Capillary electrophoresis for rapid identification of monoclonal antibodies for routine application in hospital

Research Article

Capillary electrophoresis for rapid identification of monoclonal antibodies for routine application in hospital mAbs are widely used in cancer therapy. Their compounding, performed just before their administration to patients, is executed in a production unit of the hospital. Identification of these drugs, individually prepared in bags for infusion before patient administration, is of paramount importance to detect potential mistakes during compounding stage. A fast and reliable analytical method based on CZE combined to a cationic capillary coating (hexadimethrine bromide) was developed for identification of the most widely used compounded therapeutic for cancer therapy (bevacizumab, cetuximab, rituximab, and trastuzumab). Considering the high structural and physico-chemical similarities of these mAbs, an extensive optimization of the BGE composition has been performed. The addition of perchlorate ions and polysorbate in the BGE greatly increased the resolution. To validate the method, an internal standard was used and the relative migration times (RTm) were estimated. Very satisfactory RSDs of the RTm for rituximab (0.76%), cetuximab (0.46%), bevacizumab (0.31%), and trastuzumab (0.60%) were obtained. The intraday and interday RSD of the method were less than 0.32 and 1.3%, respectively for RTm. Significant differences between theses RTms have been demonstrated allowing mAbs identification. Finally, accurate mAbs identification has been demonstrated by a blind test.

Keywords: Cationic coating / CZE / Optimization / Therapeutic monoclonal antibody / Validation DOI 10.1002/elps.201400603

1 Introduction

mAbs are widely used in cancer therapy [1, 2]. Nowadays, 14 different monoclonal antibodies for oncology indications have gained European Medicines Agency (EMA) and Food and Drug Administration (FDA) approvals while 150 mAbs are under clinical trials [3, 4]. These last years, the number of different mAbs prescribed in hospitals has considerably increased worldwide [5–7]. Before their administration to pa- tients, mAbs must be compounded under aseptic area by highly qualified persons according to manufacturers and in- stitutional guidelines. Hospitals need therefore to implement production units with a large scale of compounding activity. Consequently, quality controls (QC) have now to be set-up in hospitals to ensure the quality, safety of use and to avoid medication errors. For the regulation authorities, these com- pounded/reconstituted mAbs already approved by FDA and EMA, do not require anymore, at this stage, in-depth char- acterization such as micro-heterogeneity determination or degradation profile. If assays are generally not an issue for QC of compounded mAbs, identity checking is more challenging considering their similarities. A fast, simple, and quantita- tive method allowing identification and dosage checking is desirable for the QC of these mAbs. This excludes complex separation techniques combined for instance to MS, generally employed for mAbs characterization [8, 9]. Moreover, mini- mal or no sample pretreatment should be preferred for this specific quality control of mAbs present in the infusion bags. Most of the studies focusing on mAbs analysis after reconstitution have generally employed complex and orthog- onal analytical approaches based on chromatographic or electrophoretic methods. In addition, most of them included sample modifications prior analysis such as denaturation (by heating or SDS treatments) or enzymatic digestion [8–13]. Besides chromatographic or conventional electrophoretic techniques, an analyzer based on a combination of infrared and UV spectroscopies has been recently proposed by Bazin et al. [14]. However, the method failed to accurately identify mAbs due to the influence of the drug excipients. CE is widely used for mAbs analysis due to its high resolution power. CZE is certainly the most appropriate separation mode to achieve a fast and easy method to implement at the hospital. Although several papers have reported methods for the CZE analysis of intact and therapeutic mAbs [15–21], only one of them has achieved the separation of one mixture of two compounded mAbs, pertuzumab, and trastuzumab [22]. Adsorption in CZE is a major issue for proteins analysis. The mAbs that are “soft proteins” are even more prone to adsorption into silica capillary wall [23], rendering the mi- gration times and the EOF quite variable [24]. Several au- thors have proposed CZE methods based on dynamic or static capillary coating to overcome this problem. For in- stance Taga et al. proposed a carboxylated-modified capil- lary to analyze intact tocilizumab [16]. Neutral coatings such as polyethyleneoxide or hydrophilic cellulose polymers have been also proposed but mostly to characterize charge het- erogeneity of one given mAb [20]. Recently, Gassner et al. presented a comprehensive study comparing different static neutral and cationic coatings for intact mAbs analysis. They concluded that hydroxylpropylcellulose neutral coating gave better results than cationic coatings [21]. The peak obtained for rituximab using a polybrene-based SMIL (i.e. a triple layer of polybrene-dextranesulfate-polybrene coating) was however quite symmetrical and sharp. Besides this work, static posi- tively charged coatings have never been reported to achieve the separation of mixture of mAbs or to analyze compounded therapeutic mAbs. These coatings represent, still, an interest- ing alternative to neutral ones to prevent adsorption of mAbs to the capillary wall, improve resolution and reproducibility of the CZE methods.

