K₄Fe(CN)₆∙3H₂O, K₃Fe(CN)₆, KCl, L-tryptophan, uric acid, ascorbic acid, and chitosan were purchased from Fluka Chemika. TEMPO (2,2,6,6,-tetramethylpiperidine 1-pxyl) was purchased from Sigma-Aldrich. The buffer solution used for all experiments was phosphate buffer saline (PBS) containing 137 mM NaCl, 2.7 mM KCl, 0.01 M KH₂PO₄ and 0.01 M K₂HPO₄, pH 7. All solutions were prepared in ultrapure water (resistance 18.2 MW∙cm⁻¹) produced by a Millipore Milli-Q system. The NFC material was extracted from marram grass (Ammophila arenaria) collected in the central western area of Tunisia (Gafsa).

Fourier transform infrared (FTIR) spectra (from 4000 to 400 cm⁻¹) were obtained using a BX FTIR system spectrometer, Perkin Elmer Company, Waltham, MA, USA. KBr pellets technique was used for sample preparation.

The transmission electron microscope (TEM) image of the extracted NFCs was taken using a Jeol 200 CX transmission electron microscope at 80 kV. About 0.5 µL of diluted suspension of NFC (about 0.1%) was deposited onto a 300 mesh carbon-coated grid.

X-ray diffraction (XRD) patterns were collected by a Bruker diffractometer (D8 Advance, Germany) with Cu-Kα radiation (λ = 0.1548 nm).

Electrochemical experiments were carried out using an Autolab (PGSTAT 302 N, Eco Chemie). All measurements were performed in a dark Faraday cage at room temperature. A conventional three-electrode system consisting of a platinum counter-electrode, an Ag/AgCl/KCl as the reference electrode, and a glassy carbon electrode (GCE, f 3 mm) as the working electrode was used. Differential Pulse Voltammetric (DPV) measurements were carried out at a scan rate of 50 mV/s with an accumulation time of 20 s.

By a conventional soda-pulling process and bleaching treatment, the cellulose fibers were extracted from Marram grass, then the fibers were treated by TEMPO-mediated oxidation to generate carboxyl groups. Oxidation was carried out at pH 10 following the previously reported method (Jebali et al., 2018). Briefly, 5 g of fibers were dispersed in 500 mL sodium phosphate buffer (0.05 mol∙L⁻¹, pH 7) containing 25 mg of TEMPO and 250 mg of NaBr. Then 1.13 g of NaCl (80%) and 2 mol∙L⁻¹ of NaClO were added to the flask in one step and the suspension was left under magnetic stirring at 60˚C for 6 h. At the end of the oxidation reaction, 100 mL of ethanol was added and the obtained modified fibers were filtered. This operation was carried out several times with water to remove any salt. Then a suspension of 1.5 wt% of prepared fibers was homogenized in a high-pressure homogenizer from GEA (NS 1001 L PANDA 2 K-GEA, Italy). After six shear cycles (3 × 30 bar; 3 × 600 bar), a thick transparent NFC-COO⁻ gel was obtained [24] .

Before the electrode modification, the bare glassy carbon electrode (GCE) was polished with 0.5 µm alumina solution. The GCE was then fully rinsed with water and sonicated in ethanol and deionized water for 3 min. Next, the GCE was scanned in 0.5 M H₂SO₄, at 50 mV/s for many cycles between −0.2 V and 1.3 V.

Subsequently, 5 µL of CS (the optimized concentration, 1 mg/mL) solution was deposited on the GCE, then 5 µL of NFC (optimized concentration: 1 mg/mL) was deposited on the CS/GCE. The obtained CS/NFC modified surface was dried for 24 hrs at room temperature.

The Fourier transform-infrared spectroscopy (FTIR) spectrum of NFC is illustrated in Figure 1 and shows distinct absorption bands. The bands located at 3330 cm⁻¹ were attributed to O–H stretching vibration. The band at 2906 cm⁻¹ was assigned to the aliphatic C–H -stretching vibration and the band at 1660 cm⁻¹ associated with the O–H bending vibration of absorbed water. CNF showed a band at 1614 cm⁻¹, which was likely due to the introduction of carboxylic groups into the primary alcohol groups on the surface of cellulose nanofibrils after TEMPO mediated oxidation. The characteristic bands of the glucosic ring appear in the 1000 - 1162 cm⁻¹ region.

The morphological features of CNF were assessed by TEM observation ( Figure 2(a) ). Nanosized fibrils with a width distribution centered around 3 - 6 nm are observed which is a good indication that the individual fibrils correspond to elementary cellulose fibrils composed of altered amorphous and crystalline domains. The crystalline structure of NFC was confirmed through X-ray diffraction ( Figure 2(b) ); a strong peak is observed at 18˚ corresponding to the 002 plane [26] .

