Anal. Chem. 2000, 72, 5280-5284
Immunologic Trapping in Supported Liquid Membrane Extraction Eddie Thordarson, Jan A ° ke Jo 1 nsson, and Jenny Emne´us*
Department of Analytical Chemistry, Lund University, P.O. Box 124, SE-221 00 Lund, Sweden
To obtain a high degree of selectivity in sample preparation, supported liquid membrane (SLM) extraction was combined with immunologic recognition. The SLM employs a hydrophobic polymer for supporting the immobilization of an organic solvent, thus forming a nonporous membrane. Said membrane separates the aqueous sample on one side (donor) from a receiving aqueous phase on the other (acceptor). The extraction involves the partitioning of neutral compounds between the sample solution, continuously pumped alongside the membrane, and the membrane. From the membrane, reextraction takes place into a second aqueous phase containing antibodies specific for the target compound(s). Hence, there is a formation of an antibody-antigen complex at the heart of the sample preparation (ImmunoSLM). When the immunocomplex forms, the antigen can no longer redissolve in the organic membrane, thus being trapped in the acceptor. Consequently, the concentration gradient of free antigen over the membrane is ideally unaffected, this being the driving force for the process. With a surplus of antibody, the concentration of analyte in the receiving phase will easily exceed the initial sample concentration. In this work, the so formed immunocomplex was quantified on-line, using a fluorescein flow immunoassay in a sequential injection analysis (SIA) setup. The outlined ImmunoSLM-SIA scheme was successfully applied for the extraction of 4-nitrophenol from spiked water solutions as well as from a spiked wastewater sample, indicating that the immunoextraction can be suitable when dealing with difficult matrixes. The supported liquid membrane (SLM) technique utilizes liquid-liquid extraction from a flowing aqueous sample (donor) to an organic liquid, immobilized in the pores of a porous hydrophobic polymer, with subsequent reextraction into a receiving aqueous phase. These characteristics render the technique superior with regard to enrichment capacity, selectivity, and degree of cleanup as compared to, for example, solid-phase extraction, dialysis, and filtration.1 The technique has been a major focus in our research group for many years and has been applied to a number of different compound classes.2 Extraction by these means is feasible as long as the target molecules are uncharged in the donor and by some means charged, i.e., trapped, in the (1) Jo ¨nsson, J. A° .; Mathiasson, L. Trends Anal. Chem. 1999, 18, 318-325. (2) Jo ¨nsson, J. A° .; Mathiasson, L. Trends Anal. Chem. 1999, 18, 325-334.
5280 Analytical Chemistry, Vol. 72, No. 21, November 1, 2000
acceptor. This is usually accomplished by means of varying the pH in the two aqueous phases but can also be realized through forming a chelate or complex. In the present paper, a new extraction concept that we have chosen to call ImmunoSLM is presented. Exploring antibodies as specific reagents in liquid membrane extractions promises a high degree of analyte selectivity and enrichment, together with a very efficient matrix reduction. This extends the scope of SLM to include permanently neutral compounds that earlier have not been feasible to enrich. Moreover, it offers the possibility for on-line connection to very sensitive and selective immunoassay detection methods. Analyte-specific antibodies (Ab) are introduced as trapping reagents in the SLM acceptor. In the donor, the pH is adjusted so that the analyte is uncharged and thus extractable into the organic membrane. The uncharged analyte (Ag) is extracted from the donor to the acceptor down its concentration gradient. The gradient is upheld by the binding of Ag to the analyte-specific Ab in the acceptor, thus forming an antibody-antigen (Ab-Ag) complex at the heart of the sample preparation procedure. The configuration supports either group or analyte specificity, depending on the type of Ab used, i.e., mono- or polyclonal antibodies, where the latter is a heterogeneous population of antibodies with often broad analyte selectivity, whereas monoclonal antibodies can exhibit a higher degree of selectivity. Good trapping capacity is expected because of the normally high affinity of Ab toward its Ag. The quantification of an Ab-Ag interaction in an immunoassay is commonly performed by detecting a specific marker molecule, also called a tracer (denoted with an asterisk), attached either to the antibody (Ab*) or to the antigen (Ag*). The earliest reports involved