We report here a CZE method employing a cationic polymer polybrene (hexadimethrine bromide), to separate with high reproducibility a complex mixture of four reconstituted therapeutic anticancer mAbs widely used in hospital (Cetux- imab, Rituximab, Bevacizumab, and Trastuzumab), and allow their identification with a high specificity, based on their rel- ative migration time. This was particularly challenging if we consider that mAbs share very close physicochemical prop- erties (Mw, charge, glycosylation) and migration behaviors. Cetuximab and rituximab are two chimeric mAbs while beva- cizumab and trastuzumab, are two partially humanized IgG1. A particular attention has been paid to limit their adsorption using a wall coating strategy and by optimizing rinse proto- cols. The method was then validated and applied as a blind test to identify individual mAbs based on their relative migra- tion time directly in infusion bags.

2 Materials and methods
2.1 Chemicals

Orthophosphoric acid from Merck, Millipore (Darmstadt, Germany), sodium hydroxide from ProLabo (Fontenay-sous Bois, France), and acetic acid from Sigma Aldrich (Saint Quentin, France) were of analytical grade. Hexadimethrine bromide or polybrene (PB), (MW: 374 g/mol), polysorbate 20 (Mw: 1228 g/mol), SDS, citric acid, formic acid, and thiourea were purchased from Sigma Aldrich (Saint Quentin,France). L-Glutamine and L-histidine were obtained from Merck Millipore (Darmstadt, Germany). Polysorbate 80 (MW: 1310 g/mol) and sodium citrate were from Cooper (Melun, France). Ultrapure water to prepare all buffers was provided by a MilliQ purification station from Millipore (Bedford, USA). All solutions and buffers have been filtered through a
0.2 µm membrane (Millipore) prior their use.

2.2 Sample preparation

EMA approved monoclonal antibodies, Bevacizumab (Avastin×R –25 mg/mL), Trastuzumab (Herceptin×R – 21 mg/mL), Rituximab (Mabthera×R –10 mg/mL), were pur- chased from Roche, Boulogne-Billancourt, France). Cetux- imab (Erbitux×R –5 mg/mL) was from Merck-Serono (Lyon, France). Polyolefin Freeflex bags containing 0.9% NaCl were from Fresenius Kabi (Fresenius Kabi, France). MAbs were separately compounded in 0.9% NaCl under aseptic condi- tions according to manufacturer guidelines at fixed concen- trations (bevacizumab: 0.5 mg/mL, cetuximab: 1.5 mg/mL, rituximab: 1.0 mg/mL, trastuzumab: 1.2 mg/mL). The an- alyzed samples were collected from compounded solutions stored and protected from light at 4°C. Mixtures of mAbs were prepared freshly before each analysis by mixing equal volumes of each already compounded mAbs. 30 µL of in- ternal standard (glutamine at 1.9 mg/mL) were added to the mAbs solution (individual or mixture) prior to their analysis.

