Evolving Bioanalytical Methods for the Cardiovascular Drug Bosentan

D. Dell 1. / B. Lausecker2 / G. Hopfgartner 2 / P.L. M.van Giersbergen 3/J. Dingernanse 3 1 Consultant forActelion Pharmaceuticals Ltd.,4123 AIIschwil,Switzerland 2 Department of Non-Clinical Drug Safety, F.Hoffmann-La Roche, Basel,Switzerland 3 Department of Clinical Pharmacology, Actelion Pharmaceuticals Ltd.,AIIschwil,Switzerland

Experimental

Key Words Column liquid chromatography-tandem mass spectrometry Column switching Bioanalytics Bosentan

Summary The history of the development of analytical methods for the determination of the cardiovascular agent bosentan and its three major metabolites in biological fluids is described. Before the advent of LC-MS-MS,conventional HPLC-UV methodology was adequate for the determination of the drug alone in samples generated from toxicology studies.The progressive changes in the methods described reflect improvements in MS instrumentation and software, the need for measuring metabohtes, as well as the availability of deuterated internal standards. Finally, three LC-MS-MS methods were chosen as being suitable for routine use. The most sophisticated of these (E), involving protein precipitation, liquid-liquid extraction, and on-line solid phase extraction by automated column switching, allowed the simultaneous and sensitive determination of the drug and three metabolites in a variety of biological matrices from several species to support pharmacokinetic studies.Two further, simpler, MS methods were developed (Fand G) which, compared with the sophisticated method, focussed on reduced manual effort and reduced instrumental run cycle times. The limits of quantitation were belween 1 and 2 ng mL 1 for bosentan and the three metabolites.

HPLC-UV When methods first needed to be developed, in the early 1990s', LC-MS-MS was not available to the bioanalytical department at Roche, charged with the project. A conventional approach, involving liquid/liquid extraction of bosentan at basic pH, followed by reverse-phase HPLC and UV detection was applied. This first method had a quantification limit (LLOQ) of 25 ng mL 1. A solid phase extraction step was added later to improve the LLOQ to 5 n g m L 1. The method was adequate to monitor drug concentrations from nonclinical studies, and from clinical studies in which high doses (500 mg) were administered. The mean precision and inaccuracy were, respectively, 8.2% and 1.6% at 5.0 ng mL 1,2.5% and 1.5% at 20 ng mL 1, and 1.4% and 5.1% at 100.0 ng mL 1L. % ,o f ~ / NI@N~ooS"N

Introduction Bosentan (Figure 1) is a non-peptidic dual endothelin receptor antagonist, which is under development for the treatment of pulmonary arterial hypertension and

Presented at: 14th International Bioanalytical Forum: Sensitive Bioanalysis in Anti-cancer and other Drug Areas, Guildford, UK, Jul 3 6, 2001 Original 0009-5893/00/02

chronic heart failure [1]. To support both non-clinical and clinical pharmacokinetic studies, analytical methods were needed for the determination of the drug and its three major metabolites in plasma and other biological matrices (Figure 1). The development period included the advent of LC-MS-MS in the Roche bioanalytical department and the rapid technical improvements during the early years of this technique.

R3

$ 03.00/0

R3

Compound Bosentan IS Bosentan

R1 CH3 CH3

R2 R 3 IM+H] + DCH 3 H 552 DCH 3 2H 556

Metabolitel IS Metabolite I Metabollte I1 IS Metabolite El Metabollte I]] IS Metabolite m

CHzOH CHzOII [CH 3 liCH3 [CHeOH i

0CH 3 H DCHa 2H DH H OH ~tt DH !H

]CH2OH OH

~H

568 572 538 542 554 558

Figure 1. Structure of the analytes.

