*Article* **Constructing a Carbon-Encapsulated Carbon Composite Material with Hierarchically Porous Architectures for Efficient Capacitive Storage in Organic Supercapacitors**

**Rene Mary Amirtha 1,† , Hao-Huan Hsu 1,† , Mohamed M. Abdelaal 1,2 , Ammaiyappan Anbunathan 1 , Saad G. Mohamed 2 , Chun-Chen Yang 1,3,4 and Tai-Feng Hung 1, \***


**Abstract:** Hierarchical porous activated carbon (HPAC) materials with fascinating porous features are favored for their function as active materials for supercapacitors. However, achieving high massloading of the HPAC electrodes remains challenging. Inspired by the concepts of carbon/carbon (C/C) composites and hydrogels, a novel hydrogel-derived HPAC (H-HPAC) encapsulated H-HPAC (H@H) composite material was successfully synthesized in this study. In comparison with the original H-HPAC, it is noticed that the specific surface area and pore parameters of the resulting H@H are observably decreased, while the proportions of nitrogen species are dramatically enhanced. The free-standing and flexible H@H electrodes with a mass-loading of 7.5 mg/cm<sup>2</sup> are further prepared for electrochemical measurements. The experiments revealed remarkable reversible capacitance (118.6 F/g at 1 mA/cm<sup>2</sup> ), rate capability (73.9 F/g at 10 mA/cm<sup>2</sup> ), and cycling stability (76.6% of retention after 30,000 cycles at 5 mA) are delivered by the coin-type symmetric cells. The cycling stability is even better than that of the H-HPAC electrode. Consequently, the findings of the present study suggest that the nature of the HPAC surface is a significant factor affecting the corresponding capacitive performances.

**Keywords:** supercapacitors; hierarchical porous activated carbon; hydrogel; composite materials; clean energy technology

#### **1. Introduction**

Electrochemical-based energy storage devices such as metal–ion batteries/capacitors and supercapacitors are recognized as alternative choices for electricity storage owing to their high flexibility, remarkable reversibility, and simple maintenance as compared to other electric storage technologies [1,2]. Recently, preparing electrodes with high-mass loading has attracted much attention because the active material ratio in devices should be increased as much as possible to provide high total capacitances and gravimetric or volumetric energy densities [3–5]. However, challenges associated with this target still remain, especially in employing the hierarchical porous activated carbon (HPAC) as the active material for supercapacitors. This can be attributed to its huge specific surface area (typically more than 1000 m2/g), leading to the limited mass loading (normally 1 to

**Citation:** Amirtha, R.M.; Hsu, H.-H.; Abdelaal, M.M.; Anbunathan, A.; Mohamed, S.G.; Yang, C.-C.; Hung, T.-F. Constructing a Carbon-Encapsulated Carbon Composite Material with Hierarchically Porous Architectures for Efficient Capacitive Storage in Organic Supercapacitors. *Int. J. Mol. Sci.* **2022**, *23*, 6774. https://doi.org/ 10.3390/ijms23126774

Academic Editor: Ana María Díez-Pascual

Received: 22 May 2022 Accepted: 15 June 2022 Published: 17 June 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

2 mg/cm<sup>2</sup> by a doctor blade method) [6]. Therefore, developing an HPAC that retains the distinctive textural properties and increases the mass loading of the resulting electrode is of interest and highly desirable [3,4,7–9].

Carbon/carbon (C/C) composites are demonstrated to possess a variety of characteristics, i.e., high specific strength, remarkable electrical and thermal conductivities, and excellent dimensional stability [10]. Given the diverse properties, they are beneficial in the field of biomedical, automobile industries, and aeronautics. To realize the C/C composites, it is reported that polymer infiltration pyrolysis and chemical vapor infiltration were commonly adapted [11,12]. As a result, compact and dense C/C composites were obtained, particularly from the repeatedly manufacturing processes. Such a configuration would be beneficial to enhance the mass loading, but not favorable in terms of electrolyte penetration and ionic transportation.

