Biosensors and Bioelectronics 66 (2015) 169–176
Contents lists available at ScienceDirect
Biosensors and Bioelectronics journal homepage: www.elsevier.com/locate/bios
Graphene-based rapid and highly-sensitive immunoassay for C-reactive protein using a smartphone-based colorimetric reader Sandeep Kumar Vashist a,b,n, E. Marion Schneider c, Roland Zengerle a,b, Felix von Stetten a,b, John H.T. Luong d a
HSG-IMIT-Institut für Mikro- und Informationstechnik, Georges-Koehler-Allee 103, 79110 Freiburg, Germany Laboratory for MEMS Applications, Department of Microsystems Engineering-IMTEK, University of Freiburg, Georges-Koehler-Allee 103, 79110 Freiburg, Germany c Sektion Experimentelle Anaesthesiologie, University Hospital Ulm, Albert Einstein Allee 23, 89081 Ulm, Germany d Innovative Chromatography Group, Irish Separation Science Cluster (ISSC), Department of Chemistry and Analytical, Biological Chemistry Research Facility (ABCRF), University College Cork, Cork, Ireland b
art ic l e i nf o
a b s t r a c t
Article history: Received 23 September 2014 Received in revised form 7 November 2014 Accepted 12 November 2014 Available online 15 November 2014
A novel immunoassay (IA) has been developed for human C-reactive protein (CRP), an important biomarker and tissue preserving factor for infection and inflammation. Graphene nanoplatelets (GNP) and 3-aminopropyltriethoxysilane (APTES) were admixed and covalently attached to a polystyrene basedmicrotiter plate (MTP), pretreated with KOH. The resulting surface served as a stable layer for the covalent attachment of the anti-human CRP antibody. The IA procedure was based on the one-step kinetics-based sandwich IA employing a minimum number of process steps, whereas the enzymatic reaction solution was monitored by a smartphone-based colorimetric reader. With a limit of detection and a limit of quantification of 0.07 ng mL 1 and 0.9 ng mL 1, it precisely detected CRP spiked in diluted human whole blood and plasma as well as the CRP levels in clinical plasma samples. The results obtained for “real-world” patient samples agreed well with those of the conventional immunosorbent assay and the clinically-accredited analyzer-based IA. The antibody-bound GNP-functionalized MTPs retained its original activity after 6 weeks of storage in 0.1 M PBS, pH 7.4 at 4 °C. & 2014 Elsevier B.V. All rights reserved.
Keywords: Graphene C-reactive protein One-step kinetics-based sandwich immunoassay Smartphone-based colorimetric reader
1. Introduction As a member of a class of acute-phase reactants that mediates innate and adaptive immunity (Clyne and Olshaker, 1999; Kushner and Somerville, 1970; Vashist, 2013), CRP plays an important role in the host defense by binding to phosphocholine and related molecules on microorganisms. Indeed, CRP has been considered as a biomarker of inflammation rather than infection and extends its application for several clinical and pathological conditions (Table 1). The routine CRP measurement is important to identify states of inflammation and judge the efficacy of treatment intervention. In normal human sera, CRP ranges from 0.8 to 480 m g mL 1 (Vashist, 2013), compared to 10–40 mg mL 1 of the patients with mild inflammation and viral infection. For active inflammation and bacterial infections, and severe bacterial n Corresponding author at: HSG-IMIT-Institut für Mikroformationstechnik, Georges-Koehler-Allee 103, 79110 Freiburg, Fax: þ49 761 20373299. . E-mail address: [email protected] (S.K. Vashist).
http://dx.doi.org/10.1016/j.bios.2014.11.017 0956-5663/& 2014 Elsevier B.V. All rights reserved.
und InGermany.
infections, CRP levels are 40–200 mg mL 1, and 4 200 mg mL 1, respectively (Vashist, 2013). The normal (0.2–480 mg mL 1) and high sensitivity (0.08–80 mg mL 1) are the clinically-relevant CRP concentration ranges for neonatal sepsis. The high sensitive CRP IA is performed and followed by the normal CRP IA if the CRP level is above 80 mg mL 1. The differential diagnosis of neonatal sepsis is one of the most challenging diagnostic areas since elevated CRP may indicate both sterile inflammation and infection (Mukherjee et al., 2014; Mussap et al., 2011). Different analytical methods have been developed (Algarra et al., 2013) for CRP such as immunoturbidimetry (Deegan et al., 2003; Kjelgaard-Hansen et al., 2007), ELISA (Vashist et al., 2014a, 2014b, 2014c), surface plasmon resonance (Kim et al., 2008), chemiluminescence (Islam and Kang, 2011), impedimetry (Vermeeren et al., 2011), beads (Punyadeera et al., 2011), piezoresistive cantilevers (Lee et al., 2004), reflectometric interference spectroscopy (Choi et al., 2012), electrochemistry (Bryan et al., 2013) and microfluidics (Lee et al., 2011). Nevertheless, ELISA is most widely used in clinical settings due to its high precision and sensitivity.
170
S.K. Vashist et al. / Biosensors and Bioelectronics 66 (2015) 169–176
Table 1 Clinical relevance and significance of CRP. Clinical relevance of CRP
Remarks and references
Best inflammatory marker for clinical diagnosis An early indicator of infection or inflammation
Identified by American Heart Association/Center for Disease Control (Myers et al., 2004) Elevated in neonatal sepsis, meningitis, pancreatitis and bacteremia (Chiesa et al., 2011; Dollner et al., 2001; Nguyen-Vermillion et al., 2011) Higher serum CRP levels are associated with malignant diseases and Characterized by very high mortality rates in hospitalized patients (Chundadze et al., 2010) bacterial infections The strongest predictor of cardiovascular events Bassuk et al. (2004), Hung et al. (2008), Kuo et al., (2007), Park et al. (2012), Ridker, (2003), SicrasMainar et al., (2013), Tanaka et al., (2005) Cardiac tolerance and the development of diabetes Morimoto et al. (2013), Testa et al. (2008) A biomarker for atherosclerotic cardiovascular risk Tomiyama et al. (2005). Also associated with cardiorespiratory fitness (Kuo et al., 2007) Depression and posttraumatic stress Danner et al. (2003), Raison et al. (2006), Spitzer et al. (2010) Metabolic syndrome Lin et al. (2007) Alzheimer's disease Bassuk et al. (2004) Inflammasome related diseases Dinarello et al. (2012), Dinarello and van der Meer (2013)
This article unravels a rapid and highly-sensitive IA format using a smartphone-based colorimetric reader (SBCR) (Vashist et al., 2014c) for the detection of CRP in diluted human whole blood and plasma. The assay surface expansion is realized using graphene nanoplatelets (GNPs) for covalent attachment of the capture CRP antibody to provide remarkable signal enhancement. The analytical performance of this one-step kinetics-based sandwich IA procedure is compared with that of the conventional sandwich ELISA and the clinically-accredited analyzer-based IA.
2. Materials and methods 2.1. Materials APTES (purity 98%, w/v), Tween 20, and Nunc microwell 96well polystyrene plates (flat bottom, non-treated, and sterile) were procured from Sigma-Aldrich. Bovine serum albumin (BSA), bicinchoninic acid (BCA) protein assay kit and other remaining chemicals were purchased from Thermo Scientific. The human CRP Duoset ELISA kit (cat.# DY 1707E) components were purchased from RnD Systems. Human whole blood (HQ-Chex Level 2) and human serum were obtained from Streck and Biological Specialty, respectively. Recombinant human serum albumin (HSA), human fetuin A (HFA), human lipocalin 2 (LCN2), interleukin (IL)-1β, IL-6, IL-8 and tumor necrosis factor (TNF)-α were procured from RnD Systems. The clinically-accredited CRP IA was performed using the Roche COBASs 8000 modular analyzer. The anonymized EDTA plasma samples of patients were provided by the University Hospital Ulm, Germany. The binding and washing buffers were PBS, pH 7.4 with 0.1% BSA and PBS with 0.05% Tween 20 (PBST), respectively. The working aliquots of commercial lyophilized human CRP were prepared in 20 mM Tris–HCl, pH 8.0 with 0.1% BSA (as mentioned in the product brochure), while the CRP spiked samples were prepared by spiking various CRP concentrations in a fixed dilution (1:100) of human whole blood and plasma. The patient samples (0.3 to 81 mg mL 1 CRP or above) were diluted 1:1000 and 1:4000, respectively, to accommodate the detection range of the developed IA (0.03–81 ng mL 1). Both dilutions of 1:1000 and 1:4000 were employed for each of the unknown clinical samples, as per the standard sample preparation guidelines for the clinical ELISA. GNPs (1 mg) was mixed with 1 mL of 0.25% APTES followed by sonication for 1 h. The human CRP concentrations were prepared in BSA-preblocked sample vials to minimize the analyte loss due to non-specific surface binding (Dixit et al., 2011). The UPW and PBST washings were done five times with 300 mL of the respective solutions; blocking was done with 300 mL of 5% BSA; while 100 mL was taken for 1% KOH.
