Low-Density Lipoprotein Sensor Based on Molecularly Imprinted

Subscriber access provided by The University of Liverpool

Article

Low Density Lipoprotein Sensor Based on Molecularly Imprinted Polymer Suticha Chunta, Roongnapa Suedee, and Peter Alexander Lieberzeit Anal. Chem., Just Accepted Manuscript • DOI: 10.1021/acs.analchem.5b04091 • Publication Date (Web): 08 Dec 2015 Downloaded from http://pubs.acs.org on December 9, 2015

Just Accepted “Just Accepted” manuscripts have been peer-reviewed and accepted for publication. They are posted online prior to technical editing, formatting for publication and author proofing. The American Chemical Society provides “Just Accepted” as a free service to the research community to expedite the dissemination of scientific material as soon as possible after acceptance. “Just Accepted” manuscripts appear in full in PDF format accompanied by an HTML abstract. “Just Accepted” manuscripts have been fully peer reviewed, but should not be considered the official version of record. They are accessible to all readers and citable by the Digital Object Identifier (DOI®). “Just Accepted” is an optional service offered to authors. Therefore, the “Just Accepted” Web site may not include all articles that will be published in the journal. After a manuscript is technically edited and formatted, it will be removed from the “Just Accepted” Web site and published as an ASAP article. Note that technical editing may introduce minor changes to the manuscript text and/or graphics which could affect content, and all legal disclaimers and ethical guidelines that apply to the journal pertain. ACS cannot be held responsible for errors or consequences arising from the use of information contained in these “Just Accepted” manuscripts.

Analytical Chemistry is published by the American Chemical Society. 1155 Sixteenth Street N.W., Washington, DC 20036 Published by American Chemical Society. Copyright © American Chemical Society. However, no copyright claim is made to original U.S. Government works, or works produced by employees of any Commonwealth realm Crown government in the course of their duties.

Page 1 of 10

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Analytical Chemistry

Low Density Lipoprotein Sensor Based on Molecularly Imprinted Polymer Suticha Chunta,†,‡ Roongnapa Suedee,§ Peter A. Lieberzeit*,† †

University of Vienna, Faculty for Chemistry, Department of Analytical Chemistry, Waehringer Strasse 38, A-1090 Vienna, Austria, Tel.: +43 1 4277 52341, Fax: +43 1 4277 9523, [email protected] ‡

Department of Clinical Chemistry, Faculty of Medical Technology, Prince of Songkla University, Hatyai, Songkla 90112, Thailand §

Department of Pharmaceutical Chemistry, Faculty of Pharmaceutical Sciences, Prince of Songkla University, Hatyai, Songkla 90112, Thailand

ABSTRACT: Increased level of low density lipoprotein (LDL) strongly correlates with incidence of coronary heart disease. We synthesized novel molecularly imprinted polymers (MIP) as biomimetic specific receptors to establish rapid analysis of LDL levels. For that purpose the ratios of monomers acrylic acid (AA), methacrylic acid (MAA) and N-vinylpyrrolidone (VP), respectively, were screened on 10 MHz dual-electrode quartz crystal microbalances (QCM). Mixing MAA and VP in the ratio 3:2 (m/m) revealed linear sensor characteristic to LDL cholesterol (LDL-C) from 4 to 400 mg/dL or 0.10 to 10.34 mmol/L in 100 mM PBS without significant interference: high density lipoprotein (HDL) yields 4-6% of the LDL signal, very low density lipoprotein (VLDL) 1-3%, and human serum albumin (HSA) 0-2%. LDL-MIP sensor system reveals analytical accuracy of 95-96% at 95% confidence interval with precision at 6-15%, respectively. Human serum diluted 1:2 with PBS buffer was analyzed by LDL-MIP sensors to demonstrate applicability to real-life samples. The sensor responses are excellently correlated to the results of the standard technique, namely a homogeneous enzymatic assay (R2=0.97). This demonstrates that the system can be successfully applied to human serum samples determining LDL concentrations.

