Postcolumn Renewal of Sensor Surfaces for High-Performance Liquid

May 6, 2008 - The combination of high-performance liquid chromatography (HPLC) with surface plasmon resonance (SPR) for continuous separation and ...
1 downloads 0 Views 509KB Size
Anal. Chem. 2008, 80, 4225–4230

Postcolumn Renewal of Sensor Surfaces for High-Performance Liquid Chromatography-Surface Plasmon Resonance Detection Ming Du and Feimeng Zhou* Department of Chemistry and Biochemistry, California State University, Los Angeles, California 90032 The combination of high-performance liquid chromatography (HPLC) with surface plasmon resonance (SPR) for continuous separation and label-free detection of protein samples is described. The detection was realized by electrostatic adsorption of proteins bearing positive and negative charges onto chemically modified SPR sensors in two separate SPR channels. One SPR channel is coated with carboxymethylated dextran which facilitates the detection of positively charged proteins, whereas the other, devoted to the monitoring of negatively charged proteins, is covered with ethylenediamine molecules attached onto a dextran surface. Renewal of the sensor surface in the channels can be accomplished by introducing regeneration solutions through two six-port valves. The coupled technique (HPLC-SPR) was assessed for its analytical figures of merit and applied to the quantification of lysozyme in human milk samples. Unlike the SPR detection of bulk solution refractive index changes during chromatographic peak elutions, the highest sensitivity of SPR is retained in this work since the measurement is performed at the SPR sensor surface where the evanescent field is the strongest. The renewable SPR detection of continuous separations is reproducible and versatile and does not require the separated proteins to contain chromophores or to be prelabeled with a tag (e.g., a redoxactive or fluorescent molecule). Such generality makes SPR complementary to other types of chromatographic detectors. Among the numerous detectors developed for high-performance liquid chromatography (HPLC),1–4 about a dozen are commonly used and commercially available. Chromatographic detectors can be generally classified into two types: bulk property detectors (e.g., refractive index and density detectors) and solute property detectors (e.g., absorbance and fluorescence detectors).2 Detectors in the former type are more universally applicable than those in the latter. However, their sensitivity is not high and the * Corresponding author. Phone: 323-343-2390. Fax: 323-343-6490. E-mail: [email protected]. (1) Skoog, D. A.; Holler, F. J.; Crouch, S. R. Principles of Instrumental Analysis, 6th ed.; Thomson Higher Education: Belmont, CA, 2007. (2) Scott, R. P. W. Liquid Chromatographic Detectors, 2nd ed.; Elsevier: Amsterdam,1986. (3) Yeung, E. S.; Synovec, R. E. Anal. Chem. 1986, 58, 1237A–1256A (4) Dorschel, C. A.; Ekmanis, J. L.; Oberholtzer, J. E.; Warren, F. V.; Bidlingmeyer, B. A. Anal. Chem. 1989, 61, 951A–968A 10.1021/ac702632y CCC: $40.75  2008 American Chemical Society Published on Web 05/06/2008

