Direct Ion Speciation Analysis with Ion-Selective Membranes

Sep 20, 2012 - lished to give direct information on the so-called free ion concentration .... and its square root is ideally proportional to the sampl...
0 downloads 0 Views 1MB Size
Download PDF
Article pubs.acs.org/ac

Direct Ion Speciation Analysis with Ion-Selective Membranes Operated in a Sequential Potentiometric/Time Resolved Chronopotentiometric Sensing Mode Majid Ghahraman Afshar,† Gastón A. Crespo, and Eric Bakker* Department of Inorganic and Analytical Chemistry, University of Geneva, Quai Ernest-Ansermet 30, CH-1211 Geneva, Switzerland S Supporting Information *

ABSTRACT: Ion-selective membranes based on porous polypropylene membranes doped with an ionophore and a lipophilic cation-exchanger are used here in a new tandem measurement mode that combines dynamic electrochemistry and zero current potentiometry into a single protocol. Open circuit potential measurements yield near-Nernstian response slopes in complete analogy to established ion-selective electrode methodology. Such measurements are well established to give direct information on the so-called free ion concentration (strictly, activity) in the sample. The same membrane is here also operated in a constant current mode, in which the localized ion depletion at a transition time is visualized by chronopotentiometry. This dynamic electrochemistry methodology gives information on the labile ion concentration in the sample. The sequential protocol is established on potassium and calcium ion-selective membranes. An increase of the ionophore concentration in the membrane to 180 mM makes it possible to determine calcium concentrations as high as 3 mM by chronopotentiometry, thereby making it possible to directly detect total calcium in undiluted blood samples. Recovery times after current perturbation depend on the current amplitude but can be kept to below 1 min for the polypropylene based ion-selective membranes studied here. Plasticized PVC as membrane material is less suited for this protocol, especially when the measurement at elevated concentrations is desired. An analysis of current amplitudes, transition times, and concentrations shows that the data are described by the Sand equation and that migration effects are insignificant. A numerical model describes the experimental findings with good agreement and gives guidance on the required selectivity in order to observe a well-resolved transition time and on the expected errors due to insufficient selectivity. The simulations suggest that the methodology compares well to that of open circuit potentiometry, despite giving complementary information about the sample. The tandem methodology is demonstrated in a titration of calcium with nitrilotriacetic acid (NTA) and in the direct detection of calcium in undiluted heparinized and citrated blood.

P

sense these two analytical parameters in undiluted/unmodified blood samples is still not available. Clinical standard methods are based on atomic absorption spectroscopy, volumetric EDTA titrations with murexide as an indicator, or potentiometric detection with ion selective electrodes.5−7 All the methods permit the determination of total calcium in blood, but an acid pretreatment and/or dilution of the sample is required. Depending on the selected official method, between 2 and 3 h are needed to quantify an unknown sample. Pretsch and co-workers have demonstrated the possibility to obtain information on free and total calcium ion concentrations by direct potentiometry in the presence of labile sample complexes.8−10 Diminishing the concentration of primary analyte in the inner filling solution induced strong calcium ion fluxes from the sample in the direction of this inner

otentiometric ion-selective electrodes are established tools as a routine methodology in clinical diagnostics for the determination of small target ions.1,2 In potentiometry, the recorded potential is, in the absence of current flow through the electrochemical cell (also called open circuital potential), ideally a function of the ion activity in the sample phase. 3 Potentiometric ion sensors have a unique characteristic over other electrochemical sensors, which is the ability to give direct information on so-called free or uncomplexed ions activities, even in complex matrixes such as blood or urine.4 This chemical information may be employed in conjunction with complementary analytical techniques to obtain more detailed speciation information on the sample (for instance on the levels of free and total calcium in whole blood). The generation of inexpensive, fast, and reliable chemical information on free and total calcium concentration would allow one to take more appropriate clinical decisions. Ionized calcium participates in many processes such as human growth and blood coagulation. Sensor technology that would be able to © 2012 American Chemical Society

