Development of an ImmunoFET for Analysis of Tumour Necrosis Factor- α in Artiﬁcial Saliva: Application for Heart Failure Monitoring

: Assessing tumour necrosis factor- α (TNF- α ) levels in the human body has become an essential tool to recognize heart failure (HF). In this work, label-free, rapid, easy to use ImmunoFET based on an ion-sensitive ﬁeld effect transistor (ISFET) was developed for the detection of TNF- α protein. Monoclonal anti-TNF- α antibodies (anti-TNF- α mAb) were immobilized on an ISFET gate made of silicon nitride (Si 3 N 4 ) after salinization with 11-(triethoxysilyl) undecanal (TESUD). The obtained ISFET functionalized with the mAbs (ImmunoFET) was used to detect TNF- α protein in both phosphate buffer saline (PBS) and artiﬁcial saliva (AS). The change in the threshold voltage of the gate ( ∆ V T ) showed approximately linear dependency on the concentration of the antigens in the range 5–20 pg/mL for both matrixes. The cross-selectivity study showed that the developed ImmunoFET demonstrated to be selective towards TNF- α , when compared to other HF biomarkers such as N-terminal pro-brain natriuretic peptide (NT-proBNP), interleukin-10 (IL-10), and cortisol, even if further experiments have to be carried out for decreasing possible unspeciﬁc absorption phenomena. To the best of our knowledge, this is the ﬁrst ImmunoFET that has been developed based on Si 3 N 4 for TNF- α detection in AS by electrical measurement.


Introduction
A decrease in the performance of the heart to pump and/or fill with blood is commonly known as heart failure (HF). Many different factors such as cardiac abnormalities (e.g., structural abnormalities including stiffness of ventricular chamber, functional abnormalities including mitral incompetence, etc.), coexisting conditions (e.g., diabetes, anaemia, sleep apnoea, etc.), or other factors (e.g., age, genetic influence, life style, etc.) can add up to the cause of HF [1,2]. Statistics show that approximately 26 million people worldwide are affected by HF [3], which causes high mortality rates of hospitalized elderly patients [4] and larger health expenditure. With the increasing prevalence of this cardiac syndrome [3,4], effective and early diagnosis is needed. While there are some distinct symptoms of HF, such as oedema (swelling due to fluid entrapment in the body), dyspnoea (shortness of breath), which is observed in most of the patients just before hospitalization, they are attributed to acute decompensated HF, which is a sudden aggravation of HF [5]. However, before the worsening of symptoms, HF frequently has nonspecific symptoms, making it challenging to detect it in its early stages, which is why the detection of specific biomarkers Procleaner™ (BioForce, Papenburg, Germany) was used in order to activate the ISFET's surface by creating -OH groups.

Fabrication of the ISFET Devices
The ISFET devices were fabricated as N-type FETs on 100 mm diameter silicon wafers, with a boron substrate doping of 10 15 cm −3 . First, a silicon dioxide layer was grown by wet oxidation with a thickness of 800 nm. The oxide is patterned with photolithography and wet etching, and the source and drain regions of the ISFET are doped by ion implantation of phosphorus with a dose of 4.2 × 10 15 cm −2 . The oxide layer was then eliminated from the gate region by photolithography and wet etching. The gate dielectric was added by growing a 78 nm thick silicon dioxide, followed by the deposition of a 100 nm thick silicon nitride layer by low-pressure chemical vapour deposition (LPCVD). The silicon nitride was eliminated from all areas except the gate by photolithography and dry reactive ion etching (RIE). This was followed by the opening of the contact windows by photolithography and RIE etching. The metal interconnect lines are made of 15 nm thick titanium plus 150 nm thick platinum, which were deposited by vacuum evaporation and patterned by photolithography plus lift-off. Finally, a passivation layer of 400 nm of silicon dioxide plus 400 nm of silicon nitride was deposited by plasma-enhanced chemical vapour deposition (PECVD). The passivation was patterned by photolithography plus RIE (for the silicon nitride) and wet etching (for the silicon dioxide) to open the gate area and the electrical contact pads. The gate area of the ISFETs had dimensions of 400 × 20 µm.
To package the ISFETs devices for measurements in liquid solutions, the silicon wafer was diced, and the resulting chips were glued to a printed circuit board (PCB) using an epoxy resin. The electrical contact pads were wire-bonded to the PCB by using 25 µm diameter aluminium wires. Finally, the bonding wires, the external areas of the chip and the electric tracks on the PCB, were encapsulated by the same epoxy resin to protect them from the liquid solution ( Figure 1). Finally, the reference electrode (RE) based on Ag/AgCl was made by electrochemical deposition of Ag on one of microelectrode based on platinum, using sodium nitrate (NaNO 3 ) (1 M) and silver nitrate (AgNO 3 ) (25 mM) at pH 1. The deposited silver layer was chlorinated by overnight incubation of the RE in HCl (1 M) solution. A detailed process of RE fabrication can be found in previous work by our group [40,41].

