Quantitative protein detection using single molecule imaging enzyme-linked immunosorbent assay (iELISA)

Quantitative protein detection using single molecule imaging enzyme-linked immunosorbent assay (iELISA)

Analytical Biochemistry 587 (2019) 113466 Contents lists available at ScienceDirect Analytical Biochemistry journal homepage: www.elsevier.com/locat...

4MB Sizes 0 Downloads 13 Views

Analytical Biochemistry 587 (2019) 113466

Contents lists available at ScienceDirect

Analytical Biochemistry journal homepage: www.elsevier.com/locate/yabio

Quantitative protein detection using single molecule imaging enzyme-linked immunosorbent assay (iELISA)

T

Chengcheng Wua,1, Yanke Shana,1, Xuping Wub, Shouyu Wanga,c,∗∗, Fei Liua,∗ a Joint International Research Laboratory of Animal Health and Food Safety of Ministry of Education & Single Molecule Nanometry Laboratory (Sinmolab), Nanjing Agricultural University, Nanjing, Jiangsu, 210095, China b The Second Hospital of Nanjing Affiliated to Southeast University, Nanjing, Jiangsu, 210003, China c Computational Optics Laboratory, School of Science, Jiangnan University, Wuxi, Jiangsu, 214122, China

A R T I C LE I N FO

A B S T R A C T

Keywords: Protein detection Single molecule Imaging enzyme-linked immunosorbent assay (iELISA) Total internal reflection fluorescence microscopy Porcine circovirus type 2 (PCV2) Cap protein

Protein detection is a key step in molecular biology research and is required for pathogen and protein marker testing for disease diagnostics. Here, single molecule imaging enzyme-linked immunosorbent assay (iELISA) is proposed to quantitatively measure the porcine circovirus type 2 (PCV2) Cap protein. The monoclonal antibody against PCV2 Cap protein indirectly immobilized on a polyethylene glycol (PEG) passivated slide by biotinstreptavidin interaction is used to capture the PCV2 Cap protein, and the PCV2 Cap protein can be detected in single molecule level according to the fluorescein isothiocyanate (FITC)-labeled secondary antibody using total internal reflection fluorescence microscopy. The single molecule iELISA measurements can be finished within 1 h skipping the time-consuming sample preparation procedures; moreover, it also exhibits excellent protein selectivity and anti-interference capability. With the proposed single molecule iELISA, linear relation between the fluorescent signals and logarithm of target protein concentrations is obtained with the detection limit of 7 ng/ mL. Considering its high accuracy in target protein detection with simple procedures and fast speed, it is believed single molecule iELISA can be potentially adopted in fast trace protein detection.

1. Introduction Target protein detection is the basic and critical evaluation in molecular biology research, and is of great significance in pathogen and protein marker testing for disease diagnostics as well [1–5]. In order to detect proteins, various mature approaches, such as western blotting [6], colloidal gold test strip [7] and enzyme-linked immunosorbent assay (ELISA) [8,9], etc., were designed and now have been widely used. Western blotting has rather high sensitivity and is often used for trace protein detection; however, it is often time-consuming because of its complicated sample pre-preparations and measuring operations. Target protein detection using colloidal gold test strip is rather simple and rapid, and the results can be directly observed even with naked eyes. But its sensitivity is often poor, thus its detecting results are qualitative rather than quantitative. Relying on the high selectivity of the capture antibody, ELISA is another tool in detecting the target proteins [10–14]. Traditional ELISA methods often require long time-

consuming in sample pre-preparations and only can detect the target proteins semi-quantitatively. However, the currently developed ELISA procedures are rather fast, even capable of detecting the target proteins within few minutes [4]. Moreover, ELISA techniques together with compact devices and advanced algorithms were also proposed to realize quantitative detection in high accuracy [15,16]. Among these ELISA techniques, colorimetric ELISA can detect the target according to the color generated by the reaction between the enzyme and substrate, it is often cost-effective but with poor sensitivity and long time consuming. In chemiluminescent ELISA, the target can be detected according to the light signal as the product of chemical reactions, and it often has higher sensitivity than colorimetric ELISA, but suffers from poor selectivity. Compared to colorimetric and chemiluminescent ELISA, fluorescent ELISA can reach high sensitivity and selectivity relying on the immunology, thus it is a promising technique in target protein detection. Unfortunately, fluorescent ELISA is easily affected by different interference factors which often causing poor repeatability and false



Corresponding author. Single Molecule Nanometry Laboratory (Sinmolab), Nanjing Agricultural University, 210095, China. Corresponding author. Single Molecule Nanometry Laboratory (Sinmolab), Nanjing Agricultural University, 210095 & Computational Optics Laboratory, School of Science, Jiangnan University, 214122, China. E-mail addresses: [email protected] (S. Wang), [email protected] (F. Liu). 1 These authors contributed equally to this work. ∗∗

https://doi.org/10.1016/j.ab.2019.113466 Received 1 June 2019; Received in revised form 19 September 2019; Accepted 27 September 2019 Available online 28 September 2019 0003-2697/ © 2019 Elsevier Inc. All rights reserved.

Analytical Biochemistry 587 (2019) 113466

C. Wu, et al.

high sensitivity [10,28,32]. Besides, the sample under detection can be directly measured using single molecule iELISA without time-consuming and complicated sample pre-preparation procedures required by traditional ELISA techniques. Considering the proposed single molecule iELISA can achieve quantitative protein detection but only with simple operations and high sensitivity, it is believed the proposed single molecule iELISA can be potentially adopted for fast quantitative trace protein detections.