1-ethyl-3-(3-dimethylaminopropyl)-carbodiimide hydrochloride (EDC)-activated capture anti-human CRP Ab mixed with GNPs in 0.25% (v/v) APTES, human CRP, biotinylated anti-human CRP Ab conjugated HRP-labeled streptavidin, and 3,3,5,5-tetramethylbenzidine (TMB). The assay was performed at 37 °C while the absorbance was measured by an Infinite M200 Pro microplate reader from Tecan (Austria). The Ab was conjugated to SA-HRP by adding 1 mL of biotinylated anti-CRP antibody (0.5 mg mL 1) to 1 mL of SA-HRP to 2998 mL of the binding buffer followed by 20 min of incubation at room temperature (RT). The biotinylated anti-CRP Ab concentration was 0.17 mg mL 1, while SA-HRP dilution employed was 1:3000. Samsung Galaxy SIII Mini was used for colorimetric imaging whereas iPAD Mini provided screen-based bottom illumination (Vashist et al., 2014c). 2.2. Preparation of ab-bound GNPs-functionalized MTP The MTP wells treated with 1.0% (w/v) KOH at RT for 10 min generated the hydroxyl groups for binding to APTES (Fig. 1A). The EDC-activated anti-human CRP Ab was prepared by incubating 990 mL of Ab (4 mg mL 1) with 10 mL of EDC (4 mg mL 1 in 0.1 M MES buffer, pH 4.7) for 15 min at RT. The EDC-activated Ab was mixed with GNPs (1 mg/mL) in 0.25% APTES in the ratio of 1:1 (v/v). Thereafter, the anti-human CRP Ab solution (2 mg mL 1, 0.5 mg mL 1 GNPs and 0.125% APTES) was added to the MTP wells and incubated for 30 min at RT. The resulting wells were then washed five times with PBST. Subsequently, the MTP wells were preblocked by incubating with 5% BSA for 30 min at 37 °C and washed five times with PBST. The anti-human CRP Ab-bound and BSA-preblocked MTPs were stored in 0.1 M PBS, pH 7.4 at 4 °C for 6 weeks to determine their functional stability. 2.3. Graphene-based Human CRP IA The anti-human CRP Ab-bound and BSA-preblocked MTP was provided with biotinylated anti-human CRP Ab preconjugated to SA-HRP. Human CRP (0.03–81 ng mL 1) was provided in buffer, diluted human plasma or diluted human whole blood. After 15 min of incubation at 37 °C, the MTPs were washed with PBS, added with TMB. After 4 min, the enzyme–substrate reaction was stopped by adding 50 mL of 2N H2SO4. The image of the colorimetric solution was captured by the smartphone. The absorbance of the resulting solution was also measured with a MTP reader (MTPR) at 450 nm with reference at 540 nm. All experiments were performed in triplicate with the reading of the blank (without human CRP) subtracted from all assay values. The LOD (limit of detection) is determined by 3 (standard deviation of the blank/ slope), while LOQ (limit of quantitation) is 10 (standard
S.K. Vashist et al. / Biosensors and Bioelectronics 66 (2015) 169–176
171
Fig. 1. Schematics of (A) the bioanalytical procedure for the development of Ab-bound and BSA-blocked graphene-functionalized MTP, and (B) the developed graphenebased human CRP IA.
S.K. Vashist et al. / Biosensors and Bioelectronics 66 (2015) 169–176
deviation of the blank/slope) (Vashist et al., 2014c). Various controls were also employed to determine the efficiency of BSA blocking and plausible non-specific interactions caused other biomarkers. The conventional sandwich ELISA and the clinicallyaccredited analyzer-based IA were performed using the manufacturer's guidelines. 2.4. SBCR analysis The SBCR analysis has been described by (Vashist et al., 2014c) with detailed information provided in Figs. S1–S3. The pixel intensity (PI) of the capture images was determined by Image J 1.48v (http://imagej.nih.gov/ij/) to extract the PI of blue channel images. The algorithm determines the PI of the center of each MTP well and calculates the desired PI of each MTP well as (255-PITest Conc.–PIBlank). Detailed information of the set-up and the image processing algorithm is provided in Fig. S1.
3. Results and discussion 3.1. Graphene-based human CRP IA The assessment of human CRP is of critical importance in the management of infectious and inflammatory conditions (Marnell et al., 2005), neonatal sepsis (Chiesa et al., 2011, 2004; Dollner et al., 2001; Nguyen-Vermillion et al., 2011; Tappero and Johnson, 2010), depressive and posttraumatic stress (Danner et al., 2003; Raison et al., 2006; Spitzer et al., 2010), diabetes (Lin et al., 2007; Morimoto et al., 2013; Testa et al., 2008) and cardiovascular diseases (Kuo et al., 2007; Park et al., 2012; Ridker, 2003; SicrasMainar et al., 2013; Wilson et al., 2006). The developed IA employs a highly-simplified procedure that enables the precise detection of human CRP in less than 30 min (Fig. 1B). After the hydroxylation of the polystyrene-based MTP with KOH (Park and Jung, 2002), the GNP–APTES complex is covalently attached to the treated surface owing to the binding of the alkoxy groups of APTES with the hydroxyl groups present on GNPs and the KOH-treated MTP surface via a hydrolysis-dependent reaction (Liedberg et al., 1998). This occurs simultaneously with the covalent binding of EDC-activated capture Ab to the amino groups on APTES by the formation of an amide bond. However, for the characterization of each process step by FTIR, polystyrene beads (PS) beads and IgG were selected to model the MTP well and antibody in the ATR-FTIR characterization study. As shown in Fig. 2, the hydroxylation of PS with KOH shows very little changes in the FTIR in comparison to untreated KOH. Peaks centered around 1632 cm 1 are associated with the skeletal vibrations of the graphitic sheet of GNPs (Lam et al., 2012) and those at 1154 cm 1 indicates the polymerization of APTES (Si‒O‒Si) bond confirming the functionalization of GNPs with APTES on the PS surface (Vashist et al., 2014b). Two amide bands at 1550 and 1645 cm 1 confirm the presence of EDC-activated IgG on the PS surface. The binding of Ab to APTES via EDC activation is highly specific (Vashist, 2012; Vashist et al., 2014d) with minimal non-specific reaction products. The resulting MTP was blocked by incubating with 5% BSA for 30 min. Thereafter, the procedure involved sequentially the dispensing of human CRP and biotinylated antihuman CRP Ab conjugated to SA-HRP into the MTP wells, resulting in a sandwich immune complex after 15 min of incubation at 37 °C (Fig. 1B). This was followed by washing with PBST, incubating with TMB for 4 min, stopping the enzyme–substrate reaction and performing the colorimetric readout (Fig. 3A) and analysis (Figs. S1– S3).
1645 cm-1
1550 cm-1
1154 cm-1
Intensity / a.u.
172
KOH-treated PS PS+GNP+APTES PS+GNP+APTES+Ab 2000
1800
1600
1400
1200
1000
Wavenumber / cm-1 Fig. 2. ATR-FTIR spectra pertaining to the immobilization of graphene nanoplatelets (GNP), APTES and IgG-EDC on polystyrene (PS) surface.
The developed IA exhibits a dynamic range of 0.03–81 ng mL 1 with linearity from 0.3 to 81 ng mL 1 for human CRP (Fig. 3B). The LOD, LOQ, and EC50 values were determined as 0.07 ng mL 1, 0.9 ng mL 1, and 6.6 ng mL 1 respectively (Table S1). The procedure detected the entire pathophysiological range of human CRP (0.08–480 mg mL 1) in spiked clinical samples after appropriate dilution (Fig. 3B). However, LOD and LOQ obtained for such spiked samples 0.07 ng mL 1 and 1.0 ng mL 1, and 0.08 ng mL 1 and 1.1 ng mL 1, respectively, implying a slight effect due to the sample matrices. The intraday and interday variability (five assay repeats in triplicate) in a single day and on five consecutive days, respectively, were 2.3–9.3 and 3.1–14.6, respectively. The procedure was not affected by immunological reagents used in the IA, as shown by various experimental process controls (Fig. 3C). LCN2, HFA, HSA, IL-1β, IL-6, IL-8 and TNF-α were also tested for any plausible interference. They are elevated along with CRP in patients with infections and other disorders. The pathophysiologically high concentration levels of these biomarkers were employed taking into account the dilution factor of 1:1000 used for the dilution of human whole blood or serum in the developed IA. These non-specific biomarkers provoked no interference in the determination of CRP as shown in Fig. 3C. The developed IA has a 14-fold reduced IA duration than the conventional sandwich IA procedure used in the commercial human CRP Duoset ELISA kit. Such a result could be attributed to the rapid covalent immobilization procedure and the use of the onestep kinetics-based IA format with minimal process steps (Vashist et al., 2014a). The immobilization procedure is 18-fold faster than passive adsorption (used in the commercial kit), whereas its IA duration (excluding Ab immobilization) is 20-fold shorter than that of the conventional IA (used in the commercial kit) (Tables S1–S2). The preparation of the Ab-bound and BSA-preblocked MTP takes only 1 h and 15 min, which is significantly shorter than the procedure using magnetic beads (Vashist et al., 2014a). The Ab binding to the magnetic beads takes about two days using the protocol provided by the provider. Therefore, the graphene-based CRP IA exhibited a 6-fold lower LOD and a wider dynamic range compared to the magnetic bead-based CRP IA. The improvement might be due to the very large surface area of GNPs to accommodate a significantly higher binding of the capture Ab (Fig. S4), which in turn results in significantly enhanced analyte detection and analytical sensitivity. GNPs were employed in the developed IA due to their much larger surface area, cost-effectiveness and commercial-availability in bulk with desired purity.