2.2 BGE preparation

A phosphate solution (75 mM) was mixed with an appropriate volume of sodium hydroxide 1 N to reach pH 3.0. Perchlo- ric acid (0.1 M) was added in the phosphate buffer (75 mM, pH 3.0) until the desired concentration of 0.15 mM. Polysor- bate 80 was then added to the BGE at a final concentration of 0.01% w/v.

2.2 Instrumentation

A CE P/ACE MDQ system equipped with an UV detector (Beckman Coulter, CA, USA) was used to perform all the CZE experiments. Fused silica capillaries (TSP, Polymicro Technologies, USA) of 50 µm id × 360 µm od with a total length of 60.5 cm (50 cm effective length) were used for separation. Analysis and collection of the data were obtained with 32 Karat software (Beckman Coulter, CA, USA).

2.5 Methods

2.5.1 CZE separation and polybrene coating

Uncoated fused silica capillaries have been conditioned by successive rinses with water (5 min)/NaOH 1 M (8 min)/water (5 min) at 20 psi. The separation method contained successive rinses with NaOH 0.1M for 2 min (at 20 psi) fol- lowed by 0.2% w/v PB solution in water for 2 min (at 20 psi) to coat positively the capillary. A rinse with the fresh running buffer for 3 min (at 20 psi) was done before the application of the separation voltage. Between each analysis the capil- lary was rinsed with 50 mM SDS (3 min) followed by water (5 min). The PB coating was therefore regenerated before each analysis following the same procedure. All the samples were hydrodynamically injected at 0.5 psi for 10 s. The sepa- rations were carried out under a negative voltage of –20 kV at 25°C. UV detection was made at 214 nm.

2.5.2 EOF and relative migration times measurements

The EO mobilities were evaluated by injecting thiourea (1 mg/mL) as the neutral marker at 0.5 psi for 10 s. The ratio of apparent mobilities of the mAbs to the internal stan- dard (IS) one were calculated to estimate the RTms of each mAbs.

2.5.3 Validation of the method

Intraday and intermediate precisions of the RTms have been assessed. Repeatability was evaluated from six consecutive runs of the mixture of the four mAbs that included the IS, performed the same day and using the same capillary. In- termediate precision has been evaluated by analyzing one mixture of mAbs and IS six times per day for three days (n = 18) with one new capillary each day. ANOVA was applied to evaluate the precision. XLstat software (Addinsoft, FRANCE) was used to calculate all statistical operations.

A blind test has been carried out to assess the ability of the method to identify mAbs from their own RTm. Five blinded samples (A1–A5) corresponding to each of the stud- ied mAbs and a control group have been prepared. Three blinded samples have been randomly selected and then an- alyzed three times under the optimized CE conditions. To determine the expected RTms, three consecutive runs of one mixture of the four mAbs that included the IS have been done. An agglomerative hierarchical clustering (AHC) has been performed on these RTms to assess their significance. This approach provides an interpretable visualization of the data. All data analysis and graphic were performed on XLstat (Addinsoft, France).

3 Results and discussion
3.1 Optimization of the CE conditions

In order to select the best conditions to achieve the CZE sepa- ration of the four mAbs (cetuximab, rituximab, bevacizumab, trastuzumab), counter EOF mode has been selected. In reversal EOF mode, the resolution between mAbs is expected to be higher as the mean of mAb electrophoretic mobility (µep) becomes close to that of the EOF. As adsorption of mAb, to unmodified silica capillaries has been often reported [20, 21], even under repelling conditions (i.e. at pH > pI), we have chosen a silica coating strategy. Two options were pos- sible either working at acidic pHs using a cationic coating or, at neutral or alkaline pHs using a negatively charged one. Anionic coatings such as sodium 2-acrylamido-2-methyl-1- propanesulfonate [25] or polybrene/poly(vinylsulfonate) [26] reported for the CZE of proteins generally exhibit too high electroosmotic mobilities (µeos) that are detrimental to high resolutions. According to data published by Gassner et al. [21] on µeo obtained with different charged coatings, we have considered that at low BGE pHs, the absolute µep values of the investigated mAb were closer to the µeo determined with cationic coatings. We selected therefore polybrene as a static coating. Based on a previous study carried out in our group [27], 0.2% polybrene solution was employed to coat the capillary before each separation.