positives. Therefore, improved protein detecting approach, which not only distinguishes the target protein with excellent selectivity and high sensitivity, but also provides quantitative target protein measurements with high accuracy and fast speed, is still in demands. To further improve the protein detecting sensitivity, nanoparticles were adopted to further amplify the signals thus obviously increasing the detecting sensitivity [17–24]. Unfortunately, nanoparticle based labeling not only amplifies the signals, but also amplifies the noises. Moreover, the nanoparticles are often sensitive to environments, and long-time storage and environmental changes easily induce nanoparticle coagulation which generates errors in target detections. As another tactic for sensitivity increasing, single molecule techniques [25–37], especially microwell based digital ELISA was designed for noise reduction to pursue high detecting sensitivity: the immunoreaction is often implemented outside the microwell [38–41], while the single molecule detection is implemented when the nanoparticle probes are transferred to microwells, which obviously reduce the background noises in fluorescent signal detection. Moreover, digital ELISA as a single molecule technique does not require time-consuming sample pre-preparations. However, digital ELISA requires expensive and complicated fabricating microwell array, in addition, cross reactions among antibodies often occur during the immunoreaction process and inevitably introduce detecting errors, decreasing the sensitivity and the accuracy of detection. In order to realize quantitative protein detection free of nanoparticles and complicated fabricating microwell array, we propose the single molecule imaging enzyme-linked immunosorbent assay (iELISA) method. Fig. 1(A) illustrates the principle of the single molecule iELISA, in which the Cap protein of porcine circovirus type 2 (PCV2) [42] is the target to be detected. Both the target capture and detection are implemented in the sample chamber composed of glass slide and coverslip modified with biotin-PEG. The biotin-PEG coated surface is saturated with streptavidin, and the biotinylated goat anti-mouse secondary antibody is immobilized on the coverslip surface through the biotinstreptavidin linkage. Binding with the biotinylated goat anti-mouse secondary antibody, the murine monoclonal antibody against PCV2 Cap protein is used to capture the PCV2 Cap protein. Binding with the rabbit polyclonal antibody against PCV2 Cap protein, the FITC-labeled goat anti-rabbit secondary antibody providing fluorescent signals is used to detect the target PCV2 Cap protein relying on the high specific binding between the PCV2 Cap protein and the murine monoclonal antibody, as well as that between the PCV2 Cap protein and the rabbit polyclonal antibody. Since the target PCV2 Cap protein can be detected by the fluorescein isothiocyanate (FITC)-labeled secondary antibody when FITC is excited, moreover, TIRF microscopy is used to collect these surface fluorescent signals as shown in Fig. 1(B), not only to increase the signal to noise ratio, but also to avoid the fluorescent error from the free FITC-labeled secondary antibody not connected to target protein. In this design, biotinylated antibody and FITC-labeled antibody are both secondary antibodies which are much easier and cheaper to obtain than the biotinylated or FITC-labeled monoclonal or polyclonal antibodies. This sandwich detection is a multiplex system, monoclonal antibody acts as a capture antibody and polyclonal antibody acts as a reporting antibody. In other words, other proteins could be detected by just replacing the unlabeled monoclonal and polyclonal antibodies. Therefore, considering the cost and the simplicity of applying the system to detect other targets, the four-antibody system is selected instead of the two-antibody system. Using the target recognition and counting via thresholding method as shown in Fig. 1(C), quantitative relation between the fluorescent signals and the target protein concentrations was obtained for quantitative target PCV2 Cap protein detection. It is worth noting that the proposed single molecule iELISA is quite different from the traditional ELISA approaches [43,44]. This single molecule iELISA can determine the specific target by locating the single fluorescence spot, while the traditional ELISA approaches only can reflect overall effect of the targets according to the total fluorescence intensity; therefore, the single molecule techniques often have

2. Materials and methods 2.1. Chemicals and reagents Methanol (CH3OH), acetone ((CH3)2CO), potassium hydroxide (KOH), glacial acetic acid (CH3COOH), tris (hydroxymethyl) aminomethane-hydrochloric acid (Tris-HCl), sodium bicarbonate (NaHCO3) and sodium chloride (NaCl) purchased from Sigma-Aldrich (USA) were analytical grade and used without further purification. Aminosilane (N(2-Aminoethyl)-3-Aminopropyltrimethoxysilane) was purchased from United Chemical Technology (United Kingdom); epoxy was purchased from Pattex (Germany); streptavidin, biotinylated goat anti-mouse antibody and FITC-labeled goat anti-rabbit antibody were purchased from Pierce (USA); methoxy PEG (mPEG-Succinimidyl Valerate, MW 5,000) and biotin-PEG (Biotin-PEG-SVA, MW 5,000) were purchased from Laysan (USA). Hypoxanthine aminopterin thymidine (HAT) medium and polyethylene glycol 2000 (PEG2000) were purchased from Sigma (USA). All the solutions were prepared using the distilled water from Milli-Q system (USA). T50 buffer was prepared by mixing 10 mM TrisHCl (pH = 8.0) and 50 mM NaCl. 15 positive and negative serums with and without PCV2, the N protein of porcine reproductive and respiratory syndrome virus (PRRSV) and the VP2 protein of porcine parvovirus (PPV) were given by pig farm in Jiangsu. 2.2. Antibody preparation In the proposed single molecule iELISA technique, four antibodies were used. Besides the purchased biotinylated goat anti-mouse antibody and FITC-labeled goat anti-rabbit antibody, the murine monoclonal antibody against PCV2 Cap protein and the rabbit polyclonal antibody against PCV2 Cap protein were both prepared in our laboratory. In preparation of murine monoclonal antibody against PCV2 Cap protein, the Cap protein was first used as the antigen to immunize female mice aged 6–8 weeks to acquire immunized spleen cells; then the spleen cells were fused with the myeloma cells into the hybridoma cells using PEG2000, and the hybridoma cells were cultured and selected using HAT medium; next, the positive clones were screened using the indirect ELISA method; finally, the ascites was prepared, and the monoclonal antibody was purified and identified using western blotting. While in preparation of rabbit polyclonal antibody against PCV2 Cap protein, the Cap protein was first used as the antigen to immunize male rabbits aged 2 months; then the blood was collected and the antiserum titer was determined using the indirect ELISA method; finally, the serum was isolated from the collected blood to purify the antibody. 2.3. Sample chamber fabrication The sample chamber shown in Fig. 1(B) was fabricated using a glass slide and a coverslip with simple operations. First, holes with the diameter of 0.8 mm were drilled on the glass slide for sample input and output. After slide and coverslip cleaning and amination, 98% methoxy polyethylene glycol (mPEG) and 2% biotinylated PEG (bPEG) were coated on channel surfaces in order to significantly reduce the nonspecific binding of proteins to the glass surfaces [25–27,30,34,35]. Then, the sample chamber was assembled with both the PEG modified slide and coverslip, and multiple microfluidic channels were separated by double-sided tape and sealed with epoxy for more than 5 min. 2