Received: August 3, 2012 Accepted: September 20, 2012 Published: September 20, 2012 8813

dx.doi.org/10.1021/ac302092m | Anal. Chem. 2012, 84, 8813−8821

Analytical Chemistry

■

solution. This resulted in a local ion depletion at the sample− membrane phase boundary, which was detectable by potentiometry and gave information on total calcium levels. Subsequent work aimed at improving the repeatability and reliability of this concept by controlling the above-mentioned ion fluxes by galvanostatic means. Small currents on the order of nanoamps applied to permselective ion-selective membranes were shown to enhance some analytical sensing parameters such as the limit of detection.11,12 Our group introduced the pulstrode concept, which typically uses current pulses of larger amplitude applied across ionophore-based polymeric membranes.13−15 A constant current pulse forces a defined flux of primary analyte in direction of the membrane, while the associated potential change is monitored over time. Operational reversibility was achieved by an applied potentiostatic pulse, which regenerates the membrane. This protocol used membranes containing suppressed ion-exchanger properties and gave rise to nonlinear (sigmoidal) calibration curves, in analogy to their potentiometric membrane counterparts originally introduced by Pretsch. More recently, the same class of ion-selective membranes was characterized by chronopotentiometry in a protocol analogous to that originally described by Sand at metal electrodes.16,17 The applied current results in the extraction of analyte ion from the sample into the membrane and gives rise to a localized depletion at a transition time (described by the Sand equation). This transition time is detected as an inflection point of the chronopotentiometric readout signal, and its square root is ideally proportional to the sample concentration. Even though this methodology is very attractive, it was not yet possible to apply it to the detection of total calcium in undiluted blood samples (∼2.3 mM).17 The upper limit of detection was found at ∼0.7 mM of calcium and was explained by the relatively low mobility of the components in the ion-selective membranes used at the time. Much of this limitation originates in the use of nonpermselective membranes, where a counterion must be extracted from the inner solution side of the membrane for every ion transferred at the sample side. These two counterions should not be given the time to meet within the membrane, placing limits on the choice of the membrane material. We have recently demonstrated that supported liquid permselective membranes can be used to design a reversible chronopotentiometric sensor for the determination of polycations in blood samples.18 The utilization of thin polypropylene membranes (25 μm thickness) improves the mobility of the membrane components, thereby resulting in a lower membrane resistance and higher transport efficiency. This is supported by other recent results with this class of membranes that include thin layer coulometric ion-selective electrodes,19 electrochemiluminescence imaging sensors,20,21 and membranes used in current controlled backside calibration potentiometry.22,23 We report a new tandem methodology that combines potentiometry and chronopotentiometry techniques, allowing one to measure free and total calcium concentration using the same sensing membrane. Moreover, permselective membranes based on a high concentration of calcium ionophore exhibits a linear measuring range up to 3 mM, suitable for the determination of total calcium in undiluted blood samples. Numerical simulations are in good qualitative agreement with the obtained experimental results.

Article

EXPERIMENTAL SECTION

Reagents and Solutions. Potassium tetrakis[3,5-bis(trifluoromethyl)phenyl]borate (KTFPB), sodium tetrakis[3,5-bis(trifluoromethyl)phenyl]borate (NaTFPB), tetrakis(4chlorophenyl)borate tetradodecylammonium salt (ETH 500), potassium ionophore I, calcium ionophore IV, 2-nitrophenyloctylether (o-NPOE), bis(2-ethylhexyl)sebacate (DOS), high molecular weight poly(vinyl chloride) (PVC), 2amino-2-hydroxymethyl-propane-1,3-diolhydrochloride (Tris·HCl), nitrilotriacetic acid (NTA) sodium chloride, sodium hydroxide (1 M), and tetrahydrofuran (THF) were purchased from Sigma-Aldrich (analytical grade). All experiments were performed in Tris buffer (10 mM buffer at pH 7.4 + 100 mM NaCl), unless otherwise indicated. Heparinized human blood (added heparin, 1 unit mL−1) and citrated human blood (concentrated in red cells) were obtained from Hôpitaux Universitaires de Genève (HUG). Electrochemical Equipment. A double-junction Ag/ AgCl/3 M KCl/1 M LiOAc reference electrode was used in potentiometric and chronopotentiometric measurements (Mettler-Toledo AG, Schwerzenbach, Switzerland). Electrode bodies (Oesch Sensor Technology, Sargans, Switzerland) were used to mount the polymeric membranes. A platinum working rod (3.2 cm2 surface area) was used as a counter electrode. Selectivity coefficients were determined by zero current potentiometry employing a high impedance input 16-channel EMF monitor (Lawson Laboratories, Inc., Malvern, PA). Potentiometric, chronopotentiometric, and electrochemical impedance spectroscopy measurements were performed with an Autolab PGSTAT302N (MULTI 16 module, Metrohm Autolab, Utrecht, The Netherlands) that allows one to read up to 16 working electrodes placed in the same electrochemical cell. A Faraday cage was used to protect the system from undesired noise. Membrane Preparation. Potassium PVC-based membranes were prepared in the classical manner using a 1:2 mass ratio of PVC and plasticizer (1:2). Specifically, 15 mmol kg−1 of potassium ionophore I, 5 mmol kg−1 of NaTFPB, 20 mmol kg−1 of ETH 500, 63 mg of PVC, and 127 mg of DOS were completely dissolved in THF (membrane preparation details in the Supporting Information). Porous polypropylene (PP) membranes (Celgard, 0.237 cm2 surface area and 25 μm thickness and kindly provided by Membrana, Wuppertal, Germany) were used as supporting material. The used cocktail for the impregnation of PP membranes contained all the reagents mentioned before except PVC. Membrane K1 contained 15 mmol kg−1 of ionophore I, 5 mmol kg−1 of NaTFPB, 20 mmol kg−1 of ETH 500, 190 mg of DOS, and 1 mL of THF. Calcium PP membranes were optimized in order to increase the upper limit up to 3 mM, which is the highest concentration to be found in undiluted blood. Therefore, different membranes varying the ionophore concentration were evaluated. PP-Ca1 (15:5:90; the values denote the concentrations of active ingredients in mmol kg−1: 15 mmol kg−1 of calcium ionophore IV, 5 mmol kg−1 of NaTFPB, 90 mmol kg−1 of ETH 500, and o-NPOE up to 100 mg of total cocktail amount), and in a similar manner PP-Ca2 (30:5:90), PP-Ca3 (50:5:90), PP-Ca4 (70:5:90), PP-Ca5 (90:5:90), PP-Ca6 (120:5:90), PP-Ca7 (150:5:90), and PP-Ca8 (180:5:90). 8814

