Analytica Chimica Acta xxx (2016) 1e13
Contents lists available at ScienceDirect
Analytica Chimica Acta journal homepage: www.elsevier.com/locate/aca
A novel method for serum lipoprotein profiling using high performance capillary isotachophoresis Estefanía Moreno-Gordaliza a, *, 1, Sven J. van der Lee b, Ays¸e Demirkan b, Cornelia M. van Duijn b, Johan Kuiper c, Petrus W. Lindenburg a, Thomas Hankemeier a a
Division of Analytical Biosciences, Leiden Academic Centre for Drug Research, Faculty of Science, Universiteit Leiden, Einsteinweg 55, 2300 RA Leiden, The Netherlands Department of Epidemiology, Erasmus Medical Centre, PO Box 2040, 3000 CA Rotterdam, The Netherlands c Division of Biopharmaceutics, Leiden Academic Centre for Drug Research, Faculty of Science, Universiteit Leiden, Einsteinweg 55, 2300 RA Leiden, The Netherlands b
Blood serum/plasma
+ 24 spacers + lipophilic stain cITP profiling
18 Lipoprotein peaks
HDL
Val
New isotachophoresis method for lipoprotein separation with higher particle coverage. 18 lipoprotein peaks could be separated using a fine-tuned mixture of 24 spacers. Highly reproducible separations achieved with a new stable doublycoated capillary. Capillary isotachophoresis results were compared with those of a NMRbased method. Subparticle lipoprotein changes found in a high cholesterol atherosclerosis model.
g r a p h i c a l a b s t r a c t Glu ACES MES Glucuronic acid/ octanesulfonic acid GlyGly/ MOPSO Cys Tricine AlaGly TAPSO AlaAla GlyPhe/ GlyHis TAPS 3-methyl-His Ser Gln Met His, Phe 1-methyl-His Gly
h i g h l i g h t s
LDL VLDL/chylom.
a r t i c l e i n f o
a b s t r a c t
Article history: Received 16 June 2016 Received in revised form 29 September 2016 Accepted 30 September 2016 Available online xxx
A new capillary isotachophoresis (cITP) method for lipoprotein profiling with superior lipoprotein coverage compared to previous methods has been developed, resolving twice as many lipoprotein species (18 peaks/fractions) in serum or plasma in less than 9.5 min. For this, a novel mixture of 24 spacers, including amino acids, dipeptides and sulfonic acids, was developed and fine-tuned, using predictive software (PeakMaster) and testing of spiked serum samples. Lipoprotein peaks were identified by serum-spiking with reference lipoproteins. Compatibility with common lipophilic stains for selective lipoprotein detection with either UV/Vis or laser-induced fluorescence was demonstrated. A special new capillary with a neutral coating (combining water-compatible OV1701-OH deactivation and methylation) was used for the first time for electrodriven separations, allowing very stable separations in a pH 8.8e9.4 gradient system, being functional for more than 100 injections. Excellent reproducibility was achieved, with coefficients of variation lower than 2.6% for absolute migration times. Comparison was performed with human plasma samples analyzed by NMR, leading to similar results with cITP after multivariate statistics, regarding group-clustering and lipoprotein species correlation. The new cITP method was applied to the analysis of serum samples from a LDL receptor knock-out mice model fed either a normal diet or a western-type diet. Differences in the lipoprotein levels and in the sublipoprotein types were
Keywords: Capillary isotachophoresis Lipoproteins Spacers Capillary coating Atherosclerosis Hypercholesterolemia
* Corresponding author. E-mail address: [email protected] (E. Moreno-Gordaliza). 1 Present address: Department of Analytical Chemistry, Faculty of Chemistry, Universidad Complutense de Madrid, Avda. Complutense s/n, 28040 Madrid, Spain. http://dx.doi.org/10.1016/j.aca.2016.09.038 0003-2670/© 2016 Published by Elsevier B.V.
Please cite this article in press as: E. Moreno-Gordaliza, et al., A novel method for serum lipoprotein profiling using high performance capillary isotachophoresis, Analytica Chimica Acta (2016), http://dx.doi.org/10.1016/j.aca.2016.09.038
2
E. Moreno-Gordaliza et al. / Analytica Chimica Acta xxx (2016) 1e13
detected, showing a shift to more atherogenic particles due to the high cholesterol diet. In summary, this novel method will allow more detailed and informative profiling of lipoprotein particle subtypes for cardiovascular disease research. © 2016 Published by Elsevier B.V.
1. Introduction Lipoproteins are heterogeneous blood particles delivering or uptaking specific lipids within the body, thus playing an essential role in lipid metabolism [1]. In addition, they have important roles in signalling and regulation. These particles present a superficial hydrophilic monolayer of phospholipids (PL), including apolipoproteins, and unesterified free cholesterol (FC); and a hydrophobic core, comprising cholesterol esters (ChoE) and triglycerides (TG). Based on their density, size and/or protein composition, lipoproteins are divided into five main classes: chylomicrons, very low density lipoproteins (VLDL), intermediate density lipoproteins (IDL), low density lipoproteins (LDL) -mainly apoB-containing lipoproteins-; and high density lipoproteins (HDL) -mainly apoA1containing lipoproteins. Within these classes, different subtypes with diverse lipid and protein composition, relating to different biological functions, are found. LDL has the physiological role of providing cells with cholesterol and is therefore its major carrier. High levels of LDL and, to a lesser extent, VLDL particles are associated with the development of atherosclerosis, the main cause of cardiovascular disease (CVD) [2]. Especially, small, dense LDL and modified electronegative particles such as oxidized LDL, release and accumulate cholesterol in the arterial intima, triggering plaque formation [3,4]. HDL particles, in contrast, protect against the development of atherosclerosis [5] through reverse cholesterol transport (mainly pre-b HDL and a-3-HDL), removing cholesterol from the arterial intima to the liver [6] and through their antioxidant properties [7]. Among the different approaches for atherosclerosis risk and progression prediction, routine determination of total cholesterol and TG in serum or cholesterol associated to the whole HDL or LDL fractions have been shown not to fully serve, whereas the number of atherogenic lipoprotein particles instead is better correlated [8]. Alternatively, lipoprotein subfractionation and subsequent biochemical profiling would be very useful to improve prediction of atherosclerosis risk and progression and the outcome of therapeutic approaches, since each lipoprotein subtype has a different composition and role in atherosclerosis [8]. Indeed, lipoprotein profiles have been used in computational modeling for prediction of CVD [9]. This could also help to understand the role of lipoproteins in more detail. Diverse separation methods have been described for lipoprotein fractionation, giving rise to different classifications of lipoprotein subfractions depending on the analytical separation mechanism [10]. These techniques include: (density gradient) ultracentrifugation [11], size exclusion chromatography (SEC) [12] gel electrophoresis [13], capillary zone electrophoresis [14], isotachophoresis [15], immunoaffinity chromatography [16], field-flow fractionation [17], precipitation [18], nuclear magnetic resonance (NMR) [19] or ion mobility mass spectrometry [20]. However, difficulties are encountered in most of the methods, including: high sample amount requirements, alterations in the original lipoprotein composition during the analytical procedure, limited resolution of particle subtypes, lack of robustness, lengthy procedures, or the need to use mathematical calculations for data transformation, like in NMR. In addition, several methods do not allow collecting
subfractions of interest. Table S-1 summarizes the most important characteristics (separation resolution, automation, fraction collection, possible degradation, required sample volume) of reported lipoprotein separation techniques. Ultracentrifugation is considered as the “gold standard” for lipoprotein separation, allowing fractionation of lipoprotein particle types and also subclasses according to density, which is the classical classification criteria for lipoproteins. However, its technical demands make this method non suitable for routine clinical use, although being especially useful for producing lipoprotein standards [8]. Capillary isotachophoresis (cITP) is a very powerful alternative for lipoprotein profiling due to its ability to separate several lipoprotein subparticle types in a few-minute single run. Moreover, very small sample volumes can be analyzed with cITP, also offering easy automation and on-line coupling possibilities, in contrast to some of the alternative techniques. Separation of ions takes place in a discontinuous system, where the sample is injected between a leading electrolyte (LE), with the highest mobility in the system, and a terminating electrolyte (TE), with lower mobility than those ions of interest in the sample. Such a system ensures isotachophoresis, where concentrating and self-sharpening effects take place [21]. Once equilibrium is reached, analytes