Capillary electrokinetic chromatography of insulin and related synthetic analogues

Capillary electrokinetic chromatography of insulin and related synthetic analogues

Journal of Chromatography A, 1216 (2009) 2953–2957 Contents lists available at ScienceDirect Journal of Chromatography A journal homepage: www.elsev...

274KB Sizes 0 Downloads 50 Views

Journal of Chromatography A, 1216 (2009) 2953–2957

Contents lists available at ScienceDirect

Journal of Chromatography A journal homepage: www.elsevier.com/locate/chroma

Capillary electrokinetic chromatography of insulin and related synthetic analogues K. Ortner, W. Buchberger ∗ , M. Himmelsbach Johannes-Kepler-University, Institute of Analytical Chemistry, Altenbergerstrasse 69, A-4040 Linz, Austria

a r t i c l e

i n f o

Article history: Available online 7 November 2008 Keywords: Insulin Insulin analogues Capillary electrophoresis Micellar electrokinetic chromatography Microemulsion electrokinetic chromatography Mass spectrometry

a b s t r a c t With the implementation of recombinant DNA technology in the pharmaceutical industry, some synthetic insulins have been developed in order to improve the therapy of diabetes. These analogues differ only slightly in the amino acid sequence, therefore displaying a great challenge for analytical chemistry. Within the work presented in this paper, capillary zone electrophoresis (CZE), micellar electrokinetic chromatography (MEKC) with sodium dodecylsulphate (SDS) as micelle-forming agent, and microemulsion electrokinetic chromatography (MEEKC) with microemulsions consisting of SDS, n-octane and 1-butanol were investigated for the separation of human insulin and five synthetic analogues. Best results were achieved with a solvent-modified MEKC system consisting of 100 mM sodium dodecyl sulphate and 15% acetonitrile in 10 mM borate buffer (pH 9.2). A similar system based on perfluorooctanoic acid as micelleforming agent in ammonium acetate (pH 9.2) was successfully employed for the hyphenation with a quadrupole/time-of-flight mass spectrometer via a sheath-flow interface. In this case, detection limits at 10 mg/L could be achieved. © 2008 Elsevier B.V. All rights reserved.

1. Introduction Since 1921, when Banting and Best discovered insulin as a major regulator for the human metabolism, the treatment of diabetes mellitus is an impressive success story. The initial sources of insulin for clinical use in humans were cattle, horse or hog pancreases. Insulin from these sources is effective in humans as it is nearly identical to human insulin—the porcine insulin differs in only one amino acid, whereas the bovine insulin only varies in three amino acids [1]. Differences in applicability of these insulins for individual patients have historically been due to poor purity resulting in allergic reactions towards the presence of non-insulin substances. As a result of highly developed purification techniques, e.g. liquid chromatography, the quality has improved steadily up to a purity of 99%, but minor allergic reactions still occur occasionally. The advent of recombinant DNA (rDNA) technology in the 1980s and its application in the pharmaceutical industry has made a revolutionary impact in the area of recombinant pharmaceuticals [2]. Since that time, several recombinant insulins with slightly modified structures have become available for human use, simplifying diabetes

∗ Corresponding author. Tel.: +43 732 2468 8724; fax: +43 732 2468 8679. E-mail address: [email protected] (W. Buchberger). 0021-9673/$ – see front matter © 2008 Elsevier B.V. All rights reserved. doi:10.1016/j.chroma.2008.11.008

treatment for the patients and assuring a constant blood level of insulin [3]. Human insulin is a peptide hormone consisting of two chains with 21 and 30 amino acids, respectively. These two chains are connected via two disulphide bonds (Fig. 1). Its relative molecular mass is 5808 g/mol, the isoelectric point pI 5.4–5.5. Various recombinant insulin analogues nowadays on the market are listed in Table 1. While in Insulin Lispro the primary sequence has been altered by the inversion of the order of the amino acids at position 28 and 29 in the B-chain, in Insulin Aspart and Insulin Glulisin two or three amino acids have been changed. Insulin Glargin is characterized by an extension of the B-chain with two additional arginin molecules. In Insulin Detemir a fatty acid – myristic acid – is connected to the amino acid at position B29, while B30 is missing. The various insulin analogues may either provide a slow effect which assures a constant basic supply and requires the use one or two times daily, or may show a rapid-acting behaviour and are applied directly before the meal. The latter also prevent dangerous glucose spikes in the blood circuit [4]. With the increasing number of these modified analogues, there is a rising need for efficient analysis techniques. A variety of analytical methods, which can be roughly sorted into immunochemical and instrumental analytical methods, had been applied for determination of insulin. Immunochemical methods have been used for detecting insulin in vivo early from 1950s, which mainly included radioimmunoassays [5,6], enzyme immunoassays [7,8],