3.1.1 Buffer composition optimization

Common buffers from pH 3.0 to 7.0 such as sodium citrate, sodium formate, or sodium phosphate have been tested at various ionic strengths from 25 to 100 mM. As depicted in Fig. 1, the best resolution was achieved with sodium phos- phate buffer pH 3.0. We also observed that higher ionic strengths greatly improved the peak shape and the reso- lution between the peaks (data not shown). Considering Joule heating, the conditions with sodium phosphate buffer pH 3.0 at 75 mM were selected for the following experiments. At this stage however, only three mAbs on the four studied, were partially resolved. To improve the separation an anionic ion pairing agent has been tested. Different concentrations of perchloric acid (PA) (from 0 to 1.5 mM) were added to the BGE. As shown in Fig. 2, the resolution between the two last peaks Rs3–4 increased as soon as a small amount of perchloric acid (0.15 mM) was added to the BGE. The resolution between peaks 2 and 3 was also slightly enhanced by this addition. The separation profile was greatly improved with the addition of 0.15 mM perchlorate in the BGE. To explain the observed improvement, we hypothesized that perchlorate ions electro- statically interacted either with the coating or with the mAbs modulating their solubility or mobility. To better understand the role of this ion-pairing agent, we studied the influence of PA addition on both µeo and µep of cetuximab by analyzing thiourea or cetuximab independently. Perchloric acid yielded a 7.6% increase of the µeo when added in BGE as shown in Table 1. Perchlorate anions interact probably with the poly- brene coating. However, if PA compensates the cationic charge of polybrene, we would have expected a decrease of the µeo. It is probable that this interaction modifies only slightly the surface charges but much more the thickness of the dou- ble layer explaining the observed increase of the µeo. Other anions such as phosphate anions (i.e. HPO42−) are present in the BGE but perchlorate ion, which is a stronger ion pairing agent according to Hofmeister series [28] has a greater affinity toward the amines of the PB layer. Besides, we also observed an increase of the effective mobility of the cetuximab by 22% with the addition of 0.15 mM of perchloric acid to the BGE. This could be due to a change either of the net ionic charge or the hydrodynamic radius of this mAb. Finally despite the low concentration of perchlorate added, interactions between the mAbs and perchlorate ions influence greatly their elec- trophoretic behavior, leading to an effective improvement of the mAbs separation. However, a lack of repeatability has been observed when analyzing the complex mixture of mAbs and we suspected some mAbs adsorption on the coating.

Figure 1. Comparison of various BGE for the CZE separation of a mixture of Bevacizumab (0.5 mg/mL), Cetuximab (1.5 mg/mL), Rituximab (1 mg/mL), and Trastuzumab (1.2 mg/mL) with polybrene coating (0.2%). (A) Sodium citrate 50 mM pH 4.0. (B) Sodium phosphate 50 mM pH 3.0. (C) Sodium formate 50 mM pH 5.0. (D) Sodium phosphate 50 mM pH 6.0. (E) Sodium phosphate 50 mM pH 7.0. (F) Sodium phosphate 75 mM pH 3.0. CE conditions: hydrodynamic injection (10 s/0.5 psi). Applied voltage: –25 kV. Temperature: 25°C. Detection wavelength: 214 nm.

Indeed, RSD of migration times for cetuximab reached 2.5% (n = 6) when analyzed with the BGE containing perchlo- ric acid. Thus, a neutral surfactant, polysorbate 80 has been added in the BGE. The influence of various concentrations of Polysorbate 80 (PS 80) from 0.001 to 0.1% w/v in the BGE has been investigated. With 0.1% of PS 80 in the BGE, no peak was obtained within 35 min. Table 2 shows the low RSDs of the migration times (< 0.7%) obtained for all mAbs ana- lyzed separately with the addition of 0.01% PS 80 to the BGE demonstrating the positive impact of this neutral surfactant (n
= 10). We also observed (Fig. 3A), a significant improvement of the separation by this addition. The best resolution between the mAbs mixture was obtained at the concentration of 0.01%.

