Real-time detection of cardiac troponin I and mechanism analysis of AlGaAs/GaAs high electron mobility transistor biosensor

Acute myocardial infarctions (AMIs) are one of the top five most expensive conditions during inpatient hospitalizations.¹ In 2015, about 15.9 × 10⁶ myocardial infarctions occurred worldwide. Cardiac muscle cells fracture when myocardial infarction occurs, and this process releases biomarkers into the blood, which could be measured by a blood test to diagnose myocardial infarctions.² A rapid diagnosis test is of great importance in the treatment of myocardial infarctions.³ The concentration of cardiac troponin I (cTnI) is efficient in diagnosing or excluding myocardial infarctions.⁴ According to the Guideline for the Management of ST-Elevation Myocardial Infarction, cTnI is a gold standard marker for the detection of AMI.⁵ Now, enzyme-linked immunosorbent assay (ELISA) is the most common method of measuring protein,^6–8 but this method is not quick enough, as it takes about 2 h. There are motivations to find a real-time cTnI detection method.

Field effect transistor (FET) biosensors show plenty of advantages, such as high-sensitivity, rapid response time, and label-free detection. FET biosensor research studies have been intensively enhanced in the world because of those advantages.^9–11 There are several kinds of FET-based biosensors, such as Si-FET¹² and GaN high electron mobility transistors (HEMTs).¹³ Furthermore, the FET fabrication process is compatible with micro-electronics technology, which increases the possibility of miniaturization and mass production. The FET biosensor can achieve mass production of low-cost, and it is of great medical value and market prosperity. In particular, high electron mobility transistor (HEMT) biosensors are attracting a lot of attention.¹⁴ Attributing to the wurtzite lattice structure, GaAs and GaN structures lack a symmetric center, so there is spontaneous polarization in their structure.¹⁵ This phenomenon enhances two-dimensional electron gas concentration in the channel. The two-dimensional electron gas channel is close to the surface, so it could be affected by the external factors as a sensing device.

HEMT biosensor devices are reported to realize the label-free detection of protein.¹⁶ The biosensor HEMT is, in fact, nothing other than a conventional HEMT with a float gate metal. The gate metal in the conventional HEMT is extracted out to an external circuit on which a gate voltage can be applied directly. In the biosensor HEMT, the gate metal is not extracted out, rather than modified with biomolecules as a sensing district. Electric biomolecules absorbed on the sensing gate region will induce an opposite charge in the channel of the HEMT and influence the channel conductance of the HEMT.¹⁷ Protein is charged in the solution, and when the pH of the solution is above a certain value, the protein molecule will lose H+ to be negatively charged; on the contrary, the protein absorbs H+ to be positively charged. The isoelectric point (PI) is the value that the protein is not charged, and different kinds of proteins have different PI. This suggests the possible application of this kind of biosensor as a real time cardiac troponin I detection method. Si-FET^18,19 and GaN-HEMT^20,21 biosensors have been reported to realize cardiac troponin I rapid detection. Compared to AlGaN/GaN, AlGaAs/GaAs as the second generation semiconductor has a more mature fabrication technology.

In this paper, a high-sensitivity AlGaAs/GaAs HEMT biosensor was fabricated; we first applied the AlGaAs/GaAs HEMT biosensor to measure the concentration of cardiac troponin I. We studied the range of detection and the response time of the biosensor. The result indicates that the HEMT biosensor is capable of detecting the concentration of cTnI from 1 pg/ml to 10 ng/ml. Furthermore, we further investigated the selectivity of the antibody modified HEMT biosensor by comparing the signal of bovine serum albumin (BSA), serum of healthy human (SHH), and serum of the AMI patient (SAP). We calculated the equilibrium (dissociation) constant of the cTnI and cTnI antibody. This result shows that the detection method will have the capability to realize real-time and high-sensitivity detection.

The HEMT epitaxial structure, 300 nm buffer layer/15 nm channel layer/4 nm spacer layer/Si δ-doping layer/25 nm barrier layer/30 nm cap layer, was grown by molecular beam epitaxy (MBE) on the GaAs substrate. Its epitaxial structure and fabricated process are shown in Fig. 1(a).

