Article Cite This: Anal. Chem. XXXX, XXX, XXX-XXX
pubs.acs.org/ac
Nanoelectronic Platform for Ultrasensitive Detection of Protein Biomarkers in Serum using DNA Amplification Luye Mu,*,†,⊥ Ilia A. Droujinine,*,‡,⊥ Jieun Lee,† Mathias Wipf,† Paschall Davis,† Chris Adams,§ Jennifer Hannant,§ and Mark A. Reed*,†,∥ †
Department of Electrical Engineering, Yale University, New Haven, Connecticut 06511, United States Department of Genetics, Harvard Medical School, Boston, Massachusetts 02115, United States § QuantuMDx Group, Newcastle NE1 2JQ, United Kingdom ∥ Department of Applied Physics, Yale University, New Haven, Connecticut 06511, United States ‡
S Supporting Information *
ABSTRACT: Silicon nanowire field effect transistors (NWFETs) are low noise, low power, ultrasensitive biosensors that are highly amenable to integration. However, using NWFETs to achieve direct protein detection in physiological buffers such as blood serum remains difficult due to Debye screening, nonspecific binding, and stringent functionalization requirements. In this work, we performed an indirect sandwich immunoassay in serum combined with exponential DNA amplification and pH measurement by ultrasensitive NWFET sensors. Measurements of model cytokine interleukin-2 concentrations from 200 pM were demonstrated, surpassing the conventional NWFET urease-based readout. Our approach paves way for future development of universal, highly sensitive, miniaturized, and integrated nanoelectronic devices that can be applied to a wide variety of analytes.
C
using an enzyme (such as urease) to catalyze pH change;13−15 however, the linear amplification rate of the enzyme limits the potential sensitivity of the system (Figure 1). Owing to the exponential nature of nucleic acid amplification, immunoassays can be made orders of magnitude more sensitive by replacing the conventionally used enzyme with a DNA probe16−18 in a process called immuno-polymerase chain reaction (immuno-PCR) (Figure 1). Despite its advantages, immuno-PCR has never been used to improve protein detection by NWFET devices. In this work, we performed indirect sandwich immunoassay combined with downstream PCR, with signal transduction by NWFET sensors. During DNA amplification, the attachment of each nucleotide to the nucleic acid strand resulted in the release of a proton, changing the pH of the solution (see Figure S1 for mechanism). Our ultrasensitive NWFETs with superb pH resolution allowed us to detect these minute changes and correlate them to the original protein analyte concentration. We were able to achieve highly sensitive protein detection in full serum, as demonstrated using a model cytokine interleukin-2 (IL-2), while retaining the aforementioned benefits of an electronic readout system. We showed detection of IL-2 concentrations from 200 pM, surpassing both the linear NWFET urease-based readout as well as the conventional ELISA on which the model is based.
linical monitoring of biomarkers for disease diagnosis, treatment, or forensic applications often requires the identification and quantification of low levels of proteins in small volumes of complex biological samples. Silicon nanowire field effect transistors (NWFETs) have shown superb performance as low noise, low power, and high sensitivity biosensors, and NWFETs fabricated using CMOS-compatible techniques are particularly desirable for their potential ease of integration on-chip with other electrical components (resistive heaters, temperature sensors, signal processing circuitry, etc.). Therefore, the development of highly sensitive and specific NWFETbased protein detection strategies is of high interest. Most of the work on NWFET detection of proteins thus far has been performed in low ionic strength solutions;1−3 however, detection in high ionic strength, complex physiological buffers such as blood serum remains challenging due to nonspecific adsorption, Debye screening, stringent functionalization requirements, etc., and often require desalting methods which can alter protein stability.4−7 For these reasons, an indirect sandwich immunoassay that converts the presence of the target protein to an amplified pH response is highly desirable, with the added benefit of two antibodies for increased specificity. Due to the small size of protons, pH detection is immune to the limitations associated with direct protein sensing. NWFETs have been previously used to detect pH changes due to cellular activation,8 enzyme−substrate interactions,9 and buffer exchange.10−12 Electronic enzyme-linked immunosorbent assays (ELISAs) have also been reported to detect proteins indirectly, © XXXX American Chemical Society
Received: May 28, 2017 Accepted: September 22, 2017
A
DOI: 10.1021/acs.analchem.7b02036 Anal. Chem. XXXX, XXX, XXX−XXX
Article
Analytical Chemistry
Figure 1. Schematic depiction of linearly amplifying vs exponentially amplifying immunoassays. t is time, [E] is the enzyme concentration, [DNAi] is the initial DNA concentration, and α is the thermocycling rate (cycles per time). If we assume that the concentration of the analyte of interest is equivalent in both cases, and the same linker is used to attach the enzyme and the DNA to the detection antibody, then [E] = [DNAi].