Analytical Biochemistry 587 (2019) 113466

C. Wu, et al.

Fig. 1. Scheme of single molecule iELISA detection. (A) First, biotinylated antibody, murine monoclonal antibody, target PCV2 Cap protein, rabbit polyclonal antibody and FITC-labeled antibody are added to PEG coated slide step-by-step. (B) Then, fluorescent images are collected by TIRF microscopy. (C) Finally, the fluorescent signals are recognized and analyzed via thresholding method for target PCV2 Cap protein detection.

afterwards, wash off the excess streptavidin using T50 buffer as Step 2 in Fig. 1(A). (3) Mix 10 μL biotinylated goat anti-mouse secondary antibody solution with the concentration of 40 nM, 10 μL unlabeled murine monoclonal antibody against PCV2 Cap protein solution with the concentration of 20 nM, 10 μL rabbit polyclonal antibody against PCV2 Cap protein solution with the concentration of 10 nM and 10 μL FITC-labeled goat anti-rabbit secondary antibody solution with the concentration of 2 nM together with the target PCV2 Cap protein under detection. Inject 10 μL mixed solution into the sample chamber and incubate for 10 min at room temperature, and use T50 buffer to wash off the excess reagents. The mixed sample adding approach could directly construct the target detection configuration as Step 7 in Fig. 1(A). Finally, target PCV2 Cap protein can be detected when FITC is excited with light at the wavelength of 488 nm. Though the mixed sample adding approach is simple, there are always cross reactions among antibodies, generating background fluorescent noise, which inevitably decreases the sensitivity and the accuracy of the target protein detection. In order to reduce the cross reactions between antibodies, here we adopted the step-by-step sample-adding approach. (1) Coat PEG on the glass surfaces after slide and coverslip cleaning and amination as Step 1 in Fig. 1(A). (2) Prepare 0.2 mg/mL solution of streptavidin in T50 buffer, then inject 10 μL streptavidin-T50 mixture into the sample chamber and incubate for 5 min at room temperature; afterwards, wash

Finally, since the adopted poly-tetrafluoroethylene can prevent nonspecific adsorption of protein, poly-tetrafluoroethylene tube was connected to the drilled holes for sample input and output. Fig. S1 in the Supplementary Information explains the sample chamber fabrication procedures in details. Besides, during sample adding and fluorescence detection, the sample chamber was fixed at a homemade sample stage as shown in Fig. S2 in the Supplementary Information. 2.4. Sample adding approaches According to the principle of the proposed single molecule iELISA, the samples under detection can be directly introduced into the sample chamber for measurements without any pre-preparation required by traditional ELISA techniques. Two different sample adding approaches including the mixed and the step-by-step tactics were adopted, and the schemes are simply illustrated in Fig. 1(A), and more detailed sample adding steps are explained in Figs. S3 and S4 in the Supplementary Information corresponding to the mixed and the step-by-step tactics, respectively. First, the mixed sample adding approach was implemented. (1) Coat PEG on the glass surfaces after slide and coverslip cleaning and amination as Step 1 in Fig. 1(A). (2) Prepare 0.2 mg/mL solution of streptavidin in T50 buffer, then inject 10 μL streptavidin-T50 mixture into the sample chamber and incubate for 5 min at room temperature; 3

Analytical Biochemistry 587 (2019) 113466

C. Wu, et al.

the illumination of the TIRF microscope and the recording parameters (exposure time and gain) of the EMCCD camera were unchanged during all the recording for different channels. Next, the effective fluorescent spots could be recognized according to the determined threshold value (Fig. 2(B)) from the captured fluorescent image (Fig. 2(A)). (Note that the fluorescent aggregations (often the overexposure spots) were removed as shown in Fig. 2(C) before statistical analysis on these fluorescent spots.) However, the number of the targets was not equal to the recognized number of fluorescent spots, it is because though most of the single recognized fluorescent spot represents one captured target protein according to the single molecule iELISA principle, there were still a few fluorescent clusters as the accumulation of the captured targets (often less than 5). To deal with these fluorescent clusters for highaccurate single molecule iELISA measurement, the number of the targets in each cluster was estimated according to the fluorescent intensity. After fluorescent spot recognition, the histogram corresponding to fluorescent intensity of all the recognized spots was obtained, and the statistical histogram distribution was fitted by multiple Gaussian functions, also shown in Fig. 2(D): the left-most fitted Gaussian function indicates the single-target fluorescent intensity distributions, while other fitted Gaussian functions represent multi-target cases. According to the fitted Gaussian functions, the assembled number of the targets could be estimated as shown in Fig. 2(E). Finally, the number of fluorescent signals which equals to the number of the captured targets was counted from fluorescent spots representing single target protein and those including multiple targets as shown in Fig. 2(F). Moreover, the verification on the method was also provided in Figs. S5–S7 in the Supplementary Information, which proves that the adopted thresholding method could obtain the target recognition and counting with extremely high accuracy.

off the excess streptavidin using T50 buffer as Step 2 in Fig. 1(A). (3) Inject 10 μL biotinylated goat anti-mouse secondary antibody solution with the concentration of 40 nM into the sample chamber and incubate for 10 min at room temperature; afterwards, use T50 buffer to wash off the excess antibody as Step 3 in Fig. 1(A). (4) Inject 10 μL unlabeled murine monoclonal antibody against PCV2 Cap protein solution with the concentration of 20 nM into the sample chamber and incubate for 15 min at room temperature; afterwards, wash off the excess antibody using T50 buffer as Step 4 in Fig. 1(A). (5) Inject 10 μL PCV2 Cap protein solution into the antibody coated sample chamber and incubate for 15 min; afterwards, wash off the unbound protein using T50 buffer as Step 5 in Fig. 1(A). (6) Inject 10 μL rabbit polyclonal antibody solution with the concentration of 10 nM into the sample chamber and incubate for 15 min at room temperature; afterwards, wash off the excess antibody using T50 buffer as Step 6 in Fig. 1(A). (7) Inject 10 μL FITC-labeled goat anti-rabbit secondary antibody solution with the concentration of 2 nM into the sample chamber and incubate for 10 min at room temperature; afterwards, wash off the excess antibody using T50 buffer as Step 7 in Fig. 1(A). Finally, target PCV2 Cap protein can be detected according to the fluorescent signals when FITC is excited. 2.5. Optical system A semiconductor laser (Coherent, USA) was used for generating the excited light at the wavelength of 488 nm with the power of 24 mW. A TIRF microscope (microscope system: Olympus IX81, Japan; microobjective: 100×, N.A. = 1.49, Olympus, Japan) was adopted to collect the fluorescent signals using an EMCCD camera (Evolve, USA). The exposure time of the EMCCD was set as 100 ms with the gain of 300. The detection was performed at room temperature.

