Renewable pH Cross-Sensitive Potentiometric Heparin Sensors with

method of membrane renewal. The electrically charged. H+ ionophore 5-(octadecanoyloxy)-2-(4-nitrophenylazo)- phenol (ETH 2412) is incorporated as an ...
0 downloads 0 Views 82KB Size
Download PDF
Anal. Chem. 1999, 71, 4614-4621

Renewable pH Cross-Sensitive Potentiometric Heparin Sensors with Incorporated Electrically Charged H+ Ionophores Sally Mathison and Eric Bakker*

Department of Chemistry, Auburn University, Auburn, Alabama 36849

Polymer membrane-based potentiometric sensors have been developed earlier to provide a rapid and direct method of analysis for polyions such as heparin, a natural anticoagulant administered to prevent thrombus formation during cardiovascular surgery. These heparin sensors are irreversible, requiring a membrane renewal procedure between measurements which currently prevents the sensors from being used for continuous monitoring of blood heparin. A newly developed heparin sensor is shown here to allow an alternate and more practical method of membrane renewal. The electrically charged H+ ionophore 5-(octadecanoyloxy)-2-(4-nitrophenylazo)phenol (ETH 2412) is incorporated as an additional ionophore into a heparin-sensing membrane. This membrane will respond to pH only at low H+ concentrations, while sample anions are coextracted with H+ ions into the membrane at physiological pH. In buffered samples at physiological pH, the sensors will therefore respond to heparin via an ion-exchange mechanism with chloride anions. The pH cross-sensitive heparin-sensing membranes are shown to give an excellent potentiometric response toward heparin in aqueous samples at physiological pH and Cl- levels as well as in undiluted whole blood with no loss of heparin response. The membrane renewal is accomplished by moderately increasing the pH of the sample, causing heparin to diffuse out of the membrane with H+ ions. Reproducibilities are, with less than 1 mV standard deviation, improved over the classical system. Unlike the high NaCl concentration used to strip heparin from the previously established heparin sensor, the pH change used here could ultimately be performed locally at the sample-membrane interface, allowing the sensor to be used for automated long-term monitoring of heparin in blood. A theoretical model is presented to explain the experimental results. Polymeric membrane-based potentiometric electrodes are used extensively in clinical analysis due to their direct, rapid, and selective measurements of ions in diluted and undiluted blood and plasma.1 Although they have primarily been used to accurately measure activities of simple electrolytes and gases in blood,2 the discovery that similar polymer ion-selective membranes also give (1) Bu ¨ hlmann, P.; Pretsch, E.; Bakker, E. Chem. Rev. 1998, 98, 1593. (2) Bakker, E.; Bu ¨ hlmann, P.; Pretsch, E. Chem. Rev. 1997, 97, 3083.

4614 Analytical Chemistry, Vol. 71, No. 20, October 15, 1999

a response to polyions has opened a new branch of research in the field.3,4 Potentiometric sensors of this type have been developed for heparin as well as for protamine,5,6 and the development of similar membrane-based sensors for many other polyions is in progress.7 Heparin is a natural anticoagulant with an average molecular weight of about 15 000 and an average charge of -70. It is a parenteral drug with a very rapid onset of action due to its inhibition of clotting factors near the end of the coagulation cascade. In cardiovascular surgery and other surgical procedures, heparin is administered to the patient to avoid thrombus formation during surgery.8 The heparin treatment must give blood heparin levels that result in the control of thrombosis without causing severe bleeding. Since much of the heparin is metabolized in the body, and the extent of metabolism varies with age, health, and other factors,9 the heparin levels in the patient’s blood need to be carefully and accurately monitored during surgery and recovery. Protamine is given postsurgically to neutralize heparin activity and avoid prolonged bleeding by forming an inactive complex with heparin, but the postsurgical heparin levels must be known to give an accurate dose of protamine. An ideal method of monitoring blood heparin levels would be a small sensor that could be placed in a heart-lung apparatus during surgery or in vivo along with the IV site during recovery, allowing for the constant monitoring and maintenance of therapeutic heparin levels without the inconvenience of repetitive blood extractions.10 In the past years, a polymer membrane-based ion-selective electrode (ISE) has been developed that contains the heparin carrier tridodecylmethylammonium chloride (TDMAC). With a quaternary ammonium structure very similar to Polybrene, a strong heparin antagonist, the TDMA+ ion is known to associate strongly with the negatively charged heparin molecule.11 Direct measurement of heparin with this type of sensor is more rapid (3) Ma, S.-C.; Yang, V. C.; Meyerhoff, M. E. Anal. Chem. 1992, 64, 694. (4) Fu, B.; Bakker, E.; Yun, J. H.; Yang, V. C.; Meyerhoff, M. E. Anal. Chem. 1994, 66, 2250. (5) Ma, S.-C.; Yang, V. C.; Fu, B.; Meyerhoff, M. E. Anal. Chem. 1993, 65, 2078. (6) Yun, J. H.; Meyerhoff, M. E.; Yang, V. C. Anal. Biochem. 1995, 224, 212. (7) Meyerhoff, M. E.; Fu, B.; Bakker, E.; Yun, J.-H.; Yang, V. C. Anal. Chem. 1996, 68, 168A. (8) Hirsh, J.; Dalen, J.; Deykin, D.; Poller, L. Chest 1992, 102, 337S. (9) Andrew, M.; Marzinotto, V.; Massicotte, P.; Blanchette, V.; Ginsberg, J.; Brill-Edwards, P.; Burrows, P.; Benson, L.; Williams, W.; David, M.; Poon, A.; Sparling, K. Pediatr. Res. 1993, 35, 78. (10) Espadas-Torre, C.; Meyerhoff, M. E. Anal. Chem. 1995, 67, 3108. (11) Grode, G. A.; Falb, R. D.; Crowley, J. P. J. Biomed. Mater. Res. 1972, 6, 77. 10.1021/ac990387s CCC: $18.00