To maximize the electrolyte storage and ionic conductivity within a high mass loading electrode, the hierarchical porosities including micro-, meso-, and macro-pores are crucial. For example, the charges are primarily adsorbed/desorbed inside the micropores. As for the latter two, they can contribute to (i) an electrolyte reservoir, (ii) enlarging the ionic diffusion rate, and (iii) facilitating the migration of large ions/molecules [13–16]. Undoubtedly, utilizing a template and activation are straightforward approaches for synthesizing the HPAC [14,17]. Even so, it will be highly appreciated if the greener templates and activators were chosen owing to resolve environmental issues and promote cost-effectiveness issues [9,18–21].

In our recent study, hydrogel-derived HPAC (H-HPAC) synthesized by pyrolysis of polyvinylpyrrolidone hydrogel under an argon atmosphere at 900 ◦C was successfully obtained [9]. The merits of H-HPAC can be attributed to (i) numerous water molecules encapsulated within PVP hydrogel efficiently serving as green templates, and (ii) the simultaneous function of K2CO<sup>3</sup> as an initiator for hydrogel formation and an activator to enable rich porous conformations. Accordingly, the resultant H-HPAC revealed fascinating structural features and distinguished capacitive performances for electrochemical storage applications. Inspired by the concepts of C/C composites and hydrogels, an alternative H-HPAC encapsulated H-HPAC (H@H) composite material was proposed in the present study. After systemically investigating the physicochemical and morphological properties, the H@H electrodes with a mass loading of 7.5 mg/cm<sup>2</sup> were prepared to evaluate the electrochemical performances of supercapacitors that were assembled with an organic electrolyte. Moreover, various factors such as the physicochemical and textural properties that affect the corresponding electrochemical performances were also explored. On the basis of the results and viewpoints reported here, it is reasonably anticipated that such a strategy also has general applicability to other C/C composites.

#### **2. Results and Discussion**

#### *2.1. Characterizations of Hydrogel-Derived Hierarchical Porous Activated Carbon (H-HPAC)-Encapsulated H-HPAC (H@H) Composite Material*

To explore the physicochemical and morphological properties of H@H composite material, it was systematically investigated by PXRD, Raman, SEM, TEM, BET, EA, and XPS, with the corresponding characteristics compared with those of H-HPAC. Figure 1 depicts the normalized PXRD pattern of H@H to show its crystalline structure. As can be seen, only two broad peaks, assigned to the (002) and (100) planes of carbon (JCPDS No.: 41-1487), were reflected, which was consistent with the original H-HPAC and other activated carbon materials [9,14,17,20,22]. In line with the possible formation mechanism for PVP hydrogel proposed in our previous study, the cross-linking reactions among the polymer chains were initiated by the coordination between potassium cations (K<sup>+</sup> ) and oxygen anions (O−) [9]. When the K2CO<sup>3</sup> solution was added to the PVP/H-HPAC solution, it is reasonably postulated that the K<sup>+</sup> would also interact with the H-HPAC because 4.7% of the oxygen present in the original H-HPAC was verified by elemental analysis [9]. If so, the K<sup>+</sup> coordination among the H-HPAC would be covered by the PVP hydrogel. Even so, no peaks associated with unreacted K2CO<sup>3</sup> were observed from the PXRD pattern, implying

that the purity of H@H was not affected by the presence of H-HPAC after thoroughly rinsing with DI water.

**Figure 1.** Normalized PXRD pattern of H@H composite material.

− − Raman characterization is another approach that can directly examine the crystallinity of carbonaceous materials. The normalized Raman spectrum illustrated in Figure 2 shows two distinct peaks representing the D (~1327 cm−<sup>1</sup> ) and G (~1593 cm−<sup>1</sup> ) bands. Besides, it is meaningful to discuss the intensity ratio between D and G bands (*I*D/*I*G) because the degree of defects within the carbonaceous materials can be further evaluated. The value calculated from the H@H was 1.16, the same as the H-HPAC and close to that of the A-PVP-NC (1.18) [21]. The high *I*D/*I*<sup>G</sup> ratio suggests that many defects and/or highly disordered degrees exist, as is generally observed in the carbonaceous materials with numerous functional groups [23,24]. Consequently, we could ascribe this result to the presence of heteroatoms (i.e., N and O) and lower crystallinity, as demonstrated in the original H-HPAC [9]. On the other hand, the corresponding Raman spectrum was sequentially deconvoluted into four peaks (labeled peaks (1)–(4)) since the integrated area ratio of sp<sup>3</sup> to sp<sup>2</sup> (*A*sp <sup>3</sup>/*A*sp 2 ) has been shown to provide helpful information on the nature of carbon, e.g., a low *A*sp <sup>3</sup>/*A*sp 2 ratio indicates that a large amount of carbon exists as the sp<sup>2</sup> type [25,26]. Among them, peaks (2) and (4) are associated with sp<sup>2</sup> -type carbon, whereas the others are related to sp<sup>3</sup> -type carbon [27]. The integrated area ratio of sp<sup>3</sup> to sp<sup>2</sup> (*A*sp <sup>3</sup> / *A*sp 2 ) was calculated to be 0.28, which was identical to that of H-HPAC [9]. This result signifies that the H@H still retained a high proportion of sp<sup>2</sup> -type carbons, even with intrinsically lower crystallinity.