Chromatographia Supplement Vol. 55, 2002 S-115- 05

112

9 2002 Friedr. Vieweg & Sohn Verlagsgesellschaft mbH

S-115

Table I. Summary of major characteristics of LC-MS-MS methods for bosentan. Method (with C- 18 column)

Analytes

Species

Matrix

Sample preparation

LC-MS-MS, A LC-MS-MS, B LC-MS-MS, C LC-MS-MS, D

B B B, I, II B, I, II

man man man man

plasma plasma plasma plasma

LC-MS-MS, E

B,I,II,III

man, dog, rat

LC-MS-MS, F

B,I,II,III

man, rat

Plasma, / serum, bile, liver plasma /

LC-MS-MS, G

B,I,II,III

man

plasma

pp

1/1

/ / / /

Calib range ng mL 1

LLOQ ng mL 1

SPE

IS

pH 11 pH 11 pH 2 pH 4

/ on line

pH4

,/online

Re 47-8761 DB DB DB for B; Ro47-8761forI,II DB, DI, DII, DIII

0.5 200 0.5 200 2 1000; 200-20,000 2 1000for B; 5 1000for I,II B: 1 10,000; I,II,III: 2 2000

0.5 0.5 2.0 2.0 for B; 5.0forI,II B: 1.0; I,II,III: 2.0

DB, DI, DII, DIII DB, DI, DII, DIII

1 2000

1.0 for all analytes 2.0 for all analytes

/

/ off line

,/online

B:2 10,000; I,II,III: 2 2000

pp: protein precipitation; 1/1:liquid/liquid extraction; SPE: solid phase extraction (always C-18); IS: internal standard; DB: deuterated bosentan; B: bosentan; I, II, III are the metabolites (see Figure 1); DI, DII, DIII are the deuterated metabolites.

LC-MS-MS Methods As doses were lowered in the clinic, it soon became apparent that an L L O Q of 5 ng m L 1 was inadequate. In addition, it was also deemed necessary to determine the three metabolites. Thus, all efforts were directed towards the development of an L C - M S - M S method, since we considered that this methodology presented the best chance of achieving these objectives within the shortest possible time. All methods described here used tandem MS in the electrospray ionisation positive ion mode, 2 m m narrow bore R P columns, and selected reaction monitoring. The characteristics of the various LCM S - M S methods described below are summarised in Table I. This table also shows the calibration ranges, within which the inter-assay precision and inaccuracy of the various methods were determined. F o u r of these methods have been published in detail elsewhere: Method A [2], and methods E, F, G [3].

MethodA The objective of the first L C - M S - M S method was the quantitation of bosentan only, to an L L O Q of 0 . 5 n g m L 1 in hum a n plasma, with an upper limit of quantification of 200 ng m L 1. The work up procedure involved protein precipitation, liquid/liquid extraction (dichloromethane) at basic pH, and isocratic chromatography. A run time of 3 to 4 minutes allowed the overnight analysis of about 100 samples. An analogue of bosentan, in which the tertiary butyl side chain was replaced by cyclopropyl (Re 47-8761), was used as internal standard. The mean precision and S- 116

inaccuracy ranged between 3.9% and 6.9%, and 0.5% and 5.7%, respectively.

Method B F o r this method, tetra-deuterated bosentan was synthesized for use as internal standard. In addition, an improved collision cell became available, which allowed better fragment ion resolution. In all other respects this method was identical to method A. The mean precision and inaccuracy ranged between 3.5% and 6.0%, and 4.6% and 6.5%, respectively.

Method C This method allowed, in addition to bosentan, the determination of the hydroxy metabolite (metabolite I, Figure 1) and the phenol metabolite (metabolite II, Figure 1) in human plasma. Deuterated bosentan was used as internal standard for all three analytes. Protein precipitation was followed by liquid-liquid extraction (dichloromethane) at pH 2.0, the acidic p H being necessary because of the acidic phenol metabolite. A two-step gradient was used for the chromatographic separation, using eluents, which were mixtures of 1) a m m o n i u m acetate/acetic acid, and 2) acetonitrile/methanol/ammonium acetate/acetic acid. The L L O Q was 2 ng m L 1 for all three analytes. The inter-assay precision and inaccuracy for bosentan ranged from 4.3% to 12.5% and 1.3% to 4.3%, respectively. For the two metabolites, the inter-assay precision and inaccuracy ranged from 13.2% to 15.2% and 0.5% to 9.3% (for I), and from 10.1% to 16.8% and 0.8% to 2.0% (for II), respectively.