response is subject to ambient temperature variation and fluctuations in pressure and flow rate. Since the LC detectors are not as general as their counterparts in gas chromatography, efforts in combining various analytical techniques with HPLC for sensitive detection of a wide range of compounds have continued. Surface plasmon resonance (SPR) is an optical technique that is extremely sensitive to refractive index changes at the metal/ solution interface.5–8 Some of the unique advantages of SPR include its high sensitivity, obviation of sample labeling, simplicity, and relatively low cost. SPR has been coupled to chromatographic techniques (HPLC-SPR)9–13 to detect compounds that do not exhibit strong absorption in the UV-vis range. For example, Blikstad et al. performed SPR detection of oligosaccharides chromatographically separated in the isocratic elution mode.9 More recently, Jungar et al. used such a combination to analyze carbohydrates.11 Whelan and Zare were the first to couple SPR to capillary electrophoresis and monitored the change in the refractive index during the separation of a mixture of components with concentrations at millimolar levels.14 In the same work, they also functionalized the SPR sensor with biotinylated protein A and observed its interaction with electrophoretically delivered (not separated) antibody IgG. Ly et al. recently visualized the electrophoretic migration of sample bands in an ultrathin fluidic channel that confined the SPR evanescent wave within ∼500 nm using imaging SPR.15 Although these papers elegantly demonstrated the versatility of SPR as a detector for column and electrophoretic separations, SPR has been exclusively utilized to monitor the refractive index variation of the mobile phase and, thus, can only be regarded as a variant of the conventional refractive index detector. The SPR sensitivity for monitoring solution refractive (5) Hanken, D. G.; Jordan, C. E.; Frey, B. L.; Corn, R. M. Electroanalytical Chemistry; Bard, A. J., Rubinstein, I., Eds.; Marcel Dekker, Inc.: New York, 1998; Vol. 20, pp 41-225. (6) Surface Plasmon Resonance Based Sensors; Homola, J., Ed.; Springer: Berlin, 2006; Vol. 4. (7) Willets, K. A.; Van Duyne, R. P. Annu. Rev. Phys. Chem. 2007, 58, 267– 297. (8) Phillips, K. S.; Cheng, Q. Anal. Bioanal. Chem. 2007, 387, 1831–1840. (9) Blikstad, I.; Fagerstam, L. G.; Bhikhabhai, R.; Lindblom, H. Anal. Biochem. 1996, 233, 42–49. (10) Cepria, G.; Castillo, J. R. J. Chromatogr., A 1997, 759, 27–35. (11) Jungar, C.; Strandlh, M.; Ohlson, S.; Mandenius, C.-F. Anal. Biochem. 2000, 281, 151–158. (12) Nice, E.; Lackmann, M.; Smyth, F.; Fabri, L.; Burgess, A. W. J. Chromatogr., A 1994, 660, 169–185. (13) Castillo, J. R.; Cepria, G.; de Marcos, S.; Galban, J.; Mateo, K.; Garcia, R. Sens. Actuators, A 1993, 37-38, 582–586. (14) Whelan, R. J.; Zare, R. N. Anal. Chem. 2003, 75, 1542–1547. (15) Ly, N.; Foley, K.; Tao, N. Anal. Chem. 2007, 79, 2546–2551.

Analytical Chemistry, Vol. 80, No. 11, June 1, 2008

4225

index changes is not high, because the evanescent field, enhanced by the plasmon resonance, decays exponentially into the dielectric medium in contact with the sensor. The field of the evanescent wave becomes approximately 37% (1/e) of that at the sensor at a distance of 164 nm away from the surface.5 Consequently, in terms of the detection levels, the analytes are typically separated and detected by HPLC-SPR9–13 and CE-SPR14 at millimolar concentrations, which are not lower (better) than those detectable with even insensitive refractive index detectors for column HPLC (submillimolar levels3). Some specially designed refractive index detectors can detect analyte concentrations down to submicromolar levels.16–19 Therefore, to capitalize on the high sensitivity of SPR, the detection should be carried out at or very near to SPR sensor surface. Unfortunately, the attachment of a species that first elutes out of the HPLC column onto the SPR sensor can adversely affect the specific recognition or detection of subsequent eluents. As a consequence, detection via adsorption of chromatographically separated species prevents SPR detection from being continuously performed. More importantly, SPR has not been used for the detection of separated proteins, which can nonspecifically adsorb onto an unmodified Au surface. Postcolumn derivatization and manipulation of HPLC eluents are common practices to overcome deficiencies inherent in certain detectors.20,21 In conventional flow injection SPR experiments, surface regeneration is routinely conducted for obtaining reproducible and accurate biomolecular affinity and kinetic results.6 To our surprise, performing postcolumn regeneration of the SPR sensor surface for continuous detection and improved reproducibility has not been attempted. In this work, we describe an interface that facilitates the regeneration of chemically modified SPR sensor surfaces without affecting continuous protein separations. The use of a dual-channel SPR allows the simultaneous detection of both positively and negatively charged proteins, making the combination of HPLC with SPR a rather general approach for separation of biomolecules. We show that the uninterrupted SPR monitoring of chromatographic elutents is highly complementary to other commonly used chromatographic detectors (e.g., UV-vis spectrometric detector). The analytical figures of merit and the feasibility of the HPLC-SPR technique for real sample analysis are also demonstrated. EXPERIMENTAL SECTION Chemicals and Materials. Tris(hydroxymethyl)aminomethane (Tris), cystamine dihydrochloride, ethylenediamine, N-hydroxysulfosuccinimide (NHS), N-ethyl-N′-(3-dimethylaminopropyl)-carbodiimide (EDC), ethanolamine, bovine serum albumin (BSA), lysozyme (from human milk), myoglobin (from horse heart), metallothionein I (MT, from rabbit liver), carboxymethylated dextran, and poly-L-arginine were all purchased from Sigma (St. Louis, MO). H-Lys-Cys-Thr-Cys-Cys-Ala-OH (FT) was obtained from Bachem California (Torrance, CA). Deionized water (16) Bornhop, D. J.; Dovichi, N. J. Anal. Chem. 1987, 59, 1632–1636. (17) Bruno, A. E.; Krattiger, B.; Maystre, F.; Widmer, H. M. Anal. Chem. 1991, 63, 2689–2697. (18) Woodruff, S. D.; Yeung, E. S. Anal. Chem. 1982, 54, 2124–2125. (19) Zhu, H.; White, I. M.; Suter, J.; Zourob, M.; Fen, X. Anal. Chem. 2007, 79, 930–937. (20) Liu, Y.-C.; Sowdal, L. H.; Robinson, N. C. Arch. Biochem. Biophys. 1995, 324, 135–142. (21) Lores, M.; Cabaleiro, O.; Cela, R. Trends Anal. Chem. 1999, 18, 392–400.