**Figure 2.** Fitted Raman spectrum of H@H composite material.

→ → →

To examine the morphological features of H@H, micrographs were captured using SEM and TEM. The hierarchically porous architectures constructed by interconnected carbonaceous frameworks were clearly visible from the low-magnification SEM micrographs in Figure 3a,b. It is worth mentioning that rough surfaces with numerous voids were found, as indicated by white circles in Figure 3b. The diverse porous configurations are reasonably attributed to the water molecules encapsulating within PVP/H-HPAC hydrogel being evaporated and the activation process by interacting the carbonized residues with K2CO<sup>3</sup> under 900 ◦C (i.e., K2CO<sup>3</sup> + 2C → 2K + 3CO, K2CO<sup>3</sup> → K2O + CO2, C + CO<sup>2</sup> → 2CO) [18,19]. Based on the morphologies found in SEM, it is expected that similar characteristics were also exhibited, as shown in Figure S2 and Figure 3c at different TEM magnifications. Moreover, it is seen that the short-range disorders, such as carbon lattices, highlighted by white circles were displayed in Figure 3d, which might be correlated with skeleton collapse after high-temperature pyrolysis [28].

μ μ **Figure 3.** (**a**,**b**) SEM and (**c**,**d**) TEM micrographs of the H@H composite material. Scale bar: (**a**) 10 µm, (**b**) 5 µm, (**c**) 50 nm, and (**d**) 10 nm.

Given the positive results found in SEM and TEM, it is believed that the textural characteristics of H@H would not be significantly affected. To accurately classify pores and determine the specific surface area (SSA), the nitrogen adsorption–desorption measurement was conducted, and the corresponding isotherm is shown in Figure 4. As plotted, not only a high volume of nitrogen gases were adsorbed and desorbed at low relative pressure (i.e., Type I isotherm), but also the predominant pore diameter was less than 2 nm (inset of Figure 4), confirming the microporous feature for H@H [9,21,29]. However, all values diminished except for the pore size distributions of ultramicropores as compared with H-HPAC (see Table 1). For instance, the SSA value determined by the Brunauer–Emmett– Teller (BET) method decreased by about 35%. Accordingly, the SSA values contributed by micropores and mesopores were decrease to 1246 m2/g and 45 m2/g, respectively. In particularly, the latter was reduced by even approximately 84%. When preparing the

PVP/H-HPAC composite hydrogel, these pores within the H-HPAC would be filled with the viscous PVP solution so that the PVP blocked the original pores after drying of the composite hydrogel at 120 ◦C. In addition to the issue mentioned above, the decrease in the textural parameters might also be actuated by the possible interaction between H-HPAC and K<sup>+</sup> . Such a phenomenon is rationally postulated to affect the cross-linking degree of the PVP/H-HPAC composite hydrogel, as shown by the XRD characterization, leading to fewer sites for activation.

**Figure 4.** Nitrogen adsorption–desorption isotherm of H@H composite material collected by an accelerated surface area and porosimetry system at 77 K. Inset shows the pore size distribution curve determined by the 2D-NLDFT model.


**Table 1.** Textural properties of H@H composite material and H-HPAC.

<sup>1</sup> V<sup>t</sup> : total pore (single-point) volume obtained from the amount of adsorbed nitrogen at P/P<sup>0</sup> = 0.995. <sup>2</sup> Vultra: volume of ultramicropores (pores < 0.7 nm). <sup>3</sup> Vmicro: volume of micropores (pores < 2.0 nm). <sup>4</sup> Vmeso: volume of mesopores (difference between V<sup>t</sup> and Vmicro). <sup>5</sup> The values were obtained from Ref. [9].