The electrochemical behavior of the different modified electrodes was examined by cyclic voltammetry (CV). As shown in Figure 3 , a pair of reversible and well-defined redox peaks was observed on the bare GCE, with a ΔEp of 100 mV and an intensity ratio of about 1:1 between anodic and cathodic peak currents demonstrating that the glassy carbon electrode (GCE) has a quasi-reversible electrochemical reaction with the [Fe(CN)₆]^3−/4− couple [26] . The redox currents decreased significantly after the GCE was modified with CS/NFC, proving the electrode modification.

So, with the integration of NFCs and CS, the synergistic effect then took place at this electrochemical platform, leading to increased adsorption of electrochemically active sites, and of the effective electrode area.

To determine the electrochemical behavior of 0.1 mM Trp in phosphate buffer solution (pH 7.0) on a GCE, CS/GCE, NFC/GCE, and a CS/NFC/GCE, cyclic voltammetry was investigated. From the electrochemical results shown in Figure 4(a) , it can be seen that Trp presents only one oxidation peak. No reduction peak was observed in the reverse scan, confirming that the electrochemical reaction is an irreversible process, in agreement with previous results [29] . As can be seen from Figure 4(a) , the oxidation peak at the bare GCE appears at 0.75 V, while at the CS/NFC/GCE the peak potential shifted negatively to 0.55 V and the peak current increased significantly. This observation proves that CS/NFC film can catalyze the oxidation of L-Tryptophan. The strong electrostatic attraction between CNF-COO⁻ (negatively charged) (pie < 2.8) and Trp molecule (pie = 5.89) can explain the increase in intensity. The conductive properties of NFC (2 mS/cm [30] ) explain its role in favoring charge transfer. The DPV peaks presented in Figure 4(b) confirm the results obtained by cyclic voltammetry.

By increasing the buffer solution pH from 4.5 to 8.5, the anodic potential gradually decreased which confirms the direct participation of protons in the oxidation reaction of Trp at the CS/NFC/GCE. The anodic peak potential varies linearly with pH value, following this equation:

$\text{Epa}=-55.43\text{pH}+0.98,R^2=0.996$(1)

The slope (−55.43 mV/pH) is very close to the theoretical value of 59 mV/pH, showing an equal number of exchanged electrons and protons during the oxidation process.

The peak current response of Tryp varied also with the change of pH and the maximum peak current is at pH 7.0 ( Figure 5 ); for pH higher than 7 the attractivity between Trp (pie 5.89) and NFC decreases. Subsequently, pH 7.0 was considered as the optimal electrolyte solution for the determination of L-Trp.

CVs of 0.1 mM L-Trp at the CS/NFC/GCE, with various scan rates (20 to 500 mV∙s⁻¹) were investigated. As shown in Figure 6 , the oxidation peak current for L-Tryptophan increases as the scan rate increases, suggesting that the electrocatalytic oxidation of L-Tryptophan at the CS/NFC modified GCE was a typical adsorption-controlled process. The linear relation between the oxidation peak intensity and the scan rate is the following one:

$\text{Ipa}\left(\mu \text{A}\right)=2.19+0.011v\left(\text{mV}\cdot {\text{s}}^{-1}\right),R^2=0.991$(2)

In addition, the oxidation peak potentials shift positively as the scan rate increases. So, 50 mV/s is employed as the optimal scan rate in the following experiments. The plot of peak current vs. square root of scan rate (v^1/2) is linear over the whole studied range of scan rate, indicating that the electro-oxidation of Trp at the CS/NFC/GCE was a typical diffusion-controlled process, and the equation can be expressed as:

$\text{Ipa}\left(\mu \text{A}\right)=0.35v^{1/2}+0.012,R^2=0.994$(3)

Moreover, the relationship between log (ipa) and log of the scan rate was found to be linear.

$\log \text{ipa}\left(\mu \text{A}\right)=0.47\log v-0.39,R^2=0.990$(4)

given a slope of 0.47 which is in agreement with the theoretical value of 0.5 for a diffusion-controlled process.

The number of electrons involved in the oxidation of tryptophan was calculated using the equation :

$\text{ipa}=nFQv/4RT$(5)

where ipa denotes the anodic peak current, Q denotes the amount of the integrated charge from the area of the voltammetric peak, F is the Faraday constant (96,500 C/mol), T denotes the temperature (298 K), R is the gas constant (8.314 J∙mol⁻¹) and n is the number of transferred electrons. From the slope of ip versus v, the number of electrons transferred in the oxidation of L-Trp was calculated to be 1.66. Consequently, the oxidation process of L-Trp involves two electrons and two protons, and the oxidation reaction of Trp is irreversible.

The anodic peak potential (Epa) versus the logarithm of the scan rate (logv) is a linear relationship:

$\text{Epa}=0.044\log v+0.54,R^2=0.991$(6)

Based on Laviron’s relation [31] , the linear plot of Epa versus logv reaches a slope of (2.3RT/αnF), which is equal to 0.044; the value of αn is calculated to be 1.34. According to Equation (7), the calculated α is 0.67.