■
EXPERIMENTAL SECTION NWFET Sensing Setup. A photograph of the research grade portable sensing board is shown in Figure S2a, and a schematic is shown in Figure 2a. This printed circuit board (PCB) was specifically designed as a prototype for diagnostics of potential problems; future designs will eliminate the manual components, drastically reducing the PCB size into a small chip. The PCB was powered by batteries and fully portable. The current PCB can measure up to 32 devices simultaneously. The NWFET chips were wirebonded to 68 pin leadless chip carriers (LCCs) and inserted into the LCC socket on the board for measurement. Solution was sequentially introduced via a microfluidic channel for measurement. Solution withdrawal allowed quick exchanges between samples with each measurement requiring no more than a few microliters of solution. The board was connected to a portable USB data acquisition card (DAQ; National Instruments). The DAQ supplied the drain (typically 0.1 to 0.5 V) and solution gate voltages (typically 0 to −2 V) to the chip. The drain current from each device was fed into the inverting terminal of a quad low noise op-amp (LT1125; Digikey). The op-amp circuit was designed to be a current-to-voltage converter with a gain of 106 (1 V/μA) and a low pass filter with a cutoff frequency of 1.6 Hz (R = 1 MΩ, C = 0.1 μF). The resulting voltages were fed into a 32 channel multiplexer (ADG732BSUZ; Newark). The multiplexer sequentially passed through the voltage output of each channel into the DAQ, and the signal of each channel was individually monitored and recorded. Board control and data recording were achieved via Labview programs. All time-trace data presented in this work were taken at a rate of 4 Hz/channel. NWFET Fabrication. The NWFETs were fabricated in a foundry process (IME) from 8″ silicon-on-insulator wafers (SOITech; resistivity 14−19 Ωcm) using a process similar to those reported previously.7,19 p-type nanowires were fabricated from wafers with buried oxide thickness around 150 nm and active silicon layer thickness around 140 nm. To ensure compatibility with CMOS processing, we employed top-down deep UV optical lithography combined with interferometric masking and self-limiting silicon oxidation to form the nanowire arrays. Each NWFET device contained five nanowires in parallel within a width of 100 μm with the length less than 20 μm, and individual wire widths varied depending on the wafer, but typically by no more than 180 nm. To achieve maximum pH sensitivity, we atomically deposited Al2O3 (Oxford Instruments) as the interfacial dielectric with typical thickness below 15 nm. Layer thickness and compositional purity of the dielectric layer were evaluated as a function of material thickness using Auger spectral analysis, as any defect or contamination in the sensing dielectric had adverse effects on
the NWFET performance. Aluminum source and drain leads were passivated by a layer of SiO2 to prevent leakage into solution. Post fabrication and prior to use, the devices were cleaned using acetone, 1-propanol, and ultrapure water followed by drying in a stream of nitrogen and baking at 110 °C for 30 min. Finally, devices were oxygen plasma cleaned (Henniker Plasma) at 30 W for 5 min at 10 sccm and thereafter stored under vacuum. PCR for Nanodrop, Gel Electrophoresis, and NWFET Detection. Each PCR sample consisted of 100 μL of solution containing 0.4 μM of each primer, 0.4 mM of each dNTP (New England Biolabs), 1 mg/mL of ultrapure bovine serum albumin (BSA; Thermo Fisher Scientific), 3 mM of MgCl2 (New England Biolabs), 50 mM of KCl, 0.1% of Triton-X 100, 0.8 M of betaine (Sigma-Aldrich), 100 μM of phenol red (only in nanodrop and gel electrophoresis samples), 0.05 units/μL of Taq polymerase (Takara), and various amounts of starting target DNA concentration in deionized (DI), ultrapure, DNase and RNase-free water (Sigma-Aldrich). All materials were purchased or dissolved in pure DI water. The ingredients were combined to