4226

Analytical Chemistry, Vol. 80, No. 11, June 1, 2008

Scheme 1. Schematic Representation of the Interface for Coupling HPLC to a Dual-Channel SPRa

a The valve at the top is shown in the load position (SPR detection), whereas that at the bottom is in the injection position (surface regeneration).

was collected from a Millipore water purification system (Millipore, Billerica, MA). Human milk was purchased from Mother’s Milk Bank (Presbyterian/St. Luke’s Medical Center, Denver, Colorado) and stored at -20 °C prior to use. Instruments. HPLC. Size exclusion liquid chromatography was performed with a Varian 9010 HPLC system (Varian Inc., Walnut Creek, CA) equipped with a 9050 UV-vis detector and a GFC 3000 column (Phenomenex Inc., Torrance, CA). The mobile phase was a 50 mM Tris-HCl solution prepared by adding 0.3 M HCl to the Tris solution until the final pH became 6.2. The mobile phase and the acid/base regeneration solution (0.1 M NaOH and 0.1 M HCl) were filtered with a 0.20 µm filter and degassed before use. SPR. SPR measurements were conducted with a BI SPR 1000 system (Biosensing Instrument, Tempe, AZ) equipped with a polyetheretherketone (PEEK) flow cell with two sets of inlet and outlet ports. Mounted onto the flow cell by two screws was a poly(dimethylsiloxane) (PDMS) gasket embossed with two fluidic channels. The channel width and depth were 2 and 0.254 mm, respectively, and the separation between the two channels was 1.5 mm. These fluidic channels were in precise alignment with two separate position-sensitive photodetectors. Calibration was performed using ethanol to adjust the two detectors to the same sensitivity. Interface for Coupling SPR to HPLC. Coupling SPR to HPLC was accomplished by connecting the column outlet to a tee connector which splits the mobile phase into two streams flowing to the two SPR channels (Scheme 1). Two six-port injectors (Valco Instruments, Houston, TX) were positioned between the tee connector and the SPR fluidic channels for postcolumn regeneration of the SPR sensor surface. Since the loop volumes (20 µL) are much smaller than the HPLC column, switching the valves between the load and injection positions did not disproportionate the two flowing streams in the two SPR channels or affect the SPR baseline signals. To minimize dispersion of the separated protein peaks, the lengths of the narrow-bore Teflon tubings (i.d. ) 0.005 in., Upchurch Scientific, Oak Harbor, WA) were kept short: 10 cm from the column to the valves and 25 cm from the valves to the SPR flow cell. The total volume from the column to the cell was estimated to be less than 40 µL.