Functionalization of ISFET Surface with Antibodies
Before the functionalization, the chip, containing four working ISFETs, was thoroughly rinsed with ultrapure water and dried using nitrogen gas (N2) flow. The ISFET was then placed in a UV/Ozone ProCleaner TM for 45 min to remove any remaining organic contaminants as well as to activate the surface of the gate made of silicon nitride (Si3N4) with hydroxyl groups (-OH). Afterwards, the surface was functionalized with 11-(triethoxysilyl) undecanal (TESUD) using the vapour phase method: ISFET with activated surface was placed inside a desiccator that was saturated with TESUD [42][43][44][45] for 45 min. Later, the sample was placed inside a furnace (100 °C) for 1 h to ensure TESUD adhesion onto the gate surface of ISFET through the hydroxyl groups, followed by cleaning in pure ethanol to remove excess TESUD. The ISFET was then covered with 10 μg/mL anti-TNFα mAbs diluted in a mixture of a 10 mM phosphate buffer saline (PBS) pH 7.4 and 4 mM sodium cyanoborohydride (NaBH3CN) in PBS and left overnight for mAbs immobilization during which Abs covalently bond to TESUD. Finally, after removal of the excess solution of mAbs by cleaning the chips with PBS and drying with N2 flow, unreacted chains of TESUD were capped by the addition of ethanolamine (Et-NH2). Briefly, the surface of ISFET with immobilized mAbs (ImmunoFET) was covered with 1% of Et-NH2 diluted in the previously mentioned mixture of 4 mM NaBH3CN in PBS for 30 min, after which the ISFET was once again cleaned and dried using PBS and N2 flow and afterwards placed into containers with PBS solution for storage at 4 °C.

Preparation of Artificial Saliva (AS)
Artificial saliva (AS) was prepared as described elsewhere [11,16,17]. Briefly, AS preparation was carried out by dissolving 4 mg/mL of urea, 4 mg/mL of mucin from porcine stomach, 0.6 mg/mL of sodium phosphate dibasic (Na2HPO4), 0.3 mg/mL of sodium bicarbonate (NaHCO3), 0.6 mg/mL of anhydrous calcium chloride (CaCl2), 0.4 mg/mL of potassium chloride (KCl), and 0.4 mg/mL of sodium chloride (NaCl) in Millipore Milli-Q nanopure water (resistivity > 18 MΩ cm) produced by a Millipore Reagent Water System. Dissolving was aided by placing the solution in ultrasonic bath for 20 min to ensure the solution homogeneity. Afterwards, 0.1 M of sodium hydroxide (NaOH) was added into the solution to obtain pH with the value of 7.2. The prepared AS was stored at 4 °C until further use.

Functionalization of ISFET Surface with Antibodies
Before the functionalization, the chip, containing four working ISFETs, was thoroughly rinsed with ultrapure water and dried using nitrogen gas (N 2 ) flow. The ISFET was then placed in a UV/Ozone ProCleaner TM for 45 min to remove any remaining organic contaminants as well as to activate the surface of the gate made of silicon nitride (Si 3 N 4 ) with hydroxyl groups (-OH). Afterwards, the surface was functionalized with 11-(triethoxysilyl) undecanal (TESUD) using the vapour phase method: ISFET with activated surface was placed inside a desiccator that was saturated with TESUD [42][43][44][45] for 45 min. Later, the sample was placed inside a furnace (100 • C) for 1 h to ensure TESUD adhesion onto the gate surface of ISFET through the hydroxyl groups, followed by cleaning in pure ethanol to remove excess TESUD. The ISFET was then covered with 10 µg/mL anti-TNF-α mAbs diluted in a mixture of a 10 mM phosphate buffer saline (PBS) pH 7.4 and 4 mM sodium cyanoborohydride (NaBH 3 CN) in PBS and left overnight for mAbs immobilization during which Abs covalently bond to TESUD. Finally, after removal of the excess solution of mAbs by cleaning the chips with PBS and drying with N 2 flow, unreacted chains of TESUD were capped by the addition of ethanolamine (Et-NH 2 ). Briefly, the surface of ISFET with immobilized mAbs (ImmunoFET) was covered with 1% of Et-NH2 diluted in the previously mentioned mixture of 4 mM NaBH 3 CN in PBS for 30 min, after which the ISFET was once again cleaned and dried using PBS and N 2 flow and afterwards placed into containers with PBS solution for storage at 4 • C.