form a Mastermix, and the pH of the mix was tuned to approximately 8.5 using fine-increment pH paper (VWR) before adding in phenol red and Taq polymerase. Phenol red was added in some of the samples to provide a visual indication as well as a quantitative absorbance analysis of the pH change but was left out of the samples for NWFET measurement in case it interfered with sensing. Phenol red supplementation up to 0.3% (8.47 mM) was found to not interfere with amplification.20 The Mastermix was then divided into PCR tubes, and various amounts of DNA were added to the individual tubes. PCR was set to run using the following conditions: 10 min of denaturation at 95 °C, 55 cycles of 30 s at 95 °C, 30 s at 48 °C, and 2 min at 72 °C. Thermocycling was paused temporarily at the end of 30 cycles to remove the corresponding 30 cycle tubes, and the program was resumed for the 55 cycle tubes. Immuno-PCR. The capture antibody, detection antibody, and IL-2 were purchased from Affymetrix and were the same as those used in ELISA. Robostrips (Analytik Jena) for immunoPCR were coated with 100 μL/well of IL-2 capture antibody in coating buffer overnight at 4 °C. The wells were washed once for 1 min with 200 μL/well of wash buffer (1× phosphate buffered saline (PBS), 0.05% Tween-20) and 3 times for 1 min each with 1× PBS. Blocking was done with 200 μL of the Blocking Solution (Candor) at room temperature for 2 h with high-speed shaking. After washing once for 1 min with wash buffer, varying concentrations of IL-2 antigen were prepared in 100% fetal bovine serum (FBS), and 100 μL of the antigen solution was added per well and incubated for 3.5 h at room B
DOI: 10.1021/acs.analchem.7b02036 Anal. Chem. XXXX, XXX, XXX−XXX
Article
Analytical Chemistry
Figure 2. Measurement setup and device characteristics. (a) Schematic of portable sensing board for simultaneous device measurement. See also Figure S2 for NWFET sensing setup photograph and packaged device. (b) Optical image of NWFET devices on a sensor chip. The beige-colored sections are aluminum source and drain metallization, blue sections are Al2O3, and ochre sections are passivating SiO2. Scale bar is 200 μm. (c) Drain current and transconductance vs gate voltage scans of 15 devices on one chip. (d) pH sensing over the range of interest from a representative device. pH of the flowed solutions is shown on top. (e) Signal responses to gate steps on a representative device. Changes in applied gate voltage are shown on top.
temperature with gentle shaking. The wells were washed 4 times for 1 min each with wash buffer, and 100 μL of biotinylated detection antibody was added and incubated for 1 h at room temperature with gentle shaking. The wells were washed 4 times for 1 min each with wash buffer, and 100 μL of 1 μg/mL streptavidin in diluent (1× PBS, 0.05% Tween-20, 3% biotin-free BSA) was added and incubated for 30 min at room
temperature with gentle shaking. The wells were washed 6 times for 1 min each, followed by the addition of 100 μL of 40 nM biotinylated reporter DNA in diluent. After 30 min of incubation at room temperature with shaking, the wells were washed 8 times for 1 min each, and 100 μL of PCR Mastermix (same recipe as in PCR for NWFET detection, but without DNA) was added to each well, and 30 cycles of PCR were C
DOI: 10.1021/acs.analchem.7b02036 Anal. Chem. XXXX, XXX, XXX−XXX
Article
Analytical Chemistry
Figure 3. NWFET pH detection from PCR amplification and colorimetric verification. (a) NWFET detection and (b) Nanodrop absorbance ratio at 30 and 55 cycles. Dotted line depicts signal level of the no DNA control. The error bars represent standard error of the mean (SEM) from 3 Nanodrop measurements or 13 device measurements. (c) Representative example of a real-time trace from NWFET detection at 30 cycles. (d) Representative photographs of samples after PCR amplification. Phenol red indicator allowed for visualization of pH changes and served as positive control verification for NWFET detection.