S.K. Vashist et al. / Biosensors and Bioelectronics 66 (2015) 169–176
173
Fig. 3. (A) SBCR employed for the developed graphene-based human CRP IA. Reproduced with permission from Elsevier B.V. (Vashist et al., 2014c) (B) Detection of CRP in PBS (10 mM, pH 7.4), diluted human plasma and diluted human whole blood. (C) Experimental process controls being employed to evaluate the efficiency of blocking and nonspecific interactions between immunoassay components, and non-specific interactions with other biomarkers. Anti-CRP1 and anti-CRP2 are capture and detection antibodies, respectively. The control proteins employed are human serum albumin (HSA), human lipocalin-2 (LCN2), human fetuin A (HFA), interleukin (IL)-1 beta, IL-6, IL-8 and tumor necrosis factor (TNF)-alpha. (D) Storage stability of anti-human CRP capture Ab-bound graphene-functionalized MTPs stored at 4 °C in PBS for 6 weeks. All experiments were performed in triplicate with the error bars representing the standard deviation.
They provide higher signal enhancement with superior bioanalytical performance than other materials, such as polystyrene beads, magnetic dynabeads, agarose and multiwalled carbon nanotubes (unpublished results). The developed IA procedure employs a washing step after the immune complex formation to counteract non-specificity, which makes it challenging to adapt to point-ofcare (POC) IA formats. Indeed, this procedure has been implemented in our inexpensive and disposable EasyELISA and fullyautomated centrifugal microfluidics-based LabDisk platforms (unpublished results). The sample pretreatment by appropriate dilution, as employed in our developed IA, is another challenge for POC applications. However, it is still a necessary step for the assay using commercial POC immunodiagnostic kits. The clinical analysis of CRP is usually carried out by latex agglutination or nephelometry based assays with lower detection ranges (in mg mL 1), similar to that of phosphocholine and Ophosphorylethanolamine based CRP IAs. Other IAs based on
electrochemical detection, surface plasmon resonance, nanoparticles, nanocomposites, chemiluminescence, total internal reflection and micromosiac IAs enhance sensitivities down to ng mL 1. A dual signal enhancement approach with magnetic nanoparticles enables the detection of CRP from 1.18 ng mL 1 to 11.8 mg mL 1 but requires several hours (Yang et al., 2014). The three-line lateral flow strip format has improved detection range of 1 ng mL 1–500 mg mL 1 but takes ∼2 days to prepare a strip (Kyoung Oh et al., 2014). A chemiluminescence resonance energy transfer-based IA using graphene detects CRP in the range of 1.6–100 ng mL 1 but the overall preparation and analysis time exceeds several hours (Lee et al., 2012). However, despite the large number of IAs demonstrated, only very few have shown the precision similar to clinically-accredited and commercial IAs, real sample trials, potential for whole blood analysis, desired high reproducibility and stability. Our novel IA procedure has clearly demonstrated all these important bioanalytical features, thereby
174
S.K. Vashist et al. / Biosensors and Bioelectronics 66 (2015) 169–176
Table 2 Determination of CRP in the EDTA plasma samples of patients by the clinically-accredited Roche COBASs 8000 modular analyzer-based IA, the developed graphene-based IA and the conventional sandwich ELISA-based IA (as used in the commercial human CRP Duoset ELISA kit). Samples
1 2 3 4 5 6 7 8 9 10 11 12
Clinically-accredited
Developed
Determined conc. (ng mL 1)
Determined conc. (ng mL 1)
Percentage recovery
Determined conc. (ng mL 1)
Percentage recovery
4.1 10.1 14.1 35.4 45.7 63.7 80.4 128.7 228.3 251.3 270.7 301.8
4.8 10.5 14.6 34.8 45.2 63.1 81.7 130.2 225.2 247.8 275.2 296.1
117.1 104.0 103.5 98.3 98.9 99.1 101.6 101.2 98.6 98.6 101.7 98.1
4.5 10.9 14.7 35.0 44.9 62.8 82.2 132.1 226.6 246.9 277.6 295.7
109.7 107.9 104.2 98.9 98.2 98.6 102.2 102.6 99.2 99.6 102.5 98.0
showing its immense potential for clinical diagnostics. It detects CRP in the range of pg mL 1 to ng mL 1, which is significantly lower than most of the analytical techniques used for the determination of CRP in blood (Algarra et al., 2013). Moreover, it is a
Conventional
highly-simplified and cost-effective IA procedure with the analysis time of just 30 min. It detects the complete clinically-relevant CRP range in serum and whole blood with remarkable bioanalytical performance.
Fig. 4. (A) Correlation of the developed graphene-based human CRP IA using SBCR with the conventional human CRP IA using MTPR for the detection of CRP-spiked in diluted human whole blood. (B) Detection of CRP in PBS (0.1 M, pH 7.4) by the developed graphene-based IA based on the use of different smartphone models. (C) Determination of the CRP concentration in the unknown EDTA plasma samples of patients by the developed graphene-based IA using various smartphone models based on their specific calibration curves. (D) An overlay plot of the developed graphene-based human CRP IAs based on the use of SBCR and MTPR, and employing the normalized signals (the signal response obtained for a particular human CRP concentration divided by the signal response obtained for the highest human CRP concentration).
S.K. Vashist et al. / Biosensors and Bioelectronics 66 (2015) 169–176
3.2. Determination of bioanalytical parameters The results obtained by the developed IA for anonymized EDTA plasma samples of patients compared well with the clinically-accredited Roche COBASs 8000 modular analyzer-based IA and the conventional sandwich ELISA (Table 2). The percentage recoveries of CRP by the developed IA, using the CRP values determined by the clinically-accredited IA as standard, were in the range of 98.1– 117.0, in agreement with the conventional sandwich ELISA (98.0– 109.7). Similarly, the CRP level spiked in diluted human whole blood obtained by the developed IA was comparable to that of the conventional sandwich ELISA (Fig. 4A). The determination of CRP concentrations in unknown EDTA plasma samples of patients by the developed IA also agreed well with the conventional and clinically-accredited analyzer-based IAs with percentage recoveries in the range of 93.7–103.4 (Table S3). The anti-human CRP Ab-bound and BSA-blocked MTPs, stored at 4 °C in 0.1 M PBS, pH 7.4, retained their functional activity for up to 6 weeks (Fig. 3D), attesting the leach-proof covalent binding of the capture Ab and its high functional stability. This makes the developed IA ideal for clinical diagnostics, where the Ab-bound MTPs are routinely used on a large scale and need to be stored for up to 4 weeks to facilitate rapid CRP detection. The developed Ab immobilization procedure was highly reproducible as there was no obvious batch-to-batch variability for various preparations of GNP-anti-CRP Ab when tested for the detection of 9 ng mL 1 CRP (Fig. S5). The effect of different smartphone models on the performance of the developed IA was evaluated by employing iPhone 4 and iPhone 5s apart from Samsung Galaxy SIII mini. As expected, the signals for different CRP concentrations in the developed IA were varying for the various smartphone models due to their different camera specifications (Fig. 4B). However, the determination of CRP in the unknown EDTA plasma samples of patients by each smartphone model was highly reproducible by using its respective calibration curve (Fig. 4C). There was no difference in the bioanalytical performance of the IA using SBCR or MTPR, as shown by the overlay plot of the normalized signals obtained from both readers (Fig. 4D). Apparently, SBCR is an ideal and cost-effective device for POC colorimetric detection of IAs. The overall set-up is ∼US$ 660; consisting of a Samsung Galaxy SIII mini (US$ 243), iPAD mini (US$ 399), and an inexpensive dark hood and a base holder assembly (US$ 20). It is ∼30-fold cheaper than the standard MTPR that costs ∼US$ 20,000. In addition, as almost everyone has a gadget and a smartphone nowadays, the use of SBCR will incur negligible cost to the end-users. Moreover, smartphones have become ubiquitous and are equipped with all the advanced features desired for mobile and personalized healthcare monitoring and management. These features include high resolution imaging, secure data storage, wireless connectivity, cloud computing, text alerts, spatiotemporal mapping and capabilities for robust analysis via an integrated smart application. Our ongoing efforts are dedicated to the development of a smart application for automated image analysis on the smartphone itself, controlling the smartphone's camera exposure and illumination of the iPAD mini's screen, and the development of a more compact SBCR. The ongoing development in smartphone technologies, cloud computing and software will further facilitate the improvement of the SBCR's performance.
4. Conclusion The developed IA employs inexpensive GNPs to provide signal enhancement based on the high loading of the capture Ab due to the significantly increased surface area of GNPs. The use of our
175
recently developed low-cost SBCR (Vashist et al., 2014c) for colorimetric readout instead of the expensive conventional MTPR makes it an ideal rapid IA format for decentralized and remote settings. The use of the ubiquitous smartphone in SBCR further provides all the advanced features that are desired for mobile and personalized healthcare monitoring and management (Vashist et al., 2014e). Especially in the case of newborns and children with inflammasome-associated diseases, the monitoring of CRP may be extremely helpful to predict a fever episode and prevent or at least attenuate the disease severity. Moreover, continuous monitoring of metabolic diseases may be of high relevance in a homecare setting (Gugapriya et al., 2014; Tsimikas et al., 2006). Overall, the high sensitivity assay format opens a new field of application in hospital-independent health management using CRP as a marker for sterile inflammation and metabolic stress.
Acknowledgments We thank PD Dr. Eberhard Barth for anonymizing and providing us the left-over EDTA plasma samples of patients treated by the intensive care unit at the University Hospital Ulm, Germany to validate the developed CRP immunoassay. We also thank Dr. Edmond Lam of the National Research Council Canada in Montreal for conducting the FTIR experiment.
Appendix A. Supplementary materials Supplementary data associated with this article can be found in the online version at http://dx.doi.org/10.1016/j.bios.2014.11.017.