dx.doi.org/10.1021/ac302092m | Anal. Chem. 2012, 84, 8813−8821

Analytical Chemistry

■

Article

RESULTS AND DISCUSSION The main goal of this paper is to provide a sensing tool that may accomplish a direct ion speciation in important samples, i.e., blood, by potentiometry and chronopotentiometry. For this purpose, we introduce here for the first time permselective membranes that are able to operate in both electrochemical modes. Figure 1 schematically illustrates the operational

at the membrane side of the same interface, but the latter change is small compared to the initial analyte concentration in the membrane and has only a small influence on the observed potential. The potential is recorded as a function of time (chronopotentiometry) during the constant current pulse and is chiefly dependent on the aqueous concentrations at the sample−membrane interface. The transition time (τ) is found as the inflection point of the chronopotentiometric signal. After this transition time, the selectivity is expected to break down, and a background cation is extracted along with the analyte ion to maintain the imposed ion flux. This mixed ion response now results in a decreased membrane potential. After each chronopotentiometric determination, a potentiostatic pulse is applied for 30 s at the open circuital potential determined before the current pulse. This is to regenerate the membrane to the unperturbed situation shown in Figure 1a. Potassium ion-selective membranes based on a plasticized PVC membrane were first characterized as an initial model system. Figure 2a shows the potential variation as a function of time for different potassium concentrations in the range of 0.01−0.07 mM at a cathodic current of −5 μA. The transition time is visualized as the maximum of the time derivative of the potential in Figure 2b. The observed transition time increases with higher potassium ion concentrations. The relationship between the square root of transition time (τ1/2) and added potassium concentration at different applied current levels is shown in Figure 3a. As expected from the Sand equation, a linear behavior is found for all applied currents from −4.0 to −17.5 μA. The data do not display a deviation from linearity with increasing applied current values. However, the sensitivity (Δτ1/2/Δc) significantly decreases with increasing applied current amplitudes. The higher the driving force, the faster the observed depletion, giving rise to a more apparent overlap of the transition times under these conditions. Figure 3b confirms the expected linear relationship between the inverse square root of transition time and the applied current at constant levels of potassium concentration. This suggests that diffusion is the predominant mechanism of transport for the conditions used here and that effects of electrical migration can be neglected. Even though migration is expected to become more dominant at high current amplitudes, there is no evidence of this in the current range used here. The influence of lipophilic salts is well established, and the addition of ETH 500 on this methodology was also evaluated. Indeed, the presence of ETH 500 is required to avoid a high membrane resistance, which would otherwise result in a large ohmic potential drop. Membranes without added ETH 500 did not exhibit a measurable transition time with this methodology at current amplitudes larger than −15 μA (data not shown). Numerical simulations of potassium ion-selective membranes under galvanostatic conditions were performed (see the theory in the Supporting Information), giving predictions that are in good agreement with observed experimental data, see Figure 2. In a first approach, the model is used with the aim of obtaining semiquantitative information about position and height of the peaks at the transition time. The potential and its time derivative are plotted in Figure 2c,d as a function of time. The shape of the predicted curves is very similar to the experimental ones, with potential steps of around 200 mV in both cases. Note that the experimental data show a decreasing baseline potential change, even in the absence of analyte (red line, Figure 2a), which is not predicted by the model. It is attributed to non-Faradaic charging currents,24 and the experimental slope

