Automated Column Liquid Chromatographic Determination of a Basic

Automated Column Liquid Chromatographic Determination of a Basic Drug in Blood Plasma Using the Supported Liquid Membrane Technique for Sample ...
0 downloads 0 Views 862KB Size
Anal. Chem. 1994,66,4490-4497

Automated Column Liquid Chromatographic Determination of a Basic Drug in Blood Plasma Using the Supported Liquid Membrane Technique for Sample Pretreatment Boel Lindegird, Henrik Bjark, Jan Ake J&nsson,*Lennart Mathiasson, and Anne=MarieOlssono Department of Analytical Chemistry, University of Lund, P.O. Box 124,S-221 00 Lund, Sweden

The determination of the basic drug Amperozide, one of its main metabolites, and a third related compound in human blood plasma is described. All three compounds include an amine functionality. The process is governed by a moditled ASTED instrument (Gilson). The dialysis unit is exchanged for a supported liquid membrane (SLM), which gives a simultaneous cleanup and enrichment of the analytes. The ASI'ED is coupled to column liquid chromatographic equipmentwith W detection. An efficient cleanup is achieved, and there is no visible difference between the chromatograms after enrichment of a plasma blank and an aqueous blank. To increase the recovery, the same sample can be passed by the membrane several times, while the receiving phase is kept stagnant. An enrichment with three passages takes 1520 min. The limit of detection for Amperozide in blood plasma is then -30 ng/mL. The use of the SLM technique for estimation of the degree of protein binding is discussed. Preliminary results agree reasonably with measurements with other techniques. The optimization of the enrichment with supported liquid membranes is described and different membrane supports are compared. When biological samples are analyzed, a cleanup procedure is usually necessary before any chromatographic determination. The most frequently used methods are liquid-liquid solvent extraction, solid phase extraction, SPE, or protein precipitation.' The former often gives a good cleanup from the matrix, but is either laborious or difficult to automate, and occasionally problems with emulsions can occur. The use of large amounts of organic solvents should also be avoided for environmental and health reasons. SPE is a convenient method?a but suitable packing materials are not always available, and the cleanup is sometimes not sufficiently efficient. With protein precipitation the sample is diluted, which increases the detection limit,and many endogenous compounds are still present in the sample after the removal of the proteins. A new pretreatment method is the use of restricted 5 Bioanalytical Department, CNS, Pharmacia Therapeutics, $223 63 Lund, Sweden (1) Mehta, A. C. Talunta 1986,33,67-73. (2) Nielen, M. W. F.; Frei, R W.; Brinkman, U. A. T. In Selective sample handling and detection in high-pe~onnanceliquid chromatogrophy, Frei, R W., Zech, K, Eds.; Elsevier: Amsterdam, 1988; Chapter 1, p 5 (3) McDowall, R D.; Pearce, J. C.; Murkitt, G. S. Trends Anal. Chem. 1989,8, 134-140.

4490 Analytical Chemistry, Vol. 66,No.24, December 15, 1994

access packing^.^-^ In this method the sample is loaded on a precolumn where the large proteins are not retained. The analytes enter the small pores of the packing and can, after washing, be eluted with a suitable solvent. One way to utilize the efficient cleanup of liquid-liquid extraction with minimal amounts of organic liquid and still be able to automate is to use supported liquid membranes (SLM) Both basic and acidic compounds can feasibly be enriched with the SLM technique. The pH of the sample is adjusted so that the analyte compounds are uncharged. This permits them to pass through an organic liquid, immobilized in a porous membrane, into an aqueous solution on the other side. The pH in this solution is such that the analyte molecules are ionized and thus prevented from reentering the membrane. The process is chemically equivalentto liquid-liquid extraction, followed by backextraction to an aqueous phase with a different pH. Only small, uncharged molecules can pass through the organic liquid membrane, while macromolecules are excluded. A chromatogram obtained after workup of a blank plasma sample with the SLM technique is virtually identical to a blank chromatogram from an aqueous solution? The liquid membrane unit can easily be incorporated in a robotic equipment for automation. In this case, an instrument for automated sequential trace enrichment of dialysates (ASTED) was used, directly connected to column liquid chromatographic (CLC) equipment. The technique was evaluated for spiked blood plasma samples. The basic drug Amperozide, one of its main metabolites, and a third related compound (F'harmacia Therapeutics, Lund, Sweden) were chosen as Model substances. Amperozide is a psychopharmacon, which has shown promising properties in the treatment of persons suffering from schizophrenia.1° EXPERIMENTAL SECTION