For assessing coronary heart disease (CHD) and monitoring its treatment it is essential to determine serum concentrations of low density lipoprotein (LDL): It has been proven that this parameter corresponds to higher incidence of CHD.1,2 In clinical laboratory diagnosis, the cholesterol content of LDL (LDL-C) has long been the standard approach to approximate the concentration of LDL particles. The reason for this is that a large number of population studies indicate that elevated LDL-C is the most useful prognostic parameter in assessing the risk for developing CHD.3 Actual LDL particle concentrations are also accessible via nuclear magnetic resonance (NMR) spectrometry4 and ion mobility spectrometry.5 However, especially NMR is rather costly. Furthermore, there are no systematic, independent data confirming clear correlation between LDL particle numbers and CHD due to high analysis costs.3 In clinical analysis, the LDL-C concentration in serum is indirectly calculated by the so-called Friedewald equation:6,7 LDL-C = TC − HDL-C − (TG/5) Here, TC is total cholesterol, HDL-C is high density lipoprotein cholesterol, TG is triglyceride. However, this calculation only leads to valid results when taking the serum from patients after more than 12 hours of fasting to

eliminate postprandial chylomicrons, which are secreted after intestinal cells absorb triglyceride in food. They represent the first step in lipid resorption, but are cleared from blood circulation within 12 hours after a meal. If the non-fasting triglyceride value is used in the Friedewald equation, LDL cholesterol will thus be underestimated. Moreover, the overall triglyceride concentration must be lower than 400 mg/dL, otherwise the factor “TG/5” does not reasonably estimate the concentration of very low density lipoprotein (VLDL). This approach to obtain the level of LDL-C is prone to substantial statistical errors, because it considers three different analyses and their respective confidence intervals.8 LDL-C can also be measured directly by a homogeneous enzymatic colorimetric assay using cholesterol esterase, cholesterol oxidase and peroxidase to generate a quinone pigment quantified by UV-Vis photometry. However, this approach requires tedious sample preparation to remove the other lipoprotein classes from human serum before determining LDLC. Before actual measurement, LDL needs to be precipitated with additional reagents, such as heparin, polyvinyl sulfate or dextran sulfate. In a different approach one can use polyclonal goat antibodies to human apoA1 and apoE bound to latex beads or cyclodextrin sulfate to block or remove unwanted species (chylomicrons, HDL, VLDL, and intermediate density lipoprotein (IDL)) from serum.9

ACS Paragon Plus Environment


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Such limitations of established techniques to determine LDL-C fuel the desire for easy-to-use, inexpensive, and reasonably accurate measuring systems, such as sensors. Current lipoprotein sensors usually comprise various types of apolipoprotein antibodies10,11 as recognition materials. Albeit their high selectivity, such systems are limited in terms of stability, long-term storage, and comparably high cost of antibodies. Therefore, biomimetic recognition elements have attracted substantial interest to circumvent constraints of biogenic sensor layers.12,13 Molecularly imprinted polymers (MIP) are a popular example of such artificial, biomimetic recognition.14,15 In MIP synthesis, functional monomers covalently or non-covalently interact with template species to create highly selective recognition sites. After polymerization, the template is subsequently removed, revealing cavities which have complementary size and shape and thus form the binding site within the polymer.13,14,16 To date, MIP have been reported as selective recognition elements for a range of bio-analytes, such as proteins,17,18,19 bacteria,20 viruses,21 and yeasts.22 Some MIP have also been reported for targeting the main lipid constituents of lipoproteins, for instance phospholipid,23 and cholesterol.24 However, LDL consists of a complex assembly based on esterified cholesterol and triglyceride surrounded by a molecular surface containing a mixture of phospholipids, free cholesterol, and apolipoprotein. To the best of our knowledge, currently no MIPs exist for such complex lipoprotein aggregates. The work presented herein hence inherently opens up a novel field in the area of molecular imprinting that is potentially useful in clinical analysis.

EXPERIMENTAL SECTION Apparatus. Dual gold-electrode patterns, comprising diameters of 4 and 5 mm on rear and sample side, respectively, were screen-printed by depositing brilliant gold paste (Heraeus; 12%) onto 10 MHz AT-cut quartz blanks (diameter 13.8 mm, Great Microtama Industries, Surabaya, Indonesia) to yield quartz crystal microbalances (QCM). After burning at 400 °C for 4 hours to remove organic residues, both resonance frequency and damping of bare QCM were determined via Agilent 8712ET network analyzer.19 To be acceptable for the use in a sensor, we selected QCM transducers with less than -5 dB damping on the maximum. Reagents. Human serum albumin (HSA) was purchased from Millipore (MA, USA). Acrylic acid (AA), methacrylic acid (MAA), N-vinylpyrrolidone (VP), dimethyl sulfoxide (DMSO), and agarose powder were obtained from VWR International (Vienna, Austria); N,N′(1,2-dihydroxyethylene) bisacrylamide (DHEBA), 2,2′azobis (isobutyronitrile) (AIBN), sodium bromide (NaBr), and Sudan Black B from Sigma-Aldrich (Steinheim, Germany). Tris(hydroxymethyl)-aminomethan (Tris), ethylenediamine tetraacetic acid (EDTA) were purchased from Merck (Darmstadt, Germany). Dextran sulfate sodium salt was obtained from Alfa Aesar (Karlsruhe, Germany). Sodium chloride (NaCl) was obtained from Applichem (Darmstadt, Germany). Acetic acid was purchased

