

ual Y

R2 =0.27p<0.05

Independent Test set

Training set

-0.5 0 0.5 1 1.5

-0.5

0

0.5

1

1.5

Log(actual quinine ratio)

Pre

dic

ted

y

Predicted versus actual plot, from 8-variate model (blue triangles = independent test set)

R2 =0.21p<0.005

A B

Predicted Y

Act

ual Y

Pre

dict

ed Y

Log (actual quinine ratio)

Independent test set

Training set

• Both NMR and UPLC based quinine ratios can be predicted modelling profiles from pre-urines. The Pearson correlation coefficients (r2) are

high for this type of data and the p values are highly significant especially for the UPLC data

MLR predictive models based on NMR (A) and UPLC (B) quinine

ratio

MLR: Multi Liner Regression NMR: Nuclear Magnetic Resonance; PLS-DA: Partial Least Square-Discriminant Analysis; UPLC: Ultra Performance Liquid Chromatography

Main outcomes:• It is possible to detect urinary and plasmatic responses to St John’s Wort and quinine by 1H NMR

• More importantly good prediction of the CYP3A4 induction response can be obtained using the healthy individual’s metabolite levels from pre-urine spectra

• High resolution signals of quinine at 8.74 ppm and 3-OH quinine at 8.72 ppm were used to calculate 3OHQ/Q

• UPLC measurements of quinine and 3-OH quinine were performed on 367 samples

Run on 600 MHz NMR spectrometer with cryoprobe

8.74 ppm

ppm

Figure 2. Example of 6 post-urine NMR spectra

8.72 ppm3-Hydroxy-Quinine

Quinine

A B