Apparatus. A modified ASTED instrument (Gilson Medical Electronics, Villiers-le-Bel, France) was used for the sample pretreatment. The original dialysis membrane holder was ex(4) Beth, M.; Unger, K K; Tsyurupa, M. P.; Davankov, V. A. Chromatogruphia 1993,36, 351-355. (5) F'inkerton, T. J. Chromatogr. 1991,544, 13-23. (6) Unger, K K Chromatographia 1991,31,507-511. (7) Jonsson, J. A.;Mathiasson, L. Trends Anal. Chem. 1992,11, 106-114. (8) Audunsson, G. Anal. Chem. 1986,58, 2714-2723. (9) LindegHrd, B.; Jonsson, J. A,; Mathiasson, L. /. Chromatogr. 1992,573, 191-200. (10) Axelsson, R; Nilsson, A; Christensson, E.; Bjork, A. Psychopharmacology 1991,104, 287-292.

0 1994 American Chemical Society 0003-2700/94/0366-4490$04.50/0

Figure 1. Membrane unit. From ref 12, with permission from Elsevier Science Publishers B.V., Amsterdam.

changed for a custommade holder for the supported liquid membrane, shown in Figure 1. Two different holders were used, one made of poly(vinylidene dinuoride) O F ) and the other of titanium. The PVDF holder consisted of two blocks, which had identical machined grooves with the dimensions 0.1 x 2.5 x 40 mm3, forming a channel on each side of the membrane with a nominal volume of 10 pL each. The corresponding channels of the titanium holder had the dimensions 0.15 x 2.0 x 40 mm3, that is, 12 pL. The liquid membrane was placed between the two blocks, which were tightly clamped together with six bolts. A porous poly(tetratluoroethene) (P”FE)-membrane, Model TE35 from Schleicher and Schuell (Dassel, Germany), was used as support of the immobilized organic liquid. This membrane is supported by an integral polyester backing in the form of a thin felt. Five other membranes were also tested, and the physical parameters for all membranes are listed in Table 1. The membranes were cut to fit the holders (-6 x 46 “2) and soaked for 15 min in dihexyl ether (sometimes with trioctylphosphine oxide added as an extractant). The liquid membrane was mounted in the holder, and a washing sequence (regeneration 1 x 3 mL) was run to wash away excess solvent from the membrane surfaces. The orientation of the membrane is important, and in most cases, it was placed with the backing against the donor channel (see The Membrane). The original SPE cartridge in the ASTED was replaced with a stainless steel injection loop for direct injection onto the CLC equipment. Loops with a volume of either 22 or 60 pL were used. An additional volume is needed to house the 10p L acceptor plug, partly as dispersion occurs when the plug is transported to the loop. With the larger injection loop, higher peak heights were obtained, probably due to remaining substances that may be slow to wash out from the acceptor channel or the membrane.” Because of this, the 60 p L loop was used in most cases, when not otherwise stated. Possibly, dispersion also occurs during the enrichment, particularly as the volume of the acceptor channel can change with the pressure in the donor channel when the sample is alternatingly pumped and sucked. The equipment consisted of an HPLC pump (Model 2150, LKB, Bromma, Sweden), a guard column (10 mm x 2.0 mm i.d., packed

with 5 pm ODS W-18, Upchurch ScienMc Inc., Oak Harbor, WA), an analytical column (200 mm x 2.1 mm i.d., Mhalysentechnik, Mainz, Germany) packed with 5 pm Kromasil CU (Eka-Nobel AB, Bohus, Sweden), a variablewavelength UV detector (Spectroflow, Kratos, Ramsey, NJ), and a stripchart recorder (Model 2210, LI(B, Bromma, Sweden). The mobile phase flow rate was usually 0.20 mL/min, and the UV detector was set at 265 nm. Baseline separation of the compounds studied was obtained in -15 min. The guard column was added to protect the analytical column from plasma in case of a breakthrough of the liquid membrane. It was found to be very efficient, and it did not noticeably increase the band broadening. The mobile phase for the liquid chromatography was ammonium phosphate buffer (usually pH 7.8) in methanol (2278). To prevent bubble formation, the mobile phase was degassed with helium or by sonication. Operation. The operation of the system shown in Figure 2 is described in detail in ref 12. To a 0.5 mL plasma sample (1) is automatically added 50 pL of 2.5 mol/L NaOH and 0.45 mL of basic donor buffer (2). This gives a pH of 13, which neutralizes the amine drug molecules and decreases the viscosity, respectively. After mixing, a 800 pL aliquot is slowly pumped with the syringe pump (3) through the donor channel (4)of the membrane unit into a large coil (5). The uncharged amines can penetrate the liquid membrane (6) into the stagnant acidic acceptor phase (7), where they are protonated and trapped. To increase the recovery, the sample in the coil (5) can be sucked back through the donor channel for repeated extraction of the same sample. The donor channel is thereafter washed with the donor buffer, and the entire acceptor phase volume containing the trapped amines is pumped into the injection loop (8) of the HPLC and injected on the column (9). Finally, both the donor and the acceptor channels are washed with the buffers (2) and (10) (“regeneration”). The donor solution (2) contained 12.5 mmol/L EDTA (to keep the metals present in blood plasma in solution) and NaOH to give a pH of 10.0. The acceptor solution (10) contained 2.5 mmol/L HzS04 to give a pH of usually 2.5. Chemicals. Amperozide ( 4 [4,4bis(4fluorophenyl)butyl]-Nethyl-1-piperazinecarboxamide,I), a major metabolite (4[4,4bis(4fluorophenyl)butyl]-1-piperazinecarboxamide,II) , and a related compound ( 4 [4,4bis(4fluorophenyl)butyl]-N-butyl-1-piperazine carboxamide, III), were obtained as hydrochlorides from Pharmacia Therapeutics. Their purity was ~ 9 9 % .