Page 2 of 10

from Carl Roth (Karlsruhe, Germany). All reagents were of analytical or highest synthetic grade commercially available. Isolation of Lipoproteins. All standard lipoprotein solutions, i.e. VLDL, LDL, and HDL, were isolated from a total of 10 pooled human sera by gradient density ultracentrifugation25 on a Beckman Coulter Optima L-100 XP ultracentrifuge with a fixed angle rotor type 100 Ti 100,000 rpm. Samples were fractionated as follows: 4 ml serum was filled into a polypropylene centrifuge tube (V=6.5 ml). Then 2 ml 0.195 M NaCl solution with density 1.006 g/ml was layered on the top of the serum. After centrifugation for 10 hours at 80,000 rpm, the VLDL fraction had migrated into the top layer. The bottom layer containing LDL, HDL, and other serum proteins was transferred to another centrifuge tube, to which we added 2 ml 0.195 M NaCl-2.44 M NaBr solution of density 1.063 g/ml. After mixing and centrifugation for 14 hours at 80,000 rpm, the top layer comprising LDL was drawn off. The bottom layer was moved to a new tube filled with a 2 ml 0.195 M NaCl-7.65 M NaBr solution of density 1.478 g/ml. Centrifugation for 10 hours at 80,000 rpm yielded the HDL fraction in the top layer. VLDL, LDL, and HDL fractions were identified and characterized by 0.5% agarose gel electrophoresis on Bio-Rad sub-cell GT electrophoresis systems as follows: 0.5 g of agarose powder was added to 100 ml of 1x Tris-acetate-EDTA (TAE) buffer (40mM of Tris, 20mM of acetic acid, and 1mM of EDTA), and heated to 100°C until the solution became clear and colorless. The gel precursor solution was poured into the gel tray already containing the comb necessary for generating sample wells in gel and polymerized for 30 minutes. The gel was then placed into the electrophoresis chamber and filled with 1x TAE running buffer up to the top. Solutions containing 5 µl of each lipoprotein fraction and 2 µl of loading dye (glycerol and bromophenol blue) were loaded into sample wells followed by applying 170V for 45 minutes. After electrophoresis, the lipids in the gel were stained with 0.4% Sudan Black B in a mixture of acetone, 17.4 M acetic acid, and water at the ratio 4:3:13 overnight. The gel was then de-stained by washing three times with a mixture of 150 ml 17.4 M acetic acid, 200 ml acetone, and 650 ml water. The cholesterol concentration of each nonstained fraction was determined using homogeneous enzymatic colorimetry on Roche Hitachi 917 chemistry autoanalyzer to obtain lipoprotein standard solutions with known LDL-C concentration. As previously mentioned, there are hardly any standard methods for determining intact LDL in blood. Therefore, all sensor signals in our study are calibrated against LDL-C rather than actual LDL concentration, because it makes data comparable to existing diagnostic methods. LDL-Template Stamp Preparation. Previous studies reported that LDL has oblate spheroid structure with a diameter of 28.9 ± 9.2 nm and a height of 8.7 ± 2.0 nm.26,27 This size is inherently useful for stamp imprinting. LDL template stamp was prepared by coating 5 µL LDL standard corresponding to 600 mg/dL LDL-C onto 5 × 5 mm clean, untreated glass substrates followed by sedimenta-


Page 3 of 10

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


tion at 4oC for 30 minutes. After removing excess solvent by spinning, the resulting stamp was ready for imprinting. Optimization of LDL-MIP Synthesis. Starting point of this work were polymers of previously published MIP for bovine serum albumin (BSA).19 However, that material proved unsuitable for LDL, so we varied the ratio of the functional monomers, i.e. AA, MAA, and VP. All polymers contained 70% (w/w) cross linker. They were prepared by dissolving 15 mg functional monomer mixture containing varying amounts of AA, MAA and VP, respectively, 35 mg DHEBA, and 2.4 mg AIBN in 300 µl DMSO. This solution was sonicated for 5 minutes and pre-polymerized under UV at 365 nm, 180 W, for 20 minutes. Subsequently, 5 µl of this pre-polymer were spin-coated at 3,000 rpm for 2 minutes to cover the entire sample side of a dualelectrode QCM, i.e. both electrodes. Then, the LDL template stamp was pressed into the film on top of one electrode to yield the MIP. The polymer on the second electrode remained untreated and resulted in non-imprinted polymer (NIP). The overall setup is sketched in Figure 1. After that the QCM was kept in the oven at 50oC for 12 hours to complete polymerization thermally. The template was removed by continuous stirring for 20 minutes in each of the following three solutions/solvents: 10% aqueous solution of acetic acid followed by 0.1% sodium dodecyl sulfate (SDS) solution, and deionized water, respectively.