are separated into adjacent zones arranged according to their electrophoretic mobilities. The analyte concentrations can be described according to the Kohlrausch regulating function [21]. So-called spacer molecules, possessing intermediate mobilities in comparison to those of the analyte zones, can be added to the sample in order to separate analytes into discrete peaks, thus greatly improving the separation resolution. With respect to cITP of lipoproteins, a method developed by Schmitz et al. [22], and further improved [15,23,24] was applied routinely in dozens of studies related to cardiovascular disease [3,25,26]. These cITP methods employ anionic mode separation in the presence of 9 spacer molecules [15] achieving 8 discrete lipoprotein peaks. For selective detection of lipoproteins, lipophilic staining reagents are available: NBD-C6-Ceramide [15,27] (for laser induced fluorescence (LIF) detection), and Sudan Red 7B [24], or Sudan Black B [27] (for UV/vis detection). Lipoprotein ITP separations are challenging as they should be performed in anionic mode at basic pH and coated capillaries are required for prevention of lipoprotein adsorption and for EOF suppression [28]. However, long-term stability of capillary coatings at basic pH can be critical [24,29]. However, in order to consolidate cITP as a reference technique for lipoprotein profiling, a significant improvement in separation resolution and therefore in sublipoprotein fractionation is still needed, which would have tremendous benefits for clinical analysis and diagnostics. In conclusion, there is still a pressing demand for improved separation methods to enable more detailed fractionation of sublipoprotein populations in serum with the possibility to subsequently apply another biochemical analysis. Herein, a new cITP method is reported for higher resolution and reproducibility of lipoprotein profiling in serum/plasma than reported before [15]. A novel 24 spacers mixture was fine-tuned using software-based (Peakmaster) [30] prediction of electrophoretic behavior and tested on human serum. The separation of 18 lipoprotein peaks was
Please cite this article in press as: E. Moreno-Gordaliza, et al., A novel method for serum lipoprotein profiling using high performance capillary isotachophoresis, Analytica Chimica Acta (2016), http://dx.doi.org/10.1016/j.aca.2016.09.038
E. Moreno-Gordaliza et al. / Analytica Chimica Acta xxx (2016) 1e13
achieved, doubling the amount of peaks resolved by previous ITP methods, and using a new highly stable coated capillary. The method was successfully applied on human plasma, and the results compared with those obtained by NMR analysis of the lipoproteins. Finally, to demonstrate the excellent potential of our method, lipoprotein profiling was applied to serum samples of an atherosclerosis LDL receptor knock-out (LDLr/) mice model. 2. Materials and methods 2.1. Chemicals All reagents used were purchased from Sigma-Aldrich (Zwijndrecht, The Netherlands) unless otherwise stated. A pH-Meter (HI4521, Hanna Instruments, Nieuwegein, The Netherlands) was used for pH adjustments. Milli-Q water (Merck Millipore, Amsterdam, The Netherlands) was used for preparation of all aqueous solutions. Tricine was purchased from Janssen Chimica. GlyGly was purchased from Acros Organics. L-Aminoacids (L-cysteine (Cys), Lhistidine monohydrochloride (His), L-glutamine (Gln), L-glycine (Gly), L-glutamic acid (Glu), L-methionine (Met), L-phenylalanine (Phe), L-serine (Ser), L-valine (Val)) used were obtained from Fluka (Zwijndrecht, The Netherlands). 2.2. Animals Female, 10e12 weeks old, LDLr/ mice (The Jackson Laboratory, Bar Harbor, ME, USA) were used and kept under standard laboratory conditions. The animals were fed either a normal chow diet -RM1- (Special Diets Services, Witham, Essex, UK) (week 0), or fed for 3, 6 or 9 weeks a Western-type diet (WD) containing 0.25% cholesterol and 15% cocoa butter (Special Diet Services, Witham, Essex, UK). 6 animals per time point (weeks 0, 3, 6, 9) were bred and a total of 24 animals were used. Water and food were provided ad libitum. Blood was obtained from the mice by tail vein bleeding. All animal work was approved by Leiden University regulatory authority and was performed complying with the Dutch government guidelines. 2.3. Serum/plasma samples Human blood serum from healthy donors was used for method development and was obtained either frozen from Sigma-Aldrich, or fresh from Sanquin blood supply (Amsterdam, The Netherlands). Fresh serum for lipoprotein isolation was kept at 4 C until use. Ten individuals were selected from the study population of Erasmus Rucphen Family in order to facilitate the comparison of cITP profiling to an established commercial 1H-NMR method. 5 female and 5 male participants within the age range of 60e67, representing either low (1.3 mmol L1) levels of serum HDL cholesterol, were included in the comparison. None of the participants neither had known diseases such as diabetes, nor used lipid lowering medication. The venous blood samples of the ten individuals were collected after at least 8 h of fasting during the ERF cohort recruitment period (2002e2004), followed by simultaneous serum and EDTA-plasma isolation. Samples were stored at 80 C until measurements, yielding storage times ranging from to 7e9 years for the serum 1H-NMR analysis, and 11e13 years for the plasma cITP-based profiling. The samples have been previously included in a larger genome-wide association study which can be referred for further methodological information [31]. The ERF study protocol was approved by the medical ethics board of the Erasmus MC Rotterdam, The Netherlands. Finally, serum was also obtained from blood taken from the mice
3
model fed either the normal diet (week 0) or a Western-type diet (3, 6 and 9 weeks). 2.4. Preparation of fresh reference lipoprotein fractions Lipoprotein isolation was carried out by density gradient centrifugation as previously described [32] (see Supplementary Material for details). Separated lipoproteins (lipoprotein deficient serum (LPDS), HDL, LDL, VLDL/chylomicrons) were isolated and kept at 4 C until use. Isolated LDL was dialyzed against phosphate buffered saline (PBS) solution containing 10 mM EDTA (pH 7.4) for 24 h at 4 C, and oxidized by incubation at 37 C for 20 h with 10 mM CuSO4, as previously described [33]. Alternative lipoproteins (VLDL and chylomicrons) were purchased from Athens Research and Technology (GA, USA). 2.5.
1
H-NMR
NMR experiments in serum samples were acquired on a 600 MHz Bruker Avance II spectrometer (Bruker BioSpin, Karlsruhe, Germany), as previously described [31]. The NMR spectra were then deconvoluted using a commercial algorithm patented by the Bruker Corporation NMR services. Lipoproteins and their subclasses were defined according to their density and particle size (https://www. bruker.com/fileadmin/user_upload/8-PDF-Docs/MagneticResonance/NMR/brochures/lipo-analysis_apps.pdf) (see Supplementary Material for further details). 2.6. Capillary isotachophoresis 2.6.1. cITP-UV/vis 20 mL human serum/plasma was stained with 10 mL 1% Sudan Black B (SBB) (or alternatively Sudan Red 7B (SR7B)) in ethyleneglycol for 1 h at 4 C in the dark. 25 mL pre-stained serum/ plasma were diluted with 88.8 mL of leading electrolyte (LE) and mixed with 18.8 mL of spacers solution. The new spacers solution was prepared by mixing the following volumes of 10 mg L1 aqueous solutions: 100 mL N-(2-acetamido)-2-aminoethane sulfonic acid (ACES), 50 mL glucuronic acid, 50 mL octanesulfonic acid, 80 mL N-[tris(hydroxymethyl)methyl]-3-amino-2-hydroxy propanesulfonic acid (TAPSO), 120 mL N-[tris(hydroxymethyl) methyl]-3-aminopropanesulfonic acid (TAPS), 80 mL Gln, 80 mL Lmethionine (Met), 50 mL Gly, 60 mL Ser and 13.4 mL of: 2-[N-morpholino]ethanesulfonic acid (MES), glycyl-L-glycine (GlyGly), 3morpholino-2-hydroxypropanesulfonic acid (MOPSO), Tricine, Cys, L-alanyl-glycine (AlaGly), L-alanyl-L-alanine (AlaAla), glycyl-Lphenylalanine (GlyPhe), glycyl-L-histidine (GlyHis), His, Phe, Glu, Val, 3-methyl-L-His and 1-methyl-L-His. During optimization, different combinations of spacers were assessed and tests with other alternative spacers compounds were carried out, using 3-(Nmorpholino)propanesulfonic acid (MOPS), bicine, glycyl-leucine (GlyLeu), glycyl-proline (Gly-Pro), L-asparagine (Asp), L-tryptophan (Trp), glycyl-sarcosine, g-glutamyl-alanine (g-GluAla). The LE contained 10 mM HCl and 0.35% hydroxypropylmethylcellulose (HPMC), adjusted with 2-amino-2-methyl-1,3-propanediol (ammediol) to pH 8.8. The terminating electrolyte (TE) contained 20 mM L-alanine adjusted to pH 9.4 with ammediol solution. Alternative LE/TE compositions explored during optimization consisted on the following aqueous solutions: 10 mM HCl, 0.35% HPMC for the LE and 20 mM L-Ala for the TE, both adjusted with ammediol to either pH 9.0 or pH 9.5. An Agilent G1600AX CE system equipped with a Diode Array Detector (DAD) and a deuterium lamp was used for cITP analysis. Coated fused silica capillaries (BGB Analytik AG, Switzerland; OV1701-OH coating þ 0.4 mm C1 coating (100% methylated)) with