‘a

n

l

CH(CH2)a N-N-C-R

I:

R = -NHCHzCH3

11: R = - N H z

They are all tertiary amines, and Amperozide has a pKa of 6.8. The partition coefficient between the organic liquid in the membrane and the donor phase, KD,was determined by shaking the donor buffer, containing the three compounds (20 pg/mL of each), with dihexyl ether and then injecting the aqueous phase into the liquid chromatograph. The values obtained were 110 for Amperozide, 6 for the metabolite, and 2200 for the third com(12) Jonsson, J.

(11) Shen, Y.; Gronberg, L.; Jtinsson, J. A. Anal. Chim. Acta 1994,292,31-39.

0 t

A.; Mathiasson, L;Lindeglrd, B.; Trocewicz, J.; Olsson, A-M.

J Chromatogr. A 1994,665, 259-268. Analytical Chemistty, Vol. 66, No. 24, December 15, 1994

4491

Table 1. InvestigatedMembranes

membrane

material

PTFE ETFE PTFE

TE35=

Fluoropore F G P fluoropore FHUP Durapore GVHP

material of backing polyester polyethene

thickness of membrane (um)

PVDF

SM 11807‘

PTFE PTFE

SpectrapoP

polypropylene

-60 60 60 125 65 75

thickness of backing (um) 180 115

-150

average pore size (um)

porosity

0.2 0.2 0.5 0.22 0.2 0.02

0.6-0.8 0.70 0.85 0.75 -0.70 0.50

Schleicher & Schuell GmbH, Dassel, Germany. Millipore Corp., Bedford, MA. Sartorius GmbH, GBttingen, Germany. Spectrum Medical,

Los Angeles, Ch

i---73W--1

5

U

Figure 2. Experimental setup. For details, see the text.

pound. The metabolite with its polar primary amine group is thus most difficult to extract into the liquid membrane. All three of them adsorb very strongly to all sorts of surfaces. They also exhibit a very high degree of protein binding; Amperozide is reported to be protein bound to 97%in human blood serum.13 Stock solutions of the drug compounds were prepared in water (200 pg/mL) and were stable for several months when kept in a refrigerator. All other solutionswere kept at ambient temperature. The organic solvent used was dihexyl ether ( S i a , St. Louis, MO) or a solution of 5% (w/w) of trioctylphosphine oxide (Fluka, Buchs, Switzerland) in dihexyl ether. Piperazine and triethylenetetramine were purchased from Fluka, and ammonia was from Riedel-de Ha&n(Seelze, Germany). All chemicals were of analysis grade or better. The water was purified using a Milli-Q unit (Millipore, Bedford, MA). Plasma Samples. Heparinized blank plasma was obtained from Kabi Pharmacia and kept frozen at -20 “C. The plasma was thawed during the night and kept at room temperature (20 “C) during the day of analysis. The blank plasma was spiked with appropriate amounts of the stock solutions. Appropriate precautions necessary for safe handling of blood products should be observed. THEORY

The efficiency of the liquid membrane extraction depends on a number of different parameter~.~,~J~ These include flow rate, dimensions of the channels of the membrane holder, the membrane porosity and thickness, chemical composition of the phases, and kinetic and thermodynamic properties. Basically, two main sets of conditions for SLM extraction can be distinguished. With donor-controlled conditions the rate of extraction is controlled by (13) Internal information from Kabi Pharmacia. (14)Jbnsson, J. A.; Liivkvist, P.; Audunsson, G.;Nihre, G.Anal. Chim. Acfu 1993, 277, 9-24.