Figure 1. Schematic illustration of the calcium sensing mechanism proposed here. (a) The sensing membrane contains ionophore (L), cation exchanger (R−) and lipophilic salt (R+R−). The calcium concentration in the membrane before electrochemical perturbation is denoted with a dotted line. For the same situation, the calcium concentration in the aqueous phase (bulk) is close to that in the phase boundary (bp). (b) An applied current provokes a defined calcium flux across the permselective membrane. This calcium flux is described by Fickian diffusion and can be fully sustained up to a transition time τ. At this time calcium is depleted at the sample side of the phase boundary and results in an observed potential change. The associated accumulation of calcium on the left side of the membrane during the pulse is stabilized by supramolecular interaction with the ionophore.

principle of calcium ion selective membrane sandwiched between an inner solution and a sample solution. The membrane contains an electrically neutral ionophore (L), a cation exchanger (R−), a lipophilic salt (R+R−), and plasticizer. The potential difference between the indicator electrode (a Ag/ AgCl element immersed in the inner solution of the membrane containing a 10 mM calcium chloride concentration) and a commercial reference electrode placed into the sample phase is related to analyte concentration changes in the course of the potentiometric experiment (see Figure 1a). It is particularly important to emphasize that the presence of the cation exchanger allows one to avoid the extraction of counterions, in analogy to established ion-selective membranes, keeping the primary analyte activity in the membrane phase ideally indifferent to that of the sample. As with other ISEs, a Nernstian response slope toward the analyte ion is expected with these types of membranes. When a galvanostatic pulse of ∼5 s duration is applied across the electrochemical cell (see Figure 1b), the transport of the analyte from the sample into the membrane and from its backside to the inner solution is imposed with a defined flux. The analyte depletion at the aqueous side of the sample− membrane phase boundary is accompanied by an accumulation 8815

dx.doi.org/10.1021/ac302092m | Anal. Chem. 2012, 84, 8813−8821

Analytical Chemistry

Article

Figure 2. Chronopotentiometric response for PVC potassium selective membranes. (a) Potential decay as a function of time for different potassium levels at an applied cathodic current of −5 μA. The red (left most) line corresponds to the background electrolyte in the absence of potassium (10 mM NaCl). (b) Observed time derivatives of the chronopotentiometric responses upon successively increasing the final potassium concentration from 0.01 to 0.07 mM. (c) Predicted chronopotentiometric response for experimental data shown in part a. Used parameters: membrane area (0.0323 dm2), potassium diffusion coefficient (1.96 × 10−7 dm2 s−1), selectivity coefficient (−log KKNapot = −4.5), background electrolyte concentration (0.01 mol dm3), and applied current (−5 × 10−6 A). (d) Predicted time derivatives of the chronopotentiometric responses on successively increasing the final potassium concentration from 0.02 to 0.07 mM from part c.

with ion-selective electrodes predicts an error of 1% for the same selectivity. The poorest selectivity where a transition time could still be resolved in the simulation corresponded to a value of log(aI/KIJpotaJzI/zJ) = 1.40 and gave an error of 6.7% (potentiometry predicts a 4% error). Figure 4c directly compares the relative errors for concentrations found on the basis of chronopotentiometry (simulated data and fit) and potentiometry (calculated based on the Nicolsky equation). One may therefore reasonably state that the chronopotentiometry methodology is expected to give similar systematic errors to the same membranes interrogated by traditional direct potentiometry. The observed linear range for potassium-selective plasticized PVC membranes is rather narrow (0.01−0.07 mM at −5 μA). Subsequently, materials such as polypropylene membranes that are thinner and allow for a faster diffusion of the components were considered as suitable candidates in order to extend the upper detection limit (see below). Figure 1S (see the Supporting Information) shows the transition peaks for polypropylene based potassium ion-selective membrane at