the use of radioactive isotopes such as 1H3, 53I125, or 6C14. The current trend is however to use nonisotopic markers such as fluorescent compounds (as in this work) or enzymes, e.g., horseradish peroxidase, alkaline phosphatase, acetylcholine esterase, or β-galactosidase.3 Microtiter plate configurations of the enzyme-linked immunosorbent assay (ELISA) and variations thereof are the most commonly employed immunoassays for detection of Ab-Ag complex formation. Alternatives to these batch assays are flow immunoassays or flow injection immunoassays (FIIA).4 As the names imply, these assays are performed in a flow and generally benefit from faster kinetics (shorter diffusion paths or convection) and less need for intermediate washing steps. As such, these are more amenable for automation. In the present work, the model analyte was 4-nitrophenol (denoted 4-NP or Ag) (3) Diamandis, E. P.; Christopoulos, T. K. Immunoassay; Academic Press: San Diego, 1996. (4) Fintschenko, Y.; Wilson, G. S. Mikrochim. Acta 1998, 129, 7-18. 10.1021/ac0005013 CCC: $19.00
© 2000 American Chemical Society Published on Web 10/05/2000
and consequently anti-4-NP antibodies (Ab) are used in the acceptor. The quantification of the formed Ab-Ag is performed on-line using a previously developed fluorescein flow immunoassay for 4-NP,5 redesigned to suit a sequential injection analysis (SIA) format. An inherent drawback with all immunoassays is that the chemical environment suitable for antibodies is limited with regard to pH, ionic strength, organic content, and temperature, since the degree of analyte recognition and position of equilibrium will depend on these parameters. The validation of immunoassays used for real samples by correlating with classical analytical techniques has shown good results, albeit generally generating positive errors for enzymatic methods.6-9 These errors are caused by interfering components in the sample matrix. The overall effect on the analytical signal can be due to, among others, binding of antigen to the matrix, nonspecific binding or specific binding of (usually) structurally related compounds to the antibody (cross-reactivity), and inhibition/activation of enzymatic catalysis.8 Consequently, there is a need for a tailor-made sample matrix in order for a specific antibody to behave optimally. Even if optimum performance is not an issue, the antibody obliges a sample-to-sample uniformity in the matrix to the end of assay repeatability and linearity. Considering this, FIIA has the upper hand to the batch assays as it offers the possibility of including an on-line sample preparation step. To this end, sample preparation techniques employing membranes often exhibit excellent matrix reduction10 and thus suit to serve the outlined prerequisites. In the present work, polyclonal anti-4-NP antibodies were used to extract 4-NP. A shortcoming when working with polyclonal antibodies against small molecules such as 4-NP is that the crossreactivity tends to be somewhat high. For the antibodies used in this work, cross-reactivity for an array of compounds was established earlier.11 With SLM, the pH of the donor solution can in many cases be adjusted to discard possibly interfering solutes; hence there is selectivity to gain at both ends of the extraction with ImmunoSLM. EXPERIMENTAL SECTION Chemicals and Solutions. Dr. M.-P. Marco (Department of Biological and Organic Chemistry, IIQAB-CSIC, Barcelona, Spain) kindly provided us with ammonium sulfate-precipitated polyclonal antibodies raised against 4-NP (rabbit anti-4-NP, As35) as well as a carboxylic derivative of 4-NP ((S)-hydroxyl-5nitrobenzoyl)propionic acid (HOM), used for synthesis of the tracer. Fluorescein thiocarbamyl ethylenediamine (EDF) was synthesized from fluorescein isothiocyanate, isomer I (Sigma Chemical Co., St. Louis, MO) as described by Pourfarzaneh et al.12 The (S)-(hydroxyl-5-nitrobenzoyl)propionic acid derivative was (5) Nistor, C.; Oubin ˜a, A.; Marco, M.-P.; Barcelo´, D.; Emne´us, J. Anal. Chim. Acta, in press. (6) Ferguson, B. S.; Kelsey, D. E.; Fan, T. S.; Bushway, R. J. Sci. Total Environ 1993, 132, 415-428. (7) Abad, A.; Montoya, A. Anal. Chim. Acta 1995, 311, 365-370. (8) Gasco´n, J.; Oubina, A.; Ferrer, I. Anal. Chim. Acta 1996, 330, 41-51. (9) Hall, J. C.; Van Deynze, T. D.; Struger, J.; Chan, C. H. J. Environ. Sci. Health 1993, B28 (5), 577-598. (10) van de Merbel, N. C. J. Chromatogr. 1999, 856, 55-82. (11) Oubina, A.; Ballesteros, B.; Galve, R.; Barcelo, D.; Marco, M.-P. Anal. Chim. Acta 1999, 387, 255-266. (12) Pourfarzaneh, M.; White, G. W.; Landon, J.; Smith, D. S. Clin. Chem. 1980, 26, 730-733.