Please cite this article in press as: E. Moreno-Gordaliza, et al., A novel method for serum lipoprotein profiling using high performance capillary isotachophoresis, Analytica Chimica Acta (2016), http://dx.doi.org/10.1016/j.aca.2016.09.038
4
E. Moreno-Gordaliza et al. / Analytica Chimica Acta xxx (2016) 1e13
internal diameter: 180 mm, outer diameter: 365 mm; total length: 35 cm and effective length: 27.5 cm; were used for cITP separation. Pre-conditioning of the capillary was 10 min LE flush. Hydrodynamic injection was performed with 50 mbar for 15 s; lipoprotein separation was performed at 10 kV. The capillary, samples and buffers were used at 20 C. Detection took place at 254 nm, 280 nm (for protein detection), 570 nm (for SBB-stained lipoproteins) or 520 nm (for SR7B-stained lipoproteins). The capillaries were postconditioned after each run by flushing at 940 bar: 1.5 min LE, 0.2 min Milli-Q H2O, 0.4 min 1.5% Triton X-100 reduced, 0.4 min Milli-Q H2O, 2 min LE. After operation, capillaries were flushed with Milli-Q H2O for 1 min, air-blown dried and stored at RT. 2.6.2. cITP-LIF For cITP analysis with LIF detection, 5 mL serum was diluted with 17.5 mL LE and incubated at RT for 1 min with 11.22 mL NBD-C6Ceramide (Molecular Probes, Life Technologies) 0.1 mg mL1 solution prepared in anhydrous ethyleneglycol:methanol 9:1. The stained serum was mixed with 63.2 mL LE and 38.2 mL of the optimized spacers solution as described above. A ProteomeLab PA800 CE instrument (Beckman Coulter, Fullerton, CA, USA) equipped with a LIF detector was employed. Sample injection was performed at 0.8 psi for 13.6 s. The same capillaries as those used for UV/vis detection were employed, with a total length of 36 cm and length to the detector of 22.8 cm in this case. For LIF detection, excitation wavelength was set at 488 nm and emission detection at 520 nm. Preconditioning, postconditioning and separations were done at the same conditions as cITP-UV/vis at 10.3 kV.
surface phospholipids and proteins. This heterogeneity leads to different size/charge ratio and electrophoretic mobility, which is reflected in the separation achieved. Using Peakmaster, the electrophoretic behavior of the 9 spacers was predicted and this information was used to find more potential anionic spacers for an improved separation. Table S-1 shows the meff values calculated at pH 8.8, 9.0, 9.2, 9.4 and 9.5 in the absence of EOF for 61 selected compounds present in PeakMaster database, covering a wide range of mobilities, including those of the 9 spacers. In a non-linear gradient pH system, meff prediction is more complex as it should be calculated according to the pH in the position of each spacer in the gradient. As an approximation for preselection, computed meff values in an intermediate pH (with respect to those of the LE and TE) were considered, followed by experimental testing. In view of the computed data and previous literature information on spacers [24,29], 27 new spacers were selected: Asp, GABA, Glu, MES, g-GluAla, GlyGly, MOPS, MOPSO, Cys, Bicine, Tricine, AlaGly, AlaAla, GlyPhe, GlyLeu, GlyHis, GlyPro, GlySar, Asn, 3-methyl-His, Thr, Tyr, His, Phe, 1-methyl-His, Trp and Val. The selected compounds were individually tested in combination with the starting mixture of 9 spacers for cITP separation of serum in a gradient pH 8.8/9.4 system. Fig. S-2 shows typical examples of the outcome of the individual tests. Fig. 1 summarizes the experimental positions of the 27 selected spacers (red labels) found during cITP separation of human serum. General good agreement was found between experimental positions of spacers and computed meff values at intermediate pH 9.2. As can be seen, a number of compounds with intermediate mobilities compared to those of the starting 9 spacers were found, being potential candidates for separation improvement.
2.7. Electrophoretic mobility calculations 3.2. Optimization of cITP lipoprotein profiling Theoretical effective electrophoretic mobilities were computed using Peak Master v.5.3 [30] for compounds included in the software database. Basing on absolute mobilities (mo) and pKa of anions [34,35], effective mobilities (meff) were computed to predict electrophoretic behavior at a certain pH [36,37]. 2.8. Statistics Data analysis and statistics were performed using Metaboanalyst v.3.0 [38] and IBM SPSS Statistics (v 22). Prior to analysis, generalized logarithm transformation and data range scaling (mean-centered, divided by each variable's range) were performed. 3. Results and discussion 3.1. Influence of spacers on cITP separation The spacer molecules present during cITP separations have a major impact on the resolution that can be achieved and careful selection is mandatory. In a first series of experiments, the position of the 9 spacers (ACES, glucuronic acid, octanesulfonic acid, TAPSO, €ttcher et al. [15] used in TAPS, Ser, Gln, Met and Gly) proposed by Bo the current standard methods for lipoprotein cITP profiling was determined. This was done by studying the effect on the original separation of the removal one of the 9 spacers each time (Fig. S-1). Fig. 1 summarizes the location found for the 9 spacers during serum cITP separation in a gradient pH system at 8.8e9.4 (blue labels). 8 separated lipoprotein peaks were observed (Fig. 1b and S-1a), which are consistent with the literature [15]. According to previous works, these correspond to HDL (peaks 1,2,3), chylomicrons/remnants/VLDL/IDL (peaks 4 and 5) and LDL (peak 6, 7, 8). Each lipoprotein class and subtype presents different size and protein and lipid composition, influencing the charge of the particle, especially
In view of the results from section 3.1, a novel combination of 24 spacers for enhanced cITP separation of lipoproteins was designed. Fig. 2 displays the improved serum isotachopherogram when the new spacers mixture was used in a pH 8.8/9.4 system. A much higher resolution than that observed in Fig. 1 can be seen both for the total protein isotachopherogram (Fig. 2a) and the lipoprotein fractionation, showing a total of 18 separated peaks (Fig. 2b). The positions of the 24 spacers are displayed in Fig. 2a. The new lipoprotein separation is superior to previous methods described [3,15,24,29], with i.e. twice as many well-resolved peaks as reported until now. This is key to increase the information extracted from a profiling analysis and to identify differences in lipoprotein subparticles. For the identification of lipoprotein peaks, freshly isolated human lipoproteins (HDL, LDL, VLDL/chylomicrons; see Experimental section) were spiked to reference human serum. Fig. S-3 shows compared isotachopherograms where increased peak areas in spiked serum with respect to non-spiked serum indicate the presence of the added lipoprotein species (HDL, VLDL and LDL in Figs. S-3a, S-3b and S-3c, respectively). Additional spiking experiments were performed with fully oxidized LDL and commercially obtained VLDL and chylomicrons. These allowed locating oxLDL peaks and discerning between chylomicrons and VLDL peaks (present in the same band after density gradient centrifugation, considering its limited resolution). The new peaks found are assigned in Fig. S-3 and all the lipoprotein species summarized in Fig. 2b. HDL is mainly distributed within the first 8 peaks with some contribution to peaks 10 and 11, being its presence in peak 9 possible. Chylomicrons are mainly present in peak 11 with some contribution to peaks 9 and 10. VLDL was mainly found in peaks 11e12 with a slight amount in peak 10. Finally, LDL is present in peaks 13e18, being fully oxidized LDL mainly located in peak 12. No
Please cite this article in press as: E. Moreno-Gordaliza, et al., A novel method for serum lipoprotein profiling using high performance capillary isotachophoresis, Analytica Chimica Acta (2016), http://dx.doi.org/10.1016/j.aca.2016.09.038
E. Moreno-Gordaliza et al. / Analytica Chimica Acta xxx (2016) 1e13
5
€ ttcher et al. [15], monitoring A) absorbance at 280 nm for proteins and B) absorbance at 570 nm Fig. 1. cITP separation of SBB-stained human serum using conditions reported by Bo for specific lipoproteins detection. The position of the spacers used is indicated with blue dashed lines. The experimental position of 27 alternative spacer compounds tested is indicated with red dashed lines. Hydrodynamic injection was performed at 50 mbar for 15s (9 mg total protein). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
significant oxidation effects are expected during sample preparation. However, future experiments might consider the location of other oxidized or glycated particles which might be endogenously present in the serum/plasma samples. The main lipoprotein distribution is in agreement with results previously described by € ttcher [15] and Zhang [39], but a larger number of well-resolved Bo species was found in this study. A much better resolution of HDL and LDL species than previous methodologies [3,14,15,22e24,26,27,29,39] was attained with this method, using a low-microliter range serum volume. The influence of the pH on the separation was also assessed.