of 25 mV/s translates into a capacitance of 200 nF. The position (transition time) and the height of the peak (which is a function of the membrane selectivity) in the time derivative plot (Figure 2d) are in close agreement with the experimental data set. As suggested in the theoretical section of the Supporting Information, resolving the differential diffusion equation for this particular case by numerical simulation and using the correct boundary conditions, it is possible to estimate the chronopotentiometric behavior of the membrane as a function of the experimental selectivity coefficient. This allows one to estimate the required minimum selectivity to obtain a measurable analytical signal. Figure 4 shows how diminishing the membrane selectivity results in an increasingly less distinct peak maximum of the time derivative of the potential transient. Figure 4a,b plots the calculated transition time as a function of zI/zJ −1 selectivity, shown as aI(Kpot IJ aJ ) . An error in the square root of the transition time (and hence equal to the error in the concentration measurement) of just −0.5% is predicted for zI/zJ log(aI/Kpot IJ aJ ) = 2.0. In comparison, direct potentiometry 8816

dx.doi.org/10.1021/ac302092m | Anal. Chem. 2012, 84, 8813−8821

Analytical Chemistry

Article

Figure 3. (a) Observed linear calibration curve of the square root of the transition time τ1/2 as a function of potassium concentration at the indicated current levels (−4 to −17.5 μA). (b) Observed linear calibration curve of the inverse square root of the transition time τ−1/2 as a function of applied current at the indicated potassium concentrations (0.01−0.07 mM).

two current levels (−5 and −10 μA). Unfortunately, using these types of membranes the linear range was increased up to 0.11 mM, which is only slightly higher than the value obtained with PVC at the same current level (0.07 and 0.09 mM at −5 and −10 μA, respectively). Note, however, that supported membranes for calcium were successfully used to extend the use of these sensors at higher current amplitudes (to about −100 μA, see below). A linear relationship between the square root of transition time and the concentration was found in both cases (see the inset of Figure 1S in the Supporting Information). Unfortunately, an increase of the current amplitude to −10 μA did not result in a measurable higher concentration of potassium. Instead, an unspecific peak at 2.2 s (Figure 1S-b in the Supporting Information) appears at all concentration levels. Changes in the background electrolyte concentration (1 order of magnitude concentration increase or decrease) did not alter the observed trend (data not shown). This evidence suggests that the nature of the peak is related to a limited transport process confined to the membrane. As suggested earlier, a local ionophore depletion at the sample/

Figure 4. (a) Predicted chronopontentiometric responses as a function zI/zJ −1 of the membrane selectivity, given as r = (aI/Kpot IJ aJ ) , where I and J are the analyte and interfering ion with the associated charges zI and zJ, respectively, and Kpot IJ is the selectivity coefficient. Under a value of r = 20, the transition time can no longer be quantified. (b) Selectivity dependent time derivatives for the data given in part a. (c) Associated relative errors for the concentrations found on the basis of chronopotentiometry (simulated data) and compared to associated errors predicted in zero current potentiometry as calculated based on the Nicolsky equation (solid line).

membrane phase by the inward potassium flux may explain the findings,16 but perhaps the operational limit placed by a limited solubility of the potassium ionophore I is responsible for 8817

dx.doi.org/10.1021/ac302092m | Anal. Chem. 2012, 84, 8813−8821

Analytical Chemistry

Article

Figure 5. Observed calcium calibration curves for the optimized polypropylene calcium membrane PP-Ca8 sequentially measured by (a) potentiometry and (b) chronopotentiometry.

observed behavior. Therefore, the ionophore composition was optimized in order to obtain higher upper detection limits. In view of demonstrating an application of practical importance, these steps were performed for the detection of calcium, and further work on potassium was not pursued. With respect to realizing a direct potentiometry detection mode in sequence with the chronopotentiometry methodology, polypropylene membranes were found to be about an order of magnitude faster to reach the same open circuital potential before and after the galvanostatic pulse than plasticized PVC. For plasticized PVC membranes, excessively large potential values of 7−8 V were sometimes found at higher current amplitudes. This characteristic makes these materials less suitable to achieve a sequential switching between potentiometric and chronopotentiometric sensing modes. In view of demonstrating a case of practical relevance, further work focused on the detection of ionized and total calcium by a sequential potentiometry/chronopotentiometry protocol. Human undiluted blood samples contain between 2.2 and 2.8 mM of calcium. The required upper detection limit for calcium by chronopotentiometry is rather high, and a careful optimization of the ionophore content in calcium-selective membranes was performed. Table 1S in the Supporting Information summarizes the obtained results for eight different compositions. All characterized membranes displayed a sampleindependent peak that limits the detection of higher sample concentrations with that membrane. By increasing the ionophore concentration in the organic phase, a higher upper detection limit was indeed obtained (see Figure 2S in the Supporting Information, optimization of membrane composition). Figure 5 shows the chronopotentiometric and potentiometric behavior of the optimized calcium membrane (PP-Ca8). The measurements were performed in physiological conditions. The average potentiometric response (n = 4) measured at a time of 30 s after current perturbation is shown in Figure 5a. The observed slope is 2 mV lower for these electrodes than the control electrode where no current perturbation took place (26.04 mV). The small difference in the slopes may be explained due to changes in the calcium levels in the membrane phase provoked by the applied current. Nonetheless, the displayed variation does not seem to have a deteriorating