References Algarra, M., Gomes, D., Esteves da Silva, J.C., 2013. Clin. Chim. Acta 415, 1–9. Bassuk, S.S., Rifai, N., Ridker, P.M., 2004. Curr. Probl. Cardiol. 29 (8), 439–493. Bryan, T., Luo, X., Bueno, P.R., Davis, J.J., 2013. Biosens. Bioelectron. 39 (1), 94–98. Chiesa, C., Natale, F., Pascone, R., Osborn, J.F., Pacifico, L., Bonci, E., De Curtis, M., 2011. Clin. Chim. Acta 412 (11–12), 1053–1059. Chiesa, C., Panero, A., Osborn, J.F., Simonetti, A.F., Pacifico, L., 2004. Clin. Chem. 50 (2), 279–287. Choi, H.W., Sakata, Y., Kurihara, Y., Ooya, T., Takeuchi, T., 2012. Anal. Chim. Acta 728, 64–68. Chundadze, T., Steinvil, A., Finn, T., Saranga, H., Guzner-Gur, H., Berliner, S., Justo, D., Paran, Y., 2010. Clin. Biochem. 43 (13–14), 1060–1063. Clyne, B., Olshaker, J.S., 1999. J. Emerg. Med. 17 (6), 1019–1025. Danner, M., Kasl, S.V., Abramson, J.L., Vaccarino, V., 2003. Psychosom. Med. 65 (3), 347–356. Deegan, O., Walshe, K., Kavanagh, K., Doyle, S., 2003. Anal. Biochem. 312 (2), 175–181. Dinarello, C.A., Simon, A., van der Meer, J.W., 2012. Nat. Rev. Drug Discov. 11 (8), 633–652. Dinarello, C.A., van der Meer, J.W., 2013. Semin. Immunol. 25 (6), 469–484. Dixit, C.K., Vashist, S.K., MacCraith, B.D., O'Kennedy, R., 2011. Analyst 136 (7), 1406–1411. Dollner, H., Vatten, L., Austgulen, R., 2001. J. Clin. Epidemiol. 54 (12), 1251–1257. Gugapriya, T., Karthick, S., Nagarjuna, B., 2014. J. Clin. Diagn. Res. 8 (4), CC01–CC04. Hung, J., Knuiman, M.W., Divitini, M.L., Langton, P.E., Chapman, C.L., Beilby, J.P., 2008. C-reactive protein and interleukin-18 levels in relation to coronary heart disease: prospective cohort study from Busselton Western Australia. Heart Lung Circ. 17 (2), 90–95. Islam, M.S., Kang, S.H., 2011. Talanta 84 (3), 752–758. Kim, H.C., Lee, S.K., Jeon, W.B., Lyu, H.K., Lee, S.W., Jeong, S.W., 2008. Ultramicroscopy 108 (10), 1379–1383. Kjelgaard-Hansen, M., Martinez-Subiela, S., Petersen, H.H., Jensen, A.L., Ceron, J.J., 2007. Vet. J. 173 (3), 571–577. Kuo, H.K., Yen, C.J., Chen, J.H., Yu, Y.H., Bean, J.F., 2007. Int. J. Cardiol. 114 (1), 28–33. Kushner, I., Somerville, J.A., 1970. Biochim. Biophys. Acta 207 (1), 105–114. Kyoung Oh, Y., Joung, H.-A., Soo Han, H., Suk, H.-J., Kim, M.-G., 2014. Biosens. Bioelectron. 61, 285–289. Lam, E., Chong, J.H., Majid, E., Liu, Y., Hrapovic, S., Leung, A.C., Luong, J.H.T., 2012. Carbon 50 (3), 1033–1043. Lee, J.H., Yoon, K.H., Hwang, K.S., Park, J., Ahn, S., Kim, T.S., 2004. Biosens. Bioelectron. 20 (2), 269–275.
176
S.K. Vashist et al. / Biosensors and Bioelectronics 66 (2015) 169–176
Lee, J.S., Joung, H.A., Kim, M.G., Park, C.B., 2012. ACS Nano 6 (4), 2978–2983. Lee, W.B., Chen, Y.H., Lin, H.I., Shiesh, S.C., Lee, G.B., 2011. Sens. Actuators B – Chem. 157 (2), 710–721. Liedberg, B., Cooper, J., Cass, T., Ligler, F., 1998. Immobilized Biomolecules in Analysis: A Practical Approach. Oxford University Press, Oxford, UK. Lin, M.S., Shih, S.R., Li, H.Y., Chuang, L.M., Chang, T.C., Wei, J.N., Shieh, G.J., Chien, Y.F., Hua, C.H., Hwang, J.J., 2007. Diabetes Res. Clin. Pract. 77 (2), 286–292. Marnell, L., Mold, C., Du Clos, T.W., 2005. Clin. Immunol. 117 (2), 104–111. Morimoto, H., Sakata, K., Oishi, M., Tanaka, K., Nakada, S., Nogawa, K., Suwazono, Y., 2013. Diabetes Metab. 39 (1), 27–33. Mukherjee, A., Davidson, L., Anguvaa, L., Duffy, D.A., Kennea, N., 2014. Arch. Dis. Child Fetal Neonatal Ed. , http://dx.doi.org/10.1136/archdischild-2014-306349. Mussap, M., Noto, A., Fravega, M., Fanos, V., 2011. J. Matern. Fetal Neonatal Med. 24 (S2), 12–14. Myers, G.L., Rifai, N., Tracy, R.P., Roberts, W.L., Alexander, R.W., Biasucci, L.M., Catravas, J.D., Cole, T.G., Cooper, G.R., Khan, B.V., Kimberly, M.M., Stein, E.A., Taubert, K.A., Warnick, G.R., Waymack, P.P., CDC, AHA, 2004. Circulation 110 (25), e545–549. Nguyen-Vermillion, A., Juul, S.E., McPherson, R.J., Ledbetter, D.J., 2011. J. Pediatr. 159 (1), 121–126. Park, H.E., Cho, G.Y., Chun, E.J., Choi, S.I., Lee, S.P., Kim, H.K., Youn, T.J., Kim, Y.J., Choi, D.J., Sohn, D.W., Oh, B.H., Park, Y.B., 2012. Atherosclerosis 224 (1), 201–207. Park, S.J., Jung, W.Y., 2002. J. Colloid Interface Sci. 250 (1), 93–98. Punyadeera, C., Dimeski, G., Kostner, K., Beyerlein, P., Cooper-White, J., 2011. J. Immunol. Methods 373 (1-2), 19–25. Raison, C.L., Capuron, L., Miller, A.H., 2006. Trends Immunol. 27 (1), 24–31. Ridker, P.M., 2003. Am. J. Cardiol. 92 (4B), 17K–22K. Sicras-Mainar, A., Rejas-Gutierrez, J., Navarro-Artieda, R., Blanca-Tamayo, M., 2013. Eur. Psychiat. 28 (3), 161–167.
Spitzer, C., Barnow, S., Volzke, H., Wallaschofski, H., John, U., Freyberger, H.J., Lowe, B., Grabe, H.J., 2010. J. Psychiatr. Res. 44 (1), 15–21. Tanaka, A., Shimada, K., Sano, T., Namba, M., Sakamoto, T., Nishida, Y., Kawarabayashi, T., Fukuda, D., Yoshikawa, J., 2005. J. Am. Coll. Cardiol. 45 (10), 1594–1599. Tappero, E., Johnson, P., 2010. Newborn Infant Nurs. Rev. 10 (4), 209–217. Testa, R., Bonfigli, A.R., Sirolla, C., Marra, M., Boemi, M., Mari, D., Sacchi, E., Dolci, A., Catalano, A., Procopio, A., Ceriello, A., 2008. Nutr. Metab. Cardiovasc. Dis. 18 (3), 220–226. Tomiyama, H., Okazaki, R., Koji, Y., Usui, Y., Hayashi, T., Hori, S., Yamashina, A., 2005. Atherosclerosis 178 (1), 187–192. Tsimikas, S., Willerson, J.T., Ridker, P.M., 2006. J. Am. Coll. Cardiol. 47 (S8), C19–31. Vashist, S.K., 2012. Diagnostics 2 (3), 23–33. Vashist, S.K., 2013. J. Basic Appl. Sci. 9, 496–499. Vashist, S.K., Czilwik, G., van Oordt, T., von Stetten, F., Zengerle, R., Marion Schneider, E., Luong, J.H.T., 2014a. Anal. Biochem. 456, 32–37. Vashist, S.K., Marion Schneider, E., Lam, E., Hrapovic, S., Luong, J.H.T., 2014b. Sci. Rep. 4, 4407. Vashist, S.K., van Oordt, T., Schneider, E.M., Zengerle, R., von Stetten, F., Luong, J.H.T., 2014c. Biosens. Bioelectron. , http://dx.doi.org/10.1016/j.bios.2014.08.027. Vashist, S.K., Lam, E., Hrapovic, S., Male, K.B., Luong, J.H.T., 2014d. Chem. Rev. 114 (21), 11083–11130. Vashist, S.K., Schneider, E.M., Luong, J.H.T., 2014e. Diagnostics 4 (3), 104–128. Vermeeren, V., Grieten, L., Vanden Bon, N., Bijnens, N., Wenmackers, S., Janssens, S. D., Haenen, K., Wagner, P., Michiels, L., 2011. Sens. Actuat. B – Chem. 157 (1), 130–138. Wilson, A.M., Ryan, M.C., Boyle, A.J., 2006. Int. J. Cardiol. 106 (3), 291–297. Yang, S.F., Gao, B.Z., Tsai, H.Y., Fuh, C.B., 2014. Analyst 139 (21), 5576–5581.