Preparation of Artificial Saliva (AS)
Artificial saliva (AS) was prepared as described elsewhere [11,16,17]. Briefly, AS preparation was carried out by dissolving 4 mg/mL of urea, 4 mg/mL of mucin from porcine stomach, 0.6 mg/mL of sodium phosphate dibasic (Na 2 HPO 4 ), 0.3 mg/mL of sodium bicarbonate (NaHCO 3 ), 0.6 mg/mL of anhydrous calcium chloride (CaCl 2 ), 0.4 mg/mL of potassium chloride (KCl), and 0.4 mg/mL of sodium chloride (NaCl) in Millipore Milli-Q nanopure water (resistivity > 18 MΩ cm) produced by a Millipore Reagent Water System. Dissolving was aided by placing the solution in ultrasonic bath for 20 min to ensure the solution homogeneity. Afterwards, 0.1 M of sodium hydroxide (NaOH) was added into the solution to obtain pH with the value of 7.2. The prepared AS was stored at 4 • C until further use.

Electrochemical Characterisation of ISFET
To investigate the performance and sensitivity of ISFET before the functionalization with mAbs, one ISFET was tested using different pH solutions. Briefly, pH of a mixture of 5 mM of tris-(hydroxymethyl) methylamine and 0.4 M of magnesium nitrate Mg(NO 3 ) 2 was measured using a pH meter. Afterwards, the solution was divided into three smaller ones. Their pH was adjusted using hydrochloric acid (HCl) and then measured again, giving final values of pH 7.01, 5.00, and 3.26.
Electrical measurements were performed using LabVIEW software. ISFET was connected to external chassis (National Instruments, PXI-1031) (connected to the personal computer), included with a digital multimeter and data acquisition module, through a shielded connector block (National Instruments, SCB-68) (See Figure 2). The reference voltages were applied using an Ag/AgCl reference electrode, as described elsewhere [46,47]. The ISFET transfer characteristics (drain current, I DS , vs. gate-to-source voltage, V GS ) were measured at a drain-to-source voltage (V DS ) of 500 mV varying the V GS from −0.5 V to 1.5 V value and vice versa for I DS dependence on V DS . Evaluations were performed minimum of 10 times on 4 ImmunoFET electrodes, during each experiment. After each experiment, the ISFET was cleaned with ultrapure water and carefully dried before immersing into solution with different analyte concentrations.

Electrochemical Characterisation of ISFET
To investigate the performance and sensitivity of ISFET before the functionalization with mAbs, one ISFET was tested using different pH solutions. Briefly, pH of a mixture of 5 mM of tris-(hydroxymethyl) methylamine and 0.4 M of magnesium nitrate Mg(NO3)2 was measured using a pH meter. Afterwards, the solution was divided into three smaller ones. Their pH was adjusted using hydrochloric acid (HCl) and then measured again, giving final values of pH 7.01, 5.00, and 3.26.
Electrical measurements were performed using LabVIEW software. ISFET was connected to external chassis (National Instruments, PXI-1031) (connected to the personal computer), included with a digital multimeter and data acquisition module, through a shielded connector block (National Instruments, SCB-68) (See Figure 2). The reference voltages were applied using an Ag/AgCl reference electrode, as described elsewhere [46,47]. The ISFET transfer characteristics (drain current, IDS, vs. gate-to-source voltage, VGS) were measured at a drain-to-source voltage (VDS) of 500 mV varying the VGS from −0.5 V to 1.5 V value and vice versa for IDS dependence on VDS. Evaluations were performed minimum of 10 times on 4 ImmunoFET electrodes, during each experiment. After each experiment, the ISFET was cleaned with ultrapure water and carefully dried before immersing into solution with different analyte concentrations.

Detection and Characterization of Antigens (Ag) Using ImmunoFET
Detection of antigens was performed using the same characterization method and software as described before for pH measurement. The scheme of ImmunoFET is shown in Figure 3.