Procedures. SPR Sensor Fabrication, Chemical Modifications, and Characterization. A Au film (50 nm thick) was deposited onto a 2 nm Cr adhesive layer coated onto a clean BK7 glass cover slide using an electron beam evaporation system (CHA Industries, Fremont, CA). Modification of the SPR gold surface with carboxymethylated dextran is similar to our published procedure.22 Briefly, cystamine self-assembled monolayers were formed by casting onto each Au film 0.8 mL of 20 mM cystamine dihydrochloride solution overnight at ambient temperature. After rinsing with water, 0.8 mL of a solution containing 4.35 mg/mL carboxymethylated dextran, 0.1 M NHS, and 0.4 M EDC were spread onto the gold film and the dextran film attachment was allowed to proceed for 3 h. The performance of the as-prepared sensor films compares well with that of the commercial CM5 chips (Biacore Life Sciences, GE Healthcare, Piscataway, NJ) on a Biacore X System (GE Healthcare) in terms of protein binding and surface regeneration. The average thickness of seven dry carboxymethylated dextran films, measured by a phase-modulated ellipsometer (Beaglehole Instruments Ltd., Wellington, New Zealand), was 3.0 nm with a percent relative standard deviation (% RSD) of 7.4%. A contrast between images of bare and dry dextran-covered Au surfaces, collected with an atomic force microscope (AFM) (Asylum Research, Santa Barbara, CA), revealed that the Au grain boundaries on the dextran-covered surface were still discernible but had become less distinct. The typical undulation of the dextran-covered Au surface (±4 nm) is greater than that of the bare Au surface (±2 nm). These AFM results are consistent with the ellipsometric data and suggest that the thickness of the dextran film is thin and relatively uniform. The small % RSD in the thickness variation suggests that dextran films can be reproducibly attached to the Au surface using the protocol described above. To derivatize the dextran film with ethylenediamine, a sensor covered with carboxymethylated dextran was mounted onto the SPR flow cell and a solution comprising 0.1 M NHS and 0.4 M EDC was injected into the fluidic channel through a six-port valve to activate the surface. This was followed by injecting a borate buffer (50 mM sodium borate, pH 8.5) containing 0.1 M ethylenediamine. After this step, the surface was deactivated by injecting 1 M ethanolamine solution for 7 min. Homogenization of Human Milk and Quantification of Lysozyme. To minimize adsorption of fat or other species from the human milk samples in the column and/or onto the SPR sensor surface, 20 mL aliquots of thawed samples were adjusted to pH 4.6 with 30% acetic acid. This was followed by centrifugation at 189 000 rpm at 4 °C for 1 h. The precipitate was washed with 30% acetic acid solution, and the wash solution was pooled with the supernatant. The combined solution was freeze-dried to 1.0 mL to which saturated (NH4)2SO4 solution was added until its final concentration was 3.0 M. After centrifugation at 8000 rpm at 4 °C, the supernatant was passed through a 3000 Da cutoff filter (YM-3, Millipore). The solution above the filter membrane was pipetted out and combined with water used to wash the filter. Upon freeze-drying, the final solution was analyzed by HPLC-SPR. RESULTS AND DISCUSSION The interface shown in Scheme 1 is developed for continuous separation and simultaneous detection of positively and negatively (22) Yao, X.; Li, X.; Toledo, F.; Zurita-Lopez, C.; Gutova, M.; Momand, J.; Zhou, F. Anal. Biochem. 2006, 354, 220–228.