© 1999 American Chemical Society Published on Web 09/11/1999

Figure 1. Relevant extraction processes that dictate the potentiometric behavior for classical (left) and pH cross-sensitive (right) heparin membranes. Square boxes indicate species in the organic phase, with R+, heparin-selective anion exchanger; HL, protonated H+ ionophore; L-, deprotonated H+ ionophore; and Hepn-, heparin with charge n-. Shown, from top to bottom, are the approximate membrane compositions for membranes in contact with heparin-free solutions, heparin-containing solutions, high saline solutions for heparin stripping, and high pH solutions for heparin stripping. The last process is only possible with membranes that contain a suitable additional H+ ionophore.

than current methods of measurement and avoids the inaccuracy caused by indirect assays. Although the potentiometric heparin response is intrinsically dependent on the sample activity of other ions such as chloride, thiocyanate, and bicarbonate, making sensor calibration difficult, the interferences could be experimentally corrected for if the concentrations of these interfering ions are determined simultaneously. In this manner, direct measurements of heparin may be performed. An alternate method of heparin analysis that avoids interference from other ions involves the use of the heparin sensor as an end point indicator for the titration of the heparinized blood with protamine.12 Ion-selective sensors for polyions have a response mechanism that differs from that of conventional ISEs for the detection of small ions. Polyion-sensitive membranes are typically conditioned in a solution containing only background electrolyte with no polyion present. This behavior has been investigated12 and the response mechanism has been developed and explained specifically for the heparin sensor established by the groups of Meyerhoff and Yang.4,13 Upon exposure of the membrane to a sample containing the polyion, a nonequilibrium pseudo-steady-state response is established which is due to the polyion slowly and almost irreversibly exchanging with the chloride ions in the membrane. In effect, this results in an activity change of the chloride ion in the phase boundary region of the membrane. The measured potential shows an apparent super-Nernstian slope that allows analytical measurements to be performed. This response is reproducible if the membrane is initially void of the polyion at the sample-membrane interface; i.e., the activity decrease of the chloride ion in the membrane remains the same for a given sample heparin concentration. A reproducible behavior of the heparin-sensing ISE may be accomplished only by removing the heparin from the membrane (12) Fu, B.; Bakker, E.; Wang, E.; Yun, J. H.; Yang, V.; Meyerhoff, M. E. Electroanalysis 1995, 7, 823. (13) Fu, B.; Bakker, E.; Yang, V. C.; Meyerhoff, M. E. Macromolecules 1995, 28, 5834.

surface between measurements by placing the electrodes in a high concentration of the background electrolyte, typically 2 M NaCl, so the Cl- will completely displace the heparin from the membrane surface (see Figure 1).14 A disadvantage of this is that the sensor cannot be used for continuous monitoring in blood because the high chloride concentration needed for membrane renewal will never be present at the sensor surface in the blood stream and appears impractical for many automated situations. This paper introduces a versatile, convenient way to improve the reversibility of potentiometric heparin sensors. The membrane contains an additional H+ ionophore that renders the ISE pH crosssensitive. Heparin removal may now be accomplished by sample pH changes, either in the sample bulk or locally at the sensing surface (see Figure 1). The selectivity and sensitivity of the heparin sensor are otherwise fully maintained. THEORETICAL SECTION Classical Heparin Sensor. The potentiometric heparin sensor developed by the groups of Meyerhoff and Yang is based on a solvent polymeric membrane doped with the chloride salt of the heparin-selective lipophilic tridodecylmethylammonium ion. Heparin is extracted into the membrane according to the following ion-exchange process4 (see Figure 1):