Detection and Characterization of Antigens (Ag) Using ImmunoFET
Detection of antigens was performed using the same characterization method and software as described before for pH measurement. The scheme of ImmunoFET is shown in Figure 3.
Various concentrations of TNF-α antigen (Ag) were tested: 5, 10, 15, 20 pg/mL. Before each detection, ImmunoFET was dried with N 2 flow and then incubated inside the solution of Ags with the required concentration at 4 • C for 30 min. The ImmunoFET was then carefully washed with PBS solution and placed inside the beaker with 7 mL of PBS (pH 7.4) for analysis. The sensor was then rinsed with PBS, incubated with higher Ags concentration, and the signal was measured again. For the detection of TNF-α in AS (pH 7.2), the same steps were repeated, except for the dilution of TNF-α protein in AS instead of PBS. After the incubation of ImmunoFET in AS with Ags, the sensor was cleaned with PBS and then immerged into the beaker with PBS for the electrical measurements. TNF-α detection in both PBS and AS was assessed by performing measurements on three chips for each medium. For the interference study in PBS, different heart failure biomarkers were used: N-terminal pro-brain natriuretic peptide (NT-proBNP), cortisol, and interleukin-10 (IL-10). For each biomarker, a newly functionalised chip was used, as they are not reusable due to a strong Abs-Ags bond that it is difficult to break. Various concentrations of TNF-α antigen (Ag) were tested: 5, 10, 15, 20 pg/mL. Before each detection, ImmunoFET was dried with N2 flow and then incubated inside the solution of Ags with the required concentration at 4 °C for 30 min. The ImmunoFET was then carefully washed with PBS solution and placed inside the beaker with 7 mL of PBS (pH 7.4) for analysis. The sensor was then rinsed with PBS, incubated with higher Ags concentration, and the signal was measured again. For the detection of TNF-α in AS (pH 7.2), the same steps were repeated, except for the dilution of TNF-α protein in AS instead of PBS. After the incubation of ImmunoFET in AS with Ags, the sensor was cleaned with PBS and then immerged into the beaker with PBS for the electrical measurements. TNF-α detection in both PBS and AS was assessed by performing measurements on three chips for each medium. For the interference study in PBS, different heart failure biomarkers were used: N-terminal pro-brain natriuretic peptide (NT-proBNP), cortisol, and interleukin-10 (IL-10). For each biomarker, a newly functionalised chip was used, as they are not reusable due to a strong Abs-Ags bond that it is difficult to break.

Results and Discussion
The electrical characterization has shown that all the ISFETs have good characteristics as field-effect transistors. The devices have shown good operating characteristics as shown in Figure 4. The drain current and the transconductance (Gm) versus the gate-tosource voltage curves were shifted to positive bias as the pH increased ( Figure 4A), with a threshold voltage (VT) of 52.5 mV/pH response that was obtained after the extraction of the VT vs. pH (S = VT/pH). The Si3N4 behaviour can be explained assuming that the insulator surface contains two different sites: Si-OH groups, amphoteric sites, resulting from the oxidation of the surface and amine groups, Si-NH2, that are basic sites [48]. On a completely oxygen-free Si3N4 surface, each surface nitrogen can also be assumed to be bonded to three silicon atoms, forming a group with a lone electron pair on the nitrogen atom, Si3N. This group is similar to z tertiary amine in an organic compound. The electron pair on the nitrogen atom is the only one available for a hydrolysis reaction, which would proceed as follows: Si3N + H2O Si2NH + Si-OH The secondary amine group can then be further hydrolysed to form a primary amine and another silanol group.