Constant pH lipoprotein cITP separation at either pH 9.0 (Fig. S-4) or 9.5 (not shown) resulted in a lower resolution of lipoproteins: more co-migration of LDL and VLDL particles was observed in comparison to the gradient pH separation. Therefore the combination of the gradient pH system (8.8/9.4) (used in the method by €ttcher et al. [15]) and the 24 spacers was selected as optimal. Bo Fig. 3 illustrates the influence of both sample volume injected and spacers concentration on the lipoprotein separation efficiency. The injected sample volume has a great influence in the separation outcome, since the protein components in the sample also act as internal spacers. Fig. 3a shows how, when the injected sample
Please cite this article in press as: E. Moreno-Gordaliza, et al., A novel method for serum lipoprotein profiling using high performance capillary isotachophoresis, Analytica Chimica Acta (2016), http://dx.doi.org/10.1016/j.aca.2016.09.038
6
E. Moreno-Gordaliza et al. / Analytica Chimica Acta xxx (2016) 1e13
Fig. 2. cITP separation of SBB-stained human serum using a novel mixture of 24 spacers at gradient pH 8.8/9.4. A) Signal at 280 nm for serum proteins along with the position of the spacers. B) 18 separated lipoproteins detected at 570 nm. Peak assignments from spiking experiments with freshly isolated lipoproteins are included in B). Hydrodynamic injection was performed at 50 mbar for 15s (9 mg total protein).
amount was increased from 9 to 30 mg total protein, resulting the amount of spacers 3.3-fold increased as well, the zone widths for the protein components led to higher lipoprotein resolution. A similar result was achieved when only the spacer mixture concentration in the sample was increased 3-fold (Fig. 3b). These options to improve peak resolution are especially interesting when fraction collection is desired. In this case, baseline resolution was achieved for peaks 1e2 and 6e18, which also makes this more advantageous for lipoprotein quantification with more reliable peak area integration. Regarding peak integration, it is advisable to perform a perpendicular drop integration method on the drawn baseline from the peak valleys (if unresolved). This can be best manually performed for such unresolved peaks or at least manually corrected after automatic integration. However, in some cases, if the split in peaks 3e5 is unclear, for minimizing errors in area calculation, it might be better to integrate peaks 3e5 altogether.
3.3. Capillary coating and separations reproducibility The performance of specially coated capillaries, i.e. OV1701-thin film deactivation stable against water combined with a 0.4 mm thick polydimethylsiloxane coating, for cITP lipoprotein separations was investigated. This special coating was designed for producing more stable capillaries that can better tolerate and withstand aqueous high pH conditions. Previous studies reported stability issues with coatings used for lipoprotein separations by cITP due to hydrolysis of the silanized groups at high pH. For instance, capillaries coated with polyacrylamide were stable for only around 50 runs [24] and polydimethylsiloxane-coated capillaries had to be recoated after every run [29]. In this study, the doubly-coated capillaries were stable for at least 100 runs under high pH conditions (pH 8.8e9.4). Moreover, with these capillaries, low coefficients of variation (CV) were obtained for lipoproteins migration times during repeated €ttcher et al. [15] (1.3e2.0%) or injections both for the method by Bo using our new method with the mixture of 24 spacers (2.4e2.6%),
Please cite this article in press as: E. Moreno-Gordaliza, et al., A novel method for serum lipoprotein profiling using high performance capillary isotachophoresis, Analytica Chimica Acta (2016), http://dx.doi.org/10.1016/j.aca.2016.09.038
E. Moreno-Gordaliza et al. / Analytica Chimica Acta xxx (2016) 1e13
7
Fig. 3. cITP lipoprotein profiling of SBB-stained human serum under optimized conditions. A) Injection of 30 mg total protein (50 mbar, 50s). B) Injection of 9 mg total protein (50 mbar, 15s) and 3-fold increased spacer concentration.
as summarized in Table S-2. The variation of the peak areas within a day was satisfactory, with CVs lower than 5%. The use of HPMC in the LE also allows creating an extra dynamic coating, ensuring suppression of the electroosmotic flow and analyte-wall interactions as previously reported, contributing as well to separation repeatability. Fig. S-5 also shows the variation of migration times for representative lipoprotein peaks (1, 11 and 17) during repeated runs. CVs for relative migration times (with respect to that of peak 1) were lower than 1%. Between-capillary migration time reproducibility was also good (CV within the 1.0e2.1% range (n ¼ 3)). Finally, samples were shown to be stable for at least 3 days after preparation when stored at 4 C in-between analyses, and at least for a month at 20 C. Excellent long-term reproducibility is therefore attained when separating serum in doubly-coated capillaries using the optimized 24 spacers mixture in the pH 8.8/9.4 system, compared to previously reported results [40]. 3.4. Staining methods for lipoprotein profiling SR7B and SBB were compared for specific lipoprotein staining with UV/vis detection, as can be seen in Fig. S-6. Although both stains could be used, SBB-stained lipoproteins presented higher
signal intensity than those stained with SR7B. In the reminder of this work, SBB staining was used. Fig. S-7 demonstrates that the improved profiling method can be transferred to cITP-LIF analysis using NBD-C6-Ceramide-stained serum. Adaptation of the method is straightforward, by replacing the staining agent and an appropriate sample dilution due to the higher sensitivity of LIF. The same lipoprotein profile as observed with UV/vis detection was obtained. This is of great interest as it would allow the use of a fluorescent internal standard like 5-carboxyfluorescein for more accurate quantification purposes [27]. Furthermore, it confirms the reproducibility of the method in a quite different electrophoretic system. 3.5. Comparison of cITP and NMR profiling To validate our cITP method, analysis of human plasma samples from 10 different individuals selected as described in the experimental section, was carried out and the results were compared with quantitative NMR measurements previously performed [31]. The NMR method is based on a deconvolution algorithm of the overlay of the 1H-NMR spectra of the different lipoprotein subparticle types, analogously to previously reported methods [41], where lipids are the main contributors to the signal. Lipid (FC, Total Cholesterol -C-, PL, TG) and apolipoprotein (apo A1, apoA2, apoB)
Please cite this article in press as: E. Moreno-Gordaliza, et al., A novel method for serum lipoprotein profiling using high performance capillary isotachophoresis, Analytica Chimica Acta (2016), http://dx.doi.org/10.1016/j.aca.2016.09.038
8
E. Moreno-Gordaliza et al. / Analytica Chimica Acta xxx (2016) 1e13
levels were obtained in each lipoprotein subparticle type. Fig. S-8 shows clustering of two main groups of 5 samples each (1e5: group 1; and 6e10: group 2) after Partial Least Square-Discriminant Analysis (PLS-DA) of the NMR data. Successful cross validation of the model is shown in Fig. S-8b and Table S-4. As indicated by the Variable Importance in Projection (VIP) scores displayed in Fig. S8c, representing the main components leading to group separation, the samples differ in HDL subparticle composition (mostly a decrease for very large and large HDL particles and to a lesser extent, medium size HDL in Group 2 and in LDL particles (basically with an associated increase in TG in Group 2. Fig. 4 shows successful comparison of cITP and NMR data. Comparable to the NMR results, higher cITP areas were visible for peaks in the HDL region for Group 1, while certain LDL peaks presented lower areas compared with Group 2. Fig. 4a shows the overlap of the lipoprotein isotachopherograms for two representative samples from each of the two groups. Moreover, Fig. 4b
shows clustering of the samples into the same two groups after PLS-DA analysis of cITP peak areas (Fig. 4b) as was observed for the same analysis of the NMR data (Fig. S-8). Cross validation of the PLS-DA model is displayed in Fig. 4c and Table S-5, indicating a correct group separation. Moreover, analysis of the VIP scores reveals that the most important components leading to group separation belong to the first 5 ITP lipoprotein HDL peaks, and peaks 12e15 (attributed mainly to LDL), as can be seen in Fig. 4d. Additional t-tests confirmed that the most significant differences found between the two groups with this cITP method concern peaks 1, 13, 14 and 15 in the cITP separation with a p-value
Contents lists available at ScienceDirect
Analytica Chimica Acta journal homepage: www.elsevier.com/locate/aca
A novel method for serum lipoprotein profiling using high performance capillary isotachophoresis Estefanía Moreno-Gordaliza a, *, 1, Sven J. van der Lee b, Ays¸e Demirkan b, Cornelia M. van Duijn b, Johan Kuiper c, Petrus W. Lindenburg a, Thomas Hankemeier a a
Division of Analytical Biosciences, Leiden Academic Centre for Drug Research, Faculty of Science, Universiteit Leiden, Einsteinweg 55, 2300 RA Leiden, The Netherlands Department of Epidemiology, Erasmus Medical Centre, PO Box 2040, 3000 CA Rotterdam, The Netherlands c Division of Biopharmaceutics, Leiden Academic Centre for Drug Research, Faculty of Science, Universiteit Leiden, Einsteinweg 55, 2300 RA Leiden, The Netherlands b
Blood serum/plasma
+ 24 spacers + lipophilic stain cITP profiling
18 Lipoprotein peaks
HDL
Val
New isotachophoresis method for lipoprotein separation with higher particle coverage. 18 lipoprotein peaks could be separated using a fine-tuned mixture of 24 spacers. Highly reproducible separations achieved with a new stable doublycoated capillary. Capillary isotachophoresis results were compared with those of a NMRbased method. Subparticle lipoprotein changes found in a high cholesterol atherosclerosis model.