2954

K. Ortner et al. / J. Chromatogr. A 1216 (2009) 2953–2957

Fig. 1. Structure of human insulin.

and luminescence immunoassay [9]. Instrumental methods are mainly based on high-performance liquid chromatography (HPLC) which is the method of choice for the analysis of pharmaceutical preparations as well as biological samples [10–17]. Capillary electrophoresis (CE) has become an attractive alternative for the analysis of therapeutic peptides and proteins and is known for its high separation efficiency, fast analysis time, and simple setup. The separation of insulins by capillary zone electrophoresis (CZE), has been reported [18–20], but this technique has not yet been applied to recombinant insulin analogues. Pseudostationary phases like micelles (micellar electrokinetic chromatography, MEKC) or microemulsion droplets (microemulsion electrokinetic chromatography, MEEKC) can broaden the field of applications in CE. Partitioning equilibria between the micelles in MEKC or the droplets in MEEKC and the aqueous carrier electrolyte provide additional separation mechanisms besides pure electrophoresis. MEKC has been in use for separations of peptides for some time [21] and has also demonstrated its potential for peptide mapping [22,23]. MEEKC has frequently been employed for separation of low- to medium-mass-analytes [24], whereas its use for high-molecular-mass molecules has been limited [25]. The aim of the work presented in this paper was the investigation of the potential of MEKC and MEEKC for separation of closely related peptides like human insulin and its synthetic analogues with quite similar amino acid sequences, and the comparison of the results with those obtained by CZE. Electrokinetic separation techniques offer new selectivities due to the increased number of parameters

Table 1 Differences in the amino acid sequence of various insulins. Position in the A- or B-chain A21

B3

B28

B29

B30

Insulin Human Insulin Lispro Insulin Aspart Insulin Glulisin Insulin Glargin

Asn Asn Asn Asn Gly

Asn Asn Asn Lys Asn

Pro Lys Asp Pro Pro

Lys Pro Lys Glu Lys

Thr Thr Thr Thr Thr

Insulin Detemir

Asn

Asn

Pro

Lys



B-chain extended with 2 Arg at B30 Myristic acid at B29

The bold words highlight those amino acids that are different in comparison with human insulin.

which can be varied. The present work shows the systematic optimization of carrier electrolyte compositions, which results in an excellent separation of the analytes despite their chemical similarities. In addition, experiments on hyphenation of MEKC with mass spectrometry (MS) employing a volatile surfactant were included in order to facilitate the identification of the various analytes. MEKC has already been described for analysis of insulin in an oil formulation [26], but separation of insulin analogues have not yet been reported. 2. Materials and methods 2.1. Chemicals Boric acid, ammonia, disodium hydrogenphosphate, 1-butanol and formic acid were purchased from Merck (Darmstadt, Germany), citric acid and acetonitrile were supplied by Fisher Scientific (Schwerte, Germany), ammonium acetate was from Acros (Geel, Belgium). Sodium dodecyl sulphate (SDS) and isopropanol were obtained from Fluka (Buchs, Switzerland), perfluorooctanoic acid and Brij 35 were from Sigma–Aldrich (Steinheim, Germany), noctane was supplied by Baker (Griesheim, Germany) and sodium hydroxide by ACM (Vienna, Austria). All reagents used were of analytical grade. Water from a Milli-Q (Millipore, Bedford, MA, USA) purification system was used throughout the work. Commercially available pharmaceutical formulations of the different insulins were used for method development. Human insulin (Insulin Actrapid), Insulin Detemir (Insulin Levemir) and Insulin Aspart were obtained from Novo Nordisc (Bagsvaerd, Denmark), Insulin Glulisin (Apidra) and Insulin Glargin (Insulin Lantus) from Sanofi Aventis (Bridgewater, NJ, USA), and Insulin Lispro (Humalog) from Eli-Lilly (Indianapolis, IN, USA). Each insulin preparation contained 100 international units (I.U.) of insulin or insulin analogue, which correspond to approximately 3.5 mg/mL insulin (in the case of human insulin, Insulin Glargin, Insulin Aspart, Insulin Glulisin, and Insulin Lispro) or 14.2 mg/mL (in the case of Insulin Detemir), respectively. 2.2. Instrumentation CE separations were carried out on an Agilent 3D CE system (Agilent, Waldbronn, Germany) with diode-array detection. Fused