Results and Discussion
The electrical characterization has shown that all the ISFETs have good characteristics as field-effect transistors. The devices have shown good operating characteristics as shown in Figure 4. The drain current and the transconductance (G m ) versus the gate-tosource voltage curves were shifted to positive bias as the pH increased ( Figure 4A), with a threshold voltage (V T ) of 52.5 mV/pH response that was obtained after the extraction of the V T vs. pH (S = ∆V T /∆pH). The Si 3 N 4 behaviour can be explained assuming that the insulator surface contains two different sites: Si-OH groups, amphoteric sites, resulting from the oxidation of the surface and amine groups, Si-NH 2 , that are basic sites [48]. On a completely oxygen-free Si 3 N 4 surface, each surface nitrogen can also be assumed to be bonded to three silicon atoms, forming a group with a lone electron pair on the nitrogen atom, Si 3 N. This group is similar to z tertiary amine in an organic compound. The electron pair on the nitrogen atom is the only one available for a hydrolysis reaction, which would proceed as follows: When the VGS was kept at constant 0.5 V value ( Figure 4B) IDS as a function of VDS increased when pH was reduced. For different ionic strengths plateau of IDS reached values of 53.4 μA for pH 7.01, 61.1 μA for pH 5.00, and 76.6 μA for pH 3.26. Lower current values for higher pH clearly indicated lesser ionic activity on the gate surface compared to acidic solution. The obtained results, when the ISFET was analysed at different pH values, allowed us to confirm the sensitivity of the device to different proton concentrations. The results were in perfect agreement with those described in the literature [49,50]. TNF-α detection was performed on a newly functionalised chip. ImmunoFET was firstly characterised in PBS buffer with no Ags presence, and then, it was incubated with TNF-α Ags spiked in PBS for 30 min. The response characteristics of the ImmunoFET are shown in Figure 5, displaying a positive shift of VGS with the binding of TNF-α protein in increasing the concentration buffer. The secondary amine group can then be further hydrolysed to form a primary amine and another silanol group. Si 2 NH + H 2 O → Si-NH 2 + Si-OH When the V GS was kept at constant 0.5 V value ( Figure 4B) I DS as a function of V DS increased when pH was reduced. For different ionic strengths plateau of I DS reached values of 53.4 µA for pH 7.01, 61.1 µA for pH 5.00, and 76.6 µA for pH 3.26. Lower current values for higher pH clearly indicated lesser ionic activity on the gate surface compared to acidic solution. The obtained results, when the ISFET was analysed at different pH values, allowed us to confirm the sensitivity of the device to different proton concentrations. The results were in perfect agreement with those described in the literature [49,50].
TNF-α detection was performed on a newly functionalised chip. ImmunoFET was firstly characterised in PBS buffer with no Ags presence, and then, it was incubated with TNF-α Ags spiked in PBS for 30 min. The response characteristics of the ImmunoFET are shown in Figure 5, displaying a positive shift of V GS with the binding of TNF-α protein in increasing the concentration buffer.
ues, allowed us to confirm the sensitivity of the device to different proton concentrations. The results were in perfect agreement with those described in the literature [49,50]. TNF-α detection was performed on a newly functionalised chip. ImmunoFET was firstly characterised in PBS buffer with no Ags presence, and then, it was incubated with TNF-α Ags spiked in PBS for 30 min. The response characteristics of the ImmunoFET are shown in Figure 5, displaying a positive shift of VGS with the binding of TNF-α protein in increasing the concentration buffer. After Abs functionalisation, the extracted VT shifted to negative values, when compared to the ImmunoFET measurements without Abs functionalisation. This allows us to confirm successful Abs integration on the gate surface. However, this shift of VT values increased again with the rising concentration of antigens ( Figure 5A), suggesting that due to the specific interaction between the anti-TNF-α mAbs and their corresponding Ags, gate potential is altered when more Ags are captured by Abs. This was further confirmed by performing the analyses of IDS dependency on the VDS ( Figure 5B). The plateau of IDS current decreased from the initial 61.0 μA for ImmunoFET before the incubation with TNF-α Ags to a value of 56.2 μA for ImmunoFET after the last incubation at 20 pg/mL of TNF-α Ags. The obtained results suggest that in the PBS solution the TNF-α proteins are negatively charged, thus the VT increases with the rising concentration of Ags are similar to the results obtained previously for the ISFET characterization at different pH values. After Abs functionalisation, the extracted V T shifted to negative values, when compared to the ImmunoFET measurements without Abs functionalisation. This allows us to confirm successful Abs integration on the gate surface. However, this shift of V T values increased again with the rising concentration of antigens ( Figure 5A), suggesting that due to the specific interaction between the anti-TNF-α mAbs and their corresponding Ags, gate potential is altered when more Ags are captured by Abs. This was further confirmed by performing the analyses of I DS dependency on the V DS ( Figure 5B). The plateau of I DS current decreased from the initial 61.0 µA for ImmunoFET before the incubation with TNF-α Ags to a value of 56.2 µA for ImmunoFET after the last incubation at 20 pg/mL of TNF-α Ags. The obtained results suggest that in the PBS solution the TNF-α proteins are negatively charged, thus the V T increases with the rising concentration of Ags are similar to the results obtained previously for the ISFET characterization at different pH values.