Contents lists available at ScienceDirect
Biosensors and Bioelectronics journal homepage: www.elsevier.com/locate/bios
Graphene-based rapid and highly-sensitive immunoassay for C-reactive protein using a smartphone-based colorimetric reader Sandeep Kumar Vashist a,b,n, E. Marion Schneider c, Roland Zengerle a,b, Felix von Stetten a,b, John H.T. Luong d a
HSG-IMIT-Institut für Mikro- und Informationstechnik, Georges-Koehler-Allee 103, 79110 Freiburg, Germany Laboratory for MEMS Applications, Department of Microsystems Engineering-IMTEK, University of Freiburg, Georges-Koehler-Allee 103, 79110 Freiburg, Germany c Sektion Experimentelle Anaesthesiologie, University Hospital Ulm, Albert Einstein Allee 23, 89081 Ulm, Germany d Innovative Chromatography Group, Irish Separation Science Cluster (ISSC), Department of Chemistry and Analytical, Biological Chemistry Research Facility (ABCRF), University College Cork, Cork, Ireland b
art ic l e i nf o
a b s t r a c t
Article history: Received 23 September 2014 Received in revised form 7 November 2014 Accepted 12 November 2014 Available online 15 November 2014
A novel immunoassay (IA) has been developed for human C-reactive protein (CRP), an important biomarker and tissue preserving factor for infection and inflammation. Graphene nanoplatelets (GNP) and 3-aminopropyltriethoxysilane (APTES) were admixed and covalently attached to a polystyrene basedmicrotiter plate (MTP), pretreated with KOH. The resulting surface served as a stable layer for the covalent attachment of the anti-human CRP antibody. The IA procedure was based on the one-step kinetics-based sandwich IA employing a minimum number of process steps, whereas the enzymatic reaction solution was monitored by a smartphone-based colorimetric reader. With a limit of detection and a limit of quantification of 0.07 ng mL 1 and 0.9 ng mL 1, it precisely detected CRP spiked in diluted human whole blood and plasma as well as the CRP levels in clinical plasma samples. The results obtained for “real-world” patient samples agreed well with those of the conventional immunosorbent assay and the clinically-accredited analyzer-based IA. The antibody-bound GNP-functionalized MTPs retained its original activity after 6 weeks of storage in 0.1 M PBS, pH 7.4 at 4 °C. & 2014 Elsevier B.V. All rights reserved.
Keywords: Graphene C-reactive protein One-step kinetics-based sandwich immunoassay Smartphone-based colorimetric reader
1. Introduction As a member of a class of acute-phase reactants that mediates innate and adaptive immunity (Clyne and Olshaker, 1999; Kushner and Somerville, 1970; Vashist, 2013), CRP plays an important role in the host defense by binding to phosphocholine and related molecules on microorganisms. Indeed, CRP has been considered as a biomarker of inflammation rather than infection and extends its application for several clinical and pathological conditions (Table 1). The routine CRP measurement is important to identify states of inflammation and judge the efficacy of treatment intervention. In normal human sera, CRP ranges from 0.8 to 480 m g mL 1 (Vashist, 2013), compared to 10–40 mg mL 1 of the patients with mild inflammation and viral infection. For active inflammation and bacterial infections, and severe bacterial n Corresponding author at: HSG-IMIT-Institut für Mikroformationstechnik, Georges-Koehler-Allee 103, 79110 Freiburg, Fax: þ49 761 20373299. . E-mail address: [email protected] (S.K. Vashist).
http://dx.doi.org/10.1016/j.bios.2014.11.017 0956-5663/& 2014 Elsevier B.V. All rights reserved.
und InGermany.
infections, CRP levels are 40–200 mg mL 1, and 4 200 mg mL 1, respectively (Vashist, 2013). The normal (0.2–480 mg mL 1) and high sensitivity (0.08–80 mg mL 1) are the clinically-relevant CRP concentration ranges for neonatal sepsis. The high sensitive CRP IA is performed and followed by the normal CRP IA if the CRP level is above 80 mg mL 1. The differential diagnosis of neonatal sepsis is one of the most challenging diagnostic areas since elevated CRP may indicate both sterile inflammation and infection (Mukherjee et al., 2014; Mussap et al., 2011). Different analytical methods have been developed (Algarra et al., 2013) for CRP such as immunoturbidimetry (Deegan et al., 2003; Kjelgaard-Hansen et al., 2007), ELISA (Vashist et al., 2014a, 2014b, 2014c), surface plasmon resonance (Kim et al., 2008), chemiluminescence (Islam and Kang, 2011), impedimetry (Vermeeren et al., 2011), beads (Punyadeera et al., 2011), piezoresistive cantilevers (Lee et al., 2004), reflectometric interference spectroscopy (Choi et al., 2012), electrochemistry (Bryan et al., 2013) and microfluidics (Lee et al., 2011). Nevertheless, ELISA is most widely used in clinical settings due to its high precision and sensitivity.
170
S.K. Vashist et al. / Biosensors and Bioelectronics 66 (2015) 169–176
Table 1 Clinical relevance and significance of CRP. Clinical relevance of CRP
Remarks and references
Best inflammatory marker for clinical diagnosis An early indicator of infection or inflammation
Identified by American Heart Association/Center for Disease Control (Myers et al., 2004) Elevated in neonatal sepsis, meningitis, pancreatitis and bacteremia (Chiesa et al., 2011; Dollner et al., 2001; Nguyen-Vermillion et al., 2011) Higher serum CRP levels are associated with malignant diseases and Characterized by very high mortality rates in hospitalized patients (Chundadze et al., 2010) bacterial infections The strongest predictor of cardiovascular events Bassuk et al. (2004), Hung et al. (2008), Kuo et al., (2007), Park et al. (2012), Ridker, (2003), SicrasMainar et al., (2013), Tanaka et al., (2005) Cardiac tolerance and the development of diabetes Morimoto et al. (2013), Testa et al. (2008) A biomarker for atherosclerotic cardiovascular risk Tomiyama et al. (2005). Also associated with cardiorespiratory fitness (Kuo et al., 2007) Depression and posttraumatic stress Danner et al. (2003), Raison et al. (2006), Spitzer et al. (2010) Metabolic syndrome Lin et al. (2007) Alzheimer's disease Bassuk et al. (2004) Inflammasome related diseases Dinarello et al. (2012), Dinarello and van der Meer (2013)
This article unravels a rapid and highly-sensitive IA format using a smartphone-based colorimetric reader (SBCR) (Vashist et al., 2014c) for the detection of CRP in diluted human whole blood and plasma. The assay surface expansion is realized using graphene nanoplatelets (GNPs) for covalent attachment of the capture CRP antibody to provide remarkable signal enhancement. The analytical performance of this one-step kinetics-based sandwich IA procedure is compared with that of the conventional sandwich ELISA and the clinically-accredited analyzer-based IA.
2. Materials and methods 2.1. Materials APTES (purity 98%, w/v), Tween 20, and Nunc microwell 96well polystyrene plates (flat bottom, non-treated, and sterile) were procured from Sigma-Aldrich. Bovine serum albumin (BSA), bicinchoninic acid (BCA) protein assay kit and other remaining chemicals were purchased from Thermo Scientific. The human CRP Duoset ELISA kit (cat.# DY 1707E) components were purchased from RnD Systems. Human whole blood (HQ-Chex Level 2) and human serum were obtained from Streck and Biological Specialty, respectively. Recombinant human serum albumin (HSA), human fetuin A (HFA), human lipocalin 2 (LCN2), interleukin (IL)-1β, IL-6, IL-8 and tumor necrosis factor (TNF)-α were procured from RnD Systems. The clinically-accredited CRP IA was performed using the Roche COBASs 8000 modular analyzer. The anonymized EDTA plasma samples of patients were provided by the University Hospital Ulm, Germany. The binding and washing buffers were PBS, pH 7.4 with 0.1% BSA and PBS with 0.05% Tween 20 (PBST), respectively. The working aliquots of commercial lyophilized human CRP were prepared in 20 mM Tris–HCl, pH 8.0 with 0.1% BSA (as mentioned in the product brochure), while the CRP spiked samples were prepared by spiking various CRP concentrations in a fixed dilution (1:100) of human whole blood and plasma. The patient samples (0.3 to 81 mg mL 1 CRP or above) were diluted 1:1000 and 1:4000, respectively, to accommodate the detection range of the developed IA (0.03–81 ng mL 1). Both dilutions of 1:1000 and 1:4000 were employed for each of the unknown clinical samples, as per the standard sample preparation guidelines for the clinical ELISA. GNPs (1 mg) was mixed with 1 mL of 0.25% APTES followed by sonication for 1 h. The human CRP concentrations were prepared in BSA-preblocked sample vials to minimize the analyte loss due to non-specific surface binding (Dixit et al., 2011). The UPW and PBST washings were done five times with 300 mL of the respective solutions; blocking was done with 300 mL of 5% BSA; while 100 mL was taken for 1% KOH.