The fabrication process is as follows: The first step of photolithography is to form bench, the region where the photoresist does not cover is etched, and this process is to isolate a single device. The second step of photolithography is to fabricate the source and drain electrodes, which is performed by electron beam evaporation with Ge/Au/Ni/Au; then, the wafer is sent to the process of rapid thermal annealing to form an Ohmic contact. The third step of photolithography is to fabricate the gate electrode, which is performed by electron beam evaporation with Ti/Au. The AlGaAs/GaAs HEMT has been fabricated successfully. The device works in an aqueous environment. To further protect the device, the devices are covered with SiO₂. The fourth step of photolithography is to expose the source–drain and gate regions.

In our work, we have produced two HEMT structures for analyzing the performance of biosensors, named biosensor HEMT (Bio-HEMT) and conventional HEMT (Con-HEMT) device. Figure 1(b) shows Bio-HEMT and Con-HEMT optical microscopy images. Figure 1(c) shows the device structure and sensing district modification schematic. Sensing district modification is a critical process of the HEMT biosensor. In the HEMT biosensor, the Au gate usually connects with biomolecules via mercaptan acid.^14,16 In this kind of method, one side of the molecule is thiol and the other side is carboxyl, so this molecule can connect the Au gate and protein directly. In our study, we choose 6-mercaptohexanoic acid to connect the Au gate and cTnI antibody.

N-hydroxysuccinimide (NHS) and N-(3-dimethylaminopropyl)-N′-ethylcarbodiimide-hydrochloride (EDC) were purchased from Aladdin. Phosphate buffered saline (PBS), 6-mercaptohexanoic acid, and the cTnI and cTnI antibody produced in rabbit were purchased from Sigma. EDC, NHS, PBS, and 6-mercaptohexanoic acid were dissolved in deionized water; the cTnI and cTnI antibody were dissolved in 0.01 mM PBS solution. To immobilize the cTnI-antibody, the chip was immersed in 6-mercaptohexanoic acid (10 mM in deionized water) for 24 h to form a self-assembled monolayer. Then, it was applied to the activation solution, which was mixed by 20 mM EDC and 50 mM NHS 1:1 by volume for 15 min. This process increases the stability of the active intermediate.²² After the chip was submerged in the 100 ng/ml cTnI-antibody solution for 2 h, immobilization of the cTnI antibody on the gate metal was completed. After gate modification, we used this device to measure the concentration of cTnI solutions. cTnI solutions (from 1 pg/ml to 10 ng/ml) were dipped to the gate separately. The gate was washed by the PBS solution after the measurement of each concentration. At each concentration, we measured the electrical characteristics of the biosensor HEMT. The electrical characterization of the biosensor HEMT was measured at 25 °C by the CHI 660D Electrochemical Workstation.

In order to analyze the sensitivity of the HEMT to the cTnI antigen, we compared the response of the sensor before and after gate functionalization. Figure 2(a) shows the electrical characteristics of the Bio-HEMT without gate modification for detecting different concentrations of the cTnI antigen. Obviously, the drain current (I_SD)–drain bias (V_SD) characteristics of the Bio-HEMT sensor do not vary with cTnI concentrations. This result indicates directly that physical absorption of cTnI on the gate is negligible and has little influence to gate. After the gate was functionalized by the cTnI antibody, we detected again the drain–source current of the devices as a function of the bias voltage from 0 V to 1 V, with concentrations ranging from 1 pg/ml to 10 ng/ml. As shown in Fig. 2(b), when the concentration of the cTnI antigen solution increases, I_SD decreases with a total decrease of 11%. For example, at V_SD = 0.2 V, I_SD is 326.3 µA at 0 pg/ml, 310.1 µA at 1 pg/ml, 303.5 µA at 10 pg/ml, 299.5 µA at 100 pg/ml, 294.2 µA at 1 ng/ml, and 290.2 µA at 10 ng/ml. When the concentration of cTnI increases, more cTnI biomolecules connect to the gate and influence the conductance of the channel. Compared to the Bio-HEMT without gate modification, the current signals of the gate modified Bio-HEMT change significantly. Therefore, the result indicates that the AlGaAs/GaAs HEMT biosensor is able to detect cTnI at a relatively broad range of concentrations, varying from 1 pg/ml to 10 ng/ml. The limit of detection could be as low as 1 pg/ml.