Figure 2. (A) Effect of the concentration of perchloric acid in the running BGE on the resolution (Rs) of the peaks of the mixture of mAbs. The running buffer was phosphate 75 mM pH 3.0 associ- ated to various concentration of perchloric acid (0–1.5 mM). Solid diamonds: resolutions between peaks 1 and 2; solid squares: peaks 2 and 3; solid triangles: peaks 3 and 4. (B) CZE profile of the mixture of the mAbs with 0.15 mM perchloric acid in the BGE. Other CE conditions as in Fig. 1 a change of 4.3%. As this variation could lead to quite high RSD of Tm (0.29–0.48%), we introduced an internal stan- dard (L-glutamine) at 1.9 mg/mL in the analyzed samples. We compared the precision of both Tm and relative Tm from six consecutive runs of the mAbs mixtures. As expected, the RTm allowed a substantial decrease of the RSD values that reached a twofold factor for three mAbs (0.12–0.32%).

3.2 Validation of the method

The optimized method was then validated in terms of repro- ducibility, specificity, and discrimination power. Tests were also performed to confirm its potential as an identification method able to discriminate the four mAbs.To evaluate the specificity of the method, all expected excipients of the four commercial mAbs have been ana- lyzed (citric acid, glycine, L-histidine, polysorbate 20 and 80, and sodium citrate). The results have shown that none of them interfere with mAbs peaks. This demonstrates the appropriateness of the method to analyze mAbs in presence of their own excipients.We also evaluated the discrimination power of the method by analyzing the four mAbs separately ten consec- utive times. ANOVA tests based on Levene’s tests demon- strated significant difference in RTm of the four mAbs (p < 0.023). Indeed, the results on the RSD of RTm that are gathered in Table 2 show good RSD less than 1%. There- fore, we can conclude that RTm of each studied mAb is spe- cific and significantly different from that of another mAb.

3.1.2 Reproducibility studies

The µ EOF repeatability was assessed under these optimal conditions. It was calculated from three consecutive runs per- formed both before and after six consecutive mAbs analyses. The µEO decreased from 3.94 × 10− 04 (before mAbs runs) to 3.77 × 10−04 cm2V−1s−1 (after mAbs runs) representing This discrimination based on their relative migration time is therefore “exploitable“for an identity control of a given com- pounded mAb.Intraday precision was evaluated from one sample of mAbs mixture analyzed six consecutive times, the same day, and using the same capillary. Intermediate precision has been evaluated by analyzing six times per day and for three days (n = 18) one Mab mixture, each day the capillary was replaced. Mean relative migration times and their RSD are reported in Table 2. RSDs less than 0.32% were obtained for the intraday precision of the relative migration times. The RSDs of the mean RTm for intermediate precision were less than 1.3% for the four mAbs. These results highlight the low variability of the method.

Figure 3. (A) Effect of the concentration of the polysorbate 80 (PS) added to the BGE on the mAbs separation. Other CE conditions as in Fig. 2. (B) Dendogram of the agglomerative hierarchical clustering (AHC) of the mAbs and the blinded samples A2 and A3 based on their mean relative migration times. Automatic truncation (dotted line) selected to distinguish the four groups. (C) Electrophoregram of bevacizumab (0.5 mg/mL), cetuximab (1.5 mg/mL), rituximab (1 mg/mL), and trastuzumab (1.2 mg/mL) analyzed separately and mixed with the glutamate used as internal standard (*). Other CE conditions as in Fig. 2.