an ideal material for this application due to its near-Nernstian pH sensitivity. Device sensitivity was assessed by flowing buffers of pH 4, 6, and 8, sequentially over the NWFETs (Figure 2d). The surface potential signal (Δψ) was extracted by taking the difference between the measured current value (Id) and the current value at time 0 (Id0) and dividing by the device transconductance (gm):
carried out with the same conditions as those described for PCR for NWFET detection. All other experimental descriptions are presented in detail in Supporting Information.
■
RESULTS AND DISCUSSION
Each NWFET device has a multifingered structure, with five parallel fingers for improved device stability and uniformity across the chip. A close-up of representative NWFET devices on a chip is shown in Figure 2b, and a scanning electron micrograph of a single nanowire is shown in Figure S2c. In a typical experiment, 8−16 devices were measured simultaneously to obtain an average reading of the pH signal. Figure 2c presents the gating behavior from 15 devices on one chip, showing excellent uniformity. Measurements were taken in 1× PBS with the solution gate voltage swept from 0 to −2 V then back to 0 V. Negligible hysteresis was observed in the NWFET devices. NWFETs are pH sensitive devices owing to the amphoteric nature of their sensing dielectric. Under acidic conditions, the surface increases in protonation, and the surface potential increases, whereas under basic condition, the opposite happens. The change in surface potential acts as an external gate, changing the current through the underlying NWFET. Al2O3 is
Δψ =
(Id − Id0) gm
(1)
which has been shown to be an effective method of normalizing variations between devices.21,22 We obtained a pH sensitivity of 59.4 ± 2.2 mV/pH (9 devices), close to the Nernst limit of 59.1 mV/pH at room temperature. Apart from sensitivity, it was also important to characterize the electrical noise of our devices to understand the absolute limit of performance. We conducted noise measurements (Figure S3) to estimate the theoretical limit of detection for these devices. For a signal-to-noise ratio (SNR) of 1, the smallest detectable voltage was determined to be 0.13 ± 0.01 mV (as calculated using equations provided in Supplementary Methods). This corresponded to a minimum detectable pH change of 2.2 × 10−3 units, demonstrating the excellent resolution of these devices. To verify this, we also characterized the minimum detectable voltage of our system by D
DOI: 10.1021/acs.analchem.7b02036 Anal. Chem. XXXX, XXX, XXX−XXX
Article
Analytical Chemistry
Figure 4. Assay comparison between exponential amplification (immuno-PCR) vs linear amplification (ELISA-urease) for protein detection in serum with NWFETs. (a) Procedure for stepwise coupling of DNA to antibody−antigen complex during immuno-PCR. (b) IL-2 detection through immuno-PCR. The error bars represent SEM from 15 devices. (c) IL-2 detection through ELISA-urease (note different x axis scale). The error bars represent SEM from 16 devices. The cyan region represents the dynamic range specified by the IL2 ELISA kit (see Figure S5).
would happen in immuno-PCR, we presynthesized and purified a 1074 base pair biotinylated sequence from the pUC19 plasmid and performed PCR with nonbiotinylated primers. For NWFET measurements, approximately 15 μL of each sample was sequentially withdrawn from the PCR tubes into the microchannel over the NWFETs at a flow rate of 20 μL/ min, separated by small air gaps to prevent mixing of the solutions, and measured on the NWFET setup. The change in surface potential was extracted as described in eq 1, with Id being the average current signal for each sample and Id0 being the current signal for the no DNA sample after 30 cycles, and plotted in Figure 3a. Lower cycle numbers (30 cycles) did not amplify the smaller (