(1/n)Hepn-(aq) + R+(org) + Cl-(org) (1/n)Hepn-(org) + R+(org) + Cl-(aq) (1) with the ion-exchange constant Kexch

Kexch ) ([Hepn-]/aHep)1/n (aCl/[Cl-])

(2)

where Hepn- is heparin with charge n-, R+ is the lipophilic tetraalkylammonium ion, and (org) and (aq) denote the organic (14) Mathison, S.; Bakker, E. J. Pharm. Biom. Anal. 1999, 19, 163.

Analytical Chemistry, Vol. 71, No. 20, October 15, 1999

4615

Figure 2. Predicted potential change with and without 10-5 M heparin in the sample (top) and heparin concentration in the membrane (bottom) for the classical heparin sensor. A high saline solution is required to quantitatively remove heparin from the membrane. Other parameters (see eq 2): Kexch ) 218, [R+] ) 0.005 M, and n ) 70.

and aqueous phases. Ion concentrations in the organic and activities in the aqueous phases are denoted as [Ion] and aIon, respectively. To simplify comparisons, the basic assumptions used here are identical to those used in the previous treatment for the classical heparin sensor.4,13 The exchange constant shown in eq 2 can be estimated from potentiometric experiments on heparin electrodes as described.4 The typically observed 60-mV maximum heparin response for a 10-5 M heparin concentration then gives an estimated Kexch ) 2.2 × 102 (see Figure 2). The response mechanism of this sensor has been established.4 It relies on a steady-state ion-exchange process where the sample heparin concentration is sufficiently low so that it becomes significantly depleted within the aqueous diffusion layer at the sample-membrane interface. This unique characteristic allows the sensor to function with a much improved sensitivity as compared to what would be predicted classically on the basis of the Nernst equation. Unfortunately, heparin will accumulate in the membrane upon prolonged contact with heparin-containing samples and introduce memory effects. The polyion must be removed from the membrane by driving the equilibrium reaction 1 to the left side (see Figure 1). Even a large sample heparin concentration decrease, however, would not be sufficient to accomplish this because of the very high charge of the polyion. A billionfold heparin concentration decrease, for example, would increase the membrane chloride concentration only by a factor of about 1.3 if n ) 70. Therefore, heparin has historically been displaced from the membrane by exposure to a concentrated 4616

Analytical Chemistry, Vol. 71, No. 20, October 15, 1999

saline solution, with cCl ) 2 M. The effect of changing sample chloride concentrations on the heparin extraction behavior of the membrane can be conveniently calculated by extending the previously reported model4 for describing the heparin response of these membranes. Specifically, eq 2 is combined with the membrane charge balance ([R+] ) (1/n)[Hepn-] + [Cl-]) to eliminate [Cl-] in both equations. Figure 2 (bottom) shows how the heparin concentration in the membrane is, at equilibrium, dependent on the sample chloride concentration. The corresponding expected EMF change is also shown in Figure 2 (top), which was calculated in analogy to the reference.15 Evidently, the use of concentrated salt solutions is necessary to quantitatively remove heparin from the membrane between measurements. This is not convenient for routine applications since the ionic strength of the stripping solutions would be extremely high, leading to complications in handling these fluids and in avoiding carry-over problems from the stripping solutions to the samples. In principle, stripping could be accomplished with a moderate concentration of a more lipophilic anion, for example, perchlorate. That anion would, however, be difficult to exchange again with chloride prior to a practical measurement so that this possibility is not very useful. Design of pH Cross-Sensitive Heparin Sensors. Heparinresponsive membranes may be doped with additional components to make them cross-sensitive to another sample ion that can be more conveniently controlled than chloride. In this context, a pHcontrolled uptake and stripping of heparin may be most promising since the sample pH can be conveniently adjusted with the help of buffers. A pH cross-sensitive membrane would uptake heparin on the basis of a coextraction equilibrium, where hydrogen ions and heparin anions would simultaneously extract into the membrane. Membranes containing electrically neutral carriers such as tridodecylamine, however, would likely exhibit a number of complications that would make their use here inadequate. These membranes require the addition of lipophilic anionic sites such as potassium tetrakis(4-chlorophenyl)borate, which make the incorporation of TDMAC as the heparin-sensitive reagent problematic. Such membranes with excess TDMAC would certainly exhibit anion response behavior and would be pH insensitive. A different class of H+ carriers has recently been studied in potentiometric membranes to give optimum response to pH only if they contain lipophilic cationic sites R+.16,17 These ionophores are electrically neutral when protonated and negatively charged when deprotonated, and the cationic sites provide the countercations of the deprotonated ionophore L-. The heparin coextraction process would be given as (see Figure 1)