influence on the accuracy of the potentiometry results. The control electrode showed a repeatability of recorded potentials of 0.05−0.08 mV (SD, n = 10) under ambient laboratory conditions for a 2.50 mM calcium sample, whereas for electrodes subjected to chronopotentiometric perturbation a ∼2-fold higher variation was found at 0.10−0.15 mV under otherwise the same conditions. The recovery time to reach the same open circuit potential as the one observed before the applied current was found to be strongly dependent on the magnitude of the current. At lower currents (−10−80 μA), the recovery time is quite rapid (5−10 s). At higher currents (−100−150 μA), the recovery times become longer, reaching values of 60−90 s. These times could be further reduced with a dedicated potentiostatic or galvanostatic optimization, as recently suggested elsewhere.25,26 As described in the theory, the local depletion is established on the basis of diffusion process in the aqueous layer. Mass transport induced by additional migration processes could alter the linear relationship between current, concentration, and square root of transition time (see Figure 3S in the Supporting Information, raw data and optimization of applied current). This was evaluated by normalizing the square root of the transition times by multiplying them with the applied current and plotting these values as a function of the analyte concentration. Ideally, the slope now only depends on the area of the membrane and the diffusion coefficient of calcium in the aqueous phase. Figure 5b demonstrates that the experimental data form indeed a linear relationship and are well described by the Sand equation. There is a slight deviation from ideal behavior for the two highest concentrations, but the obtained diffusion coefficient is 1.23 × 10−5 cm2 s−1 and is in agreement with reported values in the literature.27 In an attempt to evaluate the feasibility to determine free and total calcium with the same membrane, free and total calcium levels were investigated in the presence of the ligand nitrilotriacetate (NTA), which is known to form labile CaNTA complexes. Figure 6a shows the depletion signal for different calcium levels as a function of pulse time. After a 3 mM concentration of calcium has been incrementally added to the sample, titration with NTA is performed as shown in Figure 6b. The transition peaks move toward lower calcium concentrations, decreasing from 3 mM to 1.7 mM. This 8818

dx.doi.org/10.1021/ac302092m | Anal. Chem. 2012, 84, 8813−8821

Analytical Chemistry

Article

Figure 7. Determination of total and free calcium in undiluted heparinized blood. (a) Chronopotentiometric mode: Using standard addition calibration to minimize matrix effects, the peaks shift to higher transition times when the calcium concentration is increased, as expected. The red line (left most curve) corresponds to undiluted blood without adding calcium. Inset: Observed linear calibration curve of the square root of the transition time τ1/2 as a function of added calcium concentration (from 0.1 to 0.45 mM). (b) Potentiometric mode: Determination of free calcium by external calibration. The red point corresponds to the value found in the blood sample.

Figure 6. Determination of free and total calcium concentration by the proposed methodology. (a) Calibration curve for incremental additions of calcium as measured by chronopotentiometry, (b) subsequent titration with NTA as measured by chronopotentiometry. (c) Potentiometric (left y-axis scale) and chronopotentiometric (right y-axis scale) responses as a function of increasing NTA additions. Error bars correspond to triplicate measurements from three different membranes used in the same experiment.

Figure 7a shows the standard addition calibration (τ1/2 (s) = 1.36 + 0.77cCalcium [mM]), giving a total calcium concentration of 1.8 ± 0.1 mM (SD, n = 3). The total calcium concentration in the same blood sample was obtained as 1.9 ± 0.1 mM (SD, n = 3) by EDTA titration after 10-fold dilution and using murexide as a visual indicator as a comparative, established methodology. This is in quantitative agreement with the total calcium levels observed with the chronopotentiometric protocol on the undiluted sample. In comparison, an external calibration (calcium chloride in water) gave a calcium concentration of 1.7 ± 0.2 mM (SD, n = 3). Matrix effects appear therefore to be reasonably unimportant. Free calcium was also determined by direct potentiometry with the same membranes, as shown in Figure 7b. External potentiometric calibration (emf (mV) = 163.1 + 25.1 log cCalcium) predicts 0.7 ± 0.1 mM of ionized calcium which is somewhat smaller than typically expected (∼1 mM), which may be due to the weak interactions between ionized calcium and heparin. The same protocol was employed to measure a different sample of human blood treated with citrate. The predicted total calcium in this sample with the proposed methodology was 2.2 ± 0.3 mM (see Figure 5S in the Supporting Information). Using EDTA titration (see above), total calcium concentration