Figure 1. Schematics of an immunoSLM extraction.
reacted with EDF, forming the tracer molecule (essentially a fluorescent antigen, Ag*). Phosphate buffer saline (PBS) at physiological pH was prepared from NaH2PO4 and Na2HPO4 (Merck, Darmstadt, Germany) by mixing their respective 0.15 M solutions until reaching pH 7.4. Working solutions of antibodies (Ab) and tracer (Ag*) were prepared in PBS. Solutions of Ab were stable at 4 °C whereas new tracer dilutions from the stock solution had to be prepared on a daily basis. The tracer was known to be light sensitive, so all vials containing tracer were wrapped in aluminum foil. A secondary effluent wastewater sample (BAR 3) was obtained from a tannery industry outside Barcelona and required no preparation prior to SLM extraction except for spiking with 4-NP. The sample had undergone some initial treatment in the wastewater treatment plant before collection. The sample was yellow to grayish with an unpleasant odor. Dihexyl ether (97%, Sigma-Aldrich, Steinheim, Germany) was used to impregnate the membrane used for extractions. Apparatus. The experimental setup for the analysis equipment included an in-house manufactured SLM unit and a six-port multiposition valve (VICI, Valco Instruments Co. Inc., Houston, TX). A peristaltic pump (Minipuls 3, Gilson, Villiers-le-Bel, France) was use for pumping the sample in the donor channel, and a syringe pump (Kloehn Co. Las Vegas, NV) served for sequential aspiration and elution of acceptor content and tracer. Acceptor and tracer were reacted in a short tubing (0.5 mm i.d., 51 cm), and surplus of tracer was removed in a precolumn made in-house from PEEK (2 mm i.d., 2 cm), slurry-packed with ADS-C18 restricted access (RA) material (Merck). Detection was made with a fluorescence detector (L-7480, Hitachi, Tokyo, Japan), excitation at 490 nm and emission/detection at 515 nm. SLM Unit. The arrangement comprised a porous PTFE membrane (TE 35, Schleicher & Schuell, Dassel, Germany, thickness 60 µm with a 180-µm-thick supporting polyester backing, 0.2-µm-i.d. pore size and 60-80% porosity), with immobilized dihexyl ether inside the pores. The membrane was clamped between a PTFE and a PEEK piece separating two identical channels, serving as donor and acceptor containers, each holding ∼10 µL. The devise was held together by 10 screws, leaving it gas- and liquid-tight. ImmunoSLM-SIA Procedure. The principle behind the SLM extraction is visualized in Figure 1. As is shown, the sample is continuously processed, i.e., pumped alongside the immobilized liquid membrane. Target analytes (antigen, Ag) dissolve into and diffuse through the membrane and finally bind to the antibodies Analytical Chemistry, Vol. 72, No. 21, November 1, 2000
5281
pipetting different amounts of 4-NP to an optimized amount of Ab, until the reaction reaches equilibrium. An optimized amount of Ag* was added and the reaction mixture left to equilibrate, following aspiration and analysis of 60 µL in the SIA system.