(1/n)Hepn-(aq) + H+(aq) + L-(org) + R+(org) (1/n)Hepn-(org) + R+(org) + HL(org) (3) with the coextraction constant

Kcoex ) ([Hepn-]/aHep)1/n ([HL]/aH[L-])

(4)

The extracted heparin would coordinate to cationic sites such as (15) Bakker, E.; Xu, A.; Pretsch, E. Anal. Chim. Acta 1994, 295, 253. (16) Mi, Y.; Bakker, E. J. Electrochem. Soc. 1997, 144, L27. (17) Mi, Y.; Green, C.; Bakker, E. Anal. Chem. 1998, 70, 5252.

Figure 3 also shows that the upper detection limits of the two response functions, as defined by IUPAC,15,18 are important in planning optimal electrode configurations (shown as dashed lines in Figure 3 (top)). Indeed, quantitative stripping of heparin will occur at pH values just above the detection limit of the lower trace shown in Figure 3 (top). On the other hand, the maximum EMF difference between a sample containing a chloride background only and one containing heparin will be observed at pH values below the upper detection limit for the upper trace in Figure 3 (top). In practice, therefore, the sample and stripping solution pH values should preferably be adjusted above and below these two critical values. Fortunately, the upper detection limit of pHresponsive electrodes can be conveniently predicted in analogy to a simple model that was developed earlier for neutral carrierbased systems.15 The response of charged carrier-based pH electrodes can, in analogy to neutral carrier-based systems,15 be described by the phase boundary potential at the sample-membrane interface:

E)K+

Figure 3. Predicted potential change (top) and heparin concentration in the membrane (bottom) for a pH cross-sensitive heparin sensor with and without 10-5 M heparin in the sample as a function of the sample pH. Quantitative removal of heparin is expected above pH 9.5. Other parameters (see eqs 2 and 4): Kcoex ) 1.1 × 109, Kexch ) 110, [R+] ) 0.005 M, LT ) 0.010 M, n ) 70, and aCl ) 0.1 M. Vertical dotted lines indicate the two upper detection limits calculated according to eqs 9 and 12.

tridodecylmethylammonium, which shows optimum selectivity properties for heparin. A membrane with this composition would most likely retain its optimal heparin selectivity over other sample anions at low pH; increasing the sample pH would then rapidly reverse equilibrium 3 to remove heparin from the membrane (see Figure 1). As above, the equilibrium concentration of heparin in the membrane as a function of pH can be calculated by considering eq 4, the mass balance for the ionophore (LT ) [L-] + [HL]), where LT is the total concentration of the ionophore, and the charge balance ([R+] ) [L-] + (1/n)[Hepn-]). The resulting plot, together with the respective calculated EMF functions, is shown in Figure 3 for one realistic case. Since Kcoex is directly dependent on the basicity of the H+ ionophore, it was chosen for the plot as 1.1 × 109 to obtain simulated response functions that closely mimic the experimental results with ETH 2412-based membranes. If anion interference occurs around pH 8, heparin measurements are possible under physiological conditions. Since the upper trace in Figure 3 (top) shows the pH response function with 0.1 M chloride in the background, the difference in the two traces at low pH shows the maximum attainable EMF difference if heparin is present in the sample. Above pH 9.5, anion interference no longer occurs and any previously extracted heparin and/or chloride ions will be removed from the membrane. This is more specifically shown in Figure 3 (bottom) where the calculated heparin concentration in the membrane eventually drops to zero as the pH is increased from 7.5 to 9.5.

