Anal. Chem. 2003, 75, 3506-3511
The XT-Tube Extractor: A Hollow Fiber-Based Supported Liquid Membrane Extractor for Bioanalytical Sample Preparation Ove B. Jonsson, Ulrika Nordlo 1 f, and Ulrika L. Nilsson*
Department of Analytical Chemistry, Stockholm University, Sweden
A new supported liquid membrane extractor for bioanalytical sample preparation is presented. The extractor consists of a polypropylene hollow fiber mounted inside a PTFE tube by means of a cross-connector and a teeconnector. All parts are commercially available, inexpensive, and easily assembled. An organic solvent in the pores of the fiber forms a liquid membrane that separates the sample, which is pumped along the outside of the fiber, from the acceptor phase, which is pumped inside. The length of the hollow fiber may easily be varied to meet different demands on extractive surface and extract volumes. To test the system, the strongly acidic plasticizer/ flame retardant metabolite diphenyl phosphate ester (DPhP), with a pKa value of 0.26, was extracted from urine. DPhP was protonated using 4 M hydrochloric acid and extracted into an acceptor phase at pH 9. Thirty extractions were made with the same liquid membrane without any decrease in extraction efficiency and with a relative standard deviation 90% of the total amount of extracted DPhP was recovered within the first fraction. With a 5-cm hollow fiber, this volume is only 4.3 µL, which is somewhat difficult to handle in an off-line analysis step. Since the extraction efficiency achieved with different lengths of hollow fibers was to be investigated using off-line CE-UV detection, the size of the extract (microliters) was always twice the length of the fiber (cm), corresponding to ∼2.3 inner volumes. Therefore, the extracts from a 5-cm fiber were 10 µL, which together with 10 µL of volumetric standard, presented a large enough volume to be handled by the CE autoinjector. The quotient between the original sample volume and the extract volume multiplied by the extraction recovery determines the analyte enrichment obtained from the extraction system. For example, a sample volume of 500 µL and an extract of 50 µL together with a 50% recovery results in an enrichment factor of 5. For this reason, it is advantageous to keep the extract volume as small as possible. Donor Phase Composition. Since DPhP and DTP have pKa values of 0.26 and 0.40, respectively, a high concentration of a strong acid is required to protonate the phosphoric acid functionality to get neutral extractable molecules. On the other hand, adding a large proportion of acid will dilute the sample and may also result in an extreme environment that adversely affects the injector and the SLM system. The effect on extraction efficiency of different HCl concentrations, together with different organic solvents forming the liquid membrane, is presented in Table 2. Under these conditions, the only solvent that produced a stable liquid membrane with acceptable recoveries was 6-undecanone, and as can be seen, 4 M was the most effective concentration of HCL, providing a recovery of 64%. Dihexyl ether resulted in much lower recoveries than for 6-undecanone and was not suitable for this application. The extracts achieved using 1-octanol were all strongly acidic, which indicates a leakage of donor phase through the liquid membrane. 3510
Analytical Chemistry, Vol. 75, No. 14, July 15, 2003
The protonation of the diphosphate esters with 4 M HCl results in an extremely acidic and chemically aggressive environment for the equipment, mainly the injector. Therefore, to minimize the exposure time, HCl was added only to the sample and not to the carrier solution, although keeping an acidic environment close to the sample plug might enhance the degree of protonation of the analyte and thus the extraction efficiency in the mixing zones (i.e., the ends of the sample plug). Instead, water was used to transport the sample. Acceptor Phase. Because most of the extracts were analyzed by CE-UV, the extract had to be compatible with the CE separation. A 15 mM borate buffer of pH 9.2 (native pH) was found to enable fast separation of the investigated compounds, as well as providing acceptable peak shapes. The acceptor phase chosen was therefore a 1 mM borate buffer at pH 9.2. A higher ionic strength of the acceptor phase would probably promote the analyte trapping