Figure 2. Instrumental setup of the SIA system: (A) gastight vial containing antibody at 1-2 psi positive pressure, (B) SLM unit, (C) peristaltic pump, (D) vial containing Ag*, (E) syringe pump, (F) mixing tubing of 0.5-mm i.d., holding 100 µL, (G) RA column, (H) fluorescence detector, and (MPV) multiposition valve.
in the acceptor. Charged molecules are not extractable, and neutral compounds, not binding to the antibodies, will not be enriched. Compounds, other than the Ag that are dissolved in the membrane (i.e., neutrals) do not, to our experience, interfere with the ongoing extraction, nor with the following. The result is rejection of the vast majority of matrix constituents and an increase in antigen concentration in the acceptor. The acceptor is kept stagnant during the extraction and thereafter removed for identification. The multiposition valve (MPV) in Figure 2 enabled the assay to be carried out in an SIA mode. The Ab-containing solution was prepared off-line and contained in a 2-mL plastic vial (A). This vial was sealed with a rubber plug through which PTFE tubing was drawn. Helium at 1-2 psi positive pressure was applied to this gastight arrangement, preventing the membrane from being sucked against the wall of the SLM unit (B) upon aspiration of the acceptor, which would otherwise interrupt the flow. The acceptor channel was initially filled by aspirating 30 µL of the Ab solution from vial A. A peristaltic pump (C) was used to deliver the sample through the donor channel at 1 mL‚min-1. After a 10min extraction, the acceptor content (Ab and Ab-Ag) and tracer (Ag*) from vial D were alternately aspirated with a syringe pump (E), in 5-µL increments at 50 µL‚min-1 by manually switching the MPV. This was continued until 30 µL of the acceptor (essentially 3 times the acceptor nominal volume) and 30 µL of Ag* were mixed in the outlet tubing (F). The plug, now containing Ab-Ag + Ab-Ag* + Ag* was left to equilibrate for 1 min and thereafter dispensed through the ADS-C18 RA column (G) at 0.5 mL‚min-1. Here, the excess of free tracer (Ag*) was trapped in the hydrophobic cavities of the RA material. The eluting plug now only contains Ab-Ag + Ab-Ag* and the subsequent fluorescence detection (H) of the Ab-Ag* fraction yielded an indirect quantification of the extracted 4-NP. Immediately after finishing the extraction, the peristaltic pump flushes reagent water through the donor for 5 min, to remove not-extracted 4-NP from the membrane. The ImmunoSLM-SIA assay was completed on-line within 15 min (10 min for extraction and ∼4 min for sequential aspiration/mixing and incubation). The ImmunoSLM-SIA procedure was compared with off-line SIA (excluding the SLM part). Off-line SIA was performed by first 5282
Analytical Chemistry, Vol. 72, No. 21, November 1, 2000
RESULTS AND DISCUSSION An initial concern in designing the ImmunoSLM was whether the antibody would be adsorbed in the acceptor channel or by any means rendered inactive toward the antigen. This was tested by comparing the signals obtained from sequential on-line mixing with Ag* for aliquots of Ab solution, either having stood in the SLM acceptor channel or not. In the former case, an Ab solution stood still in the acceptor for 30 min and thereafter, alternately with the Ag*, sequentially aspirated, incubated for 1 min, and detected. In the other case, Ab contained in a plastic vial was sequentially aspirated with Ag*, incubated for 1 min, dispensed, and quantified. No difference was observed in comparing these signals, indicating no problems with inactivation or adsorption of Ab. On-line mixing of Ab and Ag* was accomplished by alternately aspirating small increments of Ab and Ag* by switching the MPV in Figure 2 between the Ab and Ag* solutions. To assess the maximum volume increments allowed, while still preserving complete mixing, signals from on-line runs were compared to those obtained from direct aspiration and elution of an off-line preincubated Ab/Ag* solution of the same concentrations. There was no difference in signal between the sequential aspiration in 2.5- and 5-µL increments, showing adequate mixing in the valve and tubing. The largest