RT [L ]aH ln F [HL]

(5)

where E is the membrane potential, aH is the sample hydrogen ion activity, R, T, and F are the gas constant, absolute temperature, and Faraday constant, respectively, and K combines all other constant potential contributions. A Nernstian response is expected when [L-] and [HL] are relatively sample-independent and can be approximated by [R+] and LT - [R+], respectively:

EH ) K +

[R+]aH RT ln F L - [R+]

(6)

T

In the pH region of anion interference with samples containing chloride ions only, the membrane composition at the samplemembrane interface will consist of fully protonated ionophore and cationic sites that are counterbalanced by extracted chloride ions. Inserting the respective coextraction constant into eq 5 and for this composition therefore yields a Nernstian chloride response ECl at low pH values:

ECl ) K +

Ka [R+] RT ln F kHkCl aCl

(7)

where

kHkCl/Ka ) [HL][Cl-]/aH[L-]aCl

(8)

Since the upper detection limit is given by the hydrogen ion activity where both Nernstian response curves intersect, it is calculated by setting eqs 6 and 7 equal (18) Buck, R. P.; Lindner, E. Pure Appl. Chem. 1995, 66, 2527.

Analytical Chemistry, Vol. 71, No. 20, October 15, 1999

4617

aH(UDL) )

Ka LT - [R+] kHkCl aCl

(9)

In complete analogy, the equilibrium response function for heparin in the region of anion interference can be written as

EHep ) K +

( )

Ka [R+]/n RT ln F kH(kHep)1/n aHep

1/n

(10)

where the coextraction constant is given by eq 4:

kH(kHep)1/n/Ka ) Kcoex

(11)

The combination of eqs 6 and 10 therefore gives the upper detection limit for heparin-containing samples:

aH(UDL) )

Ka

( )

kH(kHep)1/n

[R+]/n aHep

1/n

LT - [R+] [R+]

(12)

The difference in the two upper detection limits can be directly translated into the expected EMF difference for samples with and without the polyion heparin by considering the Nernst equation:

∆EMF )

( )

aCl [R+] RT aH(UDL,Hep) RT ln ln ) + F F aH(UDL,Cl) Kexch[R ] naHep

1/n

(13) where Kexch (see eq 2) is given by kHep1/n/kCl. This relationship is formally identical to the one developed earlier for the classical heparin sensor.4 In the region of anion interference, therefore, both systems should behave identically. Equations 9 and 13 may give guidance in designing optimum membrane compositions for polyion measurements. On one hand, the overall potential change in the presence of polyions should be maximized according to eq 13. This may be mainly accomplished with a lower sample chloride activity and a larger concentration of lipophilic anion exchanger [R+]. The upper detection limit of the pH response function, on the other hand, may be increased with increasing acidity of the H+ ionophore (Ka) and increasing concentration of ionophore (LT). Of course, the composition of the membrane must be optimized to guarantee high sensitivity to heparin under steady-state measuring conditions, which is analogous to the classical heparin sensor composition.4 A higher plasticizer content, for example, will shift the useful steady-state measuring range to lower heparin concentrations. EXPERIMENTAL SECTION Reagents. The membrane components tridodecylmethylammonium chloride (TDMAC), 5-(octadecanoyloxy)-2-(4-nitrophenylazo) phenol (chromoionophore IV, ETH 2412), bis(2ethylhexyl) sebacate (DOS), and high-molecular-weight poly(vinyl chloride) (PVC), as well as tetrahydrofuran (THF), heparin sodium salt (from bovine intestinal mucosa), sodium chloride, NaOH, and HCl were purchased in the highest quality available from Fluka Chemika-Biochemika (Ronkonkoma, NY). Phosphate buffers (pH 4618