capacity and thereby somewhat improve the extraction efficiency, but on the other hand, it would also act negatively on the stacking effect in the CE injection, with poor peak shape as a result. Carryover. The construction of the XT-tube extractor, with the extra inlet position on the cross-connector, makes it easy to clean the donor channel with solvents differing in composition from the donor phase. This can be done either by adding a modifier, for example, a buffer, to the carrier flow, or by stopping the carrier flow and applying a separate washing solution. A great advantage of washing the donor channel with a solution having a pH similar to that of the acceptor phase instead of using the donor phase is that any target molecule diffusing out of the organic phase becomes charged and flushed to waste. By using the acceptor solution for washing both the acceptor channel and the donor channel, as described above, carryover could not be detected either by CE-UV or LC/ESI-MS. Figure 4 shows blank extracts, that is, nonspiked urine samples extracted directly after spiked urine samples. In the case of diphosphate esters, this washing procedure, for reasons not yet clarified, also had a positive influence on extraction recoveries as well as on repeatability. Validation. In a validation study, the recoveries from 30 repeated extractions using the same 15-cm fiber and liquid membrane were determined by CE-UV detection. The urine samples, spiked with 1.69 µg/mL DPhP and 2.98 µg/mL DTP, contained 4 M HCl, and the flow rate was 100 µL/min. The reasons for choosing a fiber length of 15 cm instead of 25 cm were the higher repeatability achived with the shorter fiber and that the smaller size (resulting in a smaller extract volume) should be more appropriate in a situation in which the extractor is coupled online to LC. The average recovery of DPhP, when corrected to the volumetric standard (aspartame), was 46%, with an RSD of 8.5%, and the corresponding values for DTP were 72 and 4.7%. The RSD for DPhP when corrected to the surrogate standard (DTP) areas was 6.7%. To investigate if there was any trend in extraction efficiency over time, DPhP recoveries from the thirty consecutive extractions were plotted as a function of time. The 95% confidence interval for the slope, achieved by linear regression, was determined to 0.0902 ( 0.177 (i.e., including zero), which confirms that the extraction recovery was stable for at least 30 extractions under these chemically extreme conditions. The LOD, calculated as three times the amplitude of the noise, for DPhP in
The calibration curves obtained using CE-UV and LC/ESIMS instruments were linear in the range 0.5-20 and 0.05-10 µg/ mL, respectively, with correlation coefficients (R2) higher than 0.99. The linear range of the entire method, including SLM extraction and detection using CE-UV or LC/MS, was tested for sample concentrations of 0.05-10 µg/mL. With CE-UV, the linear range was 0.5-10 µg/mL with an R2 value of 0.99, and for LC/ ESI-MS, the linear range was 0.05-5 µg/mL with an R2 value of 0.99.
Figure 4. CE-UV (210 nm) electropherograms of a blank urine extract and a urine sample spiked with 1.7 µg/mL DPhP and 3.0 µg/mL DTP (A), and LC/ESI-MS chromatograms in SIM mode of a blank urine extract and a urine sample spiked with 0.29 µg/mL DPhP and 0.25 µg/mL DTP (B). The ions m/z 249.2 and 277.2 correspond to the quasimolecular ions [M - H]- of DPhP and DTP, respectively.
urine was 0.18 µg/mL using CE-UV and 0.014 µg/mL using LC/ MS. In addition to a lower LOD, MS also has the advantage of enabling identification of the analytes and, thereby, a more accurate determination, as compared to UV. In the case of coupling, this should be more straightforward to LC than to CE. Furthermore, UV detection limits the applicability to molecules containing a chromophore, altogether this makes LC/ESI-MS a better choice. Figure 4A demonstrates typical CE-UV electropherograms, and Figure 4B shows LC/ESI-MS chromatograms in SIM mode. As can be seen in the LC/ESI-MS chromatogram, the different DTP isomers (ortho, meta and para) gave rise to a double peak, whereas the CE separation resulted in a single peak because of a less structurally dependent separation mechanism. According to the results presented in Figure 3, these experiments were performed at nonequilibrium conditions, with a residence time of