contribution to this efficient mixing is thought to be convection, because of momentarily induced eddies as a result of the transient subambient pressure arising upon aspirating the sample through the tubing. However, with 10-µL increments the signal decreased; therefore, 5-µL increments were chosen for the assays. Analyzing the following 30-µL fraction assessed that the transfer of acceptor solution (10 µL) with pumping of three acceptor volumes was indeed quantitative. Having analyzed the extract and washed the donor channel for 5 min, the Ab now in the acceptor was left stagnant for 30 min, and thereafter analyzed in the same way. The measured signal was compared with one where Ag* was mixed with Ab not having passed through the SLM unit. There was no difference between these peaks, showing no sign of carry-over effects. This was shown for extractions up to 5 µg‚mL-1 4-NP, which would clearly show if any memory effects were present. A solution of Ag* resulting in a signal ∼100 times the baseline noise was used as the working tracer concentration ([Ag*]w) in order to obtain an assay as sensitive as possible with an acceptable linear range. Once the [Ag*]w was established, the optimal antibody concentration was found by adding different amounts of Ab to aliquots of Ag* solution. The antibody concentration was increased until the signal reaches a plateau, indicating that all Ag* is bound to the antibody binding sites. Plotting the signals obtained from these incubations against the [Ab] yielded an antibody dilution curve for the system (Figure 3). An [Ab] binding 50% of the tracer was selected as the working antibody concentration ([Ab]w) and was found to be 150 µg‚mL-1. The optimized [Ag*]w and [Ab]w were then used to generate analyte calibration curves, either by incubating Ag and Ab off-line or by ImmunoSLM
Figure 3. Antibody dilution curve. The relative response is the resulting signal from the Ab-Ag* incubations divided by the signal obtained for the free Ag* alone.
Figure 4. Calibration curve for the off-line incubations of 4-NP from 10 pg‚mL-1 to 20 µg‚mL-1. Each point represents an average of three injections. RSD values for all points ranged between 0.8 and 4.6%, except for the highest concentration that had an RSD of 14.1%. The relative responses are obtained by dividing the signals with the blank, i.e., a sample free from Ag.
extraction by pumping of 4-NP in the donor channel of the SLM unit with Ab in the acceptor. The calibration plot for the off-line incubations is shown in Figure 4. Triplicate analysis in each point of the calibration curve yielded RSD values ranging between 0.8 and 4.6% with the exception of the point of lowest 4-NP concentration that showed an RSD of 14.1%. The results from the ImmunoSLM-SIA of spiked reagent water are shown in Figure 5. Also inserted in Figure 5 is a point for extraction of a spiked wastewater sample, a smelly, yellow secondary wastewater effluent from a tannery industry in Barcelona (BAR 3). The BAR 3 extraction yielded a signal 98% of the one obtained from extracting reagent water spiked with the same 4-NP concentration. This is an indication that it is most likely feasible to extract from very difficult matrixes, which obviously is something we expected with the ImmunoSLM technique. The ImmunoSLM-SIA assay was completed fully on-line with a total time of 15 min, with few preceding manual steps. The curves in Figure 4 and Figure 5 respectively are not fully comparable as they are registered under somewhat different premises. With the off-line incubations (no SLM extraction), the aspirated 60-µL plug is a homogeneous mixture, where 30 µL of Ag* simply is added to 30 µL of preincubated Ab-Ag solution using a micropipet. For the on-line extractions, only the front of the aspirated plug contains a high concentration of Ab-Ag (remembering that the acceptor nominal volume is 10 µL), which
Figure 5. Calibration curve for the semiautomatic ImmunoSLMSIA system: ([) on-line assay of spiked reagent water, 4-NP concentration from 10 pg‚mL-1 to 5 µg‚mL-1; (2) on-line assay of a wastewater sample spiked with 0.4 µg‚mL-1 4-NP. Relative responses as in Figure 4.