K. Ortner et al. / J. Chromatogr. A 1216 (2009) 2953–2957

silica capillaries used in this work (50 ␮m I.D. × 375 ␮m O.D.) were obtained from Polymicro Technologies (Phoenix, AZ, USA). New capillaries were cut to a total length of 48.5 cm (UV detection) or 75 cm (MS detection), respectively, and conditioned by flushing with 0.5 M NaOH, water, and background electrolyte (BGE) for 30 min each. As a daily routine, the capillary was flushed with 0.5 M NaOH, water, and BGE for 15 min each before the first run. Prior to each run, the capillary was flushed with BGE for 3 min. The separation voltage was +30 kV and the capillary temperature was 20 ◦ C. Sample injection was carried out by hydrodynamic injection, 50 mbar for 5 s (UV detection) or 10 s (MS detection), respectively. MS detection was performed on an Agilent Q-TOF mass spectrometer equipped with an electrospray ionization (ESI) source. All analyses were done in the positive ionization mode. Nitrogen was used as a drying gas at a temperature of 250 ◦ C and a flow rate of 4 L/min. The voltage set for the MS capillary was 3.5 kV. The scanning mass range included the mass-to-charge ratios (m/z) from 1450 to 1550. The CE was hyphenated with the MS via an orthogonal interface with sheath liquid, which consisted of isopropanol/water (50:50, v/v) with 0.1% formic acid and was delivered via a syringe pump (Modell 22, Harvard Apparatus, South Natick, MA, USA) at 10 ␮L/min. 2.3. Samples and carrier electrolytes For all experiments the commercial samples of the different insulins were diluted with background electrolyte, resulting in a solution of 70 mg/L of each insulin. Buffers were prepared by dissolving appropriate amounts of boric acid, ammonium acetate, disodium hydrogenphosphate or citric acid in water, and adjusting the pH with aqueous sodium hydroxide. MEKC-solutions were prepared by dissolving SDS or perfluorooctanoic acid in the suitable buffer. In the case of perfluorooctanoic acid, the used buffer was ammonium acetate and the pH was readjusted with ammonia at the end of the preparation procedure. The sheath liquid for the CE–MS interface was prepared by adding 1 mL formic acid to 1 L isopropanol/water (50:50, v/v). The preparation of the microemulsions included mixing of surfactant, oil phase and short-chain alcohol with the appropriate buffer, sonication for 30 min, and filtration through a 0.45 ␮m filter. 3. Results and discussion First experiments were carried out employing capillary zone electrophoresis, which is known as a simple and fast separation technique, although with the major disadvantage of possible peptide adsorption on the inner capillary wall. Different buffers, such as sodium borate, ammonium acetate, sodium phosphate and ammonium citrate with a concentration of 10 mM and a pH of 9.2 and varying concentrations of acetonitrile were used as BGEs. The pH was varied over a range of 7.7–9.2 (lower pH values were not investigated in order to prevent difficulties with precipitation of the analytes during the run because of the poor solubility of Insulin Glargin at neutral pH). Even though the results of the injection of single insulins were promising and comparable with results reported in the literature [19], no separations could be obtained for all six analytes in one run (Fig. 2a). The use of different buffers did not show any effect on the separation. With increasing acetonitrile content there was some peak broadening but no better separation of the insulins. In addition, a separation of peaks corresponding to phenol and m-cresol was observed which are used as preservatives in the commercial insulin solutions. MEKC is a capillary electrophoresis technique that utilizes buffer solutions containing surfactants above their critical micelle con-

2955

Fig. 2. Comparison of different CE separation techniques for separation of insulins. Capillary: 40 cm effective length, 50 ␮m I.D. (total length 48.5 cm). Separation voltage: +30 kV. Temperature: 20 ◦ C. Injection: 50 mbar, 5 s. Detection: UV at 210 nm. Analyte concentrations: 70 mg/L each. Peaks: 1 = phenol, 2 = cresol, 3 = Insulin Aspart, 4 = human insulin, 5 = Insulin Lispro, 6 = Insulin Glulisin, 7 = Insulin Glargin, 8 = Insulin Detemir. (a) CZE (BGE: 10 mM borate buffer, pH 9.2); (b) MEKC (BGE: 50 mM SDS in 10 mM borate buffer, pH 9.2); (c) MEEKC (BGE: 3% SDS, 0.8% n-octane, 6.6% 1-butanol in 10 mM borate buffer, pH 9.2); (d) solvent-modified MEKC (BGE: 50 mM SDS, 15% acetonitrile in 10 mM borate buffer, pH 9.2).