Figure 1. Schematic of LDL-MIP QCM sensor

Polymer Characterization. Aside of QCM measurements, AFM was utilized to characterize MIP and NIP. For that purpose they were synthesized on glass plates and then assessed using a Bruker Instruments NanoScope 9 in contact mode with a silicon nitride cantilevers at 1 Hz scan rate (i.e one line per second). Sensor Measurements. For sensor measurements, dual-electrode QCM were placed into a custom-made measuring cell consisting of poly-dimethyl siloxane (PDMS) as described earlier.28 This measuring cell was linked to a customized oscillator circuit connected to a frequency counter (Agilent 53131A). Real-time frequency shifts as a function of time of two channels at 25 oC were read out via GPIB/USB interface and a LabView routine. All QCM measurements were started by obtaining baseline signal via exposing the sensor to 180 µL of 100 mM PBS, pH 7.4, in stopped flow until reaching equilibrium. During subsequent measurements, we exposed the sensor to 180 µL of solutions containing different concentrations of LDL-C (50-600 mg/dL) in 100 mM PBS, respectively, in stopped flow until reaching equilibrium. This was followed by washing steps with 10% aqueous solution of acetic acid followed by 0.1% sodium dodecyl sulfate (SDS)

solution, and deionized water (10 minutes each at a flow rate of 0.46 mL/min).19 During washing, no sensor signal was recorded, because conductivity of those solutions strongly differs from those of the samples leading to abnormally large frequency responses.29 We characterized QCM in terms of limit of detection, accuracy, precision, analytical sensitivity, and selectivity. For determining accuracy we undertook recovery test at clinically “normal” and “high” LDL-C concentrations: Different volumes of a standard LDL-C solution at a concentration of 500 mg/dL (namely 10 µL, 50 µL, and 100 µL) were spiked to 150 µL of 75 mg/dL of LDL-C solution to reach final concentrations at 102, 181, and 245 mg/dL, respectively. The resulting frequency shifts were compared with the values resulting from calibration. For precision test, the sensor data of LDL-C standards containing 102 and 200 mg/dL, respectively, were recorded three times each. Selectivity of LDL-MIP sensors were examined by exposing them to physiologically normal and high concentrations of possible interfering species that can be found in human serum, namely: HDL-C at concentrations of 40 and 200 mg/dL, VLDL-C at 20 and 80 mg/dL, and HSA at 500 and 1000 mg/dL, respectively. For clinical sample analysis, the effect of human serum to QCM sensors was studied. Two types of matrices, namely 100 mM PBS, and LDL-free serum, were used to obtain baseline signal before applying human serum diluted in each PBS or LDL-free serum (1 part serum + 1 part solvent). LDL-free serum was prepared by precipitating with a solution containing 20 g/L dextran sulfate (MW 500,000) and 2 mol/L MgCl2 in water. Briefly, we added 250 µL of MgCl2 solution and 250 µL of dextran sulfate solution to 5 ml serum following a routine that can be found in literature.30 After incubating for 15 minutes, the solution was centrifuged at 4,000 rpm for 20 minutes. LDL precipitated in the bottom fraction. The supernatant fraction was dialysed against Tris-HCl-NaCl buffer, pH 7.6 at 4oC overnight followed by 1%BaCl2 in 1%NaCl at 4oC overnight31 to remove excess precipitation agent. This resulted in LDL-free serum. Furthermore, we validated the results of LDL-MIP sensors by comparing their results to those obtained by a homogeneous enzymatic colorimetric assay on an autoanalyzer routinely used in clinical analysis. Ten human blood samples were taken at the Faculty of Medical Technology, Prince of Songkla University. Then, serum samples were isolated from whole blood by centrifugation at 3,500 rpm for 5 minutes. The amount of LDL-C of all serum samples was determined using both methods. The only difference was that for measuring with LDL-MIP sensors, all sera had to be diluted with 100 mM PBS at 1:2 prior to measurement.