g r a p h i c a l a b s t r a c t Glu ACES MES Glucuronic acid/ octanesulfonic acid GlyGly/ MOPSO Cys Tricine AlaGly TAPSO AlaAla GlyPhe/ GlyHis TAPS 3-methyl-His Ser Gln Met His, Phe 1-methyl-His Gly
h i g h l i g h t s
LDL VLDL/chylom.
a r t i c l e i n f o
a b s t r a c t
Article history: Received 16 June 2016 Received in revised form 29 September 2016 Accepted 30 September 2016 Available online xxx
A new capillary isotachophoresis (cITP) method for lipoprotein profiling with superior lipoprotein coverage compared to previous methods has been developed, resolving twice as many lipoprotein species (18 peaks/fractions) in serum or plasma in less than 9.5 min. For this, a novel mixture of 24 spacers, including amino acids, dipeptides and sulfonic acids, was developed and fine-tuned, using predictive software (PeakMaster) and testing of spiked serum samples. Lipoprotein peaks were identified by serum-spiking with reference lipoproteins. Compatibility with common lipophilic stains for selective lipoprotein detection with either UV/Vis or laser-induced fluorescence was demonstrated. A special new capillary with a neutral coating (combining water-compatible OV1701-OH deactivation and methylation) was used for the first time for electrodriven separations, allowing very stable separations in a pH 8.8e9.4 gradient system, being functional for more than 100 injections. Excellent reproducibility was achieved, with coefficients of variation lower than 2.6% for absolute migration times. Comparison was performed with human plasma samples analyzed by NMR, leading to similar results with cITP after multivariate statistics, regarding group-clustering and lipoprotein species correlation. The new cITP method was applied to the analysis of serum samples from a LDL receptor knock-out mice model fed either a normal diet or a western-type diet. Differences in the lipoprotein levels and in the sublipoprotein types were
Keywords: Capillary isotachophoresis Lipoproteins Spacers Capillary coating Atherosclerosis Hypercholesterolemia
* Corresponding author. E-mail address: [email protected] (E. Moreno-Gordaliza). 1 Present address: Department of Analytical Chemistry, Faculty of Chemistry, Universidad Complutense de Madrid, Avda. Complutense s/n, 28040 Madrid, Spain. http://dx.doi.org/10.1016/j.aca.2016.09.038 0003-2670/© 2016 Published by Elsevier B.V.
Please cite this article in press as: E. Moreno-Gordaliza, et al., A novel method for serum lipoprotein profiling using high performance capillary isotachophoresis, Analytica Chimica Acta (2016), http://dx.doi.org/10.1016/j.aca.2016.09.038
2
E. Moreno-Gordaliza et al. / Analytica Chimica Acta xxx (2016) 1e13
detected, showing a shift to more atherogenic particles due to the high cholesterol diet. In summary, this novel method will allow more detailed and informative profiling of lipoprotein particle subtypes for cardiovascular disease research. © 2016 Published by Elsevier B.V.
1. Introduction Lipoproteins are heterogeneous blood particles delivering or uptaking specific lipids within the body, thus playing an essential role in lipid metabolism [1]. In addition, they have important roles in signalling and regulation. These particles present a superficial hydrophilic monolayer of phospholipids (PL), including apolipoproteins, and unesterified free cholesterol (FC); and a hydrophobic core, comprising cholesterol esters (ChoE) and triglycerides (TG). Based on their density, size and/or protein composition, lipoproteins are divided into five main classes: chylomicrons, very low density lipoproteins (VLDL), intermediate density lipoproteins (IDL), low density lipoproteins (LDL) -mainly apoB-containing lipoproteins-; and high density lipoproteins (HDL) -mainly apoA1containing lipoproteins. Within these classes, different subtypes with diverse lipid and protein composition, relating to different biological functions, are found. LDL has the physiological role of providing cells with cholesterol and is therefore its major carrier. High levels of LDL and, to a lesser extent, VLDL particles are associated with the development of atherosclerosis, the main cause of cardiovascular disease (CVD) [2]. Especially, small, dense LDL and modified electronegative particles such as oxidized LDL, release and accumulate cholesterol in the arterial intima, triggering plaque formation [3,4]. HDL particles, in contrast, protect against the development of atherosclerosis [5] through reverse cholesterol transport (mainly pre-b HDL and a-3-HDL), removing cholesterol from the arterial intima to the liver [6] and through their antioxidant properties [7]. Among the different approaches for atherosclerosis risk and progression prediction, routine determination of total cholesterol and TG in serum or cholesterol associated to the whole HDL or LDL fractions have been shown not to fully serve, whereas the number of atherogenic lipoprotein particles instead is better correlated [8]. Alternatively, lipoprotein subfractionation and subsequent biochemical profiling would be very useful to improve prediction of atherosclerosis risk and progression and the outcome of therapeutic approaches, since each lipoprotein subtype has a different composition and role in atherosclerosis [8]. Indeed, lipoprotein profiles have been used in computational modeling for prediction of CVD [9]. This could also help to understand the role of lipoproteins in more detail. Diverse separation methods have been described for lipoprotein fractionation, giving rise to different classifications of lipoprotein subfractions depending on the analytical separation mechanism [10]. These techniques include: (density gradient) ultracentrifugation [11], size exclusion chromatography (SEC) [12] gel electrophoresis [13], capillary zone electrophoresis [14], isotachophoresis [15], immunoaffinity chromatography [16], field-flow fractionation [17], precipitation [18], nuclear magnetic resonance (NMR) [19] or ion mobility mass spectrometry [20]. However, difficulties are encountered in most of the methods, including: high sample amount requirements, alterations in the original lipoprotein composition during the analytical procedure, limited resolution of particle subtypes, lack of robustness, lengthy procedures, or the need to use mathematical calculations for data transformation, like in NMR. In addition, several methods do not allow collecting
subfractions of interest. Table S-1 summarizes the most important characteristics (separation resolution, automation, fraction collection, possible degradation, required sample volume) of reported lipoprotein separation techniques. Ultracentrifugation is considered as the “gold standard” for lipoprotein separation, allowing fractionation of lipoprotein particle types and also subclasses according to density, which is the classical classification criteria for lipoproteins. However, its technical demands make this method non suitable for routine clinical use, although being especially useful for producing lipoprotein standards [8]. Capillary isotachophoresis (cITP) is a very powerful alternative for lipoprotein profiling due to its ability to separate several lipoprotein subparticle types in a few-minute single run. Moreover, very small sample volumes can be analyzed with cITP, also offering easy automation and on-line coupling possibilities, in contrast to some of the alternative techniques. Separation of ions takes place in a discontinuous system, where the sample is injected between a leading electrolyte (LE), with the highest mobility in the system, and a terminating electrolyte (TE), with lower mobility than those ions of interest in the sample. Such a system ensures isotachophoresis, where concentrating and self-sharpening effects take place [21]. Once equilibrium is reached, analytes are separated into adjacent zones arranged according to their electrophoretic mobilities. The analyte concentrations can be described according to the Kohlrausch regulating function [21]. So-called spacer molecules, possessing intermediate mobilities in comparison to those of the analyte zones, can be added to the sample in order to separate analytes into discrete peaks, thus greatly improving the separation resolution. With respect to cITP of lipoproteins, a method developed by Schmitz et al. [22], and further improved [15,23,24] was applied routinely in dozens of studies related to cardiovascular disease [3,25,26]. These cITP methods employ anionic mode separation in the presence of 9 spacer molecules [15] achieving 8 discrete lipoprotein peaks. For selective detection of lipoproteins, lipophilic staining reagents are available: NBD-C6-Ceramide [15,27] (for laser induced fluorescence (LIF) detection), and Sudan Red 7B [24], or Sudan Black B [27] (for UV/vis detection). Lipoprotein ITP separations are challenging as they should be performed in anionic mode at basic pH and coated capillaries are required for prevention of lipoprotein adsorption and for EOF suppression [28]. However, long-term stability of capillary coatings at basic pH can be critical [24,29]. However, in order to consolidate cITP as a reference technique for lipoprotein profiling, a significant improvement in separation resolution and therefore in sublipoprotein fractionation is still needed, which would have tremendous benefits for clinical analysis and diagnostics. In conclusion, there is still a pressing demand for improved separation methods to enable more detailed fractionation of sublipoprotein populations in serum with the possibility to subsequently apply another biochemical analysis. Herein, a new cITP method is reported for higher resolution and reproducibility of lipoprotein profiling in serum/plasma than reported before [15]. A novel 24 spacers mixture was fine-tuned using software-based (Peakmaster) [30] prediction of electrophoretic behavior and tested on human serum. The separation of 18 lipoprotein peaks was