charged proteins. We used carboxymethylated dextran and its ethylenediamine derivative to modify Au films for SPR detection based on the consideration that (1) the hydrogel-like dextran layer is highly versatile and robust for reliable protein immobilization and surface regeneration and (2) the dextran layer is rather thin and thus well-matched with the penetration depth of the evanescent field.6,23 The commercial success of the SPR chips covered with various types of dextran films and the excellent agreement between SPR and solution-based methods in affinity measurements can be attributed to these characteristics.6,23 When one SPR channel is covered with carboxymethylated dextran, the carboxylic acid groups are deprotonated (negatively charged) at the pH used for the HPLC separations (pH 6.2), rendering this channel suitable for detection of positively charged proteins thorough electrostatic attraction. In the other channel, the carboxymethylated dextran film is derivatized with ethylenediamine, which is protonated at the same pH and can be used for the detection of negatively charged proteins. Figure 1A is a chromatogram corresponding to the separation and UV-vis spectrometric detection of BSA, myoglobin, lysozyme, and MT. Notice that the lysozyme peak appears as a shoulder of the MT peak. Even after the MT concentration was decreased by 10 times, only a small lysozyme peak appeared in the chromatogram (not shown). The retention times (tR) for BSA, myoglobin, and MT were estimated to be 140, 214, and 378 s, respectively. The elution order is consistent with the protein molecular weights (66 430, 16 700, 13 900, and 6500 Da for BSA, myoglobin, lysozyme, and MT, respectively). However, the baseline resolution (Rs) between BSA and myoglobin and that between lysozyme and MT are 1.2 and 0.56, respectively. Since both Rs values are less than 1.5, the number necessary for obtaining baseline-resolved chromatographic peaks,1 the separation is not considered as successful. Shown in Figure 1B are chromatograms (SPR signal ∆θ vs time) simultaneously acquired from the two SPR channels. Four points are noteworthy. First, unlike the bell-shaped peaks in Figure 1A, the HPLC-SPR peaks are sigmoidal, which are analogous to those in typical SPR sensorgrams for adsorption processes.6 We found that regeneration can be carried out 10 s after the SPR signal had leveled off (i.e., 10 s after an adsorption equilibrium had been reached). For the four proteins studied, the baseline peak width ranges from 76 to 90 s for the peaks shown in Figure 1B. Such widths compare well to those peaks at the baseline in Figure 1A (72, 78, and 168 s estimated for BSA, myoglobin, and MT, respectively). Second, the isoelectric points (pIs) of BSA, myoglobin, lysozyme, and MT are 5.1, 7.9, 11.0, and 5.0, respectively. At the mobile phase pH (6.2), BSA and MT are negatively charged and can be detected at the positively charged (ethylenediamine) surface in channel 1, whereas myoglobin and lysozyme are positively charged and can be adsorbed onto the negatively charged carboxymethylated dextran film in channel 2. Third, the widths of our HPLC-SPR peaks become even narrower and the retention times are shorter if the eluents are not split by a tee connector (Figure 1, parts C and D). This is because the flow rate inside a single SPR channel becomes twice as high as those in the two SPR channels. The tR values, determined from the times (23) Liedberg, B.; Nylander, C.; Lundstrom, I. Biosens. Bioelectron. 1995, 10, i–ix.

Analytical Chemistry, Vol. 80, No. 11, June 1, 2008

4227

Figure 1. Chromatograms showing (A) the separation and UV-vis spectrometric detection at 280 nm of 75.8 µM BSA, 46.0 µM myoglobin, 0.9 µM lysozyme, and 430.0 µM MT; (B) separation and SPR detection of the same mixture. SPR channel 1 was covered with ethylenediaminederivatized carboxymethylated dextran (black curve), and SPR channel 2 was modified with carboxymethylated dextran (red curve). (C) Elution of all four proteins to channel 1 and (D) elution of all four proteins to channel 2. The arrows in panels B, C, and D indicate the times when the postcolumn SPR channel regenerations were made. A flow rate of 1 mL/min was used for all the separation experiments.

when the SPR signals began to level off, are 143 s (BSA), 214 s (myoglobin), 330 s (lysozyme), and 386 s (MT), which are highly comparable to those deduced from the UV-vis spectrometric detection (Figure 1A). Finally, the ability to perform rapid SPR detection with good chromatographic baseline resolution is a result of the instantaneous regeneration processes. The injections of NaOH into channel 1 and HCl into channel 2 generated two negative peaks (spikes), which can be attributed to the large differences in the refractive indexes between the mobile phase and the NaOH and HCl solutions. Notice that the regeneration peaks are only 8-11 s wide. It is particularly noteworthy that the rapid protein adsorption obviates the necessity of flowing the entire protein peak through the SPR cell for analyte quantification. The facile postcolumn regeneration provides the flexibility of cutting off the analyte elution peak. In other words, as soon as the adsorption equilibrium is established (i.e., the entire eluted peak has not passed through the SPR channel), renewal can be initiated. This feature largely reduces the band broadening caused by slow protein adsorption or binding rate, making the peak widths narrower than or comparable with those of the bell-shaped peaks in typical chromatograms. We assessed the reproducibility, detection level, sensitivity, and dynamic range of HPLC-SPR. Figure 2 displays the linear portions of the calibration curves, and the inset depicts ∆θ versus protein concentration over wider concentration ranges. The ranges of the % RSD values across the concentrations in the linear portions (i.e., Figure 2) are 1.8-7.2% for BSA, 3.6-7.4% for myoglobin, 2.2-7.6% for lysozyme, and 2.5-7.1% for MT. These % RSD values are quite reasonable, suggesting that HPLC-SPR 4228