Interference study for ImmunoFET immobilized with anti-TNF-α mAbs were performed to investigate the selectivity for TNF-α protein. The ∆V T (change in threshold voltage) for various concentrations was calculated by subtracting V T obtained after the incubation in Ags solution by the initial V T value of ImmunoFET (referred as 0 pg/mL value in the Figure 5 graphs) before. The preliminary results for different biomarkers, which could be found in saliva as HF biomarkers, are shown in Figure 6. The change in threshold voltage for TNF-α protein is clearly dependant on the concentration. It can be seen that the ∆V T approximately linearly increases with the rising amount of antigen present during the incubation step with ImmunoFET. However, changes in threshold voltage have also been observed after incubation of the ImmunoFET into solutions containing NT-proBNP, cortisol, and IL-10. Nevertheless, the changes in the threshold voltage observed due to these biomarkers were random and not influenced by the concentration of the specific biomarker, as it can be seen from Figure 6. Therefore, these signals from interfering molecules can be attributable to adsorption phenomena occurring on the ImmunoFET surface. A possible explanation might be that the remaining hydroxyl groups after functionalization may limit the sensor response to the targeted biomarker [51]. The incubation step in Et-NH 2 had to cap unreacted chains of TESUD to prevent nonspecific bonding at the detection stage, since ethanolamine is one of the most commonly used blocking agents for protein coupling via amino groups. Further experiments on the use of different blocking agents (i.e., triethylene glycol monoamine [52]) can be carried out for both reducing unspecific absorption phenomena and improving selective detection of target HF biomarkers.
the ∆VT approximately linearly increases with the rising amount of antigen present during the incubation step with ImmunoFET. However, changes in threshold voltage have also been observed after incubation of the ImmunoFET into solutions containing NT-proBNP, cortisol, and IL-10. Nevertheless, the changes in the threshold voltage observed due to these biomarkers were random and not influenced by the concentration of the specific biomarker, as it can be seen from Figure 6. Therefore, these signals from interfering molecules can be attributable to adsorption phenomena occurring on the ImmunoFET surface. A possible explanation might be that the remaining hydroxyl groups after functionalization may limit the sensor response to the targeted biomarker [51]. The incubation step in Et-NH2 had to cap unreacted chains of TESUD to prevent nonspecific bonding at the detection stage, since ethanolamine is one of the most commonly used blocking agents for protein coupling via amino groups. Further experiments on the use of different blocking agents (i.e., triethylene glycol monoamine [52]) can be carried out for both reducing unspecific absorption phenomena and improving selective detection of target HF biomarkers. Figure 6. Interference study of ΔVT of ImmunoFET with immobilized anti-mTNF-α for TNF-α, NT-proBNP, cortisol, and IL-10 detection.
The detection of TNF-α was further investigated by changing the matrix of Ags. Instead of diluting TNF-α proteins in PBS for the incubation with ImmunoFET, AS was used. Transfer curves of TNF-α antigen detection in AS are presented in Figure 7. It was observed that the change in the matrix influences the obtained signal. From both, the IDS dependence on the VGS ( Figure 7A) and on VDS ( Figure 7B) observed signal was the opposite compared to incubation in PBS solution. The detection of TNF-α was further investigated by changing the matrix of Ags. Instead of diluting TNF-α proteins in PBS for the incubation with ImmunoFET, AS was used. Transfer curves of TNF-α antigen detection in AS are presented in Figure 7. It was observed that the change in the matrix influences the obtained signal. From both, the I DS dependence on the V GS ( Figure 7A) and on V DS ( Figure 7B) observed signal was the opposite compared to incubation in PBS solution.