Please cite this article in press as: E. Moreno-Gordaliza, et al., A novel method for serum lipoprotein profiling using high performance capillary isotachophoresis, Analytica Chimica Acta (2016), http://dx.doi.org/10.1016/j.aca.2016.09.038
E. Moreno-Gordaliza et al. / Analytica Chimica Acta xxx (2016) 1e13
achieved, doubling the amount of peaks resolved by previous ITP methods, and using a new highly stable coated capillary. The method was successfully applied on human plasma, and the results compared with those obtained by NMR analysis of the lipoproteins. Finally, to demonstrate the excellent potential of our method, lipoprotein profiling was applied to serum samples of an atherosclerosis LDL receptor knock-out (LDLr/) mice model. 2. Materials and methods 2.1. Chemicals All reagents used were purchased from Sigma-Aldrich (Zwijndrecht, The Netherlands) unless otherwise stated. A pH-Meter (HI4521, Hanna Instruments, Nieuwegein, The Netherlands) was used for pH adjustments. Milli-Q water (Merck Millipore, Amsterdam, The Netherlands) was used for preparation of all aqueous solutions. Tricine was purchased from Janssen Chimica. GlyGly was purchased from Acros Organics. L-Aminoacids (L-cysteine (Cys), Lhistidine monohydrochloride (His), L-glutamine (Gln), L-glycine (Gly), L-glutamic acid (Glu), L-methionine (Met), L-phenylalanine (Phe), L-serine (Ser), L-valine (Val)) used were obtained from Fluka (Zwijndrecht, The Netherlands). 2.2. Animals Female, 10e12 weeks old, LDLr/ mice (The Jackson Laboratory, Bar Harbor, ME, USA) were used and kept under standard laboratory conditions. The animals were fed either a normal chow diet -RM1- (Special Diets Services, Witham, Essex, UK) (week 0), or fed for 3, 6 or 9 weeks a Western-type diet (WD) containing 0.25% cholesterol and 15% cocoa butter (Special Diet Services, Witham, Essex, UK). 6 animals per time point (weeks 0, 3, 6, 9) were bred and a total of 24 animals were used. Water and food were provided ad libitum. Blood was obtained from the mice by tail vein bleeding. All animal work was approved by Leiden University regulatory authority and was performed complying with the Dutch government guidelines. 2.3. Serum/plasma samples Human blood serum from healthy donors was used for method development and was obtained either frozen from Sigma-Aldrich, or fresh from Sanquin blood supply (Amsterdam, The Netherlands). Fresh serum for lipoprotein isolation was kept at 4 C until use. Ten individuals were selected from the study population of Erasmus Rucphen Family in order to facilitate the comparison of cITP profiling to an established commercial 1H-NMR method. 5 female and 5 male participants within the age range of 60e67, representing either low (1.3 mmol L1) levels of serum HDL cholesterol, were included in the comparison. None of the participants neither had known diseases such as diabetes, nor used lipid lowering medication. The venous blood samples of the ten individuals were collected after at least 8 h of fasting during the ERF cohort recruitment period (2002e2004), followed by simultaneous serum and EDTA-plasma isolation. Samples were stored at 80 C until measurements, yielding storage times ranging from to 7e9 years for the serum 1H-NMR analysis, and 11e13 years for the plasma cITP-based profiling. The samples have been previously included in a larger genome-wide association study which can be referred for further methodological information [31]. The ERF study protocol was approved by the medical ethics board of the Erasmus MC Rotterdam, The Netherlands. Finally, serum was also obtained from blood taken from the mice
3
model fed either the normal diet (week 0) or a Western-type diet (3, 6 and 9 weeks). 2.4. Preparation of fresh reference lipoprotein fractions Lipoprotein isolation was carried out by density gradient centrifugation as previously described [32] (see Supplementary Material for details). Separated lipoproteins (lipoprotein deficient serum (LPDS), HDL, LDL, VLDL/chylomicrons) were isolated and kept at 4 C until use. Isolated LDL was dialyzed against phosphate buffered saline (PBS) solution containing 10 mM EDTA (pH 7.4) for 24 h at 4 C, and oxidized by incubation at 37 C for 20 h with 10 mM CuSO4, as previously described [33]. Alternative lipoproteins (VLDL and chylomicrons) were purchased from Athens Research and Technology (GA, USA). 2.5.
1
H-NMR
NMR experiments in serum samples were acquired on a 600 MHz Bruker Avance II spectrometer (Bruker BioSpin, Karlsruhe, Germany), as previously described [31]. The NMR spectra were then deconvoluted using a commercial algorithm patented by the Bruker Corporation NMR services. Lipoproteins and their subclasses were defined according to their density and particle size (https://www. bruker.com/fileadmin/user_upload/8-PDF-Docs/MagneticResonance/NMR/brochures/lipo-analysis_apps.pdf) (see Supplementary Material for further details). 2.6. Capillary isotachophoresis 2.6.1. cITP-UV/vis 20 mL human serum/plasma was stained with 10 mL 1% Sudan Black B (SBB) (or alternatively Sudan Red 7B (SR7B)) in ethyleneglycol for 1 h at 4 C in the dark. 25 mL pre-stained serum/ plasma were diluted with 88.8 mL of leading electrolyte (LE) and mixed with 18.8 mL of spacers solution. The new spacers solution was prepared by mixing the following volumes of 10 mg L1 aqueous solutions: 100 mL N-(2-acetamido)-2-aminoethane sulfonic acid (ACES), 50 mL glucuronic acid, 50 mL octanesulfonic acid, 80 mL N-[tris(hydroxymethyl)methyl]-3-amino-2-hydroxy propanesulfonic acid (TAPSO), 120 mL N-[tris(hydroxymethyl) methyl]-3-aminopropanesulfonic acid (TAPS), 80 mL Gln, 80 mL Lmethionine (Met), 50 mL Gly, 60 mL Ser and 13.4 mL of: 2-[N-morpholino]ethanesulfonic acid (MES), glycyl-L-glycine (GlyGly), 3morpholino-2-hydroxypropanesulfonic acid (MOPSO), Tricine, Cys, L-alanyl-glycine (AlaGly), L-alanyl-L-alanine (AlaAla), glycyl-Lphenylalanine (GlyPhe), glycyl-L-histidine (GlyHis), His, Phe, Glu, Val, 3-methyl-L-His and 1-methyl-L-His. During optimization, different combinations of spacers were assessed and tests with other alternative spacers compounds were carried out, using 3-(Nmorpholino)propanesulfonic acid (MOPS), bicine, glycyl-leucine (GlyLeu), glycyl-proline (Gly-Pro), L-asparagine (Asp), L-tryptophan (Trp), glycyl-sarcosine, g-glutamyl-alanine (g-GluAla). The LE contained 10 mM HCl and 0.35% hydroxypropylmethylcellulose (HPMC), adjusted with 2-amino-2-methyl-1,3-propanediol (ammediol) to pH 8.8. The terminating electrolyte (TE) contained 20 mM L-alanine adjusted to pH 9.4 with ammediol solution. Alternative LE/TE compositions explored during optimization consisted on the following aqueous solutions: 10 mM HCl, 0.35% HPMC for the LE and 20 mM L-Ala for the TE, both adjusted with ammediol to either pH 9.0 or pH 9.5. An Agilent G1600AX CE system equipped with a Diode Array Detector (DAD) and a deuterium lamp was used for cITP analysis. Coated fused silica capillaries (BGB Analytik AG, Switzerland; OV1701-OH coating þ 0.4 mm C1 coating (100% methylated)) with
Please cite this article in press as: E. Moreno-Gordaliza, et al., A novel method for serum lipoprotein profiling using high performance capillary isotachophoresis, Analytica Chimica Acta (2016), http://dx.doi.org/10.1016/j.aca.2016.09.038
4
E. Moreno-Gordaliza et al. / Analytica Chimica Acta xxx (2016) 1e13