Analytical Chemistry, Vol. 80, No. 11, June 1, 2008

Figure 2. Calibration curves for lysozyme, BSA, MT, and myoglobin. The following equations were obtained from linear regressions: ∆θ ) 2.1[C] + 0.048 (R2 ) 0.998) for lysozyme, ∆θ ) 1.2[C] - 0.020 (R2 ) 0.995) for BSA, ∆θ ) 0.42[C] - 0.050 (R2 ) 0.999) for MT, and ∆θ ) 0.32[C] - 0.090 (R2 ) 0.999) for myoglobin. Each concentration was repeated three times, and the error bar shows the relative standard deviation. The inset shows the relationships between ∆θ and [C] over wider concentration ranges.

is reproducible. The lowest protein concentrations detected are 0.7, 1.0, 8.0, and 12.5 µM for lysozyme, BSA, MT, and myoglobin, respectively. These concentrations are 3-4 orders of magnitude lower (better) than results from previously reported HPLC-SPR studies that monitored the refractive index changes in the mobile phase.9–13 Such detection levels are comparable to those achievable by UV-vis spectrometry.3,24 The dual-channel SPR instrument we used is capable of resolving a SPR signal change of ca.

Figure 3. Chromatograms of 72.3 µM poly-Arg and 160.0 µM FT recorded with (A) a UV-vis spectrometric detector and (B) the SPR detector. The SPR flow channel used for the detection was covered with a carboxymethylated dextran film. The arrows indicate the times when regeneration solutions were injected.

0.1 mDeg,25 which corresponds to about 1.4 × 10-6 refractive index units. Thus, the detection level is better than that of the conventional refractive index detectors used for column chromatography.3 The slopes (sensitivities) for the four proteins are in the order of lysozyme (2.1) > BSA (1.2) > MT (0.42) > myoglobin (0.32). The linear plots suggest that amounts of the adsorbed proteins are proportional to the protein concentrations in the eluents (i.e., linear adsorption isotherms26). Thus, a higher protein concentration should facilitate the adsorption of more protein molecules onto the sensor surface and the attainment of the full monolayer surface coverage. We note that the SPR sensitivity, detection level, and peak width, in addition to being governed by the protein adsorption coefficient and/or binding constant, are dependent on factors such as the polarizability of the protein, the number of ionizable amino acid residues on the surface of the protein (i.e., protein pI value), and the size of the protein. Thus, different protein molecules at the same concentration will exhibit different signal intensities. The dynamic ranges were found to be 2.0-240 µM for BSA, 25-1200 µM for myoglobin, 2.0-200 µM for lysozyme, and 10-800 µM for MT. In the inset of Figure 2, although the proportionality between the SPR signal and the protein concentration in the eluent band has become smaller at protein concentrations ∼1 mM, the SPR signal does not level off at the full surface coverage that is expected to attain at a high protein concentration. This is because when the protein concentration in the solution bulk is high, the refractive index variation in the eluent has become so large that it now can be readily detected by the SPR (i.e., the evanescent field is considerably perturbed by the large solution refractive index variation). As mentioned in the introduction, compounds that either do not possess chromophores or do not absorb strongly in the UV-vis range might be difficult to detect via absorbance measurement. The chromatogram of a mixture of poly-L-arginine (polyArg) and a small peptide, H-Lys-Cys-Thr-Cys-Cys-Ala-OH (FT), is an example. As can be seen from Figure 3A, elution of polyArg, which does not absorb strongly in the UV-vis range, appeared as an ill-defined and weak peak before the FT elution (tR ) 318 s) and resulted in an unsatisfactory Rs value (∼0.99). In contrast, the chromatogram recorded with the SPR detector (24) LaCourse, W. R. Anal. Chem. 2000, 72, 37R–51R (25) Zhai, P.; Guo, J.; Xiang, J.; Zhou, F. J. Phys. Chem. C 2007, 111, 981–986. (26) Bard, A. J.; Faulkner, L. R. Electrochemical Methods. Fundamentals and Applications; John Wiley & Sons: New York, 2001.