The plateau shift of I DS dependence on the V DS was observed to be lower compared to measurements in PBS, the change being only by 0.5 µA between the reference and highest concentration of Ags in AS. Similar to the results shown in Figure 5, functionalisation of ImmunoFET with antibodies caused the V T to shift to negative values. However, the V T continued to shift to negative values with the rising concentration of TNF-α protein in AS, contrary to V T for TNF-α diluted in PBS. Nevertheless, the absolute values of the ∆V T shift remained approximately the same (Figure 8). A similar influence of the change in matrix on the signal direction was already observed by Pulikkathodi et al. [53], who obtained an opposite response of circulating tumour cell detection in PBS and cell culture mediums. The author and his co-authors observed that this influence is due to divalent cation (Ca 2+ and Mg 2+ ) presence in the culture cell medium. They noted that the response is greatly dependant on the concentration of ions Ca 2+ and Mg 2+ with which the trans-membrane proteins interact, changing the potential of the cell membrane. Therefore, it is likely that Ca 2+ presence in AS may interact with TNF-α proteins, charging them positively, thus inducing inverse response when compared to PBS solution without Ca 2+ . The plateau shift of IDS dependence on the VDS was observed to be lower compared to measurements in PBS, the change being only by 0.5 μA between the reference and highest concentration of Ags in AS. Similar to the results shown in Figure 5, functionalisation of ImmunoFET with antibodies caused the VT to shift to negative values. However, the VT continued to shift to negative values with the rising concentration of TNF-α protein in AS, contrary to VT for TNF-α diluted in PBS. Nevertheless, the absolute values of the ∆VT shift remained approximately the same (Figure 8). A similar influence of the change in matrix on the signal direction was already observed by Pulikkathodi et al. [53], who obtained an opposite response of circulating tumour cell detection in PBS and cell culture mediums. The author and his co-authors observed that this influence is due to divalent cation (Ca 2+ and Mg 2+ ) presence in the culture cell medium. They noted that the response is greatly dependant on the concentration of ions Ca 2+ and Mg 2+ with which the trans-membrane proteins interact, changing the potential of the cell membrane. Therefore, it is likely that Ca 2+ presence in AS may interact with TNF-α proteins, charging them positively, thus inducing inverse response when compared to PBS solution without Ca 2+ .  The plateau shift of IDS dependence on the VDS was observed to be lower compared to measurements in PBS, the change being only by 0.5 μA between the reference and highest concentration of Ags in AS. Similar to the results shown in Figure 5, functionalisation of ImmunoFET with antibodies caused the VT to shift to negative values. However, the VT continued to shift to negative values with the rising concentration of TNF-α protein in AS, contrary to VT for TNF-α diluted in PBS. Nevertheless, the absolute values of the ∆VT shift remained approximately the same (Figure 8). A similar influence of the change in matrix on the signal direction was already observed by Pulikkathodi et al. [53], who obtained an opposite response of circulating tumour cell detection in PBS and cell culture mediums. The author and his co-authors observed that this influence is due to divalent cation (Ca 2+ and Mg 2+ ) presence in the culture cell medium. They noted that the response is greatly dependant on the concentration of ions Ca 2+ and Mg 2+ with which the trans-membrane proteins interact, changing the potential of the cell membrane. Therefore, it is likely that Ca 2+ presence in AS may interact with TNF-α proteins, charging them positively, thus inducing inverse response when compared to PBS solution without Ca 2+ . The comparison between electrical measurements performed in this work and other similar techniques based on electrochemical signal measurement is shown in Table 1. It should be noted that other established systems show a significantly lower limit of detection values. Nevertheless, even though lower concentrations than 5 pg/mL were not tested, the ImmunoFET sensor system in this work provides the ability to detect TNF-α antigens for the detection of HF. Moreover, this system has facile use and size and strong potential as a point of care testing system.

Conclusions
In this work, ImmunoFET for the detection of TNF-α protein was developed by immobilizing monoclonal anti-TNF-α antibodies onto the gate surface made of Si 3 N 4 . The performance of the biosensor was investigated by measuring the shift in the threshold voltage of the gate and the drain-to-source current. Analyses of the electrical characterization of ImmunoFET showed a fast and direct response to the change in antigen concentration. It can be concluded that the developed ImmunoFET based on ISFET with surface modification using anti-TNF-α mAbs is highly sensitive to TNF-α p r o teins. The shift in ∆V T for the concentration of TNF-α being in the range of 5 to 20 pg/mL was approximately linear in both PBS and AS the antigen matrix. While the absolute values of ∆V T shift were approximately the same for both matrixes, the direction of the signal was found to be the opposite. This suggests that proteins possibly interact with ions in the AS, which then changes the electrical charges associated to the molecule, giving the contrary results when compared to PBS measurements. More analyses and improvements are still needed to be performed, such as antibody blocking by BSA, which would allow us to better assess the selectivity of the sensors, and detection in real saliva, which is a much more complex matrix than AS. Nevertheless, the developed ImmunoFET showed great potential for direct and fast detection of TNF-α protein in saliva samples for heart failure diagnosis and monitoring.