$\alpha =47.7/\left(\text{Epa}-\text{Epa/}2\right)$(7)

where Epa and Epa/2 represent the formal potential and half of the formal potential value respectively.

The electrochemical performance of Trp was investigated by Differential Pulse Voltammetry (DPV) on bare GCE and two modified electrodes: NFC/GCE and CS/NFC/GCE in PBS (pH 7.0) at 50 mV/s. Figure 7(a) displays the DPV signal for different concentrations of Trp at the bare GCE. The oxidation peak current was observed at 0.75 V. The peak current response of the electrochemical sensor presents a linear variation with the concentration of L-Tryptophan, over the range from 5 × 10⁻⁴ to 10⁻² mM, with a sensitivity of 8.15 µA∙mM⁻¹.

Figure 7(b) shows the DPV for different concentrations of Trp at the NFCs-modified GCE; The corresponding peak current at 0.64 V of L-Trp increases proportionally with the increase in concentration of L-Trp. The plot of the oxidation peak current versus concentration exhibits good linearity in the range of 7.5 × 10⁻⁴ mM to 10⁻² mM with a sensitivity of 33.14 µA∙mM⁻¹. A detection limit of 0.75 µM was obtained.

Figure 7(c) shows the DPV for different concentrations of Trp at the CS/NFCs modified GCE in the range of 4 × 10⁻⁴ mM to 10⁻² mM; it was found that the anodic peak current shifts to the negative side and appeared at 0.55 V with an improved anodic peak response. The DPV curves clearly show that CS/NFC/GCE displays a proportionally increasing current response with increasing L-Trp concentration. The corresponding sensitivity of 140.0 µM∙mM⁻¹ is three times higher than the sensitivity of the NFC/GCE and a lower detection limit of 0.2 µM was obtained.

Figure 7(d) shows the plots of the L-Trp oxidation peak current versus L-Trp concentration on bare GCE, NFC/GCE, and CS/NFC/GCE. The excellent electrochemical sensing property was attributed to the CS/NFC film. The RSD on the measurements was 1%.

Furthermore, the analytical parameters of the proposed green nanocomposite CS/NFC were compared with the other modified electrodes, and the values are listed in Table 1 . In comparison, our proposed sensor exhibits excellent electrocatalytic activity for the electrochemical determination of Trp: the lowest potential value was required for the Trp oxidation and the higher sensitivity was obtained.

(a) GCE, (b) CPE, (c) graphite electrode, (d) DPV, (e) GC.

The stability of the modified electrode was studied by investigating the oxidation current of 0.1 mM of L-Trp in pH 7.0 PBS after storage in a refrigerator at 4˚C for one month; the catalytic current response was maintained at 91.3% of its initial value after one month.

The repeatability of the CS/NFCs/GCE was estimated by measuring the current response of 0.1 mM L-Trp for 4 modified electrodes. The relative standard deviation (RSD) was 2.6% (n = 4).

To testify to the selectivity of the sensor, an interference experiment was carried out by recording the electrochemical responses of 0.1 mM Trp on the CS/NFCs/ GCE in the electrochemical cell in the presence of potential interfering substances. Several amino acids including leucine, isoleucine, proline, lysine, aspartic acid, valine, and glutamine were used to examine the interference with L-Trp; no significant change in the Trp peak current in the presence of these interfering amino acids was found.

Figure 8 shows the simultaneous detection of Trp with two kinds of important biological substances: ascorbic acid (AA) and uric acid (UA). There were no significant changes in the Trp current peak in the presence of these interfering substances. Based on these results, the CS/NFCs/GCE showed good selectivity for L-Tryptophan determination.

A CS/NFC modified glassy carbon electrode was used to detect Trp in human urine samples. The real samples were collected from healthy volunteer patients with informed consent from the volunteers based on the Declaration of Helsinki. These samples were diluted 100 times with 0.1 phosphate buffer solution (pH7) without any treatment. The standard addition method was applied, so different concentrations of L-Trp were added to the previously diluted human urine. Figure 9 shows the DPV of L-Trp in the concentration range from 8 × 10⁻³ mM to 12 × 10⁻³ mM on CS/NFC/GCE, current peaks were observed at 0.55 V.

The equation of the linear relationship between the current and the concentration of L-Trp, in the diluted human urine is: y = 140.0x + 1.02. The detected concentration of L-Tryptophan was 7.3 × 10⁻³ mM (149.1 mg/L). As reported in the literature, the normal range for L-Tryptophan in human urine is 137 - 240 mg/L. The detected value is included in the normal concentration range [40] . Consequently, the obtained results demonstrated that the developed sensing NFC matrix can present a suitable applicability to biological fluids.