Chromatographia Supplement Vol. 55, 2002

Method D In an attempt to improve the precision and accuracy of method C, the extraction procedure was modified, and two internal standards were used. Deuterated bosentan was used as internal standard for bosentan, and the cyclopropyl analogue, R e 47-8761, for the two metabolites. A mixture of dichloromethane and n-chlorobutane was used for extraction at the higher pH of 4.0. This was followed by on-line solid phase extraction and automated column switching, using eluents consisting of mixtures of acetonitrile/methanol/ammonium acetate/acetic acid. The L L O Q was 2 . 0 n g m L 1 for bosentan, and 5.0ng m L 1 for metabolites I and II. The interassay precision and inaccuracy for bosentan ranged from 1.6% to 5.3% and 1.7% to 9.1%, respectively. F o r the two metabolites, the inter-assay precision and inaccuracy ranged from 5.8% to 8.6% and 2.0% to 9.8% (for metabolite I), and from 7.5% to 10.6% and 2.0% to 12.8% (for metabolite II), respectively.

Method E At this stage of development, deuterated internal standards became available for all three metabolites. Method E is the most widely applicable procedure, and can be used for the determination of drug and all three metabolites in plasma, serum, bile and liver from man, dog, and rat. The extraction and chromatographic conditions were as described for method D. Calibration samples were always made up in plasma, but the quality control samples were prepared in the same matrix as the incurred samples to be analysed. The L L O Q was 1 . 0 n g m L 1 for bosentan, Original

and 2.0 n g m L 1 for metabolites I, II, and metabolite III, which is the hydroxyphenol substance (Figure 1). The inter-assay precision and inaccuracy for bosentan ranged from 1.1% to 9.6% and 0.0% to 8.2%, respectively. For the metabolites, the inter-assay precision and inaccuracy ranged from 1.1% to 3.4% and 0.1% to 7.3% (for I), from 0.3% to 10.1% and 0.2% to 7.3% (for II), and from 1.2% to 9.6% and 0.1% to 6.4% (for III), respectively.

Method F With outsourcing in mind, and mindful of the fact that automated off-line solid phase extraction (SPE) is widely used by contract research organisations, it was decided to develop such a method. Sample preparation involved addition of deuterated internal standards to the plasma sample, followed by protein precipitation with methanol. The supernatant was applied to SPE cartridges. After washing the cartridge with water, phosphoric acid, water/ methanol, and water again, the analytes were eluted with acetonitrile/triethylamine at pH 10. The whole procedure was automated using a Zymark RapidTrace SPE system. The eluate was evaporated and reconstituted in acetonitrile/ammonium acetate. Isocratic chromatography was performed on an end-capped reversephase column with an acetonitrile/methanol/ammonium acetate/acetic acid mixture as mobile phase. The method was applied to the determination of bosentan and the three metabolites in human, dog, rat, and mouse plasma. The LLOQ was 1.0ngmL 1 for bosentan and the three metabolites. The inter-assay precision and inaccuracy for bosentan ranged from 1.7% to 6.3% and 1.2% to 5.0%, respectively. For the metabolites, the inter-assay precision and inaccuracy ranged froln 1.5% to 1.9% and 1.0% to 4.1% (for I), from 2.0% to 4.8% and 1.4% to 3.8% (for II), and from 1.3% to 6.5% and 2.3% to 5.3% (for III), respectively.

Method G Finally, a method was developed with the objective of minimising the sample handling effort, in particular for the analysis of plasma samples from clinical studies. Sample preparation consisted only of protein precipitation with an acetonitrile/ ethanol mixture which contained the deuterated internal standards. After centrifugation and evaporation of the supernatant, the residue was dissolved in an acetoOriginal

nitrile/ammoniumacetate/acetic acid mixture prior to analysis. Automatic column switching controlled the following steps. The analytes were first retained on a trapping column (SPE), while endogenous compounds were washed out with methanol/water and then ammonium acetate/ acetic acid. The analytes were subsequently transferred to the narrow bore analytical column for separation, using two eluent components consisting of a mixture containing different proportions of acetonitrile/methanol/ammoniumacetate/acetic acid. The LLOQ was 2.0ng mL 1 for bosentan and the three metabolites. The inter-assay precision and inaccuracy for bosentan ranged from 1.1% to 3.0% and 0.8% to 9.4%, respectively. For the metabolites, the inter-assay precision and inaccuracy ranged from 1.7% to 2.1% and 0.2% to 8.8% (for I), from 1.0% to 2.4% and 0.6% to 7.9% (for II), and from 1.0% to 3.6% and 2.0% to 9.7% (for III), respectively.