RESULTS AND DISCUSSION Optimization of LDL-MIP Synthesis. LDL on the one hand has negative surface potential between -4.5 and -7.0 mV, and on the other hand contains a hydrophilic moiety comprising both negatively and positively charged side chains. Therefore copolymers containing AA, MAA, and



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

VP as functional monomers seem appropriate, because they complement positive and negative charges, respectively, and hence can be expected to show intermolecular interactions with LDL. However, the exact mixing ratio of these monomers needed to be determined. The starting point of this process was the AA:VP ratio of 2:3 (w/w)19 that had proven successful for MIP binding BSA. Figure 2 shows the QCM frequency responses of both LDL-MIP and NIP-coated electrodes when exposing them to 150 mg/dL LDL in PBS buffer, which leads to decreasing frequency. The amount of frequency shift (∆f) on the MIPcoated electrode (∆f = 420 Hz) indeed exceeds that of the NIP-coated one (∆f = 230 Hz). This corresponds to 190 Hz mass effect and hence corroborates successful imprinting. Furthermore, all bound LDL can be removed again by the washing sequence described in the experimental section.

Page 4 of 10

one order of magnitude. Therefore, this functional monomer ratio was chosen for further experiments.

Figure 2. QCM responses of MIP and NIP obtained for varying monomer ratios (left-hand axis). The right-hand axis gives the respective imprinting factors (green columns).

Nonetheless, the overall MIP effect is rather small, namely less than a factor of two between the two polymers. This probably happens due to the hydrophilic functionalities on the polymer surface: LDL aggregates contain major hydrophobic parts on their surfaces originating both from free cholesterol and hydrophobic tails of phospholipids. Therefore we replaced AA with a more hydrophobic monomer, MAA, and also varied the ratio between MAA and VP to enhance sensitivity. Figure 2 shows the outcome of this approach by summarizing the sensor responses observed for five different polymer compositions. When replacing AA by MAA while keeping the mixing ratio with VP constant at 2:3, both the MIP and NIP yield slightly higher signals. This indicates higher inherent affinity of the polymer to LDL, which is favorable for MIP development32 and indicates feasibility of the approach. However, other polymers lead to much higher sensor responses. The optimal monomer ratio turned out MAA:VP=3:2: In that case the NIP signal is less than 20% larger than before (327 Hz vs. 273 Hz), but the MIP signal increases by a factor of six to 2850 Hz. Figure 2 also shows the imprinting factors (i.e. the signal ratio MIP/NIP) for all polymers tested. For the final MIP it reaches almost


Page 5 of 10

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Figure 3. AFM images of LDL-MIP before removing the template (A), after removing the template (B), NIP (C), a cross-section profile across the indicated line A (D) and line B (E).

Polymer Characterization. Figure 3 shows AFM images of three different surfaces, namely LDL-MIP before (Figure 3A) and after (Figure 3B) removing the template, as well as the corresponding NIP (Figure 3C). Obviously, the surface of the LDL-MIP prior to washing shows a large number of circular structures whose dimensions are in the range of 48.8 ± 17.5 nm across (Figure 3D). On average this is somewhat larger than the expected oblate spheroid structure with a diameter of 28.9 ± 9.2 nm, the usual shape of LDL in physiological of blood samples. However, the difference can be explained by the experimental details: Gan et al.27 showed that LDL particles in air appear larger than in liquids, namely 45.8 ± 19.6 nm across and 8.5 ± 1.8 nm high. They attribute this phenomenon to the liquid film attached to individual LDL particles before completely drying. The diameters observed in the template stamp thus correlate to this data very well. The LDL-MIP surface reveals cavities that are on average 54.0 ± 12.2 nm across and 4.5 ± 2.3 nm deep (Figure 3E). This increase in size may be due to clustering of template LDL particles in the pre-polymer: the latter contains DMSO, which is known to promote protein aggregation.33 Nevertheless, the morphologies between MIP and NIP sufficiently differ from one another: MIP surface shows a substantial amount of cavities in the expected size range, whereas NIP surfaces completely lack those features as shown in Figure 3C. Therefore both AFM images and the QCM responses strongly corroborate successful synthesis of LDL-MIP.

Figure 4. Sensor response curve towards LDL in 100 mM PBS (A). Standard concentrations are given as LDL-C (B).