The response time is a very important indicator for early warning of acute myocardial infarction. At the V_SD of 0.2 V, the time dependence of current at different concentrations of cTnI (1 pg/ml, 100 pg/ml, and 10 ng/ml) is plotted in the insets of Fig. 3(a). When the cTnI solution is dropped onto the sensing district, the current acutely changes, and then, the current becomes stable. It is due to the mechanical vibration caused by the sudden drop of the cTnI solution. The criterion of the current stability moment is that current noise becomes less than 0.1 µA within 10 s. The response time of the cTnI detection via the AlGaAs/GaAs HEMT biosensor is shown in Fig. 3(a). The result reveals that the current becomes stable within 15 s–30 s. Therefore, this HEMT biosensor has a great advantage over the traditional measurement method (ELISA) that takes 1 h–2 h. This kind of HEMT biosensor could realize real-time detection.

To further investigate the selectivity of the biosensors, we dropped bovine serum albumin (BSA), serum of healthy human (SHH), and serum of the AMI patient (SAP) onto the sensing district. Figure 3(b) shows the electrical characteristics of the HEMT biosensor after submerging in PBS, BSA, SHH, and SAP solutions. Figure 3(c) compares the drain–source current variation amount of BSA, SHH, SAP-1, and SAP-2 solutions at a constant drain–source bias voltage of 200 mV. Since the cTnI antibody was blocked by BSA with a high concentration, no response of the target BSA was observed as anticipated. It is clear that the patient’s current variation is 4–6 times that of the normal person. The concentration of the cTnI antigen in the serum of healthy human is very low, so the current variation is insignificant. However, a large number of cTnI antigens are released from the serum of AMI patients, once SAP is introduced into the HEMT biosensor, and the current changes greatly. This result indicates that the gate modified HEMT biosensors have selectivity. According to the patient’s blood test results by ELISA, the cTnI concentration of SAP-1 is 156 pg/ml and the cTnI concentration of SAP-2 is 8.77 ng/ml. It is consistent with the fact that the antibody combines specifically with the antigen. This advantage simplifies the procedures of the sample purification, and it further reduces the time consumption of diagnosis.

The sensing mechanism of this kind of biosensor has already been investigated widely.²³ The threshold voltage (V_T) is an important parameter that influences the electron concentration in the channel,

The first term reflects the difference in work function of the gate metal (φ_m) and the semiconductor (φ_s), the second term is due to the accumulated charge in the dielectric (Q_ox) at the dielectric–semiconductor interface (Q_ss) and the depletion charge in the semiconductor (Q_B), whereas the last term determines the onset of inversion depending on the doping level of the semiconductor. When electric biomolecules are absorbed on the gate, they will induce an opposite charge in the channel. This effect will influence the threshold voltage.^17,24 The amount of threshold voltage variation ΔV_T is

where q_A is the electric charge contributed by the adsorbed analyte, K_D is the equilibrium (dissociation) constant, and [A] and [B]_max represent the concentrations of the analyte in solution and the maximum surface density of functional binding sites, respectively. Figure 4(a) shows the schematics of the Bio-HEMT mechanism.

On the chip, we measured the sensing properties of Con-HEMT after gate modification as a control group. The sensing district size and modification condition of Con-HEMT and Bio-HEMT are identical. The Con-HEMT was exposed to the cTnI solution, ranging from 1 pg/ml to 10 ng/ml. Figure 4(b) shows I_SD–V_SD of Con-HEMT. When the Con-HEMT was exposed to the cTnI solution, to our surprise, there was also a significant current variation. For example, at V_SD = 0.2 V, I_SD is 119.9 µA at 0 pg/ml, 111.8 µA at 1 pg/ml, 104.0 µA at 10 pg/ml, 101.8 µA at 100 pg/ml, 94.8 µA at 1 ng/ml, and 94.2 µA at 10 ng/ml. When the concentration of cTnI increases from 0 pg/ml to 10 ng/ml, the current decreases by 21.4%. In the result, the current variation of Con-HEMT is not equal to the current variation of Bio-HEMT, which is 11.1%.