The first method developed, for bosentan in plasma only (LLOQ 25 ngmL-1), used liquid/liquid extraction at pH 11 followed by HPLC with UV detection, and was satisfactory for the monitoring of drug plasma levels in samples from toxicology studies. A further refinement of the method involving the addition of a solid phase extraction step to the work up, improved the LLOQ to 5 n g m L -1 . However, it became apparent that an LLOQ of 2 ngmL -1 or lower was necessary to obtain useful pharmacokinetic information from clinical trials. In addition, the pharmacokineticists requested the simultaneous determination of metabolites I, II, and III. These objectives were achieved by the application of tandem mass spectrometric detection with selected reaction monitoring, in combination with narrow bore reverse phase chromatography. A number of LC-MS-MS methods were developed, the differences between the procedures reflecting improvements in instrumentation and software, availability of deuterated internal standards, the need to measure metabolites, and the need to have procedures which other laboratories could take over with minimum effort. The first LC-MS-MS method (A), was, in fact, the most sensitive, with an LLOQ for bosentan of 0.5ngmL -1. This was achieved only by the use of a laborious

sample work up procedure, involving protein precipitation and liquid/liquid extraction, and by restricting the method to the determination of bosentan only. The eventual replacement of a structural analogue (Ro 47-8761) as internal standard (IS) by deuterated bosentan did not improve the precision and accuracy significantly (method B). The first method (C) that allowed simultaneous determination of metabolites I and II required a lowering of the extraction pH to 2.0, in order to recover these metabolites; fortunately, the recovery of bosentan is not very sensitive to pH. Gradient elution was necessary to achieve an acceptable precision and accuracy. At this stage, deuterated bosentan was available, and this was used as IS for all three analytes. The inferior LLOQ ( 2 n g m L 1 for all three analytes) compared to method A was probably a consequence of not having deuterated analogues as IS for the two metabolites, as well as the need to extract at pH 2. However, as later experience showed, an LLOQ of 2 ng mL 1 was adequate for most pharmacokinetic purposes. Increasing the pH for extraction from 2.0 to 4.0, and using Ro 47-8761 as IS for metabolites I and II (keeping deuterated bosentan as IS only for the parent compound) (Method D) did not improve the precision and accuracy of the method significantly. The real breakthrough came when deuterated analogues became available for all three metabolites (Method E). Using the same extraction and chromatographic conditions as for method D, it was possible to achieve LLOQs of 1.0 ng mL 1 and 2.0ngmL 1 for drug and metabolites, respectively, as well as better precision and accuracy as compared with methods C and D. In addition this method could be applied to several species and matrices. It became necessary to increase the capacity to cope with the analysis of the thousands of samples emanating from clinical studies. Outsourcing is an option increasingly used by pharmaceutical companies to achieve this aim. The methods described were based mainly on automated on-line column switching, a technology which the Roche bioanalytical department had built up and refined over many years. It was apparent, however, that many contract research organisations preferred to use, and had more experience of, automated offline solid phase extraction techniques. It

Chromatographia Supplement Vol. 55, 2002

S-117

Discussion

Table II. Sample throughput for the three methods of choice.

MethodE Method F Method G

preparation time per sample (min)

LC-MS-MS instrument run time per sample (min)

Number of samples which can be prepared by one technician per working day

Overall instrument run time (h)

4 13a 3

9 4 11

100 120 100 150

15to 18 7 27

method F used a sequential robot for sample preparation.