4492 Analyfical Chemisty, Vol. 66,No. 24, December 15, 1994

the mass transfer in the donor phase. “his is the case when the distribution coefficient KDbetween the organic membrane phase and the donor phase is relatively large (KD> 10) for the analyte molecules. Diffusion coefficients in the phases also play a role here. With donorcontrolled conditions, the extraction efficiency should, for a given total enrichment time, increase with the flow rate of the donor buffer. On the other hand, if KDis small, or the mass transfer in the membrane is unusually low, the mass transfer in the membrane phase controls the rate of extraction (membrunecontrolled conditions). In that case, the donor flow rate per se is not important, and the extraction efficiency is determined by the total extraction time. In the present application,both donor and membranecontrolled conditions are represented by the different KD’Sof the three analytes. For authentic plasma samples, the volume is limited to -1 mL, and the concentrations of the drug compounds are low, which means that a high extraction efficiency is needed. To achieve this, the flow rate in the donor channel must be low to give a long contact time with the liquid membrane. It is also possible to increase the contact time by passing the sample by the membrane several times in either a push-pull or recycling mode. The dimensions of the channel have a large innuence on the recovery. This has been discussed earlier.12J4The depth of the donor channel should be as low as possible so that a large part of the sample is in contact with the membrane. This is particularly important when the mass transfer in the donor phase is the limiting factor (that is, KD is high). On the other hand, a very shallow channel (< 0.1 mm) is hard to machine and can cause stoppage, especially with viscous samples such as plasma. To achieve a high recovery, a large exposed membrane area is preferable, but a too wide channel can cause bulging of the membrane, while the length is limited by the back pressure arising in the channel. The exposed membrane area is also determined by the maximum acceptable volume and possible depth of the acceptor channel. RESULTS AND DISCUSSION Chr0matogra”a. Panels a and b of Figure 3 show chromate grams of Amperozide 0 , the metabolite 11, and the related

compound In, in plasma and aqueous buffer, respectively, with the subsequent blanks. The samples, 0.5 mL of 4 pg/mL (8 pg/ mL for III) , were enriched with the supported liquid membrane technique after optimization of various parameters as discussed below. The peaks appearing in the beginning of the chromatograms are caused by the CLC system and not by the liquid membrane procedure, as they turn up also in a direct injection on the column.

a

b

4 M m L in plasma

4 &mL in aqueous buffer

I

/I

III

1

III

TE36 Id1

TLIl 0.0 2.0

Amp

Met

CompIIl

I

Amp

Met

CQmplll

1

FGLP Id

FGLP (dl

I

4.0

I

I

0.0

Amp

1

Met

001 Amp

CompIII

1 1 1 , , Met

Comp Ill

FHUP

6.0

I- 4.0 2.0

0

5

10

minutes

15

0

5

10

--

IS

minutes

Figure 3. (a) Chromatogram of I (Amperozide),II (its metabolite), and 111, with the subsequent blank after enrichment from blood plasma. (b) Corresponding chromatograms after enrichment from an aqueous buffer solution. Conditions: concentration, 4 pg/mL (8pg/mL for Ill), membrane holder, Ti; membrane, TE35, backing against donor; membrane liquid, dihexyl ether; donor flow rate, 0.18 mumin; number of passages, 1; donor pH, 10.0; acceptor pH 2.5;injection loop volume, 60 pL.

As can be seen, the blank chromatograms are virtually identical, showing a very efficient cleanup of the complex plasma matrix. The peaks from the plasma sample are, however, much lower than those from the aqueous sample, where the recovery of Amperozide is nearly 30%. The decrease in recovery can be caused by the higher viscosity of the plasma, binding of the compounds to the plasma proteins, and (possibly) fouling of the membrane by the plasma. This will be further discussed later. The Membrane. Six different membranes have been examined (see Table 1). A suitable membrane should be stable, easy to handle, and have small pores to prevent the organic liquid from being pressed out of the pores causing leakage between the aqueous phases.8 It should also be thin to facilitate the mass transfer. A smooth surface prevents fouling of the membrane. The extraction efficiencies (E) using the different membranes are shown in Figure 4. When different membranes are compared, the following aspects can be considered: (a) the extraction efficiency for a polar compound (here the metabolite), which reflects the mass transfer process into and through the membrane, (b) the extraction efficiency for less polar compounds (Amperozide and III), which reflects the mass transfer in the donor phase, and (c) the reproducibility, which is innuenced by the ease of handling and by the stability of the membrane-liquid combination. The membrane that gave the highest extraction efficiency was Model TE35 from Schleicher and Schuell, which is rather thin with a fluffy,feltlike, backing that permits the molecules to pass. With this membrane, the extraction of the least polar compounds, Amperozide and 111,is higher when the backing is turned against the donor channel. This is probably due to more efficient mixing