then gets more and more diluted, while still adding a fresh solution of Ab. In principle, this means that the Ab-Ag concentration in the on-line mode is always lower than in the off-line mode for the same investigated Ag concentrations. This suggests that when solutions low in 4-NP concentration are extracted, the effect on the total signal will be reduced, or even swamped out, by the effect of titrating with Ag* on free Ab. This is one of the reasons why the calibration plot for the ImmunoSLM extraction is tilted to the right (toward higher detection limits) compared to the off-line incubations. The other reason is to be derived from the SLM theory of incomplete trapping.13,14 RA is the fraction of Ag that is free for reextraction from the acceptor solution (i.e., Ag not bound to Ab) is described by
RA ) [Ag]/([Ag] + [Ab-Ag])
(1)
For an efficient trapping, RA should approach zero. For the immunologic reaction, the equilibrium constant is K ) [Ab-Ag]/ [Ab][Ag], where [Ab] is the concentration of free antibody. By combining the expression for K with eq 1, we see that
RA ) 1/(1 + K[Ab])
(2)
which is valid at any time during the extraction. The concentration enrichment factor, Ee, is defined as Ee ) cA/cI, where cA is the acceptor concentration of analyte and cI is the sample analyte concentration. According to Chimuka et al.,14 the maximum concentration enrichment (Ee,max) can be set equal to the reciprocal of RA:
Ee,max ) 1/RA ) 1 + K[Ab]
(3)
This relationship tells us that a large K and a high [Ab] is necessary for obtaining a high Ee,max and thus a high Ag concentration in the acceptor. Assuming that only 10% of the polyclonal anti-4-NP antibodies are active toward 4-NP, and the (13) Jo ¨nsson, J. Å.; Lo ¨vkvist, P.; Audunsson, G.; Nilve´, G. Anal. Chim. Acta 1993, 227, 9-24. (14) Chimuka, L.; Megersa, N.; Norberg, J.; Mathiasson, L.; Jo¨nsson, J. Å. Anal. Chem. 1998, 70, 3906-3911.
Analytical Chemistry, Vol. 72, No. 21, November 1, 2000
5283
approximate values of K ≈ 107 and [Ab] ≈ 10-7 M, being valid for this work, yields an RA of 0.5, corresponding to an Ee,max as low as 2. This shows that the concentration in the acceptor can, at its best, be 2 times higher than in the original sample. The situation here is even worse considering that the amount of free Ab decreases continuously with time as more and more Ab binding sites are occupied by Ag. A more correct representation would thus be as in eq 4, where [Ab]t is the starting concentration
RA ) 1/(1+K([Ab]t - [Ab-Ag]))
(4)
of antibody in the acceptor. Equation 3 shows that the use of antibodies with a higher affinity for 4-NP or, alternatively, an increase in the active antibody concentration (affinity purified or monoclonal antibodies) would improve the situation. By doing so, the corresponding calibration plot would be tilted to the left, toward lower detection limits. In reality, antibodies with higher affinity for 4-NP are not, to our knowledge, available and thus not an option. The other possibility, viz. an increase in the [Ab]w, will also favorably influence the Ee,max. With an excess of antibody and an immediate, irreversible trapping of analyte, the amount of extracted analyte per unit time is constant. However, the final outcome of the assay will be a compromise between high values of Ee,max for the SLM extraction and sensitive immunoassay detection of the formed Ab-Ag complex. When the assay is
5284
Analytical Chemistry, Vol. 72, No. 21, November 1, 2000
beoing designed, the relatively high cost of antibodies should also be considered. The ImmunoSLM theory is currently being tested in our laboratory on an Ab/Ag system developed for atrazine, where the Ab affinity constant for the Ag is higher than in the present system. CONCLUSIONS A novel SLM concept (ImmunoSLM) for selective liquid membrane extraction, enrichment, and detection of 4-nitrophenol was developed. The outlined extraction scheme is believed to exhibit good performance with difficult sample matrixes, in combination with the exquisite sensitivity and selectivity provided by immunologic methods for determination. Operating the assay as a sequential injection analysis minimizes requirements for tubing, valves and connections. The SIA system showed rigorous mixing, without demanding a mixing coil, loop, or reactor, resulting in minimal dilution of the sample plug. ACKNOWLEDGMENT The authors gratefully acknowledge financial support from the Swedish Natural Science Research Council (NFR), the European Commission (EC Contracts IC15CT98-0910, IC15CT98-0138, ENV4CT97-0476). Prof. K.-S. Boos and D. Lubda, Merck, are also gratefully acknowledged for providing us with RA material. Received for review May 2, 2000. Accepted August 1, 2000. AC0005013