internal diameter: 180 mm, outer diameter: 365 mm; total length: 35 cm and effective length: 27.5 cm; were used for cITP separation. Pre-conditioning of the capillary was 10 min LE flush. Hydrodynamic injection was performed with 50 mbar for 15 s; lipoprotein separation was performed at 10 kV. The capillary, samples and buffers were used at 20 C. Detection took place at 254 nm, 280 nm (for protein detection), 570 nm (for SBB-stained lipoproteins) or 520 nm (for SR7B-stained lipoproteins). The capillaries were postconditioned after each run by flushing at 940 bar: 1.5 min LE, 0.2 min Milli-Q H2O, 0.4 min 1.5% Triton X-100 reduced, 0.4 min Milli-Q H2O, 2 min LE. After operation, capillaries were flushed with Milli-Q H2O for 1 min, air-blown dried and stored at RT. 2.6.2. cITP-LIF For cITP analysis with LIF detection, 5 mL serum was diluted with 17.5 mL LE and incubated at RT for 1 min with 11.22 mL NBD-C6Ceramide (Molecular Probes, Life Technologies) 0.1 mg mL1 solution prepared in anhydrous ethyleneglycol:methanol 9:1. The stained serum was mixed with 63.2 mL LE and 38.2 mL of the optimized spacers solution as described above. A ProteomeLab PA800 CE instrument (Beckman Coulter, Fullerton, CA, USA) equipped with a LIF detector was employed. Sample injection was performed at 0.8 psi for 13.6 s. The same capillaries as those used for UV/vis detection were employed, with a total length of 36 cm and length to the detector of 22.8 cm in this case. For LIF detection, excitation wavelength was set at 488 nm and emission detection at 520 nm. Preconditioning, postconditioning and separations were done at the same conditions as cITP-UV/vis at 10.3 kV.
surface phospholipids and proteins. This heterogeneity leads to different size/charge ratio and electrophoretic mobility, which is reflected in the separation achieved. Using Peakmaster, the electrophoretic behavior of the 9 spacers was predicted and this information was used to find more potential anionic spacers for an improved separation. Table S-1 shows the meff values calculated at pH 8.8, 9.0, 9.2, 9.4 and 9.5 in the absence of EOF for 61 selected compounds present in PeakMaster database, covering a wide range of mobilities, including those of the 9 spacers. In a non-linear gradient pH system, meff prediction is more complex as it should be calculated according to the pH in the position of each spacer in the gradient. As an approximation for preselection, computed meff values in an intermediate pH (with respect to those of the LE and TE) were considered, followed by experimental testing. In view of the computed data and previous literature information on spacers [24,29], 27 new spacers were selected: Asp, GABA, Glu, MES, g-GluAla, GlyGly, MOPS, MOPSO, Cys, Bicine, Tricine, AlaGly, AlaAla, GlyPhe, GlyLeu, GlyHis, GlyPro, GlySar, Asn, 3-methyl-His, Thr, Tyr, His, Phe, 1-methyl-His, Trp and Val. The selected compounds were individually tested in combination with the starting mixture of 9 spacers for cITP separation of serum in a gradient pH 8.8/9.4 system. Fig. S-2 shows typical examples of the outcome of the individual tests. Fig. 1 summarizes the experimental positions of the 27 selected spacers (red labels) found during cITP separation of human serum. General good agreement was found between experimental positions of spacers and computed meff values at intermediate pH 9.2. As can be seen, a number of compounds with intermediate mobilities compared to those of the starting 9 spacers were found, being potential candidates for separation improvement.
2.7. Electrophoretic mobility calculations 3.2. Optimization of cITP lipoprotein profiling Theoretical effective electrophoretic mobilities were computed using Peak Master v.5.3 [30] for compounds included in the software database. Basing on absolute mobilities (mo) and pKa of anions [34,35], effective mobilities (meff) were computed to predict electrophoretic behavior at a certain pH [36,37]. 2.8. Statistics Data analysis and statistics were performed using Metaboanalyst v.3.0 [38] and IBM SPSS Statistics (v 22). Prior to analysis, generalized logarithm transformation and data range scaling (mean-centered, divided by each variable's range) were performed. 3. Results and discussion 3.1. Influence of spacers on cITP separation The spacer molecules present during cITP separations have a major impact on the resolution that can be achieved and careful selection is mandatory. In a first series of experiments, the position of the 9 spacers (ACES, glucuronic acid, octanesulfonic acid, TAPSO, €ttcher et al. [15] used in TAPS, Ser, Gln, Met and Gly) proposed by Bo the current standard methods for lipoprotein cITP profiling was determined. This was done by studying the effect on the original separation of the removal one of the 9 spacers each time (Fig. S-1). Fig. 1 summarizes the location found for the 9 spacers during serum cITP separation in a gradient pH system at 8.8e9.4 (blue labels). 8 separated lipoprotein peaks were observed (Fig. 1b and S-1a), which are consistent with the literature [15]. According to previous works, these correspond to HDL (peaks 1,2,3), chylomicrons/remnants/VLDL/IDL (peaks 4 and 5) and LDL (peak 6, 7, 8). Each lipoprotein class and subtype presents different size and protein and lipid composition, influencing the charge of the particle, especially
In view of the results from section 3.1, a novel combination of 24 spacers for enhanced cITP separation of lipoproteins was designed. Fig. 2 displays the improved serum isotachopherogram when the new spacers mixture was used in a pH 8.8/9.4 system. A much higher resolution than that observed in Fig. 1 can be seen both for the total protein isotachopherogram (Fig. 2a) and the lipoprotein fractionation, showing a total of 18 separated peaks (Fig. 2b). The positions of the 24 spacers are displayed in Fig. 2a. The new lipoprotein separation is superior to previous methods described [3,15,24,29], with i.e. twice as many well-resolved peaks as reported until now. This is key to increase the information extracted from a profiling analysis and to identify differences in lipoprotein subparticles. For the identification of lipoprotein peaks, freshly isolated human lipoproteins (HDL, LDL, VLDL/chylomicrons; see Experimental section) were spiked to reference human serum. Fig. S-3 shows compared isotachopherograms where increased peak areas in spiked serum with respect to non-spiked serum indicate the presence of the added lipoprotein species (HDL, VLDL and LDL in Figs. S-3a, S-3b and S-3c, respectively). Additional spiking experiments were performed with fully oxidized LDL and commercially obtained VLDL and chylomicrons. These allowed locating oxLDL peaks and discerning between chylomicrons and VLDL peaks (present in the same band after density gradient centrifugation, considering its limited resolution). The new peaks found are assigned in Fig. S-3 and all the lipoprotein species summarized in Fig. 2b. HDL is mainly distributed within the first 8 peaks with some contribution to peaks 10 and 11, being its presence in peak 9 possible. Chylomicrons are mainly present in peak 11 with some contribution to peaks 9 and 10. VLDL was mainly found in peaks 11e12 with a slight amount in peak 10. Finally, LDL is present in peaks 13e18, being fully oxidized LDL mainly located in peak 12. No
Please cite this article in press as: E. Moreno-Gordaliza, et al., A novel method for serum lipoprotein profiling using high performance capillary isotachophoresis, Analytica Chimica Acta (2016), http://dx.doi.org/10.1016/j.aca.2016.09.038
E. Moreno-Gordaliza et al. / Analytica Chimica Acta xxx (2016) 1e13
5
€ ttcher et al. [15], monitoring A) absorbance at 280 nm for proteins and B) absorbance at 570 nm Fig. 1. cITP separation of SBB-stained human serum using conditions reported by Bo for specific lipoproteins detection. The position of the spacers used is indicated with blue dashed lines. The experimental position of 27 alternative spacer compounds tested is indicated with red dashed lines. Hydrodynamic injection was performed at 50 mbar for 15s (9 mg total protein). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
significant oxidation effects are expected during sample preparation. However, future experiments might consider the location of other oxidized or glycated particles which might be endogenously present in the serum/plasma samples. The main lipoprotein distribution is in agreement with results previously described by € ttcher [15] and Zhang [39], but a larger number of well-resolved Bo species was found in this study. A much better resolution of HDL and LDL species than previous methodologies [3,14,15,22e24,26,27,29,39] was attained with this method, using a low-microliter range serum volume. The influence of the pH on the separation was also assessed.