1-ethyl-3-(3-dimethylaminopropyl)-carbodiimide hydrochloride (EDC)-activated capture anti-human CRP Ab mixed with GNPs in 0.25% (v/v) APTES, human CRP, biotinylated anti-human CRP Ab conjugated HRP-labeled streptavidin, and 3,3,5,5-tetramethylbenzidine (TMB). The assay was performed at 37 °C while the absorbance was measured by an Infinite M200 Pro microplate reader from Tecan (Austria). The Ab was conjugated to SA-HRP by adding 1 mL of biotinylated anti-CRP antibody (0.5 mg mL 1) to 1 mL of SA-HRP to 2998 mL of the binding buffer followed by 20 min of incubation at room temperature (RT). The biotinylated anti-CRP Ab concentration was 0.17 mg mL 1, while SA-HRP dilution employed was 1:3000. Samsung Galaxy SIII Mini was used for colorimetric imaging whereas iPAD Mini provided screen-based bottom illumination (Vashist et al., 2014c). 2.2. Preparation of ab-bound GNPs-functionalized MTP The MTP wells treated with 1.0% (w/v) KOH at RT for 10 min generated the hydroxyl groups for binding to APTES (Fig. 1A). The EDC-activated anti-human CRP Ab was prepared by incubating 990 mL of Ab (4 mg mL 1) with 10 mL of EDC (4 mg mL 1 in 0.1 M MES buffer, pH 4.7) for 15 min at RT. The EDC-activated Ab was mixed with GNPs (1 mg/mL) in 0.25% APTES in the ratio of 1:1 (v/v). Thereafter, the anti-human CRP Ab solution (2 mg mL 1, 0.5 mg mL 1 GNPs and 0.125% APTES) was added to the MTP wells and incubated for 30 min at RT. The resulting wells were then washed five times with PBST. Subsequently, the MTP wells were preblocked by incubating with 5% BSA for 30 min at 37 °C and washed five times with PBST. The anti-human CRP Ab-bound and BSA-preblocked MTPs were stored in 0.1 M PBS, pH 7.4 at 4 °C for 6 weeks to determine their functional stability. 2.3. Graphene-based Human CRP IA The anti-human CRP Ab-bound and BSA-preblocked MTP was provided with biotinylated anti-human CRP Ab preconjugated to SA-HRP. Human CRP (0.03–81 ng mL 1) was provided in buffer, diluted human plasma or diluted human whole blood. After 15 min of incubation at 37 °C, the MTPs were washed with PBS, added with TMB. After 4 min, the enzyme–substrate reaction was stopped by adding 50 mL of 2N H2SO4. The image of the colorimetric solution was captured by the smartphone. The absorbance of the resulting solution was also measured with a MTP reader (MTPR) at 450 nm with reference at 540 nm. All experiments were performed in triplicate with the reading of the blank (without human CRP) subtracted from all assay values. The LOD (limit of detection) is determined by 3 (standard deviation of the blank/ slope), while LOQ (limit of quantitation) is 10 (standard
S.K. Vashist et al. / Biosensors and Bioelectronics 66 (2015) 169–176
171
Fig. 1. Schematics of (A) the bioanalytical procedure for the development of Ab-bound and BSA-blocked graphene-functionalized MTP, and (B) the developed graphenebased human CRP IA.
S.K. Vashist et al. / Biosensors and Bioelectronics 66 (2015) 169–176
deviation of the blank/slope) (Vashist et al., 2014c). Various controls were also employed to determine the efficiency of BSA blocking and plausible non-specific interactions caused other biomarkers. The conventional sandwich ELISA and the clinicallyaccredited analyzer-based IA were performed using the manufacturer's guidelines. 2.4. SBCR analysis The SBCR analysis has been described by (Vashist et al., 2014c) with detailed information provided in Figs. S1–S3. The pixel intensity (PI) of the capture images was determined by Image J 1.48v (http://imagej.nih.gov/ij/) to extract the PI of blue channel images. The algorithm determines the PI of the center of each MTP well and calculates the desired PI of each MTP well as (255-PITest Conc.–PIBlank). Detailed information of the set-up and the image processing algorithm is provided in Fig. S1.
3. Results and discussion 3.1. Graphene-based human CRP IA The assessment of human CRP is of critical importance in the management of infectious and inflammatory conditions (Marnell et al., 2005), neonatal sepsis (Chiesa et al., 2011, 2004; Dollner et al., 2001; Nguyen-Vermillion et al., 2011; Tappero and Johnson, 2010), depressive and posttraumatic stress (Danner et al., 2003; Raison et al., 2006; Spitzer et al., 2010), diabetes (Lin et al., 2007; Morimoto et al., 2013; Testa et al., 2008) and cardiovascular diseases (Kuo et al., 2007; Park et al., 2012; Ridker, 2003; SicrasMainar et al., 2013; Wilson et al., 2006). The developed IA employs a highly-simplified procedure that enables the precise detection of human CRP in less than 30 min (Fig. 1B). After the hydroxylation of the polystyrene-based MTP with KOH (Park and Jung, 2002), the GNP–APTES complex is covalently attached to the treated surface owing to the binding of the alkoxy groups of APTES with the hydroxyl groups present on GNPs and the KOH-treated MTP surface via a hydrolysis-dependent reaction (Liedberg et al., 1998). This occurs simultaneously with the covalent binding of EDC-activated capture Ab to the amino groups on APTES by the formation of an amide bond. However, for the characterization of each process step by FTIR, polystyrene beads (PS) beads and IgG were selected to model the MTP well and antibody in the ATR-FTIR characterization study. As shown in Fig. 2, the hydroxylation of PS with KOH shows very little changes in the FTIR in comparison to untreated KOH. Peaks centered around 1632 cm 1 are associated with the skeletal vibrations of the graphitic sheet of GNPs (Lam et al., 2012) and those at 1154 cm 1 indicates the polymerization of APTES (Si‒O‒Si) bond confirming the functionalization of GNPs with APTES on the PS surface (Vashist et al., 2014b). Two amide bands at 1550 and 1645 cm 1 confirm the presence of EDC-activated IgG on the PS surface. The binding of Ab to APTES via EDC activation is highly specific (Vashist, 2012; Vashist et al., 2014d) with minimal non-specific reaction products. The resulting MTP was blocked by incubating with 5% BSA for 30 min. Thereafter, the procedure involved sequentially the dispensing of human CRP and biotinylated antihuman CRP Ab conjugated to SA-HRP into the MTP wells, resulting in a sandwich immune complex after 15 min of incubation at 37 °C (Fig. 1B). This was followed by washing with PBST, incubating with TMB for 4 min, stopping the enzyme–substrate reaction and performing the colorimetric readout (Fig. 3A) and analysis (Figs. S1– S3).
1645 cm-1
1550 cm-1
1154 cm-1
Intensity / a.u.
172
KOH-treated PS PS+GNP+APTES PS+GNP+APTES+Ab 2000
1800
1600
1400
1200
1000
Wavenumber / cm-1 Fig. 2. ATR-FTIR spectra pertaining to the immobilization of graphene nanoplatelets (GNP), APTES and IgG-EDC on polystyrene (PS) surface.
The developed IA exhibits a dynamic range of 0.03–81 ng mL 1 with linearity from 0.3 to 81 ng mL 1 for human CRP (Fig. 3B). The LOD, LOQ, and EC50 values were determined as 0.07 ng mL 1, 0.9 ng mL 1, and 6.6 ng mL 1 respectively (Table S1). The procedure detected the entire pathophysiological range of human CRP (0.08–480 mg mL 1) in spiked clinical samples after appropriate dilution (Fig. 3B). However, LOD and LOQ obtained for such spiked samples 0.07 ng mL 1 and 1.0 ng mL 1, and 0.08 ng mL 1 and 1.1 ng mL 1, respectively, implying a slight effect due to the sample matrices. The intraday and interday variability (five assay repeats in triplicate) in a single day and on five consecutive days, respectively, were 2.3–9.3 and 3.1–14.6, respectively. The procedure was not affected by immunological reagents used in the IA, as shown by various experimental process controls (Fig. 3C). LCN2, HFA, HSA, IL-1β, IL-6, IL-8 and TNF-α were also tested for any plausible interference. They are elevated along with CRP in patients with infections and other disorders. The pathophysiologically high concentration levels of these biomarkers were employed taking into account the dilution factor of 1:1000 used for the dilution of human whole blood or serum in the developed IA. These non-specific biomarkers provoked no interference in the determination of CRP as shown in Fig. 3C. The developed IA has a 14-fold reduced IA duration than the conventional sandwich IA procedure used in the commercial human CRP Duoset ELISA kit. Such a result could be attributed to the rapid covalent immobilization procedure and the use of the onestep kinetics-based IA format with minimal process steps (Vashist et al., 2014a). The immobilization procedure is 18-fold faster than passive adsorption (used in the commercial kit), whereas its IA duration (excluding Ab immobilization) is 20-fold shorter than that of the conventional IA (used in the commercial kit) (Tables S1–S2). The preparation of the Ab-bound and BSA-preblocked MTP takes only 1 h and 15 min, which is significantly shorter than the procedure using magnetic beads (Vashist et al., 2014a). The Ab binding to the magnetic beads takes about two days using the protocol provided by the provider. Therefore, the graphene-based CRP IA exhibited a 6-fold lower LOD and a wider dynamic range compared to the magnetic bead-based CRP IA. The improvement might be due to the very large surface area of GNPs to accommodate a significantly higher binding of the capture Ab (Fig. S4), which in turn results in significantly enhanced analyte detection and analytical sensitivity. GNPs were employed in the developed IA due to their much larger surface area, cost-effectiveness and commercial-availability in bulk with desired purity.