S.0xl0 s

Bosentan

4.0x10 s

Method

E

3.0xlO s

I

2.0x10 s

1.0xl0 s

0.0-

9

'

i

9

4

"u

i

6 Time (miu)

io

Bosentan 8.0xlO 4

I

Method

G

6.0x10 4

II]~

J 4.0x104

2.0xl0 4

0.0

J

0 7.0xllP 6.0x10 4

1

9

i

-

2

J

3

9

i

4

9

i

9

-i - i,

5 6 Time (min)

9

t

9

$

99 9

IlI I

'1

-

10

f U

Bosentan Method

F

5.0xl0 ~ .~" 4.0x10 ~

1

3.0x10

4

2.0x10

~

1.0xl0

4

0.0

i

4

;

;

io

During the development of the sucessive methods, different protein precipitation reagents were used, and the liquid/liquid extraction solvent was modified. At the beginning, acetonitrile was used for protein precipitation. This method often results in a clumping of the denatured protein, which might give rise to poor recovery of the analyte. The reagents used later, ethanol/acetonitrile mixtures or methanol, result in more homogenous precipitates and, probably, more efficient analyte recoveries. In addition, separation of the supernatant from the denatured protein precipitate is easier when alcoholic reagents are used, compared to acetonitrile. The change from the use of dichloromethane alone as extraction solvent, to dichloromethane/n-chlorobutane mixtures gave rise to cleaner extracts with the latter mixture, probably as a result of the decreased polarity of the mixed solvent, compared with dichloromethane alone. In Table II, the sample throughput is compared among the three methods of choice. It is clear, that the off-line SPE method, combined with isocratic chromatography, is the most efficient of the three procedures. Total ion chromatograms for methods E, F, and G are shown in Figure 2. The co-elution of metabolites I and III for method F, and the incomplete resolution between I and III for methods E and G is immaterial, because different daughter ions are detected in the selected reaction mode for these two metabolites.

Time (miu) Figure 2. Total ion chromatograms in selected reaction mode of calibration samples containing

200ngmL 1 ofbosentan and its metabolites processed with methods E, F, andG. was decided to make such a method available (Method F) for outsourcing. An advantage of this approach was that it reduced the running time on the expensive triple stage MS instruments, compared with on-line column switching methodology, from 11 to 4 minutes. In addition, an improved LLOQ ( l n g m L 1) was obtained for the metabolites.

S- 118

Because most of the methods described were rather time consuming as far as sample work up was concerned, a procedure was developed (Method G) which obviated liquid/liquid extraction and used only protein precipitation followed by online column switching. This method was developed for plasma samples only, with an LLOQ of 2 ng mL ] for all analytes.

Chromatographia Supplement Vol. 55, 2002

Conclusions Of the seven procedures described, the last three, methods E, F, and G, are the methods of choice when bosentan and its metabolites have to be determined, with method A as an alternative when bosentan alone has to be analysed at the lowest concentration possible (0.5 n g m L 1). 1) Method E, using liquid-liquid extraction and column switching LC-MS-MS Original

can be applied to several species and matrices and is, therefore, the most versatile procedure. 2) Method G is a simpler version of method E, involving less sample handling, in which the liquid-liquid extraction step is replaced by sophisticated column switching; this method is for plasma samples only. 3) Method F represents an off-line solid phase extraction alternative, allowing reduced MS run times compared to Methods E and G.

Original

Acknowledgement

References

Figure 2 is reprinted from Journal of Chromatography B, 749, B. Lausecker, B. Hess, G. Fischer, M. Mueller, G. Hopfgartner, 'Simultaneous determination of bosentan and its three major metabolites in various biological matrices and species using narrow bore liquid chromatography with ion spray tandem mass spectrometric detection', 67 83, copyright 2000, with permission from Elsevier Science.

[1] Roux, S.; Breu, V.; Ertel, S.I.; Clozel, M. J. Mol. Med. 1999, 77, 364 376. [2] Lausecker, B.; Hopfgartner, G. J. Chromatogr. A 1995, 712, 75 83. [3] Lausecker, B.; Hess, B.; Fischer, G.; Mueller, M.; Hopfgartner, G. J. Chromatogr. B 2000, 749, 67 83.

Chromatographia Supplement Vol. 55, 2002

Received:Jun 21,2001 Revisedmanuscript received:Aug 21,2001 Accepted: Sep 7, 2001

S-119