A blind test has been carried out to confirm the ability of the optimized CE method to identify mAbs from the RTm evaluation only. Five blinded samples (A1–A5) containing in- dividually each of the studied mAbs or just NaCl 0.9% (i.e. without mAb) have been prepared. Three blinded samples have been randomly analyzed (A2, A3, and A4). Three con- secutive analyses of the four mAbs mixtures that included the IS have been also carried out to determine the expected RTms. The mean of the RTms of cetuximab, bevacizumab, trastuzumab, and rituximab were 1.49 ± 0.009, 1.57 ± 0.007,1.63 ± 0.008, and 1.66 ± 0.008, respectively. The electrophoregrams of samples A2 and A3 exhibited the expected peak of excipient (histidine), IS, and the major peak of mAb. CE anal- ysis of sample A4 showed no peak proving there was no mAb in this sample (data not shown). The RTms of samples A2 and A3 were 1.59 ± 0.002 and 1.63 ± 0.003, respectively. An agglomerative hierarchical clustering (AHC) has been per- formed on these RTms. AHC is built from dissimilarities between objects we try to gather. The objects are therefore represented as a tree called dendrogram according to their (dis) similarity. Automatic truncation of the dendogram has been selected to give automatically the number of class (or group) to retain. The aim was to gather homogenous clusters of mAbs based on the RTms and then to identify the cluster (i.e. the group of mAbs) to which A2 and A3 belong. As ex- pected, in Fig. 3B, the six mAbs constituted their own group. They have been successfully clustered in four higher groups corresponding to the four main mAbs according to the auto- matic truncation line. As the height of the branch of the den- drogram is proportionnal to the distance between two mAbs, it can be observed that A2 and A3 have been aggregated to be- vacizumab and trastuzumab, respectively.” After this test A1 and A5 have been also identified as rituximab and cetuximab following the same procedure (data not shown). The unblind- ing has revealed that A1, A2, A3, A4, and A5 were rituximab, bevacizumab, trastuzumab, NaCl 0.9%, and cetuximab, re- spectively (Fig. 3C). The use of the RTm through AHC has successfully grouped the observations allowing identification of the mAbs.

Finally, such a method may be transferred to the phar- macy hospital department where the first step is to perform a first separation of the four mAbs to evaluate the RTms and the second one to analyze the mAb to be identified based on its RTm. This could be applied routinely in a hospital pharmacy department with the following procedure: First, three runs of one mAb mixture will be performed to determine RTms for each mabs. Then, for a control/identification of one single mAb, one CE analysis will be performed in order to estimate accurately its RTm, followed by the computational AHC test.

4 Concluding remarks

In this work, optimal CE conditions for specifically identify- ing four monoclonal antibodies used in cancer therapy have been determined. The use of counter EOF under pH 3.0 com- bined to a cationic coating has been shown to allow a partial resolution of the four investigated mAbs. To achieve the required resolution, additives such as perchloric acid and polysorbate 80 (PS 80) have been added to the BGE. We showed the importance of the concentration of these additives on the electrophoretic behavior of the mAbs and their reso- lution and also on the µeo. Precise concentrations of these additives were found to be very useful not only to fully resolve the four mAbs but also to increase the reproducibility of the method. We demonstrated the interest of using an internal standard to improve mAbs identification and discrimination based on their relative migration times. The retained condi- tions with the use of an internal standard showed good inter- mediate precision with a RSD of the relative migration times lower than 1.3% and allowed an unambiguous identification of the mAbs. The power of the method was demonstrated by a blind test that identified and discriminated the four mAbs.

The authors declare no conflict of interest.