I

SM11807

I 2- -

I

4.0

1

Amp

Met

Complll

1

Figure 4. Extraction efficiencies with different membrane materials

(cf. Table 1). “ d and “a”signify that the membrane was mounted with the backing against the donor and the acceptor channels, respectively. Other conditions as in Figure 3. The standard deviation is shown as ‘41ags”on the bars.

in the donor phase because of the feltlike backing. On the other hand, a discoloring can be seen in the beginning of the channel with plasma samples, probably due to protein adsorption on the fluffy backing. The FGLP membrane has been used successfully as a liquid membrane support in other s t u d i e ~ , but ~ ~ ~inJthis ~ application it gave a lower extraction efficiency than TE35. The unlaminated membrane (FHUP) has no backing and is thus more difficult to handle. It gave, however, a higher extraction of the polar metabolite than the FGLP membrane. An explanation could be the rigid polyethene net backing of the FGLP membrane, which prevents the molecules from passing through on a relatively large part of the area. This effect should be more important for a polar compound, where the transport into and through the membrane is the determining step. Another membrane, Durapore, is also without backing, but is easier to handle as it is thicker. This is, however, a disadvantage for the transport through the membrane, which again can be seen as a low extraction efficiency of the polar metabolite. The SM11807 is similar to FHUP in physical parameters, but gave irreproducible (increasing) extraction efficiencies. The Spectrapor membrane is similar to FGLP, but the lower porosity gives a lower efficiency for the metabolite. (15) NilvC,

G.;Stebbins, R Chromatographia 1991,32, 269-278.

Analytical Chemisty, Vol. 66, No. 24, December f5, 1994

4493

Table 2. Recovery with Mfferont Amounts of TOP0 in the Membnne Liquid

recovery (%) %MPO

0

2 3.5 5 7

b

a

150

-

200

250

350

300

400

450

500

nm

Flguro 5. UV spectra of (a) the acceptor plug after enrichment of a plasma blank, (b) the acceptor plug after enrichment of an aqueous blank, and (c) a plasma blank without enrichment. All three samples are diluted 100 times with acceptor buffer.

As Model TE35 from Schleicher & Schuell gave the highest extraction efficiency, this membrane was used for the experiments, usually with the backing against the donor channel. The Membrane Liquid. Several different membrane liquids have previously been tested for enrichment of the drug compounds in aqueous solutions.12 Dihexyl ether gave the highest recovery, and it was thus chosen for the plasma samples in this study. Apart from facilitating the transport of the analytes through the membrane, the membrane liquid should also be an efficient barrier for other compounds in the plasma. This was examined by taking a UV spectrum (190-500 nm) of the acceptor plug (diluted 100 times) after enrichment of both an aqueous blank and a plasma blank (See Figure 5). No difference could be seen between the two acceptor plugs. When untreated plasma, diluted 100 times, was scanned, very large absorption was observed at the shorter wavelengths. An addition of trioctylphosphine oxide (TOPO) to the membrane liquid enhances the transport of polar compounds through the membrane. Four different concentrations of TOPO were examined, 2.0%,3.5%,5.0% and 7.0%(w/w), and 5% was found to be optimal for Amperozide (see Table 2). With this concentration, the recovery of the most polar compound, 11, was more than twice as high as when no TOPO was used. For I, the increase was lo%, and for the 111, the recovery was even lower. There are, however, certain drawbacks with TOPO, and it should be used with discernment. It might shorten the lifetime of the liquid membrane, and it has been shown to cause yet unexplained effects on the mass transfer kinetics under certain circumstances.11 To avoid such effects, most experiments were performed without TOPO. 4494 Analytical Chemistry, Vol. 66,No. 24, December 15, 1994