Analytical Chemistry, Vol. 71, No. 20, October 15, 1999

4, 7, and 10) were purchased from Fisher Scientific (Pittsburgh, PA). Aqueous solutions were prepared by dissolving the appropriate salts, acid, or base in Nanopure-purified distilled water. Human undiluted whole blood was donated to the laboratory. Membrane and Electrode Preparation. The membranes were cast by dissolving 3.4 mg (46.2 mmol kg-1) of ETH 2412 and 1.85 mg (23.1 mmol kg-1) of TDMAC, together with DOS and PVC (either 1:2 or 2:1 by weight), to give a total cocktail mass of 140 mg, in 1.5 mL of THF and pouring it into a glass ring (2.2cm i.d.) affixed onto a microscopic glass slide. The solvent was allowed to evaporate overnight to give a membrane with a thickness of about 200 µm. For each electrode, a 6-mm-diameter disk was cut from the parent membrane with a cork borer and incorporated into a Philips electrode body (IS-561, Glasbla¨serei Mo¨ller, Zu¨rich, Switzerland). Solutions of 10 mM phosphate buffer and 100 mM NaCl (PBS) adjusted to pH 7.4 with NaOH served as the inner filling solution for the assembled electrodes. The electrodes were conditioned in the same solution overnight before measurement. Blood Collection. Whole blood (60 mL) was collected from a healthy human donor into 4-mL collection tubes, each containing 0.05 mL of 15% EDTA (K2) (7.5 mg). Blood was set aside for use in the heparin calibration, and the remainder was placed in the refrigerator. The blood pH was found to be 7.5 with a calibrated pH electrode. Room-temperature blood was mixed prior to measurement by slowly inverting the tube 10 times. EMF Measurements. All membrane electrode potential measurements were performed at laboratory ambient temperature (22 ( 1 °C) in unstirred solutions versus a double-junction Teflon sleeve Ag/AgCl reference electrode (Ingold, Wilmington, MA) with 1 M LiOAc as the bridge electrolyte in a cell of the following type: Ag | AgCl | KCl | 1 M LiOAc || sample || membrane || inner electrolyte | AgCl | Ag. All potentials were measured via a Macintosh computer equipped with a LAB-MIO-16XL-42 16 bit A/D I/O board (National Instruments, Austin, TX) and up to four battery-powered fourchannel high Z interface modules with built-in low-pass filters (World Precision Instruments, Sarasota, FL) controlled by LabView software (National Instruments) at a software adjustable gain of 10. To minimize the noise, 500 consecutive EMF data values were acquired per 5-s measurement interval and averaged. Typically, the EMF value was calculated as the mean of the individual data points over the last minute of measurement. The standard deviations (s) for the reproducibility measurements were found using this calculated EMF value. When the response was traced over the entire measurement, the averaged data from each measurement interval were plotted versus time. The pH calibration curves were made by placing the conditioned sensors into a 100mL background solution of PBS adjusted to pH 11, with and without 100 units/mL heparin. A pH sensor was used to monitor the change in pH as HCl was added, and the data were saved at approximately every 0.25 pH unit down to pH 2. Heparin Calibrations. After stripping the heparin from the membrane in a pH 10 PBS solution followed by reconditioning in PBS at pH 7.4, a classical calibration curve was obtained by consecutively adding small aliquots from a heparin stock solution (1.58 × 10-4 M) to a 50-mL PBS sample maintained at physiological pH of 7.4. Blood heparin calibrations were made by adding

aliquots of a diluted heparin stock solution (3.16 × 10-5 M) to a 10-mL sample of undiluted whole blood. For each calibration curve, the change in EMF resulting after each addition was plotted against the log of the sample heparin concentration. Reproducibility and Stripping Procedure. Newly prepared and conditioned electrodes were stabilized in 50 mL of a pH 7.4 PBS background solution for 5 min, followed by a 5-min exposure to a 3.2 × 10-6 M (8 units/mL) heparin sample, either in PBS buffer or in whole blood. The heparin was then stripped from the membranes for 2.5 min using pH 10 PBS, followed by reconditioning the membranes in pH 7.4 PBS for 2.5 min. This procedure was repeated four times. After the sensor was introduced to the sample but before beginning the 5-min measurement, the sample was hand stirred to remove possible air bubbles from the sensor surface by gently swirling the sample container. This time period was measured and was approximately 15 s for the PBS and 45 s for the whole blood sample. The difference in these times was due to the higher viscosity and opaque appearance of the blood sample. RESULTS AND DISCUSSION Recently, electrically charged H+ carriers have been studied in ISE membranes and shown to be adequate for pH-sensing applications.16,17 These membranes must contain an ionic additive of the same charge as the target ion for optimum selectivity.19 Solvent polymeric membranes incorporated with either the charged carrier ETH 7075 or ETH 2412, along with the cationic additive TDMAC, exhibited Nernstian pH responses with different dynamic ranges.17 The membranes containing ETH 2412 showed an adequate pH measuring range at high pH values with an upper detection limit around pH 8 due to anion interference. The membrane is therefore expected to give a typical anion response at lower pH values where anion interference dominates the response, including the physiological pH range. Interestingly, since at low pH values all ionophore molecules would be protonated and electrically neutral, no electrostatic interactions between the ionophore molecules with either TDMA+ or heparin are expected (see Figure 1). Such a TDMAC-containing membrane is therefore expected to maintain a high heparin selectivity over chloride at physiological chloride levels. In theory, this membrane should give a response to heparin at neutral pH. This principle is well supported by the use of the same membrane components in an optical film that selectively coextracts heparin into the film at physiological pH.20 The membrane is also expected to expel heparin when the pH at the sample-membrane interface is raised to levels well within the Nernst pH response region (see Figure 3). The pH response curve for a sensor containing the pH crosssensitive heparin-sensing membrane is plotted in Figure 4 and shows a Nernst pH response up to about pH 8.5, where the onset of anion interference is observed. The top and bottom traces represent the calibrations in a PBS-buffered 0.1 M NaCl background solution and with an additional 100 units/mL heparin. At pH above 9, the response is determined by the sample pH only, and neither chloride nor heparin shows affinity for the sensing membrane. At lower sample pH, hydrogen ions are coextracted (19) Schaller, U.; Bakker, E.; Spichiger, U. E.; Pretsch, E. Anal. Chem. 1994, 66, 391. (20) Wang, E. J.; Meyerhoff, M. E.; Yang, V. C. Anal. Chem. 1995, 67, 522.