(Figure 3B) exhibited two well-defined chromatographic peaks. Poly-Arg is a biopolymer composed of many positively charged arginine residues, and FT is also positively charged because of its lysine residue. Owing to the presence of more ionizable residues and a bulkier size, poly-Arg produced a greater peak than FT. The excellent reproducibility between two consecutive separations again suggests that the surface regeneration was effective. In fact, we found that multiple separations could be performed continuously over 20 times without appreciable SPR signal degradation. Finally, to demonstrate the practical aspect of HPLC-SPR, we applied it to the quantification of lysozyme in homogenized human milk samples. As shown in Figure 4A, owing to the overlap between the broad lysozyme peak and the lactofarrin peak, baseline resolution is difficult to achieve with the UV-vis spectrometric detection. The interference from lactofarrin is absent in the HPLC-SPR chromatograms since under the same experimental conditions, lactofarrin does not adsorb onto the dextran film (chromatogram corresponding to the injection of a pure lactofarrin solution not shown). As can be seen from Table 1, the amounts of lysozyme in the milk samples, quantified by the method of standard addition (cf., the black chromatogram which corresponds to the elution of a lysozyme-spiked milk sample), are in excellent agreement with those deduced using the calibration curve in Figure 2. This suggests that the matrix effect is largely circumvented in the HPLC-SPR measurements. The quantified lysozyme concentrations are also well within the typical lysozyme content in human milk (0.09-1.5 mg/mL),27 suggesting that our measured results are accurate. CONCLUSIONS An interface has been developed for continuous SPR detection of proteins eluted out of a size exclusion liquid chromatographic column. This interface utilizes the postcolumn regeneration of the SPR sensor surfaces so that the highest sensitivity associated with the strongest SPR evanescent field at the sensor surface is retained. As a result, SPR detection at the sensor surface measures eluents at a level 3 orders of magnitude lower than SPR detection of refractive index changes in the bulk mobile phase. The detection level is also lower than that afforded by the traditional (27) Montagne, P. M.; Cuilliere, M. L.; Mole, C.; Bene, M. C.; Faure, G. Clin. Chem. 1998, 44, 1610–1615.

Analytical Chemistry, Vol. 80, No. 11, June 1, 2008

4229

Figure 4. (A) Chromatogram of lysozyme (tR ) 380 s), lactofarrin (tR ) 242 s), and an unknown species (tR ) 138 s) in a human milk sample recorded with a UV-vis spectrometric detector and (B) two consecutive chromatograms detected with the SPR detector of the same sample without (red curve) and with lysozyme spiked at a final concentration of 0.1 mg/mL (black curve). The flow rate in (A) was 1 mL/min, whereas that in the SPR channel was 0.5 mL/min due to the splitting of the column eluent into two solution streams. Table 1. Quantification of Lysozyme in Homogenized Human Milk Samples

sample

replicates

calibration method (mg/mL)

1 2

5 5

0.150 0.093

% RSD

standard addition (mg/mL)

% RSD

7.0 5.3

0.140 0.089

5.0 6.3

refractive index detector. The injection of regeneration solution through a six-port valve facilitates rapid adsorption of protein molecules onto a chemically modified sensor surface. HPLC-SPR based on this coupling interface yields highly reproducible chromatograms and is complementary to other commonly used chromatographic detectors (e.g., UV-vis spectrometric detector). The applicability of this hyphenated technique to real sample analysis is demonstrated with the quantification of lysozyme in

4230

Analytical Chemistry, Vol. 80, No. 11, June 1, 2008

human milk. The methodology described herein is robust and obviates the need for labeling analyte species with chromophores and redox-active labels. Thus, SPR, when used with postcolumn sensor surface regeneration, expands the stockpile of liquid chromatographic detectors designed for diverse analytical applications. ACKNOWLEDGMENT Partial support of this work by the RIMI Program at California State University, Los Angeles (P20-MD001824-01), an NIH-SCORE subproject (GM-08101), and an NSF-RUI Grant (No. 0555224) is gratefully acknowledged.

Received for review December 28, 2007. Accepted April 5, 2008. AC702632Y