3. Results and discussions 2.6. Target recognition and counting via thresholding method 3.1. Sample adding method selection In order to recognize and count the targets accurately from fluorescent images, the thresholding method was adopted as shown in Fig. 2, which not only extracted the fluorescent spots from the captured fluorescent image, but also quantified the target number within each recognized fluorescent spot. In each measurement with a new sample chamber, one channel was always kept empty for calibration, and the threshold value was first determined according to the recognized number of fluorescent spots in the empty channel. In our cases, the threshold value was chosen when the recognized number of fluorescent spots was 50 ± 5 in a field of view of ~80 μm × 80 μm. It is noted that

Two different approaches based on the mixed or step-by-step tactics can be used for sample adding. Compared to the step-by-step method, the mixed sample adding approach is with simpler operations. Unfortunately, there are always cross reactions among antibodies, especially during the incubation of PCV2 Cap protein in vitro, various antibodies are mixed with the target protein in a small sterile centrifuge tube, thus promoting cross reactions and then generating background fluorescent noise even in target-free cases. The noise inevitably decreases the sensitivity and the accuracy in single molecule iELISA Fig. 2. Flowchart of the target recognition and counting via thresholding method. (A) directly captured fluorescent image; (B) binary image of recognized fluorescent spots after threshold segmentation according to the evaluated threshold; (C) binary image after aggregation removal; (D) statistical histogram according to recognized fluorescent spot intensity and its Gaussian function fitting result; (E) target number estimation; (F) target counting.

4

Analytical Biochemistry 587 (2019) 113466

C. Wu, et al.

Fig. 3. Comparisons between the step-by-step and mixed sample adding approaches. Fluorescent images in the conditions of (A) blank, (B) step-by-step and (C) mixed sample adding only with antibodies dissolved in T50 buffer; and (D) blank, (E) step-by-step and (F) mixed sample adding only with antibodies dissolved in serum. White bar in (A) indicates 20 μm. (G) Quantitative comparisons on recognized numbers of fluorescent signals corresponding to the conditions of (A)–(F), respectively, and for each condition, 20 measurements were implemented for statistical analysis. The error bars in (G) denote standard deviation.

view. With the determined threshold, fluorescent signals in other channels can be determined and analyzed. For the blank conditions of the channels only washed by T50 buffer and serum, the recognized numbers of fluorescent signals (considering the fluorescent clusters) were 53 ± 2 and 130 ± 5, respectively. Compared to the T50 buffer, serum introduced more fluorescent noise. Next, reagents including antibodies but without targets were injected into the sample chambers using different sample adding tactics. When antibodies were diluted in T50 buffer, the recognized numbers of fluorescent signals were 150 ± 13 and 245 ± 14 corresponding to step-by-step and mixed sample adding tactics, respectively; when antibodies were diluted in serum, the recognized numbers were 199 ± 22 and 269 ± 13. Compared to the mixed sample adding tactic, the step-by-step sample

measurements. In order to reduce the cross reactions, the step-by-step sample adding approach is considered, and these two sample adding tactics were quantitatively compared as shown in Fig. 3. Fluorescent signals in sample channels only with T50 buffer/serum, and with reagents including all the antibodies but without target protein using both sample adding tactics were imaged in Fig. 3(A)-3(F) and analyzed in Fig. 3(G), respectively. For each case, 20 independent measurements were implemented. According to the requirement of data analysis, one channel in the sample chamber was always kept empty, and the threshold value was first determined according to the recognized fluorescent spots in a field of view with ~80 μm × 80 μm of the empty channel. In our measurements, the threshold value was chosen when the recognized number of fluorescent spots was 50 ± 5 in such field of 5

Analytical Biochemistry 587 (2019) 113466

C. Wu, et al.

Fig. 4. Detecting the PCV2 Cap protein in T50 buffer using single molecule iELISA. (A) Fluorescent images of PCV2 Cap protein with different concentrations of 56 ng/mL, 28 ng/mL, 14 ng/ mL, 9.3 ng/mL, 7 ng/mL, 5.6 ng/mL, 2.8 ng/mL and 0 ng/mL, respectively. White bar in (A) indicates 20 μm. (B) Linear relation between the logarithm of target concentrations (ranging from 7 ng/mL to 56 ng/mL) and their corresponding statistically recognized numbers of fluorescent signals from 20 independent measurements. The data listed in (B) indicates average recognized fluorescent signal number ± standard deviation, and the error bars in (B) denote standard deviation.

concentration of PCV2 Cap protein was 560 ng/mL, and it was further diluted by 10 times, 20 times, 40 times, 60 times, 80 times, 100 times and 200 times with T50 buffer, corresponding to the concentrations of 56 ng/mL, 28 ng/mL, 14 ng/mL, 9.3 ng/mL, 7 ng/mL, 5.6 ng/mL and 2.8 ng/mL, respectively. According to the step-by-step sample adding tactic, their corresponding fluorescent images were recorded using the TIRF microscope, and part of them are shown in Fig. 4(A). With the data processing based on the thresholding method, the recognized numbers of fluorescent signals were 494 ± 36, 351 ± 11, 223 ± 23, 212 ± 11, 188 ± 22, 153 ± 13, and 150 ± 17 statistically obtained from 20 independent measurements in Fig. 4(B), corresponding to different target concentrations ranging from 56 ng/mL to 2.8 ng/mL. Moreover, the recognized number of fluorescent signals in target free condition was 148 ± 12 also statistically obtained from 20 independent measurements, which is similar to that in Fig. 3. According to the recognized numbers of fluorescent signals without and with target protein, the detection limit was determined by t-test: since the p values between recognized numbers of fluorescent signals without the target and with the target concentrations of 5.6 ng/mL or 2.8 ng/mL were both higher than 0.05, while that between recognized numbers of fluorescent signals without the target and with the target concentration of 7 ng/mL was lower than 0.05, proving the detection limit of 7 ng/mL (250 pM). Besides, according to the recognized numbers of fluorescent