I

11

111

12.3 12.1 13.0 13.5 12.8

3.3 5.4 7.2

7.2 6.1 6.6 6.9 6.3

8.7

10.0

With the TE35 membrane and dihexyl ether with 5%TOP0 as membrane liquid, the membrane was stable for -60 samples. A breakthrough of the liquid membrane could be seen as markedly increased front peaks and diminished peak heights of the analytes. Donor Flow Rate and Number of Passages. With the ASTED,the lowest flow rate that can be obtained is 0.180 mW min, which gives a total enrichment time of only -5 min for a 1 mL sample. To increase the enrichment time, the sample plug can be passed by the membrane several times by alternating pushing and pulling it with the syringe pump (3 in Figure 2). Figure 6 shows the recoveries in plasma when up to 11passages are made at three different flow rates, 0.180,0.360, and 0.750 mW min. With only one passage, the recovery is only -5%, but with an addition of two more passages, the recovery is doubled. When the original recovery is so low, it is thus very advantageous to run additional passages. As can be seen, the recovery for 111is higher at a higher flow rate, with constant enrichment time, which is expected for compounds with a high KD. For those compounds, the mass transfer is mainly limited by the mass transfer in the donor phase, which is increased at higher flow rates. For the metabolite,which has a low KD,no effect of an increased flow rate at a constant enrichment time can be seen. This is in accordance with the theory when the mass transfer into the membrane is the limiting factor. Due to dispersion in the tubings, the volume of the sample plug is somewhat larger after every passage. This has been taken into account by programming the ASTED to increase the volume that passes by the membrane for every passage. Composition of the Donor Phase. In earlier experiments,12 an ammonium sulfate buffer, usually at pH 9.0, was used as donor phase to suppress adsorption of the compounds. It was, however, found that, at the basic pH used, the unprotonated ammonia could pass through the membrane. The pH of the acceptor was then increased above the critical value at which the compounds are protonated,resulting in a decrease in recovery. Without ammonia, a pH of 10.0 of the donor phase was found to be optimal. Thus, for the experiments in this paper, only 12.5 mmol/L EDTA and NaOH to give a pH of 10.0 was used as donor phase. It was also found that, with plasma samples, the adsorption of the analytes in the equipment was markedly decreased, so an addition of ammonia was no longer necessary for this purpose. When the partition coefficient KD is low, an addition of salt to the donor phase can be favorable as a salting-out effect.16 It increases, however, the viscosity and thus diminishes the mass transfer rate in the donor. To examine the effect of an addition (16)Audunsson, G.Anal. Chem. 1988,60,1340-1347.

250

20 0

-

15

I

!

/

/

i

/

0

F2: 100

50

500

250

-

200

-

50

-

0

25.0

r

20.0

1

00

-

0

1wO

1500

b

-

1

2000

2500

30W

2w0

2500

3000

I1

500

1wO

1500

I11

5W

IwO

1500

2wO

25W

3000

I

Figure 6. Recovery of 1-111 versus total enrichment time at three different donor flow rates: 0, 0.18 mumin; 0, 0.36 mumin; A, 0.75 mUmin. Other conditions as in Figure 3.

of salt, 50 pL of 2.5 mol/L sodium chloride was added to the sample when a plasma sample with 2 pg/mL of the analytes was enriched. The recovery remained constant, and no effects of the salt could be seen. Composition of the Acceptor Phase. The pH of the acceptor solution should be sufticiently low to ensure complete protonation

of the compounds as they enter the acceptor channel, thereby preventing them from reentering the liquid membrane. Theoretically, this demands a pH at least three units below the pKa of the substances that should be trapped14. The optimal pH of the acceptor solution had been examined previously,12 and it was found to be in the range of 2.5, which has been used for the experiments in this study. When large amounts of basic compounds are extracted into the small acceptor volume, there is a risk of increasing the pH, leading to a lower degree of protonation and a lower recovery. It is then necessary to use a buffer to maintain a low pH of the acceptor solution. In the case of trace analysis, however, no buffer was considered necessary. Most endogenous compounds are acidic,' and the liquid membrane forms an efficient barrier against basic endogenous compounds. The acceptor solution in this study simply consisted of diluted sulfuric acid. The Plasma Sample. Blood plasma is a very complex matrix. Its main differences compared to aqueous samples are, for our purposes, a higher viscosity (about twice the viscosity of water'? and the presence of proteins, mainly albumin, to which the analytes are bound. The protein binding, although high (97%for Amperozide13),is rather weak, which can be seen when recoveries in aqueous solutions and plasma samples are compared (see Figure 3). The recovery in plasma is 30% (15%for II) of that in aqueous solutions. If the protein binding is so strong that only the free fraction could pass through the liquid membrane, the expected recovery would be only 6%of that in aqueous solutions (as the plasma is diluted 1:l with donor buffer). It thus follows that the bond is broken and the equilibrium is shifted during the extraction. There are several ways to diminish the protein binding, for example, by dilution of the sample, by using a displacer, by adding an organic solvent, or by increasing or decreasing the pH of the solution.'* With the method used here, the analytes must be uncharged, and thus a decrease in pH is not possible to use. Neither can an addition of organic solvents be used, as it would interfere with the organic liquid in the membrane. To increase the pH of the sample, different amounts of sodium hydroxide have been added to the sample. The highest recovery was obtained with an addition of 50 pL of 2.5 mol/L sodium hydroxide to a 1 mL sample, giving a pH of 13. If not otherwise stated, all plasma samples were treated in this way. The increase in recovery is probably due to a decrease in protein binding (cf. next section) and not to a salting-out effect, as an addition of sodium chloride had no effect on the recovery (see Composition of the Donor Phase). To decrease the viscosity of the sample, the plasma is diluted 1:l with donor solution (this is done automatically by the ASTED). Further dilutions have also been tested as a means to decrease the protein binding. Table 3 shows that dilution does increase the recovery, but on the other hand, the detection limit will increase with every dilution. To circumvent this, additional extractions of the sample can be made, resulting in a prolonged analysis time. Three different compounds have been tested as displacers: ammonia, triethylenetetramine (TETA), and piperazine at a concentration of 200 pg/mL (the concentration of the analytes (17) WissenschafilicheTabellen, 7th ed.; Ciba Geigy A G Basel, Switzerland, 1973;