The absorbed charge effect could not explain the phenomenon of Con-HEMT. The gate metal is extracted out in Con-HEMT, so the absorbed charges will be screened by the metal. The absorbed charge effect will not induce an opposite charge in the channel in Con-HEMT. The schematic is shown in Fig. 4(a). There must be some other variable parameter. In Eq. (2), the only variable parameter that can be affected by the absorbed biomolecule is φ_m. We make a speculation that the work function changes once the biomolecules are absorbed on the gate metal. In fact, the phenomenon that the absorbed organic molecules will influence the metal work function has been reported.^25–27 If the analytes change the work function of the gate metal, the phenomenon will exist in both Bio-HEMT and Con-HEMT. In a word, the Con-HEMT is only influenced by the work function variation. The Bio-HEMT is influenced by both the absorbed charge effect and work function variation. The difference of signal variation between Bio-HEMT and Con-HEMT is individually influenced by the absorbed electric biomolecule effect.

We compared the relative signal change of Con-HEMT and Bio-HEMT at V_SD = 0.2 V. The relative signal variation of Bio-HEMT and Con-HEMT is presented in Fig. 4(c). The result indicates that the relative signal variation of Bio-HEMT is less than Con-HEMT at each cTnI concentration. This result indicates that the work function variation reduces channel conductance, but the absorbed electric biomolecules enhance channel conductance. Inferring from this result, cTnI is positively charged in 0.01 mM PBS (PH = 7.4), so the isoelectric point of cTnI should be above 7.4. This result is in agreement with the theoretical isoelectric point of cTnI (PI = 9.87).²⁸ 

The signal variation induced by the absorbed charge can be used to calculate the protein equilibrium (dissociation) constant. According to our research, both work function variation and absorbed electric biomolecule effect influenced the threshold voltage. We eliminated signal change induced by the work function variation via Bio-HEMT and Con-HEMT. The equilibrium (dissociation) constant of the cTnI and cTnI antibody was calculated by the difference of signal variation. The data used to calculate the equilibrium constant are presented in Table I. The calculated fitted curve is presented in Fig. 4(d). The result indicates that the equilibrium (dissociation) constant (K_D) of the cTnI and cTnI antibody is 6.486 ± 3.050 pg/ml, which is converted to be 0.27 pM. It is consistent with the recently reported values via the AlGaN/GaN HEMT biosensor (K_D = 2.62 pg/ml).²⁰ 

Compared with the previously reported cardiac troponin I detection method, the limit of detection is 5 pg/ml of the Si-FET¹⁸ and 6 pg/ml of the GaN-HEMT;²⁰ in our work, the limit of detection of cTnI by the GaAs HEMT biosensor is 1 pg/ml. Hence, the GaAs HEMT biosensor has the capability to realize cTnI high sensitivity and real-time detection.

In summary, we have designed and fabricated AlGaAs/GaAs HEMT biosensors for detecting acute myocardial infarction markers. The direct physical absorption of cTnI on the Au gate does not influence the threshold voltage. In our result, the detection range of cTnI by the gate-modified AlGaAs/GaAs HEMT biosensor is from 1 pg/ml to 10 ng/ml, and the response time was less than 30 s. In addition, by comparing the serum test results of AMI patients and healthy people, the selectivity of the HEMT biosensor modified by antibodies was verified. The detection method can achieve high sensitivity and real-time detection. Based on the result of Con-HEMT, we make a speculation that the work function variation exists in the sensing process. The author clarifies the abnormal phenomenon that the current decreases in the process of the gate combined with the positively charged cTnI biomolecules. Furthermore, the equilibrium (dissociation) constant of the cTnI and cTnI antibody is calculated (K_D = 6.486 ± 3.050 pg/ml).

J.L. and S.L. contributed equally to this work.

The authors acknowledge the support from the National Key Research and Development Program of China (No. 2017YFB0405400), the Capital Health Research and Development of Special (No. 2020-2-4084), the National Natural Science Foundation of China (Grant Nos. 61874105), and the Youth Innovation Promotion Association CAS (Grant No. Y201925).

Blood collection from patients was approved by the Ethics Review Board of Peking University People’s Hospital. All the patients provided their written informed consent.

The data that support the findings of this study are available from the corresponding author upon reasonable request.