Since this kind of devices is based on the electrical

Since this kind of devices is based on the electrical response of functionalized carbon nanotubes to antibody–antigen binding, they can be easily turned to detect other proteins by functionalizing CNTs with appropriate antibodies. Numerous studies revealed that proteins can interact in a non-specific way with the sidewalls of acid-oxidized CNTs [17–19], affecting negatively the specificity of the CNT-based immunosensors, and this has to be kept into account when working with these systems. A suitable rinsing out of the sensor after antibody–antigen binding should remove antigen in excess, not chemically bound, reducing the signal due to non-specific interaction.
Materials and methods
Results and discussion Anti-ARG1 antibody (Ab) with a 1μg/ml concentration was dispersed in a dilution buffer solution (PBS+TWEEN20 0.005%+BSA 0.4%) and a 10μl droplet was placed on the MWCNT–NHS sensitive layer and left to incubate for 2h in humid environment. The functionalized device was rinsed out in washing buffer (PBS+TWEEN20 0.005%) to remove Amyloid Beta-Peptide (1-40) not chemically bound, and incubated for further 2h with a 5μl droplet of dilution buffer containing ARG-1 with known concentrations in the range of 22–360ng/ml, followed by a careful rinsing out with DI water. The process is schematically illustrated in Fig. 5. The choice of this concentration range has been done keeping into account the ARG-1 values in serum observed in normal (<30ng/ml) or in pathologic conditions in the case of MM disease (>100ng/ml). Finally, electrical characterization was carried out and compared with the initial I–V characteristics of the device (before both CNT–antibody and antibody–antigen interactions). In Fig. 6 we report the I–V characteristics, in dry state, obtained for three groups of devices, each group showing the same initial CNT layer resistance (R0=115kΩ, 100kΩ, 22kΩ), after exposure to Anti-ARG1 with 1μg/ml concentration, followed by antibody–antigen binding reaction. In particular, in Fig. 6a and b the investigated antigen concentration values are 180ng/ml and 360ng/ml, whereas in Fig. 6c lower ARG-1 concentrations were used, that is 22ng/ml and 45ng/ml. For comparison, the curves obtained in the absence of ARG-1 (0ng/ml) are reported for each kind of device. Generally, we notice that the addition of the protein determines a decrease of the current measured in the CNT layer and the higher is the protein concentration the larger is the current decrease (resistance increase). The current (resistance) change respect to the initial value, for a given voltage VDS, can, therefore, be related to the protein concentration present in the solution drop poured on the sensitive layer of the device. Two sets of each kind (R0) of devices were tested and we observed a reproducibility within a variation below 10%. In Fig. 7 we report a summary of the bio-sensing results that can be extracted by the I–V curves shown in Fig. 6. In particular, for each device we have calculated the normalized change of the resistance ΔR/R0=(R−R0)/R0, were R0 and R are the CNT film resistance values at V=1V, respectively, before and after the addition of ARG-1 antigen and its binding with the antibody. The results reported in Fig. 7 clearly show that the sensitivity range depends on the initial resistance of the device: in order to detect very low ARG-1 concentrations uterus seems more convenient to use devices with low initial resistance of the CNT layer, whereas for large protein concentration devices with higher initial resistance should be preferred. Indeed, a lower initial device resistance means (if the same kind of CNTs are used) that a larger CNT density is present between the electrodes and on the whole a larger number of functional groups available for antibody–antigen binding. This allows reaching a higher sensitivity for low antigen concentration values. In general, a calibration curve can be obtained for sensors with different initial R0 and in the case of unknown concentrations, as in real samples, it could be useful to use an array of sensors with different initial I–V characteristics and compare their responses. This method allows to perform a cross check of the results given by the different sensors.