adding approach significantly decreased the background by reducing cross reactions; besides, background fluorescence was also removed by washing off the redundant antibodies in the sample chamber after each sample adding step. Since the number of the fluorescent signals with all the antibodies but without target is the background in single molecule iELISA application, which should be as low as possible, the step-by-step sample adding tactic was selected in single molecule iELISA. Though the PEG passivated sample chamber and the EMCCD camera were adopted to reduce the background signal, the inevitable non-specific absorption still existed thus generating background fluorescence signals. Even the step-by-step sample adding tactic was used to further reduce the background signal, the background signal could hardly be completely reduced, which is similar to other references on single molecule techniques [26,45,46]. Compared to the mixed sample adding approach, the step-by-step one has more procedures, but the whole protein detection process including sample preparation, sample adding, fluorescent image recording and fluorescent signal analysis could still be finished within 1 h. 3.2. PCV2 Cap protein detection using single molecule iELISA After the sample adding method selection, single molecule iELISA was then adopted for PCV2 Cap protein detection. Here, the original 6

Analytical Biochemistry 587 (2019) 113466

C. Wu, et al.

Fig. 5. Detecting the PCV2 Cap protein in serum using single molecule iELISA. (A) Fluorescent images of PCV2 Cap protein with different concentrations of 56 ng/mL, 28 ng/mL, 14 ng/mL, 9.3 ng/mL, 7 ng/mL, 5.6 ng/mL, 2.8 ng/mL and 0 ng/mL, respectively. White bar in (A) indicates 20 μm. (B) Linear relation between the logarithm of target concentrations (ranging from 7 ng/mL to 56 ng/mL) and their corresponding statistically recognized numbers of fluorescent signals from 20 independent measurements. The data listed in (B) indicates average recognized fluorescent signal number ± standard deviation, and the error bars in (B) denote standard deviation.

values between recognized numbers of fluorescent signals without the target and with the target concentrations of 5.6 ng/mL or 2.8 ng/mL were both higher than 0.05, while that between recognized numbers of fluorescent signals without the target and with the target concentration of 7 ng/mL was lower than 0.05. Besides, the linear fitting between the logarithm of target concentrations and recognized numbers of fluorescent signals with rather high R2 as 0.98 proves the excellent antiinterference of the proposed single molecule iELISA and reveals that the single molecule iELISA can be well adopted in practical applications. Additionally, in order to investigate the selectivity of single molecule iELISA, the N protein of PRRSV and the VP2 protein of PPV that cannot be recognized by the monoclonal and polyclonal antibodies against PCV2 Cap protein were used here as a non-specific sample. The PCV2 Cap protein, the PRRSV N protein and the PPV VP2 protein were diluted to 56 ng/mL with T50 buffer or serum, and added into sample chamber using the step-by-step sample adding strategy. In the measurements shown in Fig. 6, the recognized numbers of fluorescent signals in blank case only with T50 buffer or serum and in target free case with antibodies were rather close to those in Fig. 3. When the PCV2 Cap protein was added, the recognized numbers of fluorescent signals were 494 ± 36 and 516 ± 38 both extracted from 20 independent measurements corresponding to T50 buffer and serum conditions, while when the PRRSV N protein and the PPV VP2 protein were added, the

signals in concentrations of 56 ng/mL, 28 ng/mL, 14 ng/mL, 9.3 ng/mL and 7 ng/mL, linear relation between the logarithm of target concentrations and recognized numbers of fluorescent signals was fitted as shown in Fig. 4(B) with rather high R2 as 0.97, showing that single molecule iELISA provides a way for quantitative target protein measurements with extremely high sensitivity. Besides, to investigate the anti-interference potentials of single molecule iELISA, PCV2 Cap protein was also detected in fetal bovine serum, and the condition was used to simulate the real sample detections [47–49]. The original concentration of PCV2 Cap protein of 560 ng/mL was further diluted with serum to the concentrations of 56 ng/mL, 28 ng/mL, 14 ng/mL, 9.3 ng/mL, 7 ng/mL, 5.6 ng/mL and 2.8 ng/mL, respectively. According to the step-by-step sample adding tactic, their corresponding fluorescent images were recorded using the TIRF microscope, and part of them are shown in Fig. 5(A), and the recognized numbers of fluorescent signals from 20 independent measurements were 516 ± 38, 376 ± 38, 314 ± 20, 255 ± 20, 214 ± 14, 193 ± 18 and 202 ± 37 shown in Fig. 5(B), respectively. Moreover, the recognized number of fluorescent signals in target free condition was 189 ± 14 also from 20 independent measurements, which is similar to that in Fig. 3. Though the recognized numbers of fluorescent signals both with target and without target (background) increased, the detection limit was still kept around 7 ng/mL, since the p 7

Analytical Biochemistry 587 (2019) 113466

C. Wu, et al.

Fig. 6. Protein selectivity verification on single molecule iELISA. Fluorescent images in the conditions of (A) blank, (B) target-free, (C) PCV2 Cap protein, (D) PRRSV N protein and (E) PPV VP2 protein all with the concentrations of 56 ng/mL dissolved in T50 buffer; and (F) blank, (G) target-free, (H) PCV2 Cap protein, (I) PRRSV N protein and (J) PPV VP2 protein all with the concentrations of 56 ng/mL dissolved in serum. White bar in (A) indicates 20 μm. (K) Quantitative comparisons on recognized numbers of fluorescent signals corresponding to the conditions of (A)–(J), respectively, and for each condition, 20 measurements were implemented for statistical analysis. The error bars in (K) denote standard deviation.