S.K. Vashist et al. / Biosensors and Bioelectronics 66 (2015) 169–176
173
Fig. 3. (A) SBCR employed for the developed graphene-based human CRP IA. Reproduced with permission from Elsevier B.V. (Vashist et al., 2014c) (B) Detection of CRP in PBS (10 mM, pH 7.4), diluted human plasma and diluted human whole blood. (C) Experimental process controls being employed to evaluate the efficiency of blocking and nonspecific interactions between immunoassay components, and non-specific interactions with other biomarkers. Anti-CRP1 and anti-CRP2 are capture and detection antibodies, respectively. The control proteins employed are human serum albumin (HSA), human lipocalin-2 (LCN2), human fetuin A (HFA), interleukin (IL)-1 beta, IL-6, IL-8 and tumor necrosis factor (TNF)-alpha. (D) Storage stability of anti-human CRP capture Ab-bound graphene-functionalized MTPs stored at 4 °C in PBS for 6 weeks. All experiments were performed in triplicate with the error bars representing the standard deviation.
They provide higher signal enhancement with superior bioanalytical performance than other materials, such as polystyrene beads, magnetic dynabeads, agarose and multiwalled carbon nanotubes (unpublished results). The developed IA procedure employs a washing step after the immune complex formation to counteract non-specificity, which makes it challenging to adapt to point-ofcare (POC) IA formats. Indeed, this procedure has been implemented in our inexpensive and disposable EasyELISA and fullyautomated centrifugal microfluidics-based LabDisk platforms (unpublished results). The sample pretreatment by appropriate dilution, as employed in our developed IA, is another challenge for POC applications. However, it is still a necessary step for the assay using commercial POC immunodiagnostic kits. The clinical analysis of CRP is usually carried out by latex agglutination or nephelometry based assays with lower detection ranges (in mg mL 1), similar to that of phosphocholine and Ophosphorylethanolamine based CRP IAs. Other IAs based on
electrochemical detection, surface plasmon resonance, nanoparticles, nanocomposites, chemiluminescence, total internal reflection and micromosiac IAs enhance sensitivities down to ng mL 1. A dual signal enhancement approach with magnetic nanoparticles enables the detection of CRP from 1.18 ng mL 1 to 11.8 mg mL 1 but requires several hours (Yang et al., 2014). The three-line lateral flow strip format has improved detection range of 1 ng mL 1–500 mg mL 1 but takes ∼2 days to prepare a strip (Kyoung Oh et al., 2014). A chemiluminescence resonance energy transfer-based IA using graphene detects CRP in the range of 1.6–100 ng mL 1 but the overall preparation and analysis time exceeds several hours (Lee et al., 2012). However, despite the large number of IAs demonstrated, only very few have shown the precision similar to clinically-accredited and commercial IAs, real sample trials, potential for whole blood analysis, desired high reproducibility and stability. Our novel IA procedure has clearly demonstrated all these important bioanalytical features, thereby
174
S.K. Vashist et al. / Biosensors and Bioelectronics 66 (2015) 169–176
Table 2 Determination of CRP in the EDTA plasma samples of patients by the clinically-accredited Roche COBASs 8000 modular analyzer-based IA, the developed graphene-based IA and the conventional sandwich ELISA-based IA (as used in the commercial human CRP Duoset ELISA kit). Samples
1 2 3 4 5 6 7 8 9 10 11 12
Clinically-accredited
Developed
Determined conc. (ng mL 1)
Determined conc. (ng mL 1)
Percentage recovery
Determined conc. (ng mL 1)
Percentage recovery
4.1 10.1 14.1 35.4 45.7 63.7 80.4 128.7 228.3 251.3 270.7 301.8
4.8 10.5 14.6 34.8 45.2 63.1 81.7 130.2 225.2 247.8 275.2 296.1
117.1 104.0 103.5 98.3 98.9 99.1 101.6 101.2 98.6 98.6 101.7 98.1
4.5 10.9 14.7 35.0 44.9 62.8 82.2 132.1 226.6 246.9 277.6 295.7
109.7 107.9 104.2 98.9 98.2 98.6 102.2 102.6 99.2 99.6 102.5 98.0
showing its immense potential for clinical diagnostics. It detects CRP in the range of pg mL 1 to ng mL 1, which is significantly lower than most of the analytical techniques used for the determination of CRP in blood (Algarra et al., 2013). Moreover, it is a
Conventional
highly-simplified and cost-effective IA procedure with the analysis time of just 30 min. It detects the complete clinically-relevant CRP range in serum and whole blood with remarkable bioanalytical performance.
Fig. 4. (A) Correlation of the developed graphene-based human CRP IA using SBCR with the conventional human CRP IA using MTPR for the detection of CRP-spiked in diluted human whole blood. (B) Detection of CRP in PBS (0.1 M, pH 7.4) by the developed graphene-based IA based on the use of different smartphone models. (C) Determination of the CRP concentration in the unknown EDTA plasma samples of patients by the developed graphene-based IA using various smartphone models based on their specific calibration curves. (D) An overlay plot of the developed graphene-based human CRP IAs based on the use of SBCR and MTPR, and employing the normalized signals (the signal response obtained for a particular human CRP concentration divided by the signal response obtained for the highest human CRP concentration).
S.K. Vashist et al. / Biosensors and Bioelectronics 66 (2015) 169–176
3.2. Determination of bioanalytical parameters The results obtained by the developed IA for anonymized EDTA plasma samples of patients compared well with the clinically-accredited Roche COBASs 8000 modular analyzer-based IA and the conventional sandwich ELISA (Table 2). The percentage recoveries of CRP by the developed IA, using the CRP values determined by the clinically-accredited IA as standard, were in the range of 98.1– 117.0, in agreement with the conventional sandwich ELISA (98.0– 109.7). Similarly, the CRP level spiked in diluted human whole blood obtained by the developed IA was comparable to that of the conventional sandwich ELISA (Fig. 4A). The determination of CRP concentrations in unknown EDTA plasma samples of patients by the developed IA also agreed well with the conventional and clinically-accredited analyzer-based IAs with percentage recoveries in the range of 93.7–103.4 (Table S3). The anti-human CRP Ab-bound and BSA-blocked MTPs, stored at 4 °C in 0.1 M PBS, pH 7.4, retained their functional activity for up to 6 weeks (Fig. 3D), attesting the leach-proof covalent binding of the capture Ab and its high functional stability. This makes the developed IA ideal for clinical diagnostics, where the Ab-bound MTPs are routinely used on a large scale and need to be stored for up to 4 weeks to facilitate rapid CRP detection. The developed Ab immobilization procedure was highly reproducible as there was no obvious batch-to-batch variability for various preparations of GNP-anti-CRP Ab when tested for the detection of 9 ng mL 1 CRP (Fig. S5). The effect of different smartphone models on the performance of the developed IA was evaluated by employing iPhone 4 and iPhone 5s apart from Samsung Galaxy SIII mini. As expected, the signals for different CRP concentrations in the developed IA were varying for the various smartphone models due to their different camera specifications (Fig. 4B). However, the determination of CRP in the unknown EDTA plasma samples of patients by each smartphone model was highly reproducible by using its respective calibration curve (Fig. 4C). There was no difference in the bioanalytical performance of the IA using SBCR or MTPR, as shown by the overlay plot of the normalized signals obtained from both readers (Fig. 4D). Apparently, SBCR is an ideal and cost-effective device for POC colorimetric detection of IAs. The overall set-up is ∼US$ 660; consisting of a Samsung Galaxy SIII mini (US$ 243), iPAD mini (US$ 399), and an inexpensive dark hood and a base holder assembly (US$ 20). It is ∼30-fold cheaper than the standard MTPR that costs ∼US$ 20,000. In addition, as almost everyone has a gadget and a smartphone nowadays, the use of SBCR will incur negligible cost to the end-users. Moreover, smartphones have become ubiquitous and are equipped with all the advanced features desired for mobile and personalized healthcare monitoring and management. These features include high resolution imaging, secure data storage, wireless connectivity, cloud computing, text alerts, spatiotemporal mapping and capabilities for robust analysis via an integrated smart application. Our ongoing efforts are dedicated to the development of a smart application for automated image analysis on the smartphone itself, controlling the smartphone's camera exposure and illumination of the iPAD mini's screen, and the development of a more compact SBCR. The ongoing development in smartphone technologies, cloud computing and software will further facilitate the improvement of the SBCR's performance.
4. Conclusion The developed IA employs inexpensive GNPs to provide signal enhancement based on the high loading of the capture Ab due to the significantly increased surface area of GNPs. The use of our
175
recently developed low-cost SBCR (Vashist et al., 2014c) for colorimetric readout instead of the expensive conventional MTPR makes it an ideal rapid IA format for decentralized and remote settings. The use of the ubiquitous smartphone in SBCR further provides all the advanced features that are desired for mobile and personalized healthcare monitoring and management (Vashist et al., 2014e). Especially in the case of newborns and children with inflammasome-associated diseases, the monitoring of CRP may be extremely helpful to predict a fever episode and prevent or at least attenuate the disease severity. Moreover, continuous monitoring of metabolic diseases may be of high relevance in a homecare setting (Gugapriya et al., 2014; Tsimikas et al., 2006). Overall, the high sensitivity assay format opens a new field of application in hospital-independent health management using CRP as a marker for sterile inflammation and metabolic stress.
Acknowledgments We thank PD Dr. Eberhard Barth for anonymizing and providing us the left-over EDTA plasma samples of patients treated by the intensive care unit at the University Hospital Ulm, Germany to validate the developed CRP immunoassay. We also thank Dr. Edmond Lam of the National Research Council Canada in Montreal for conducting the FTIR experiment.
Appendix A. Supplementary materials Supplementary data associated with this article can be found in the online version at http://dx.doi.org/10.1016/j.bios.2014.11.017.