centration (CMC) to separate analytes with identical net charges. Since the size of the micelles is about 3–6 nm in diameter, insulin with a molecular weight of 5808 g/mol is slightly too large to partition into the hydrophobic core of the micelles and can only associate with micelles through hydrophobic, hydrophilic and electrostatic mechanisms [27]. Since peptides are charged under basic conditions, they exhibit their own electrophoretic mobilities. Separation is therefore based on this electrophoretic mobility and on the interaction with the micelles. The same buffers which were used in CZE were tested, but no significant improvement of the separation could be achieved. Due to the fact that ammonium acetate and sodium borate buffers showed the best peak shape, subsequent experiments were limited to these two buffers. As can be seen in Fig. 2b, with 50 mM SDS only two peaks were obtained. The next step was the development of an MEEKC system. Starting with the standard conditions for a microemulsion (3% SDS, 0.8% n-octane, 6.6% 1-butanol, 10 mM sodium borate buffer, pH 9.2 [28]), a partial separation was achieved (Fig. 2c). To improve the separation, the impact of acetonitrile was investigated. A slightly better separation was achieved with 5% acetonitrile. The major disadvantage of adding higher amounts of acetonitrile was the tedious preparation of the microemulsion. To get a clear microemulsion, solutions had to be sonicated for nearly 2 h. To speed up the separation, a part of SDS was replaced by Brij 35, a nonionic detergent, to slow the migration of the microemulsion droplet towards the inlet of the capillary. This resulted in a somewhat faster but not a really satisfactory separation. To finish the comparison of these different electrokinetic separation techniques, a forth technique was used, called solvent-modified micellar electrokinetic chromatography, which is considered to be a link between MEKC and MEEKC [29,30]. As a starting point, the MEKC system described above was employed, and different concentrations of acetonitrile and 1-butanol were added. While 1-butanol did not lead to better separations, promising results were obtained with acetonitrile which was varied from 5 to 15%. Up to 15% acetonitrile, this organic modifier has no remarkable impact on the CMC of SDS in the electrolyte [31]. The next step was the variation of the SDS concentration from 20 to 100 mM. With increasing SDS content, the migration times increased, and best separations were achieved with 35 mM SDS and 15% acetonitrile. Finally, the pH was varied over a range from 7.7 to 9.2 as it