recognized numbers of fluorescent signals were 150 ± 24 and 134 ± 23 in T50 buffer and 178 ± 29 and 149 ± 24 in serum both extracted from 20 independent measurements, which are close to the background values in Fig. 6(I). The results prove that single molecule iELISA has high selectivity in target protein detection. In addition, practical samples were also measured to verify the proposed method. 9 negative serum samples (Nos. 1–9) without PCV2 and 6 positive serum samples (Nos. 10–15) with PCV2 were detected using the proposed single molecule iELISA. For each sample, their corresponding statistically recognized numbers of fluorescent signals were still from 20 independent measurements, and the representative fluorescent images corresponding to these 15 samples are listed in Fig. 7(A). Fig. 7(B) lists their statistically recognized numbers of fluorescent signals, showing that the recognized numbers of fluorescent signals of the samples without PCV2 are much less than those of the samples with PCV2. Moreover, according to the ~200 as the background level in the serum condition determined in Fig. 5, the samples with and without PCV2 could be successfully distinguished according to Fig. 7(B). Moreover, according to the linear relation between the logarithm of target concentrations and recognized numbers of fluorescent signals in Fig. 5, the PCV2 concentrations were measured as 127.68 ± 8.94 ng/mL, 66.80 ± 7.13 ng/mL, 23.25 ± 1.96 ng/mL, 11.31 ± 1.45 ng/mL, 12.80 ± 2.23 ng/mL and 17.63 ± 0.98 ng/mL corresponding to the Sample No. 10 to 15. According to the results in Fig. 7, positive and negative samples could be accurately distinguished, proving that the proposed single molecule iELISA has excellent specificity. Moreover, according to the detecting results in Fig. 7, the fluctuation of the recognized numbers of fluorescent signals was not large, and via the detecting procedure, each detecting determination relies on the 20 independent fields of view, making the proposed single molecule

iELISA stable. In order to test the reproducibility of the proposed single molecule iELISA, we repeated the detecting on Sample No. 3 and No. 15 for another three times, each still containing 20 measurements, and the results are shown in the insert figure of Fig. 7(B). Not only the samples could be still distinguished into positive and negative samples, but also the recognized numbers of fluorescent signals were similar, proving the reproducible of the proposed single molecule iELISA. 3.3. Comparisons to commercial colloidal gold test strips and ELISA kits Besides the proposed single molecule iELISA used for PCV2 Cap protein detection, both the colloidal gold test strip and the traditional ELISA kit are mature tools in PCV2 Cap protein detection. They are cost-effective and easy to use, thus becoming widely used in many applications. Here, the performance of single molecule iELISA was also compared to the colloidal gold test strip approach (Tongdian Biotechnology, Zhejiang, China) and the traditional ELISA (Zeyu Biotechnology, Jiangsu, China). In PCV2 Cap protein detection using colloidal gold test strips, the target concentrations were diluted with serum to 8 different concentrations ranging from 0.96×103 ng/mL to 1.22×105 ng/mL. According to the protocol of the colloidal gold test strip (Fig. S8 in the Supplementary Information), all these 8 samples were measured, each sample was measured for 5 times using 5 colloidal gold test strip, and the results are listed in Fig. 8(A), revealing that its detection limit was around 3.83×103 ng/mL. In addition, different concentrations of PCV2 Cap protein from 0 ng/mL to 2000 ng/mL were also detected using the commercial ELISA kits, and each sample was measured for 5 times. According to the statistical OD450 values measured by the microplate reader (M200 pro, Tecan, Switzerland) as 0.13 ± 0.02, 0.14 ± 0.01, 0.20 ± 0.01, 0.25 ± 0.02, 0.40 ± 0.02, 8

Analytical Biochemistry 587 (2019) 113466

C. Wu, et al.

Fig. 7. Verification on single molecule iELISA using practical samples. (A) Fluorescent images of different practical samples. (B) Quantitative comparisons on recognized numbers of fluorescent signals corresponding to (A), and for each condition, 20 measurements were implemented for statistical analysis. The error bars in (B) denote standard deviation.

obtain quantitative target protein measurements with the detection limit of 7 ng/mL, which is ~550-fold and ~85-fold more sensitive than commercial colloidal gold test strip and ELISA, respectively. Besides, single molecule iELISA also exhibits excellent protein anti-interference and selectivity capability, proved by detection in serum as well as with PRRSV N protein and PPV VP2 protein, respectively. Moreover, single molecule iELISA was also verified by practical samples, proving it not only can successfully distinguish the positive and negative samples, but also have excellent specificity and reproducibility. However, compared to traditional protein detection methods, such as traditional ELISA, single molecule iELISA still relies on expensive instruments as TIRF microscope and EMCCD camera as well as complicated system preparation as PEG coating; besides, the step-by-step sample adding requires ~1 h for sample preparation, which is longer than the currently developed rapid ELISA [4]; moreover, the detection limit of the reported single molecule iELISA is lower than those ELISA using nanoparticles for signal magnifications [38–41]. However, the proposed single molecule iELISA is still potential in the future applications. First, single molecule iELISA can locate and detect the targets directly, while the traditional ELISA approaches can only reflect the

0.44 ± 0.02, 0.58 ± 0.02, 0.61 ± 0.05, 0.86 ± 0.03 and 0.97 ± 0.02 corresponding to the concentrations of 200 ng/mL, 400 ng/mL, 600 ng/mL, 800 ng/mL, 1000 ng/mL, 1200 ng/mL, 1400 ng/mL, 1600 ng/mL, 1800 ng/mL and 2000 ng/mL shown in Fig. 8(B), the detection limit of the commercial ELISA kits was around 600 ng/mL according to the average OD450 of negative control. While our proposed single molecule iELISA could reach the detection limit of 7 ng/mL, which is ~550-fold and ~85-fold more sensitive than the commercial colloidal gold test strip and ELISA kit, respectively. 4. Conclusion In order to realize quantitative target protein detection, in this paper, we design and implement single molecule iELISA which reaches target protein detection by directly imaging and analyzing fluorescent signals in single molecule level. Using the self-designed sample chamber with microfluidic channels and optimized step-by-step sample adding tactic, both the data collection and analysis can be finished within 1 h skipping the sample pre-preparation. According to its application in detecting PCV2 Cap protein, single molecule iELISA can 9