decrease in observed total calcium concentration is explained on the basis of a lower diffusion coefficient of the NTA-Ca complex relative to that of the hydrated calcium ion as suggested earlier for EDTA-Ca and DCTA-Ca systems.28 Figure 6c shows the chronopotentiometric and the potentiometric results for the same membrane. The contrasting behavior of both responses confirms that the sensor may be operated in both modes. Potentiometric titration simulation under physiological conditions is shown in Figure 4S in the Supporting Information and compares well to the experimental data for that mode of detection. Having established that permselective membranes may be used for a combined potentiometric/chronopotentiometric detection methodology in a wide calcium concentration range, undiluted heparinized human blood was evaluated using this electrochemical protocol. Figure 7a shows the time derivative of the chronopotentiometric response of calcium depletion in such undiluted whole blood. The left most peak corresponds to the unmodified blood sample. A calibration by standard addition was performed in order to avoid matrix effects. The positions of the peaks shift to higher transition times when precise aliquots of calcium are added. The inset of 8819

dx.doi.org/10.1021/ac302092m | Anal. Chem. 2012, 84, 8813−8821

Analytical Chemistry

■

was found as 2.1 ± 0.3 mM (SD, n = 5), which is also in quantitative agreement with the obtained total calcium levels by chronopotentiometric protocol. The free calcium concentration was estimated by open circuit potentiometry with the same membranes, obtaining just 0.05 mM of ionized calcium, which is explained by the complexation of calcium by citrate. While these early studies are very encouraging and suggest that total and free calcium can be determined directly in the undiluted blood sample without modification, more work would need to fully establish this new methodology for acceptance in routine clinical practice. Specifically, the results from this chronopotentiometric protocol are expected to depend on the chemically bound form of calcium in the blood sample and its associated diffusion coefficient. Hence, calcium bound to biomacromolecules is expected to diffuse more slowly than free calcium and give rise to an experimental bias, but this might be mitigated by facile dissociation of these complexes in the Nernst diffusion layer near the membrane electrode. Moreover, as with other sensor approaches, the influence of the volume occupied by the red blood cells (hematocrit) may need to be taken into account by a simultaneous conductivity measurement for accurate comparisons with measurements obtained with diluted samples or with atomic spectrometric techniques.

■

CONCLUSIONS

■

ASSOCIATED CONTENT

Article

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Notes

The authors declare no competing financial interest. † On leave from Imam Khomeini International University, Qazvin, Iran.

■

ACKNOWLEDGMENTS The authors thank the Swiss National Science Foundation and the University of Geneva for supporting this research. We thank Dr. Till Saxer and Dr. Karim Bendjelid (HUG) for collaborating on this project and for providing citrated human blood bags. M.G.A. thanks Imam Khomeini International University for an international fellowship.