2956

K. Ortner et al. / J. Chromatogr. A 1216 (2009) 2953–2957

was done within in the MEKC experiments. A pH of 9.2 yielded the optimum results. A typical separation can be seen in Fig. 2d. As mentioned above, the size of the micelles is too small to allow a separation of insulins due to partitioning. Simple partitioning would also hardly explain the observed separation of closely related analytes like human insulin and Insulin Lispro, which differ just with respect of the order of the amino acids lysine and proline at the positions B28 and B29 (see Table 1). Therefore, other types of interactions between the analytes and SDS must be responsible for the separation. It is a well-known fact that dodecyl sulphate ions associate with proteins, thereby forming negatively charged complexes. This reaction is often seen as a denaturation step that includes the unfolding of the protein. Nevertheless, this interaction with SDS does not imply that the association product would not exhibit any secondary structure [32]. It seems reasonable to assume that the number of dodecyl sulphate ions associated with different insulins as well as the secondary structure of the SDS–protein complexes (and therefore the size) may be slightly different, thereby enabling the electrophoretic separation. To improve the selectivity of the detection, the hyphenation of the MEKC separation system with mass spectrometric detection was investigated. In case of SDS which is not sufficiently compatible with MS, a partial-filling technique can be applied. Only a part of the capillary contains micelles whereas the rest contains the background electrolyte without micelles so that SDS does not enter the MS interface. A major disadvantage of this technique may be an extra band-broadening mechanism occurring at the boundary between micellar zone and buffer boundary. Therefore, the use of a volatile micelle-forming compound compatible with MS instead of SDS might be the better choice and can avoid the use of the partial-filling technique [33]. In the present work perfluorooctanoic acid (PFOA) in combination with a volatile ammonium acetate buffer (pH adjusted to 9.2 with ammonia) was investigated. Preliminary experiments with UV detection indicated that the separation was similar to that obtained with SDS micelles in a borate buffer, except for the migration order of Insulin Glargin and Insulin Detemir which was reversed. This change must be due to the different micelle-forming compound, since the mere replacement of the borate electrolyte by the ammonium acetate electrolyte did not result in any changes of migration order. Although PFOA should be volatile enough to avoid serious contamination of the MS detector, suppression of the MS signal may still occur. The possible impact of the micellar background electrolyte on the sensitivity was investigated by direct infusion of sheath liquid without or with 1% MEKC electrolyte, and containing 10 mg/L insulin. As can be seen in Fig. 3 from the intensities at m/z 1162.5 and 1452.7 (charge number 5 and 4, respectively), significant suppression does occur, but the signals were still suited for qualitative and quantitative analysis. For further optimization of the hyphenation, four different sheath liquid compositions were compared, namely isopropanol/0.1% formic acid (50:50, v/v), isopropanol/10 mM ammonium acetate pH 3.5 (50:50, v/v), isopropanol/10 mM ammonium acetate pH 6.7 (50:50, v/v) and methanol/10 mM ammonium acetate, pH 3.5 (50:50, v/v). The selection of a mixture of isopropanol/aqueous solution (50:50, v/v) was based on experience obtained in previous work on CE–MS. Employing the signal-to-noise ratio obtained in the total ion current as the selection criterion, a sheath liquid consisting of isopropanol/water (50:50, v/v) with 0.1% formic yielded the best results. The flow rate of the sheath liquid was also varied (5, 7 and 10 ␮L/min), but the effect on signal/noise was not significant (maybe the suppression effect of PFOA dominated to such an extent that the effect of different flow rates became negligible). To ensure a stable electrospray, the sheath liquid flow was set to 10 ␮L/min. Further optimization of the mass spectrometer included the gas temperature which was varied from 225 to 300 ◦ C,

Fig. 3. Mass spectra obtained by direct infusion of (a) 10 mg/L human insulin in isopropanol/water (50:50, v/v) with 0.1% formic acid, (b) 10 mg/L human insulin in isopropanol/water (50:50, v/v) with 0.1% formic acid and 1% MEKC buffer based on PFOA. (c) Detail of (b), mass range m/z 900–1500.

and the fragmentor voltage which was set from 180 to 250 kV. Best results were achieved with 250 ◦ C gas temperature and 230 kV fragmentor voltage. In Fig. 4 the calculated isotope pattern of human insulin is shown and compared with the spectrum obtained within a MEKC–MS run. This comparison clearly demonstrates that resolution and mass accuracy of the MS detection is sufficient for identification of the insulins. Under the optimized conditions, all six analytes could be separated and detected by mass-spectrometry with a limit of detection (signal/noise ratio of 3) of 10 mg/L (see Fig. 5).

Fig. 4. MS spectrum of human insulin (molecular ion with charge number of 4). (a) Calculated mass spectrum; (b) mass spectrum obtained within a CE run. Analyte concentration: 70 mg/L.

K. Ortner et al. / J. Chromatogr. A 1216 (2009) 2953–2957

2957

based on PFOA was optimized allowing detection limits in the low mg/L level. Although these detection limits may be higher than those obtained by HPLC, the CE method can be seen as a complementary technique with different separation selectivity. References [1] [2] [3] [4] [5] [6] [7] [8] [9]

Fig. 5. Separation of insulins by MEKC with (a) UV-detection, (b) mass spectrometric detection. BGE: 50 mM PFOA in 10 mM ammonium acetate buffer, pH 9.2. Capillary: 40 cm (UV) or 75 cm (MS) total length, 50 ␮m I.D. Injection: 50 mbar, 5 s (UV), 10 s (MS). All other parameters and peak assignment as in Fig. 2.