Analytical Biochemistry 587 (2019) 113466

C. Wu, et al.

Fig. 8. Sensitivity verification on single molecule iELISA by comparing PCV2 Cap protein detections in serum using commercial colloidal gold test strips and ELISA kits. (A) Detection results of PCV2 Cap protein with different concentrations of 1.22×105 ng/mL, 6.10×104 ng/mL, 3.05×104 ng/mL, 1.53×104 ng/ 3.83×103 ng/mL, mL, 7.65×103 ng/mL, 1.92×103 ng/mL and 0.96×103 ng/mL using commercial colloidal gold test strips. All the detections were repeated for 5 times. (B) Detection results (OD450) of PCV2 Cap protein with different concentrations of 2000 ng/mL, 1800 ng/mL, 1600 ng/ mL, 1400 ng/mL, 1200 ng/mL, 1000 ng/mL, 800 ng/ mL, 600 ng/mL, 400 ng/mL, 200 ng/mL and 0 ng/mL using commercial ELISA kits. The statistical data listed in (B) indicates average OD450 value ± standard deviation extracted from 5 independent measurements. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

Acknowledgements

overall effect of targets, therefore single molecule iELISA is potential to achieve higher accuracy than classical ELISA. Secondly, the sample under detection can be directly introduced into the microfluidic channel for measurements, thus skipping the time-consuming and complicated sample pre-preparation procedures required by traditional ELISA techniques. Finally, the proposed single molecule iELISA does not rely on nanoparticles, thus avoiding coagulation induced detecting error. Moreover, it also does not require specially fabricated microwell. Considering that the proposed single molecule iELISA can realize quantitative target protein detection in high accuracy, excellent selectivity and anti-interference; moreover, by future optimizing the channel modification to further decrease the non-specific binding of protein to the glass surfaces, and simplifying the sample preparation procedures to reduce the time consuming in sample adding, it is still believed that single molecule iELISA can be potentially adopted for fast trace protein quantitative detection in various fields such as in biological and medical applications.

This work was supported by grants from the National Natural Science Foundation of China (31870154, 61705092); National Key Research and Development Program (2015BAD12B01, 2018YFD0500100); Natural Science Foundation of Jiangsu Province of China (BK20170194); Jiangsu Key Research and Development Program (BE2018709) and the Priority Academic Program Development of Jiangsu Higher Education Institutions. Appendix A. Supplementary data Supplementary data to this article can be found online at https:// doi.org/10.1016/j.ab.2019.113466. References [1] A. Khalilpour, T. Kilic, S. Khalilpour, et al., Proteomic-based biomarker discovery for development of next generation diagnostics, Appl. Microbiol. Biotechnol. 101 (2015) 475–491. [2] J.F. Rusling, C.V. Kumar, J.S. Gutkind, et al., Measurement of biomarker proteins for point-of-care early detection and monitoring of cancer, Analyst 135 (2010) 2496–2511. [3] P. Zhurauski, S.K. Arya, P. Jolly, et al., Sensitive and selective Affimer-functionalised interdigitated electrode-based capacitive biosensor for Her4 protein tumour

Declaration of competing interest F. Liu and S. Wang are the co-founders of Sinmotec LLC, which commercializes the single molecule sensing and imaging techniques. 10

Analytical Biochemistry 587 (2019) 113466

C. Wu, et al.

[27] A. Jain, R. Liu, Y.K. Xiang, et al., Single-molecule pull-down for studying protein interactions, Nat. Protoc. 7 (2012) 445–452. [28] L. Li, X. Qu, J. Sun, et al., Single-molecule-counting protein microarray assay with nanoliter samples and its application in the dynamic protein expression of living cells, Biosens, Bioelectron 26 (2011) 3688–3691. [29] X. Lu, P.R. Nicovich, K. Gaus, et al., Towards single molecule biosensors using super-resolution fluorescence microscopy, Biosens. Bioelectron. 93 (2017) 1–8. [30] R. Roy, S. Hohng, T. Ha, A practical guide to single-molecule FRET, Nat. Methods 5 (2008) 507–516. [31] L. Shang, Y. Cheng, Y. Zhao, Emerging droplet microfluidics, Chem. Rev. 117 (2017) 7964–8040. [32] K. Shirai, K. Mawatari, R. Ohta, et al., A single-molecule ELISA device utilizing nanofluidics, Analyst 143 (2018) 943–948. [33] H. Song, D.L. Chen, R.F. Ismagilov, Reactions in droplets in microfluidic channels, Angew. Chem. Int. Ed. 45 (2006) 7336–7356. [34] D. Singh, T. Ha, Understanding the molecular mechanisms of the CRISPR toolbox using single molecule approaches, ACS Chem. Biol. 13 (2018) 516–526. [35] S. Syed, M. Pandey, S.S. Patel, et al., Single-molecule fluorescence reveals the unwinding stepping mechanism of replicative helicase, Cell Rep. 6 (2014) 1037–1045. [36] Y. Ding, X. Liu, J. Zhu, L. Wang, W. Jiang, Quantitative single-molecule detection of protein based on DNA tetrahedron fluorescent nanolabels, Talanta 125 (2014) 393–399. [37] S. Lee, S.H. Kang, Wide-range quantification of human thyroid-stimulating hormone using gold-nanopatterned single-molecule sandwich immunoassay chip, Talanta 99 (2012) 1030–1034. [38] L. Chang, D.M. Rissin, D.R. Fournier, et al., Single molecule enzyme-linked immunosorbent assays: theoretical considerations, J. Immunol. Methods 378 (2012) 102–115. [39] B.K. Duan, P.E. Cavanagh, X. Li, et al., Ultrasensitive single-molecule enzyme detection and analysis using a polymer microarray, Anal. Chem. 90 (2018) 3091–3098. [40] D.M. Rissin, D.R. Fournier, T. Piech, et al., Simultaneous detection of single molecules and singulated ensembles of molecules enables immunoassays with broad dynamic range, Anal. Chem. 83 (2011) 2279–2285. [41] D.M. Rissin, C.W. Kan, T.G. Campbell, et al., Single-molecule enzyme-linked immunosorbent assay detects serum proteins at subfemtomolar concentrations, Nat. Biotechnol. 28 (2011) 595–599. [42] I. Tischer, H. Gelderblom, W. Vettermann, et al., A very small porcine virus with circular single-stranded DNA, Nature 295 (1982) 64–66. [43] D. Kim, Q. Wei, D.H. Kim, et al., Enzyme-free nucleic acid amplification assay using a cellphone-based well plate fluorescence reader, Anal. Chem. 90 (2017) 690–695. [44] D.Y. Joh, A.M. Hucknall, Q. Wei, et al., Inkjet printed point-of-care immunoassay on a nanoscale polymer brush enables sub-picomolar detection of analytes in blood, Proc. Natl. Acad. Sci. U. S. A 114 (2017) E7054–E7062. [45] G. Je, B. Croop, S. Basu, et al., Endogenous alpha-synuclein protein analysis from human brain tissues using single-molecule pull-down assay, Anal. Chem. 89 (2017) 13044–13048. [46] G.I. Mashanov, D. Tacon, A.E. Knight, et al., Visualizing single molecules inside living cells using total internal reflection fluorescence microscopy, Methods 29 (2003) 142–152. [47] M. Sharafeldin, G.W. Bishop, S. Bhakta, et al., Fe3O4 nanoparticles on graphene oxide sheets for isolation and ultrasensitive amperometric detection of cancer biomarker proteins, Biosens. Bioelectron. 91 (2017) 359–366. [48] Y. Shan, Y. Zhang, W. Kang, et al., Quantitative and selective DNA detection with portable personal glucose meter using loop-based DNA competitive hybridization strategy, Sens. Actuators B 282 (2019) 197–203. [49] Y. Chen, Y. Xianyu, Y. Wang, et al., One-step detection of pathogens and viruses: combining magnetic relaxation switching and magnetic separation, ACS Nano 9 (2015) 3184–3191.