References Algarra, M., Gomes, D., Esteves da Silva, J.C., 2013. Clin. Chim. Acta 415, 1–9. Bassuk, S.S., Rifai, N., Ridker, P.M., 2004. Curr. Probl. Cardiol. 29 (8), 439–493. Bryan, T., Luo, X., Bueno, P.R., Davis, J.J., 2013. Biosens. Bioelectron. 39 (1), 94–98. Chiesa, C., Natale, F., Pascone, R., Osborn, J.F., Pacifico, L., Bonci, E., De Curtis, M., 2011. Clin. Chim. Acta 412 (11–12), 1053–1059. Chiesa, C., Panero, A., Osborn, J.F., Simonetti, A.F., Pacifico, L., 2004. Clin. Chem. 50 (2), 279–287. Choi, H.W., Sakata, Y., Kurihara, Y., Ooya, T., Takeuchi, T., 2012. Anal. Chim. Acta 728, 64–68. Chundadze, T., Steinvil, A., Finn, T., Saranga, H., Guzner-Gur, H., Berliner, S., Justo, D., Paran, Y., 2010. Clin. Biochem. 43 (13–14), 1060–1063. Clyne, B., Olshaker, J.S., 1999. J. Emerg. Med. 17 (6), 1019–1025. Danner, M., Kasl, S.V., Abramson, J.L., Vaccarino, V., 2003. Psychosom. Med. 65 (3), 347–356. Deegan, O., Walshe, K., Kavanagh, K., Doyle, S., 2003. Anal. Biochem. 312 (2), 175–181. Dinarello, C.A., Simon, A., van der Meer, J.W., 2012. Nat. Rev. Drug Discov. 11 (8), 633–652. Dinarello, C.A., van der Meer, J.W., 2013. Semin. Immunol. 25 (6), 469–484. Dixit, C.K., Vashist, S.K., MacCraith, B.D., O'Kennedy, R., 2011. Analyst 136 (7), 1406–1411. Dollner, H., Vatten, L., Austgulen, R., 2001. J. Clin. Epidemiol. 54 (12), 1251–1257. Gugapriya, T., Karthick, S., Nagarjuna, B., 2014. J. Clin. Diagn. Res. 8 (4), CC01–CC04. Hung, J., Knuiman, M.W., Divitini, M.L., Langton, P.E., Chapman, C.L., Beilby, J.P., 2008. C-reactive protein and interleukin-18 levels in relation to coronary heart disease: prospective cohort study from Busselton Western Australia. Heart Lung Circ. 17 (2), 90–95. Islam, M.S., Kang, S.H., 2011. Talanta 84 (3), 752–758. Kim, H.C., Lee, S.K., Jeon, W.B., Lyu, H.K., Lee, S.W., Jeong, S.W., 2008. Ultramicroscopy 108 (10), 1379–1383. Kjelgaard-Hansen, M., Martinez-Subiela, S., Petersen, H.H., Jensen, A.L., Ceron, J.J., 2007. Vet. J. 173 (3), 571–577. Kuo, H.K., Yen, C.J., Chen, J.H., Yu, Y.H., Bean, J.F., 2007. Int. J. Cardiol. 114 (1), 28–33. Kushner, I., Somerville, J.A., 1970. Biochim. Biophys. Acta 207 (1), 105–114. Kyoung Oh, Y., Joung, H.-A., Soo Han, H., Suk, H.-J., Kim, M.-G., 2014. Biosens. Bioelectron. 61, 285–289. Lam, E., Chong, J.H., Majid, E., Liu, Y., Hrapovic, S., Leung, A.C., Luong, J.H.T., 2012. Carbon 50 (3), 1033–1043. Lee, J.H., Yoon, K.H., Hwang, K.S., Park, J., Ahn, S., Kim, T.S., 2004. Biosens. Bioelectron. 20 (2), 269–275.
176
S.K. Vashist et al. / Biosensors and Bioelectronics 66 (2015) 169–176
Lee, J.S., Joung, H.A., Kim, M.G., Park, C.B., 2012. ACS Nano 6 (4), 2978–2983. Lee, W.B., Chen, Y.H., Lin, H.I., Shiesh, S.C., Lee, G.B., 2011. Sens. Actuators B – Chem. 157 (2), 710–721. Liedberg, B., Cooper, J., Cass, T., Ligler, F., 1998. Immobilized Biomolecules in Analysis: A Practical Approach. Oxford University Press, Oxford, UK. Lin, M.S., Shih, S.R., Li, H.Y., Chuang, L.M., Chang, T.C., Wei, J.N., Shieh, G.J., Chien, Y.F., Hua, C.H., Hwang, J.J., 2007. Diabetes Res. Clin. Pract. 77 (2), 286–292. Marnell, L., Mold, C., Du Clos, T.W., 2005. Clin. Immunol. 117 (2), 104–111. Morimoto, H., Sakata, K., Oishi, M., Tanaka, K., Nakada, S., Nogawa, K., Suwazono, Y., 2013. Diabetes Metab. 39 (1), 27–33. Mukherjee, A., Davidson, L., Anguvaa, L., Duffy, D.A., Kennea, N., 2014. Arch. Dis. Child Fetal Neonatal Ed. , http://dx.doi.org/10.1136/archdischild-2014-306349. Mussap, M., Noto, A., Fravega, M., Fanos, V., 2011. J. Matern. Fetal Neonatal Med. 24 (S2), 12–14. Myers, G.L., Rifai, N., Tracy, R.P., Roberts, W.L., Alexander, R.W., Biasucci, L.M., Catravas, J.D., Cole, T.G., Cooper, G.R., Khan, B.V., Kimberly, M.M., Stein, E.A., Taubert, K.A., Warnick, G.R., Waymack, P.P., CDC, AHA, 2004. Circulation 110 (25), e545–549. Nguyen-Vermillion, A., Juul, S.E., McPherson, R.J., Ledbetter, D.J., 2011. J. Pediatr. 159 (1), 121–126. Park, H.E., Cho, G.Y., Chun, E.J., Choi, S.I., Lee, S.P., Kim, H.K., Youn, T.J., Kim, Y.J., Choi, D.J., Sohn, D.W., Oh, B.H., Park, Y.B., 2012. Atherosclerosis 224 (1), 201–207. Park, S.J., Jung, W.Y., 2002. J. Colloid Interface Sci. 250 (1), 93–98. Punyadeera, C., Dimeski, G., Kostner, K., Beyerlein, P., Cooper-White, J., 2011. J. Immunol. Methods 373 (1-2), 19–25. Raison, C.L., Capuron, L., Miller, A.H., 2006. Trends Immunol. 27 (1), 24–31. Ridker, P.M., 2003. Am. J. Cardiol. 92 (4B), 17K–22K. Sicras-Mainar, A., Rejas-Gutierrez, J., Navarro-Artieda, R., Blanca-Tamayo, M., 2013. Eur. Psychiat. 28 (3), 161–167.
Spitzer, C., Barnow, S., Volzke, H., Wallaschofski, H., John, U., Freyberger, H.J., Lowe, B., Grabe, H.J., 2010. J. Psychiatr. Res. 44 (1), 15–21. Tanaka, A., Shimada, K., Sano, T., Namba, M., Sakamoto, T., Nishida, Y., Kawarabayashi, T., Fukuda, D., Yoshikawa, J., 2005. J. Am. Coll. Cardiol. 45 (10), 1594–1599. Tappero, E., Johnson, P., 2010. Newborn Infant Nurs. Rev. 10 (4), 209–217. Testa, R., Bonfigli, A.R., Sirolla, C., Marra, M., Boemi, M., Mari, D., Sacchi, E., Dolci, A., Catalano, A., Procopio, A., Ceriello, A., 2008. Nutr. Metab. Cardiovasc. Dis. 18 (3), 220–226. Tomiyama, H., Okazaki, R., Koji, Y., Usui, Y., Hayashi, T., Hori, S., Yamashina, A., 2005. Atherosclerosis 178 (1), 187–192. Tsimikas, S., Willerson, J.T., Ridker, P.M., 2006. J. Am. Coll. Cardiol. 47 (S8), C19–31. Vashist, S.K., 2012. Diagnostics 2 (3), 23–33. Vashist, S.K., 2013. J. Basic Appl. Sci. 9, 496–499. Vashist, S.K., Czilwik, G., van Oordt, T., von Stetten, F., Zengerle, R., Marion Schneider, E., Luong, J.H.T., 2014a. Anal. Biochem. 456, 32–37. Vashist, S.K., Marion Schneider, E., Lam, E., Hrapovic, S., Luong, J.H.T., 2014b. Sci. Rep. 4, 4407. Vashist, S.K., van Oordt, T., Schneider, E.M., Zengerle, R., von Stetten, F., Luong, J.H.T., 2014c. Biosens. Bioelectron. , http://dx.doi.org/10.1016/j.bios.2014.08.027. Vashist, S.K., Lam, E., Hrapovic, S., Male, K.B., Luong, J.H.T., 2014d. Chem. Rev. 114 (21), 11083–11130. Vashist, S.K., Schneider, E.M., Luong, J.H.T., 2014e. Diagnostics 4 (3), 104–128. Vermeeren, V., Grieten, L., Vanden Bon, N., Bijnens, N., Wenmackers, S., Janssens, S. D., Haenen, K., Wagner, P., Michiels, L., 2011. Sens. Actuat. B – Chem. 157 (1), 130–138. Wilson, A.M., Ryan, M.C., Boyle, A.J., 2006. Int. J. Cardiol. 106 (3), 291–297. Yang, S.F., Gao, B.Z., Tsai, H.Y., Fuh, C.B., 2014. Analyst 139 (21), 5576–5581.