Constant pH lipoprotein cITP separation at either pH 9.0 (Fig. S-4) or 9.5 (not shown) resulted in a lower resolution of lipoproteins: more co-migration of LDL and VLDL particles was observed in comparison to the gradient pH separation. Therefore the combination of the gradient pH system (8.8/9.4) (used in the method by €ttcher et al. [15]) and the 24 spacers was selected as optimal. Bo Fig. 3 illustrates the influence of both sample volume injected and spacers concentration on the lipoprotein separation efficiency. The injected sample volume has a great influence in the separation outcome, since the protein components in the sample also act as internal spacers. Fig. 3a shows how, when the injected sample
Please cite this article in press as: E. Moreno-Gordaliza, et al., A novel method for serum lipoprotein profiling using high performance capillary isotachophoresis, Analytica Chimica Acta (2016), http://dx.doi.org/10.1016/j.aca.2016.09.038
6
E. Moreno-Gordaliza et al. / Analytica Chimica Acta xxx (2016) 1e13
Fig. 2. cITP separation of SBB-stained human serum using a novel mixture of 24 spacers at gradient pH 8.8/9.4. A) Signal at 280 nm for serum proteins along with the position of the spacers. B) 18 separated lipoproteins detected at 570 nm. Peak assignments from spiking experiments with freshly isolated lipoproteins are included in B). Hydrodynamic injection was performed at 50 mbar for 15s (9 mg total protein).
amount was increased from 9 to 30 mg total protein, resulting the amount of spacers 3.3-fold increased as well, the zone widths for the protein components led to higher lipoprotein resolution. A similar result was achieved when only the spacer mixture concentration in the sample was increased 3-fold (Fig. 3b). These options to improve peak resolution are especially interesting when fraction collection is desired. In this case, baseline resolution was achieved for peaks 1e2 and 6e18, which also makes this more advantageous for lipoprotein quantification with more reliable peak area integration. Regarding peak integration, it is advisable to perform a perpendicular drop integration method on the drawn baseline from the peak valleys (if unresolved). This can be best manually performed for such unresolved peaks or at least manually corrected after automatic integration. However, in some cases, if the split in peaks 3e5 is unclear, for minimizing errors in area calculation, it might be better to integrate peaks 3e5 altogether.
3.3. Capillary coating and separations reproducibility The performance of specially coated capillaries, i.e. OV1701-thin film deactivation stable against water combined with a 0.4 mm thick polydimethylsiloxane coating, for cITP lipoprotein separations was investigated. This special coating was designed for producing more stable capillaries that can better tolerate and withstand aqueous high pH conditions. Previous studies reported stability issues with coatings used for lipoprotein separations by cITP due to hydrolysis of the silanized groups at high pH. For instance, capillaries coated with polyacrylamide were stable for only around 50 runs [24] and polydimethylsiloxane-coated capillaries had to be recoated after every run [29]. In this study, the doubly-coated capillaries were stable for at least 100 runs under high pH conditions (pH 8.8e9.4). Moreover, with these capillaries, low coefficients of variation (CV) were obtained for lipoproteins migration times during repeated €ttcher et al. [15] (1.3e2.0%) or injections both for the method by Bo using our new method with the mixture of 24 spacers (2.4e2.6%),
Please cite this article in press as: E. Moreno-Gordaliza, et al., A novel method for serum lipoprotein profiling using high performance capillary isotachophoresis, Analytica Chimica Acta (2016), http://dx.doi.org/10.1016/j.aca.2016.09.038
E. Moreno-Gordaliza et al. / Analytica Chimica Acta xxx (2016) 1e13
7
Fig. 3. cITP lipoprotein profiling of SBB-stained human serum under optimized conditions. A) Injection of 30 mg total protein (50 mbar, 50s). B) Injection of 9 mg total protein (50 mbar, 15s) and 3-fold increased spacer concentration.
as summarized in Table S-2. The variation of the peak areas within a day was satisfactory, with CVs lower than 5%. The use of HPMC in the LE also allows creating an extra dynamic coating, ensuring suppression of the electroosmotic flow and analyte-wall interactions as previously reported, contributing as well to separation repeatability. Fig. S-5 also shows the variation of migration times for representative lipoprotein peaks (1, 11 and 17) during repeated runs. CVs for relative migration times (with respect to that of peak 1) were lower than 1%. Between-capillary migration time reproducibility was also good (CV within the 1.0e2.1% range (n ¼ 3)). Finally, samples were shown to be stable for at least 3 days after preparation when stored at 4 C in-between analyses, and at least for a month at 20 C. Excellent long-term reproducibility is therefore attained when separating serum in doubly-coated capillaries using the optimized 24 spacers mixture in the pH 8.8/9.4 system, compared to previously reported results [40]. 3.4. Staining methods for lipoprotein profiling SR7B and SBB were compared for specific lipoprotein staining with UV/vis detection, as can be seen in Fig. S-6. Although both stains could be used, SBB-stained lipoproteins presented higher
signal intensity than those stained with SR7B. In the reminder of this work, SBB staining was used. Fig. S-7 demonstrates that the improved profiling method can be transferred to cITP-LIF analysis using NBD-C6-Ceramide-stained serum. Adaptation of the method is straightforward, by replacing the staining agent and an appropriate sample dilution due to the higher sensitivity of LIF. The same lipoprotein profile as observed with UV/vis detection was obtained. This is of great interest as it would allow the use of a fluorescent internal standard like 5-carboxyfluorescein for more accurate quantification purposes [27]. Furthermore, it confirms the reproducibility of the method in a quite different electrophoretic system. 3.5. Comparison of cITP and NMR profiling To validate our cITP method, analysis of human plasma samples from 10 different individuals selected as described in the experimental section, was carried out and the results were compared with quantitative NMR measurements previously performed [31]. The NMR method is based on a deconvolution algorithm of the overlay of the 1H-NMR spectra of the different lipoprotein subparticle types, analogously to previously reported methods [41], where lipids are the main contributors to the signal. Lipid (FC, Total Cholesterol -C-, PL, TG) and apolipoprotein (apo A1, apoA2, apoB)
Please cite this article in press as: E. Moreno-Gordaliza, et al., A novel method for serum lipoprotein profiling using high performance capillary isotachophoresis, Analytica Chimica Acta (2016), http://dx.doi.org/10.1016/j.aca.2016.09.038
8
E. Moreno-Gordaliza et al. / Analytica Chimica Acta xxx (2016) 1e13
levels were obtained in each lipoprotein subparticle type. Fig. S-8 shows clustering of two main groups of 5 samples each (1e5: group 1; and 6e10: group 2) after Partial Least Square-Discriminant Analysis (PLS-DA) of the NMR data. Successful cross validation of the model is shown in Fig. S-8b and Table S-4. As indicated by the Variable Importance in Projection (VIP) scores displayed in Fig. S8c, representing the main components leading to group separation, the samples differ in HDL subparticle composition (mostly a decrease for very large and large HDL particles and to a lesser extent, medium size HDL in Group 2 and in LDL particles (basically with an associated increase in TG in Group 2. Fig. 4 shows successful comparison of cITP and NMR data. Comparable to the NMR results, higher cITP areas were visible for peaks in the HDL region for Group 1, while certain LDL peaks presented lower areas compared with Group 2. Fig. 4a shows the overlap of the lipoprotein isotachopherograms for two representative samples from each of the two groups. Moreover, Fig. 4b
shows clustering of the samples into the same two groups after PLS-DA analysis of cITP peak areas (Fig. 4b) as was observed for the same analysis of the NMR data (Fig. S-8). Cross validation of the PLS-DA model is displayed in Fig. 4c and Table S-5, indicating a correct group separation. Moreover, analysis of the VIP scores reveals that the most important components leading to group separation belong to the first 5 ITP lipoprotein HDL peaks, and peaks 12e15 (attributed mainly to LDL), as can be seen in Fig. 4d. Additional t-tests confirmed that the most significant differences found between the two groups with this cITP method concern peaks 1, 13, 14 and 15 in the cITP separation with a p-value