biomarker detection, Biosens. Bioelectron. 108 (2018) 1–8. [4] K. Mahato, P. Chandra, Paper-based miniaturized immunosensor for naked eye ALP detection based on digital image colorimetry integrated with smartphone, Biosens. Bioelectron. 128 (2018) 9–16. [5] M.H. Akhtar, K.K. Hussain, N.G. Gurudatt, et al., Ultrasensitive dual probe immunosensor for the monitoring of nicotine induced-brain derived neurotrophic factor released from cancer cells, Biosens, Bioelectron 116 (2018) 108–115. [6] W.N. Burnette, Western blotting": electrophoretic transfer of proteins from sodium dodecyl sulfate–polyacrylamide gels to unmodified nitrocellulose and radiographic detection with antibody and radioiodinated protein A, Anal. Biochem. 112 (1981) 195–203. [7] Y. Wang, R. Deng, G. Zhang, et al., Rapid and sensitive detection of the food allergen glycinin in powdered milk using a lateral flow colloidal gold immunoassay strip test, J. Agric. Food Chem. 63 (2015) 2172–2178. [8] S.K. Arya, P. Estrela, Electrochemical ELISA-based platform for bladder cancer protein biomarker detection in urine, Biosens. Bioelectron. 117 (2018) 620–627. [9] C.P. Jia, X.Q. Zhong, B. Hua, et al., Nano-ELISA for highly sensitive protein detection, Biosens. Bioelectron. 24 (2009) 2836–2841. [10] S. Chen, M. Svedendahl, R.P. Duyne, et al., Plasmon-enhanced colorimetric ELISA with single molecule sensitivity, Nano Lett. 11 (2011) 1826–1830. [11] D. Deng, Y. Hao, J. Xue, et al., A colorimetric enzyme-linked immunosorbent assay with CuO nanoparticles as signal labels based on the growth of gold nanoparticles in situ, Nanomaterials 9 (2018) E4. [12] T.O. Kohl, C.A. Ascoli, Indirect immunometric ELISA, Cold Spring Harb. Protoc. (2017) 396–401. [13] U. Kavita, Y. Dai, L. Salvador, et al., Development of a chemiluminescent ELISA method for the detection of total anti-adeno associated virus serotype 9 (AAV9) antibodies, Hum. Gene Ther. Methods (2018) 131 hgtb.2018. [14] Y. Lv, R. Wu, K. Feng, et al., Highly sensitive and accurate detection of C-reactive protein by CdSe/ZnS quantum dot-based fluorescence-linked immunosorbent assay, J. Nanobiotechnol. 15 (2017) 35. [15] B. Berq, B. Cortazar, D. Tseng, et al., Cellphone-based hand-held microplate reader for point-of-care testing of enzyme-linked immunosorbent assays, ACS Nano 9 (2015) 7857–7866. [16] S.K. Ludwig, H. Zhu, S. Phillip, et al., Cellphone-based detection platform for rbST biomarker analysis in milk extracts using a microsphere fluorescence immunoassay, Anal. Bioanal. Chem. 406 (2014) 6857–6866. [17] A.E. Barron, C.A. Mirkin, C. Liu, A bio-barcode assay for on-chip attomolar-sensitivity protein detection, Lab Chip 6 (2006) 1293–1299. [18] Y.P. Bao, T. F Wei, P.A. Lefebvre, et al., Detection of protein analytes via nanoparticle-based bio bar code technology, Anal. Chem. 78 (2006) 2055–2059. [19] L. Cognet, C. Tardin, D. Boyer, et al., Single metallic nanoparticle imaging for protein detection in cells, Proc. Natl. Acad. Sci. U. S. A 100 (2003) 11350–11355. [20] Y. Cui, Q. Wei, H. Park, et al., Nanowire nanosensors for highly sensitive and selective detection of biological and chemical species, Science 293 (2001) 1289–1292. [21] D.A. Giljohann, C.A. Mirkin, Drivers of biodiagnostic development, Nature 462 (2009) 461–464. [22] J.M. Nam, C.S. Thaxton, C.A. Mirkin, Nanoparticle-based bio-bar codes for the ultrasensitive detection of proteins, Science 301 (2003) 1884–1886. [23] C.S. Thaxton, R. Elghanian, A.D. Thomas, et al., Nanoparticle-based bio-barcode assay redefines "undetectable" PSA and biochemical recurrence after radical prostatectomy, Proc. Natl. Acad. Sci. U. S. A 106 (2009) 18437–18442. [24] R.S. Riley, J.R. Melamed, E.S. Day, Enzyme-linked immunosorbent assay to quantify targeting molecules on nanoparticles, Methods Mol. Biol. 1831 (2018) 145–157. [25] A.Y. Husbands, V. Aqqarwal, T. Ha, et al., Planta single-molecule pull-down reveals tetrameric stoichiometry of HD-ZIPIII:LITTLE ZIPPER complexes, Plant Cell 28 (2016) 1783–1794. [26] A. Jain, R. Liu, B. Ramani, et al., Probing cellular protein complexes using singlemolecule pull-down, Nature 473 (2011) 484–488.

11