p 554. (18) Wahlund, K-G.; Arvidsson, T. f. Chrumefogr. 1983,282, 527-539.

Analytical Chemistry, Vol. 66, No. 24, December 15, 1994

4495

Table 3. Recovery with Different Dilutions of the Plasma Sample

Table 4

compd

slope”

intercepv

r

LODb

LOQ6

I I1

53.8 f 2.5 21.2 f 1.2 26.8 f 1.4

-0.1 f 1.2 0.2 & 0.6 -0.5 f 1.3

0.997 0.995 0.996

0.03 0.06 0.06

0.1 0.2 0.2

recovery (%) dilution

conc (ng/mL)

I

I1

111

2 4 8

lo00

6.9 8.9 9.6

2.7 4.1 5.4

5.0 5.8 6.0

500 250

was 1 pg/mL) . None gave rise to a noticeable increase in recovery of the three analytes. When spiked plasma samples are used, they should be as similar to real samples as possible. To examine how fast the binding of the analytes to the plasma proteins takes place, the addition of the compounds was made immediately (some seconds) before the analysis as well as several minutes in advance. As the recovery was not higher after the shorter incubation time, the binding was considered to take place instantaneously. Some plasma samples are very turbid, and a deposit of plasma substance has often been observed on the membrane where the donor enters the membrane channel. If the channel is narrow, this can cause stoppage with a breakthrough of the liquid membrane as a result, and it is thus recommended to centrifuge turbid plasma samples. Estimation of the Degree of Protein Binding. In ref 14, the influence of different forms of the analyte (ionized and neutral) on the mass transfer in the donor phase was discussed. Originally this was used to describe the influence of incomplete deprotonation of the analytes in the donor phase, but the same theory can be used to describe protein binding, given that the donor pH is such that the free drug is completely deprotonated. In this case, the parameter aD in ref 14 is the fraction of the analyte that is not protein bound (“active form”). Then, eqs 3, 5, 9, 14, 21, and 24 in ref 14 can be used to calculate aD from the observed differences between the extraction efficiency in aqueous solutions and in blood plasma. The calculation were made with the Microsoft Excel Spreadsheet program, using the “solver” option. The diffusion cm2/s. This was coefficient of Amperozide was set to 2.7 x calculated using Perrins adaptation of the Stokes-Einstein equation,Ig with the assumption that the molecule shape could be described as a prolate ellipsoid. To describe the diffusion of the inactive form (the protein-drug complex), the diffusion coefficient cm2/s, was used.20 With the conditions stated of albumin, 6 x in Figure 3, an extraction efficiency of 29%was obtained in an aqueous solution and 9% in plasma (as usual diluted 1:l with donor buffer). This leads to a calculated aD value of 0.05,Le., 95%protein binding in diluted plasma. However, in this experiment, the sample pH was 13, which is unrealistically high and could partly disrupt the protein binding. Repeating the experiment at pH 8, both in the samples and in the donor buffer, gave the extraction recoveries 26% and 5.5%in aqueous solution and in plasma, respectively. This leads to an aD value of 0.02 or 98%protein binding, which would mean 99%in undiluted plasma. The proteinbinding degree determined with other techniques13was 97%at pH 7.4. Thus, the values agree reasonably, but more experiments are needed to evaluate this principle for the determination of (19) Cussler, E. L.Dimion: Mass transfer in fluid systems, 1st ed.; Cambridge University Press: Cambridge, UK, 1984; p 119. (20) Lehninger, A. L. Biochemisfty, 2nd ed.;Worth Publishers: New York, 1981; p 176.