■

REFERENCES

(1) Bobacka, J.; Ivaska, A.; Lewenstam, A. Chem. Rev. 2008, 108, 329−351. (2) Bakker, E.; Crespo, G.; Grygolowicz-Pawlak, E.; Mistlberger, G.; Pawlak, M.; Xie, X. J. Chimia 2011, 65, 141−149. (3) Bakker, E.; Buhlmann, P.; Pretsch, E. Chem. Rev. 1997, 97, 3083− 3132. (4) Bakker, E.; Buhlmann, P.; Pretsch, E. Electroanalysis 1999, 11, 1088−1088. (5) Cali, J. P.; Bowers, G. N.; Young, D. S.; Gambino, S. R.; Gochman, N.; Mather, A.; Porter, C. J.; Pybus, J.; Rains, T. C.; Rand, R. N.; Sideman, L. Clin. Chem. 1973, 19, 1208−1213. (6) Elliott, W. E. J. Biol. Chem. 1952, 197, 641−644. (7) Kramer, B.; Howland, J. J. Biol. Chem. 1920, 43, 35−42. (8) Sokalski, T.; Ceresa, A.; Fibbioli, M.; Zwickl, T.; Bakker, E.; Pretsch, E. Anal. Chem. 1999, 71, 1210−1214. (9) Ceresa, A.; Pretsch, E.; Bakker, E. Anal. Chem. 2000, 72, 2050− 2054. (10) Ceresa, A.; Sokalski, T.; Pretsch, E. J. Electroanal. Chem. 2001, 501, 70−76. (11) Morf, W. E.; Badertscher, M.; Zwickl, T.; de Rooij, N. F.; Pretsch, E. J. Electroanal. Chem. 2002, 526, 19−28. (12) Lindner, E.; Gyurcsanyi, R. E.; Buck, R. P. Electroanalysis 1999, 11, 695−702. (13) Shvarev, A.; Bakker, E. Anal. Chem. 2003, 75, 4541−4550. (14) Makarychev-Mikhailov, S.; Shvarev, A.; Bakker, E. J. Am. Chem. Soc. 2004, 126, 10548−10549. (15) Shvarev, A.; Bakker, E. Talanta 2004, 63, 195−200. (16) Gemene, K. L.; Bakker, E. Anal. Chem. 2008, 80, 3743−3750. (17) Gemene, K. L.; Bakker, E. Anal. Chim. Acta 2009, 648, 240− 245. (18) Crespo, G. A.; Ghahraman Afshar, M.; Bakker, E. 2012, submitted to Angewandte Chemie, International Edition. (19) Grygolowicz-Pawlak, E.; Bakker, E. Anal. Chem. 2010, 82, 4537−4542. (20) Crespo, G. A.; Mistlberger, G.; Bakker, E. Chem. Commun. 2011, 47, 11644−11646. (21) Crespo, G. A.; Mistlberger, G.; Bakker, E. Electroanalysis 2012, 24, 61−68. (22) Malon, A.; Bakker, E.; Pretsch, E. Anal. Chem. 2007, 79, 632− 638. (23) Ngeontae, W.; Xu, Y.; Xu, C.; Aeungmaitrepirom, W.; Tuntulani, T.; Pretsch, E.; Bakker, E. Anal. Chem. 2007, 79, 8705− 8711. (24) Zook, J. M.; Langmaier, J.; Lindner, E. Sens. Actuators, B: Chem. 2009, 136, 410−418. (25) Zook, J. M.; Lindner, E. Anal. Chem. 2009, 81, 5146−5154. (26) Zook, J. M.; Lindner, E. Anal. Chem. 2009, 81, 5155−5164. (27) Harned, H. S.; Parker, H. W. J. Am. Chem. Soc. 1955, 77, 265− 266.

The tandem methodology introduced here provides a convenient ion speciation analysis of the sample without requiring complex instrumentation. Ionophores ordinarily used in direct potentiometry with ion-selective electrodes can be applied for the direct measurement of labile and so-called free ion concentrations. In order to guarantee rapid recovery times after perturbation, the use of doped porous polypropylene membranes is recommended here. Concentrations as high as 3 mM can be determined by chronopotentiometry by increasing the ionophore concentration in the membrane. First, examples of this methodology include the speciation of calcium during a titration with NTA and the detection of total calcium in undiluted heparinized blood samples. Matrix effects are found to be important for complex samples such as whole blood that also contain complexing agents, as variations in the diffusion coefficients for free and complexed ions may result in an uncertainty of the observed total concentration of the ion. For reliable transition times to be detected, a numerical simulation predicts that reasonable discrimination of potential interfering ions must be assured. The limit of detectability is estimated at a level of interference by ion-exchange of ∼4%. Theory also predicts that the error in concentration measurement due to ion-exchange with interfering species is no worse than that observed in direct potentiometry with the same membranes. A sequential switching between zero current potentiometry and chronopotentiometry is only possible after a recovery time of 10−90 s, the exact value of which depends on the amplitude of applied current. If more rapid sampling times are desired, they may perhaps best be achieved with two separate electrodes of otherwise the same composition.

S Supporting Information *

Theoretical section (numerical simulation), experimental details, and additional data. This material is available free of charge via the Internet at http://pubs.acs.org. 8820

dx.doi.org/10.1021/ac302092m | Anal. Chem. 2012, 84, 8813−8821

Analytical Chemistry

Article

(28) Mateo, P. L.; Hurtado, G. G.; Vidalabarca, J. B. J. Phys. Chem. B 1977, 81, 2032−2034.

8821

dx.doi.org/10.1021/ac302092m | Anal. Chem. 2012, 84, 8813−8821