5 References

[1] Berger, M., Shankar, V., Vafai, A., Am. J. Med. Sci. 2002, 324, 14–30.
[2] Reichert, J. M., Rosensweig, C. J., Faden, L. B., Dewitz, M. C., Nat. Biotechnol. 2005, 23, 1073–1078.
[3] Glassman, P. M., Balthasar, J. P., Cancer Biol. Med. 2014,11, 20–33.
[4] Reichert, J. M., mAbs 2014, 6, 799–802.
[5] Drucker, A., Skedgel, C., Virik, K., Rayson, D., Sellon, M., Younis, T., Curr. Oncol. 2008, 15, 136–142.
[6] Lopez, M. A. A., Torcal, P. A., Bandre´ s, M. A. A., Mendoza,M. A., Gracia, M. G., Prades, E. G., Eur. J. Hosp. Pharm.2012, 19, 159–159.
[7] Simoens, S., De Rijdt, T., Declerck, P., J. Pharm. Health Serv. Res. 2010, 1, 123–130.
[8] Beck, A., Wagner-Rousset, E., Ayoub, D., Van Dorsse- laer, A., Sanglier-Cianfe´ rani, S., Anal. Chem. 2013, 85, 715–736.
[9] Fekete, S., Gassner, A-L., Rudaz, S., Schappler, J., Guil- larme, D., Trends Anal. Chem. 2013, 42, 74–83.
[10] Beer, P. M., Wong, S. J., Schartman, J. P., Kulas, K. E., Hartman, C. L., Giganti, M., Retina Phila. Pa. 2010, 30, 81–84.
[11] Sreedhara, A., Glover, Z. K., Piros, N., Xiao, N., Patel, A., Kabakoff, B., J. Pharm. Sci. 2012, 101, 21–30.
[12] Paul, M., Vieillard, V., Jaccoulet, E., Astier, A., Int. J. Pharm. 2012, 436, 282–290.
[13] Visser, J., Feuerstein, I., Stangler, T., Schmiederer, T., Fritsch, C., Schiestl, M., BioDrugs 2013, 27, 495–507.
[14] Bazin, C., Vieillard, V., Astier, A., Paul, M., Ann. Pharm. Fr. 2014, 72, 33–40.
[15] He, Y., Isele, C., Hou, W., Ruesch, M., J. Sep. Sci. 2011,34, 548–555.
[16] Taga, A., Kita, S., Nishiura, K., Hayashi, T., Kinoshita, M., Sato, A., J. Sep. Sci. 2008, 31, 853–858.
[17] He, Y., Lacher, N. A., Hou, W., Wang, Q., Isele, C., Starkey, J., Anal. Chem. 2010, 82, 3222–3230.
[18] Han, H., Livingston, E., Chen, X., Anal. Chem. 2011, 83, 8184–8191.
[19] Shi, Y., Li, Z., Qiao Y., Lin, J., J. Chromatogr. B 2012, 906, 63–68.
[20] Espinosa-de la Garza, C. E., Perdomo-Abu´ ndez, F. C., Padilla-Caldero´ n, J., Uribe-Wiechers, J. M., Pe´ rez,N. O., Flores-Ortiz, L.F., Electrophoresis 2013, 34, 1133–1140.
[21] Gassner, A-L., Rudaz, S., Schappler, J., Electrophoresis2013, 34, 2718–2724.
[22] Glover, Z. W., Gennaro, L., Yadav, S., Demeule, B., Wong, P. Y., Sreedhara, A., J. Pharm. Sci. 2013, 102, 794–812.
[23] Stutz, H., Electrophoresis 2009, 30, 2032–2061.
[24] Lucy, C., Macdonald, A., Gulcev, M., J. Chromatogr. A2008, 1184, 81–105.
[25] Sun, P., Landman, A., Barker, G. E., Hartwick, R,. A., J. Chromatogr. A 1994, 685, 303–312.
[26] Bendahl, L., Hansen, S. H., Gammelgaard, B., Elec- trophoresis 2001, 22, 2565–2573.
[27] Bohoyo-Gil, D., Le Potier, I., Rivie` re, C., Klafki, H., Wiltfang, J., Taverna, M., J. Sep. Sci. 2010, 33, 1090–1098.
[28] Morris, A. L., Harrison, C. R., Electrophoresis 2013, 34, 2585–2592.