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Page 6 of 10

tially lower than for our LDL-MIP QCM sensor, but of Sensor Characteristic. Figure 4A shows sensor responses for a dual-electrode QCM coated with LDL-MIP and NIP, respectively, towards different LDL-C standards. LDL-MIP leads to substantial frequency signals in a range of 520-3370 Hz. In contrast to this the NIP-coated electrode yields slightly positive frequency shifts in a range of 10-150 Hz. The signal shapes on the MIP electrode at high LDL concentrations are unusual, because their equilibrium frequency shift is smaller, than immediately after injection. This might be a consequence of measuring in stopped flow: during injection larger numbers of LDL particles are flushed toward the affine surface followed by equilibration. First of all, this very strongly supports the suitability of molecular imprinting in generating sensor layers. Secondly, the difference in NIP electrode signal between these data and the ones shown in Figure 4A can be attributed to the different polymers that apparently have different surface roughness: Such non-Sauerbrey behavior – i.e. increasing frequency - has already been observed for other flat layers,34 especially for biospecies, such as yeasts35 and bacteria.36 In all these cases the analyte is only weakly bound on the flat surface and hence retains some mobility. Figure 4B shows the resulting LDL MIP sensor characteristic: it is linear in a concentration range 50 to 400 mg/dL (that is 1.29 to 10.34 mmol/L) LDLC with a correlation coefficient of R2=0.9949. Above 400 mg/dL (points not shown in the graph) the sensor signal reaches saturation due to the limited number of binding sites on the MIP surface. Based on 10 Hz measurement noise the respective limits of detection (LOD) and quantification (LOQ) are as follows: LOD = 4 mg/dL or 0.10 mmol/L LDL-C, LOQ = 13 mg/dL or 0.34 mmol/L. Therefore, the sensor dynamically responds to LDL-C concentrations between 4-400 mg/dL. Compared to this, detection range of 10-84 mg/dL has been reported for a QCMbased LDL-immunosensor.37 The current standard method, namely homogeneous enzymatic colorimetric assay, has a detection range of 0.2-410 mg/dL,9 however, at higher analysis costs due to sample pretreatment and reagent costs. Hence the LDL-MIP sensor performs very appreciably in physiologically relevant ranges, including both normal concentrations (below 129 mg/dL LDL-C), and abnormal ones (high level: 130-159 mg/dL LDL-C; very high level: above 159 mg/dL LDL-C). Figure 5 shows the recovery rates of the LDL-MIP sensor, which were calculated by comparing the measured concentrations to the expected concentrations of the spiked LDL-C samples. They are 95%, 96% and 96% for at concentrations of 102, 181, and 245 mg/dL LDL-C, respectively. According to literature, homogeneous assays yield recovery rates of 97-105%.38 Therefore the accuracy of our assay is inherently acceptable from the clinical point of view both for normal and high levels of LDL-C. In terms of precision, coefficients of variation (CVs) of LDL QCM measurement at concentrations of 102 and 200 mg/dL were 15% and 6%, respectively, as shown in Table 1. Within-run precision studies of homogeneous assays revealed imprecision values of 0.25-1.43%.38 This value is substan-

Figure 5. Recovery rates of LDL-MIP sensor

course results from optimized, validated, and commercially available technology. In contrast to that, the sensor assesses LDL directly, i.e. without sample pretreatment, and to the best of our knowledge is the first example of a biomimetic sensor layer addressing a lipoprotein complex. Table 1. Repeatability test

LDL-C (mg/dL)

Mean (n=3)

SD

%CV

102 200

111 189

16.75 10.61

15 6

Selectivity is a key issue for any sensor. Figure 6 displays the QCM results of both MIP and NIP in terms of relative effect compared to the respective LDL signal for a range of (lipo-)proteins that are expected in the blood stream: evidently, HDL leads to 4-6% of the LDL signal, VLDL to 1-3%, and HSA to 0-2%, which are all highly appreciable figures for a sensor. All these data refer to clinically “normal” and “high” serum concentrations of those compounds. Selectivity values fit both morphological and functional differences between the species very well: In terms of surface chemistry HDL and LDL are very similar. However, HDL particles are 21.5 ± 6.5 nm in diameter and therefore somewhat smaller than LDL (28.9 ± 9.2 nm). Despite these similarities, the MIP sensors favor LDL very strongly. VLDL particles are much larger (30-80 nm in diameter)39 and thus do not fit into the cavities generated on the MIP surface. Finally, HSA has an ellipsoid shape of 36 nm in diameter which does not fit the MIP cavities.40 As can be seen, the LDL-MIP sensor performs highly appreciably in terms of sensitivity, accuracy, precision, and selectivity when operated in PBS buffer. Application in clinical testing, however, requires measuring in serum. When exposing QCM coated with both LDL-MIP and NIP, respectively, to PBS buffer followed by switching to human serum, it turned out that the difference in viscosity leads to substantial frequency responses (see Figure S-1


Page 7 of 10

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60


Figure 6. Selectivity of LDL-MIP sensor

in the electronic supporting material). To reduce matrix effects caused by serum, we decided to dilute human serum samples with the same amount of PBS (i.e. to half its concentration). Therefore the baseline of the respective QCM measurements is defined by a solution containing 1 part LDL-free human serum and 1 part PBS buffer. Figure 7 compares the MIP sensor characteristics obtained both in diluted serum and PBS buffer, respectively. Evidently, the both regression parameters are very similar. This clearly indicates that the QCM sensor is indeed useful for assessing serum samples. Finally, Figure 8 plots LDL-MIP sensor results for reallife serum samples against the values obtained by the clinical standard method, i.e. homogeneous colorimetric assay. The data correlate linearly with R2= 0.97 and thus further support the MIP approach. Its results excellently match those of the conventional method. Therefore they constitute the first alternative method to achieve rapid one-step analysis of LDL.