[10] [11] [12] [13] [14] [15] [16]

4. Conclusions

[17] [18]

In this work the suitability of solvent-modified MEKC based on either SDS or PFOA as micelle-forming agents could be successfully demonstrated for the analysis of human insulin and its synthetic analogues. Although these compounds exhibit only minor differences in the primary structure of the peptide, a sufficient separation of six different insulins could be achieved. For two of the insulins included in this study, migration order depended upon the type of micelle-forming agent. It can be assumed that the separation is not based on simple partitioning between the background electrolyte and the micelles, but involves the formation of association complexes of various structures depending on the type of detergent. A more detailed study of these equilibria will be the topic of additional future work. For hyphenation with a Q-TOF mass spectrometer equipped with electrospray ionization, an MEKC system

[19] [20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33]

H. Ye, J. Hill, J. Kauffman, C. Gryniewicz, X. Han, Anal. Biochem. 397 (2008) 182. G.M. Bhopale, R.K. Nanda, Curr. Sci. 89 (2005) 614. A. Rolla, Am. J. Med. 121 (2008) S9. G.B. Bolli, R.D. Di Marchi, G.D. Park, S. Pramming, V. Koivisto, Diabetologia 42 (1999) 1151. R.S. Yalow, S.A. Berson, Nature 184 (1959) 1648. R.R. Bowsher, R.A. Lynch, P. Brown-Augsburger, P.F. Santa, W.E. Legan, J.R. Woodworth, R.E. Chance, Clin. Chem. 45 (1999) 104. H.V. Webster, A.J. Bone, K.A. Webster, T.J. Wilkin, J. Immunol. Methods 134 (1990) 95. L. Andersen, P.N. Jorgensen, L.B. Jensen, D. Walsh, Clin. Biochem. 33 (2000) 627. K. Zaitsu, Y. Kimura, Y. Ohba, K. Hamase, Y. Motomura, M. Itose, M. Ishiyama, Anal. Sci. 15 (1999) 871. Y. Luo, K. Huang, H. Xu, Anal. Chim. Acta 553 (2005) 64. K. Sawicka, T. Sahota, M.J. Taylor, S. Tanna, J. Chromatogr. A 1132 (2006) 117. P. Moslemi, A.R. Najafabadi, H. Tajerzadeh, J. Pharm. Biomed. Anal. 33 (2003) 45. A. Oliva, J. Farina, M. Llabres, J. Chromatogr. B 749 (2000) 25. C. Yomota, Y. Yoshii, T. Takahata, S. Okada, J. Chromatogr. A 721 (1996) 89. E.N.M. Ho, T.S.M. Wan, A.S.Y. Wong, K.K.H. Lam, B.D. Stewart, J. Chromatogr. A 1201 (2008) 183. M. Thevis, A. Thomas, P. Delahaut, A. Bosseloir, W. Schänzer, Anal. Chem. 78 (2006) 189. G. Khaksa, K. Nalini, M. Bhat, N. Udupa, Anal. Biochem. 260 (1998) 92. N.F.C. Visser, M. van Harmelen, H. Lingeman, H. Irth, J. Pharm. Biomed. Anal. 33 (2003) 451. A. Kunkel, S. Günther, C. Dette, H. Wätzig, J. Chromatogr. A 781 (1997) 445. C. Yomota, Y. Matsumoto, S. Okada, Y. Hayashi, J. Chromatogr. B 703 (1997) 139. V. Kasicka, Electrophoresis 29 (2008) 197. S.H. Kang, X. Gong, E.S. Yeung, Anal. Chem. 72 (2000) 3013. J. Huang, J. Kang, J. Chromatogr. B 846 (2007) 364. E. McEvoy, A. Marsh, K. Altria, S. Donegan, J. Power, Electrophoresis 28 (2007) 193. S.H. Hansen, Electrophoresis 24 (2003) 3900. B. Deng, Z. Liu, G. Luo, H. Ma, M. Duan, J. Pharm. Biomed. Anal. 27 (2002) 73. M.A. Strege, A.L. Lagu, J. Chromatogr. A 780 (1997) 285. K.D. Altria, Chromatographia 49 (1999) 457. S.H. Hansen, C. Gabel-Jensen, D.T.M. El-Sherbiny, Trends Anal. Chem. 20 (2001) 614. S.H. Hansen, C. Gabel-Jensen, S. Pedersen-Bjergaard, J. Sep. Sci. 24 (2001) 643. S. Lopez-Grio, J.J. Baeza-Baeza, M.C. Garcia-Alvarez-Coque, Chromatgraphia 48 (1998) 655. W. Parker, P.S. Song, Biophys. J. 61 (1992) 1435. P. Peterson, M. Jörnten-Karlsson, M. Stalebro, Electrophoresis 24 (2003) 999.