Figure 4. Chloride (upper curve) and heparin (lower curve) interference observed in the pH calibration of a PVC-DOS (2:1) heparin sensor containing the H+ ionophore ETH 2412 in PBS solution. Heparin response is observed only below pH 8.5. At pH values above 9.5, heparin may be quantitatively removed from the membrane.

with the appropriate anion into the membrane. The sample containing heparin shows an earlier onset of anion interference, which must be due to the coextraction of heparin and hydrogen ions into the membrane. The difference in EMF between the chloride and heparin interference regions suggests that the sensor will show adequate response to heparin at physiological pH values. If the sample pH is returned to a high value above 9 after heparin exposure, any chloride and heparin anions are expected to be efficiently removed from the membrane together with hydrogen counterions in a codiffusion process, suggesting that heparin may be conveniently stripped from the membrane by increasing the sample pH. The maximum response of the sensor to heparin at physiological pH is predicted by Figure 4 to be about 40 mV. The heparin calibration of the sensor in PBS solution at physiological pH is shown in Figure 5 (top) as a typical sigmoidal polyion response curve with a maximum response of approximately 60 mV. The difference between the expected and actual maximum response values is largely due to the increased background Clconcentration in the pH calibration, where HCl was titrated to the solution. The heparin calibration shown in Figure 5 contains 0.1 M NaCl and therefore is believed to be the more accurate measurement. The measuring range of the sensor covers the full therapeutic heparin dosing range (2-10 units/mL). When this measurement was repeated in PBS solution at pH 10, the sensor gave no heparin response, indicating that no polyion was being extracted into the sensing membrane at that pH (data not shown). The same experiments were also performed with membranes having a low PVC content (PVC-DOS 1:2), which reflects the membrane composition of other small ion sensors.2 As with the classical heparin sensor,4 the measuring range was shifted by 1 order of magnitude to higher concentrations, which is less practical for blood heparin measurement since that measuring range falls outside of the therapeutic range. This behavior is well Analytical Chemistry, Vol. 71, No. 20, October 15, 1999

4619

Figure 5. (Top) Heparin response curve of the pH cross-sensitive heparin sensor (see Figure 4) in pH 7.4 PBS buffer with 0.1 M NaCl background. (Bottom) (H) Repetitive exposure of a PVC-DOS (2:1) sensor to identical 8 units/mL heparin samples in PBS background at pH 7.4, (S) 2.5-min heparin stripping with pH 10 PBS, and (R) 2.5-min reconditioning in pH 7.4 PBS leads to improved reproducibility of the sensor.

understood on the basis of the general steady-state response mechanism of this type of sensor, where a diffusion coefficient change in the membrane has a significant effect on the response characteristics.4 It is known that stripping the polyion from the membrane between sample measurements will make the sensor reversible.13,14 Effective stripping of heparin from the pH cross-sensitive heparin sensor is demonstrated in Figure 5 (bottom), which shows data from an 8 units/mL heparin measurement in a PBS background solution maintained at physiological pH. The heparin sample was measured four times followed by a 2.5-min stripping period using PBS at pH 10 and by a 2.5-min reconditioning time in PBS at pH 7.4 between each measurement. The plot shows that both the baseline potential (R, top points) and the heparin response (H, middle points) are reproducible if the heparin is stripped from the membrane between measurements (S, lowest points). The drop in EMF observed upon stripping is due to the decrease in hydrogen ion concentration at the membrane surface followed by the flux of hydrogen ions and coextracted heparin ions from the membrane to the stripping solution. When allowed to recondition for 1 h at the end of the measurement, the potential eventually returned near the initial baseline potential. An analogous reproducibility measurement was performed using the established heparin sensor where the heparin was stripped from the membrane with 2 M NaCl.14 Although the 4620 Analytical Chemistry, Vol. 71, No. 20, October 15, 1999