Figure 8. Comparison of LDL-MIP sensor data (x axis) and results of the homogeneous assay (y axis)

CONCLUSIONS This work presents successful synthesis of highly selective LDL-MIPs that are suitable to detect LDL directly in serum. To the best of our knowledge this is the first MIPbased artificial recognition material targeting a lipoprotein aggregate. The LDL-MIP sensor system provides a way to develop non-fasting blood testing equipment with high performance for clinical purposes: it measures LDL directly, reveals minimal interference effects of other lipoproteins that can be found in non-fasting serum and does not require total cholesterol, HDL-C and triglyceride measurements needed for the indirect method.41 This means advantages for both patients, who can avoid fasting for 12 hours, and physicians, who obtain results faster than with current techniques. From the social point of view, such a straightforward detection system would allow easy and cost-effective testing also in developing economies that suffer very strong increase in LDL levels due to changes in lifestyle and diet. Finally, this approach is promising for further studies with other lipoproteins.

ASSOCIATED CONTENT Supporting Information Viscosity effect of PBS and human serum to LDL-MIP sensor; Figure S-1 Sensor signal of undiluted serum and diluted serum at 1:2 (PDF)

AUTHOR INFORMATION Corresponding Author * E-mail: [email protected] Tel: +43-1-4277-52341. Fax: +43-1-4277-9523. Figure 7. Sensor characteristics of LDL MIP in both PBS and LDL-free serum as running solutions, respectively

Notes The authors declare no competing financial interest.

ACKNOWLEDGMENT This project was supported by the Royal Thai Government through a Scholarship granted by the Office of the Higher Education Commission (Grant#04/2556), Thailand. We also gratefully acknowledge the Faculty of Medical Technology



1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

and the Faculty of Pharmacy, Prince of Songkla University for blood collection and ultracentrifugation.

REFERENCES (1) Barter, P. J.; Rye, K. A. Atherosclerosis 1996, 121, 1-12. (2) Matsunaga, T.; Koyama, I.; Hokari, S.; Komoda, T. J. Chromatogr. B 2002, 781, 331-343. (3) Brunzell, J. D.; Davidson, M.; Furberg, C. D.; Goldberg, R. B.; Howard, B. V.; Stein, J. H.; Witztum, J. L. J. Am. Coll. Cardiol. 2008, 51, 1512-1524. (4) Otvos, J. D.; Jeyarajah, E. J.; Cromwell, W. C. The American Journal of Cardiology 2002, 90, 22-29. (5) Caulfield, M. P.; Li, S.; Lee, G.; Blanche, P. J.; Salameh, W. A.; Benner, W. H.; Reitz, R. E.; Krauss, R. M. Clin. Chem. 2008, 54, 1307-1316. (6) Friedewald, W. T.; Levy, R. I.; Fredrickson, D. S. Clin. Chem. 1972, 18, 499-502. (7) Nauck, M.; Kramer Guth, A.; Bartens, W.; Marz, W.; Wieland, H.; Wanner, C. Clin. Nephrol. 1996, 46, 319-325. (8) Andrade, C. A. S.; Oliveira, M. D. L.; Faulin, T. E. S.; Hering, V. R.; Abdalla, D. S. P. In Biosensors for Health, Environment and Biosecurity, InTech, 2011; pp 215-240. (9) Nauck, M.; Warnick, G. R.; Rifai, N. Clin. Chem. 2002, 48, 236-254. (10) Chunta, S.; Promptmas, C.; Cherdchu, C. The International Journal of Applied Biomedical Engineering 2009, 2, 24-32. (11) Chunta, S.; Suk-Anake, J.; Chansiri, K.; Promptmas, C. Analyst 2014, 139, 4586-4592. (12) Poma, A.; Guerreiro, A.; Whitcombe, M. J.; Piletska, E. V.; Turner, A. P. F.; Piletsky, S. A. Adv. Funct. Mater. 2013, 23, 28212827. (13) Alenus, J.; Ethirajan, A.; Horemans, F.; Weustenraed, A.; Csipai, P.; Gruber, J.; Peeters, M.; Cleij, T. J.; Wagner, P. Anal. Bioanal. Chem. 2013, 405, 6479-6487. (14) Tse Sum Bui, B.; Haupt, K. Anal. Bioanal. Chem. 2010, 398, 2481-2492. (15) Algieri, C.; Drioli, E.; Guzzo, L.; Donato, L. Sensors 2014, 14, 13863-13912. (16) El Gohary, N. A.; Madbouly, A.; El Nashar, R. M.; Mizaikoff, B. Biosens. Bioelectron. 2015, 65, 108-114. (17) Wangchareansak, T.; Sangma, C.; Choowongkomon, K.; Dickert, F.; Lieberzeit, P. Anal. Bioanal. Chem. 2011, 400, 24992506. (18) Schirhagl, R.; Podlipna, D.; Lieberzeit, P. A.; Dickert, F. L. Chem. Commun. 2010, 46, 3128-3130. (19) Phan, N.; Sussitz, H.; Lieberzeit, P. Biosensors 2014, 4, 161171. (20) Schirhagl, R.; Hall, E. W.; Fuereder, I.; Zare, R. N. Analyst 2012, 137, 1495-1499. (21) Hayden, O.; Lieberzeit, P. A.; Blaas, D.; Dickert, F. L. Adv. Funct. Mater. 2006, 16, 1269-1278. (22) Hussain, M.; Wackerlig, J.; Lieberzeit, P. Biosensors 2013, 3, 89-107. (23) Jang, R.; Kim, K. H.; Zaidi, S. A.; Cheong, W. J.; Moon, M. H. Electrophoresis 2011, 32, 2167-2173. (24) Spizzirri, U. G.; Peppas, N. A. Chem. Mater. 2005, 17, 67196727. (25) Carlson, K. J. Clin. Pathol. 1973, 5, 32-37. (26) Chouinard, J. A.; Khalil, A.; Vermette, P. Microsc. Res. Tech. 2007, 70, 904-907. (27) Gan, C.; Ao, M.; Liu, Z.; Chen, Y. FEBS Open Bio 2015, 5, 276282. (28) Bajwa, S. Z.; Dumler, R.; Lieberzeit, P. A. Sens. Actuators, B 2014, 192, 522-528. (29) Rodahl, M.; Höök, F.; Kasemo, B. Anal. Chem. 1996, 68, 2219-2227.