4496 Analytical Chemistry, Vol. 66, No. 24, December 15, 1994

I11 a

Arbitrary units; 95% confidence intervals. b In micrograms per

milliliter.

protein-drug binding. The subject will be examined in more detail in a forthcoming paper. The ASTED. With the ASTED, many operations can be made automatically. In the present procedure, the only manual step is the pipeting of the plasma samples into the vials. The ASTED then dilutes the samples with donor buffer, adds sodium hydroxide from another vial, and mixes the sample. It is advantageous to neutralize the analytes as late as possible in the workup procedure (and at a constant time period before the passage by the membrane) as the adsorption on the glass vials is prominent when the compounds are uncharged.’Z If needed, the ASTED can also wash the donor channel with diluted acid, which efficiently removes potential residuals of the previous sample. On several occasions, the ASTED was run overnight and -40 plasma samples were processed. It was then noted that when the plasma was turbid, a breakthrough of the liquid membrane occurred after 20-30 samples. If the cross section of the donor channel was made larger by putting a spacer between the donor block and the membrane, the membrane lasted for twice as many samples. The breakthrough was thus thought to be caused by high pressure, forcing the organic liquid out of the membrane pores. However, since the spacer caused a lower recovery (due to a deeper channel), it is better to centrifuge the samples before analysis, which also gave twice as long lifetime of the membrane. The daily change of the membrane needed is made in 10 min. Quantitation. Calibration curves, based on peak heights of triple injections, were made in plasma with dihexyl ether as membrane liquid. Five samples in the concentration range 0.25-4 pg/mL, together with the blank, were processed with three passages by the membrane and a donor flow rate of 0.18 mL/ min. All curves showed good linearity, and the intercepts did not d ~ esigniticantly r from zero at a 95%coniidence level. Weighted regression lines were calculated by assuming that the errors are proportional to the peak heights. The results are given in Table 4.

The overall precision (relative standard deviation) at a concentration of 1 pg/mL was -3%. The limits of detection (LOD) and of quantitation (LOQ) (Table 4) were calculated from peak heights 3 and 10 times the baseline noise, respectively. With TOP0 in the membrane (see above), the LOD and LOQ for the metabolite QI) will be -2 times lower than shown in the table. The data in the table were obtained with three passages by the membrane as this gives an enrichment time approximately equal to the chromatographic separation time. If a decreased sample throughput can be accepted, a larger number of passages can be made. The recovery will then be proportionally larger (6.Figure 6), leading to lower values of LOD and LOQ. In an experiment with 11 passages by the membrane, the blank still looked the same, while higher peaks were observed, leading to a LOD of about 0.01,0.02, and 0.02 pg/mL for I-111, respectively. The total analysis time was -1 h. For bioanalysis of the present com-

pounds, however, still lower detection limits are needed, which can be achieved with an electrochemical detector.21 Work along this line is in progress. Amines in general are apt to be adsorbed on all kinds of surfaces: glass, metal, polymers, etc. The substances in this study are no exceptions. It was, however, found that the memory effects (that is, how much of a previous sample that turns up in a subsequent blank) were smaller in plasma samples than in aqueous solutions. With a plasma sample spiked with 4 pg/mL Amperozide and the metabolite and 8 pg/mL 111, the memory effects were 0%, 0%, and 1%,respectively. The fact that the calibration curves pass through the origin also indicates that the memory effects are small. CONCLUSIONS

The supported liquid membrane technique, automated using a modified ASTED instrument, provides a high degree of cleanup (21) Olsson, A.M.; Johansson, C. G.; Gorton, L. Poster, 9th International symposium on column liquid chromatography, Edinburgh, 1985.

for basic drugs in plasma samples. This is achieved with minimal amount of organic solvents and manual work. The extraction efficiencies obtained are relatively low due to protein binding. A comparison with extraction efficiencies of protein-free solutions permits an estimation of the degree of protein binding. By increasing the number of extraction passages by the membrane, the extraction ef&ciencycan be increased on the expense of the total analysis time, leading to a decreased limit of detection and quantitation. ACKNOWLEDGMENT

This work was partly supported by grants from the Swedish Natural Science Research Council and the Crafoord Foundation, Lund. We thank Dr. Jerzy Trocewicz, Lublin, Poland, and Camilla Rossborg for experimental work. Received for review March 30, 1994. Accepted August 24, 1994.s Abstract published in Advance ACS Abstracts, October 15, 1994.

Analytical Chemistry, Vol. 66, No. 24, December 15, 1994

4497