Figure 6. (Top) Heparin response curve of the pH cross-sensitive heparin sensor (see Figure 4) in human undiluted whole blood. (Bottom) (H) Repetitive exposure of a pH cross-sensitive PVC-DOS (2:1) heparin sensor to identical 8 units/mL heparin samples in undiluted whole blood, (S) 2.5-min heparin stripping with pH 10 PBS, and (R) 2.5-min reconditioning in pH 7.4 PBS buffer leads to improved reproducibility of the sensor.

stripping significantly improved the reproducibility of the sensor, giving a standard deviation (s) of (6.9 mV versus the classical sampling method with s ) (17 mV, the pH cross-sensitive heparin sensor shows a much enhanced reproducibility with s ) (0.45 mV. This behavior suggests that the pH stripping method is removing heparin from the membrane surface more efficiently than the traditional NaCl stripping procedure. It must be emphasized that the stripping time was kept short for practical reasons, i.e., to minimize the time between sample readings. Complete removal of heparin will likely require stripping times that are longer than uptake times, as recently discussed for a pulsed amperometric detection mode of these types of sensors.21 The stripping procedure is thought to be effective as long as the relevant diffusion region in the membrane is renewed. Residual heparin that has diffused further into the bulk of the membrane seems to have negligible influence on sensor performance. The sensor also responds to heparin in undiluted human whole blood, as shown in Figure 6 (top). The heparin calibration also resulted in a sigmoidal response curve with a measuring range encompassing the therapeutic heparin range. The maximum response in blood was about 30 mV, which indicates the presence of interferences from other blood components. Figure 6 (bottom) shows data from an 8 units/mL heparin measurement in human (21) Jadhav, S.; Meir, A. J.; Bakker, E. Anal. Chem., submitted.

undiluted whole blood that was repeated four times with a 2.5min stripping period using PBS at pH 10 followed by a 2.5-min reconditioning time in PBS at pH 7.4 between each measurement. The individual blood heparin measurements appear to reach steady state more rapidly than the PBS sample measurements, but this is very likely due to the longer stirring time before beginning the 5-min measurement in the blood sample (see Experimental Section). The results show remarkable measurement reproducibility, with a calculated s ) (0.39 mV. The pH cross-sensitive heparin sensor provides possibilities for automated stripping procedures that were not feasible with the original heparin sensor. We envision that uptake and removal of heparin may now be accomplished by local changes in pH at the sensing surface, irrespective of the bulk sample pH. There are several possibilities for changing the local pH at a membrane that are being investigated in this laboratory. One such method is by transmembrane diffusion of an acid from the internal reference solution to the outer membrane surface. While it is known that the transmembrane diffusion of electrolytes is an important phenomenon that may have a dramatic effect on the detection limit of conventional small ion ISEs,22 our group has also observed local changes in pH at the sensing surface that may be accomplished by adding simple, highly concentrated acids such as acetic acid to the internal solution of the sensor.23 The pH decrease is explained by a rapid diffusion of acid across the (22) Mathison, S.; Bakker, E. Anal. Chem. 1998, 70, 303. (23) Jadhav, S.; Bakker, E. Electrochem. Solid-State Lett. 1998, 1, 194.

membrane to the sample where it lowers the local pH even in well-buffered samples. Further experiments will determine whether the pH changes will occur in the physiological buffer range and whether a pH cross-sensitive heparin sensor will also display this effect. Other possibilities include electrochemical means to change the pH at the sample-membrane interface, for example, under galvanostatic control at a platinum electrode located near the heparin sensing surface. CONCLUSIONS pH cross-sensitive heparin-sensing membranes that contain electrically charged H+ ionophores such as ETH 2412 give an analytically useful response to heparin in both PBS and human undiluted whole blood. Heparin removal may be accomplished by adjusting the pH at the membrane surface to higher values, improving sensor reversibility and measurement reproducibility. The development of pH cross-sensitive heparin sensors will make it more feasible to fabricate reliable sensor devices for continuous bedside heparin monitoring. ACKNOWLEDGMENT The authors thank the National Institutes of Health (Grant R01GM58589) for financial support. Received for review April 13, 1999. Accepted August 2, 1999. AC990387S

Analytical Chemistry, Vol. 71, No. 20, October 15, 1999

4621