Page 8 of 10

(30) Finley, P. R.; Schifman, R. B.; Williams, R. J.; Lichti, D. A. Clin. Chem. 1978, 24, 931-933. (31) Warnick, G. R.; Benderson, J.; Albers, J. J. Clin. Chem. 1982, 28, 1379-1388. (32) Baggiani, C.; Giovannoli, C.; Anfossi, L.; Passini, C.; Baravalle, P.; Giraudi, G. J. Am. Chem. Soc. 2012, 134, 1513-1518. (33) Tjernberg, A.; Markova, N.; Griffiths, W. J.; Hallén, D. J. Biomol. Screening 2006, 11, 131-137. (34) Pomorska, A.; Shchukin, D.; Hammond, R.; Cooper, M. A.; Grundmeier, G.; Johannsmann, D. Anal. Chem. 2010, 82, 22372242. (35) Lieberzeit, P. A.; Schirk, C.; Glanznig, G.; Gazda-Miarecka, S.; Bindeus, R.; Nannen, H.; Kauling, J.; Dickert, F. L. Superlattices Microstruct. 2004, 36, 133-142. (36) Dickert, F.; Hayden, O.; Lieberzeit, P.; Palfinger, C.; Pickert, D.; Wolff, U.; Scholl, G. Sens. Actuators, B 2003, 95, 20-24. (37) Matharu, Z.; Bandodkar, A. J.; Sumana, G.; Solanki, P. R.; Ekanayake, E. M. I. M.; Kaneto, K.; Gupta, V.; Malhotra, B. D. The Journal of Physical Chemistry B 2009, 113, 14405-14412. (38) Sugiuchi, H.; Irie, T.; Uji, Y.; Ueno, T.; Chaen, T.; Uekama, K.; Okabe, H. Clin. Chem. 1998, 44, 522-531. (39) Davis, P. G.; Wagganer, J. D. In Lipid metabolism and health; Moffatt, R. J.; Stamford, B., Eds.; CRC Press Taylor & Francis Group: Boca Raton, 2006; pp 47-60. (40) Chicea, D.; Chicea, R.; Chicea, L. M. Rom. Rep. Phys. 2013, 65, 178-185. (41) Steiner, M. J.; Skinner, A. C.; Perrin, E. M. Pediatrics 2011, 128, 463-470.


Page 9 of 10


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 ACS Paragon Plus Environment


1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60

Table of Contents artwork

“For TOC only”


Page 10 of 10

Low-Density Lipoprotein Sensor Based on Molecularly Imprinted

Recommend Documents