The compositional information and chemical environments of H@H were identified through EA and XPS, respectively. As quantified by the former, the proportions of carbon, nitrogen, and oxygen in the as-synthesized H@H were 76.2%, 0.58%, and 4.5%, respectively. In comparison with the original H-HPAC, the carbon content was decreased (76.2% vs. 95.1%), but the nitrogen species was enhanced (0.58% vs. 0.23%), while the oxygen species was similar (4.5% vs. 4.7%). It would be attributed the variation in carbon to the cross-linking degree of PVP/H-HPAC composite hydrogel. As for the increase in the nitrogen species, the following possible reasons could be given. It is reported that the nitrogen species included in the carbon precursor/char were preferentially removed during chemical activation with K-based salts [20]. However, as previously mentioned in XRD characterization, K<sup>+</sup> would also interact with the original H-HPAC due to the presence of oxygen, reducing the concentration of K<sup>+</sup> that was coordinated to the PVP as compared to the preparation of the original H-HPAC. According to Ref. [20] and our experimental results [9,21], the higher nitrogen percentage in the H@H could be attributed to the lower interaction between the K<sup>+</sup> and carbonaceous residues.

Figure 5 provides the high-resolution XPS spectra that were analyzed using the Gaussian–Lorentzian fitting method. From the EA result, the presence of nitrogen atoms within the H@H was already demonstrated. Therefore, the C-N bonding in the C 1s spectrum did not particularly point out for better reading. As revealed in Figure 5a, the C 1s

spectrum was deconvoluted into four peaks: (1) C=C bond at 284.8 eV, (2) C-O bond at 285.9 eV, (3) C=O bond at 287.8 eV, and (4) O=C-O bond at 290.2 eV [30,31]. It is known that the first peak was assigned to the sp<sup>2</sup> -type carbon, while the rest corresponded to the contribution of sp<sup>3</sup> -type carbon [27]. As for the O 1s spectrum (Figure 5b), three peaks fitted at 531.5 eV, 533.1 eV, and 535.0 eV have appeared, representing (1) O=C-O, (2) C=O, and (3) C-O bonds, respectively [32]. Even though lower nitrogen content was shown in the EA results, the N 1s peak in the binding energy between 396 eV and 402 eV still can be detected (Figure 5c) [9,21,33].

− − − − **Figure 5.** High-resolution XPS spectra of H@H composite material: (**a**) C 1s ((1) for C=C, (2) for C−O, (3) for C=O, and (4) for O=C−O bonds), (**b**) O 1s ((1) O=C−O, (2) C=O, and (3) C−O bonds), and (**c**) N 1s.

Based on the results discussed in this section, the as-prepared H@H produced from thermal pyrolysis of the PVP/H-HPAC composite hydrogel combines various benefits, such as good purity, hierarchical porous characteristics, and high proportions of sp<sup>2</sup> -type carbons, as well as nitrogen species. Although the structural parameters were significantly altered, it is of interest to consider the influence of physicochemical and textural features of the H@H on the corresponding electrochemical performance as organic supercapacitors.

#### *2.2. Electrochemical Performances of H@H in Symmetric Supercapacitor*

To evaluate the capacitive efficiencies of the H@H electrode, coin-type symmetric cells were fabricated to conduct the cyclic voltammetry (CV) and galvanostatic charge-discharge (GCD) measurements in the voltage window between 0 V and 2.7 V by different scanning rates and current densities. Although the SSA values and relative textural parameters were less than the H-HPAC, as discussed previously, the typical curves in nearly rectangular shapes with good symmetries were reflected from the H@H (Figure 6a), despite gradually increasing the scanning rate to 10 mV/s. During CV cycling, the integral area from the cyclic voltammogram is associated with the charges adsorbed and desorbed among the active materials. Figure 6b compares the integral area of H@H and H-HPAC acquired in terms of the forward and backward scanning. As can be seen, the values linearly increased with the scanning rate. In addition, the values for the H@H electrode were superior those

Δ

for the H-HPAC electrode; this was correlated with the different mass loading (7.5 mg/cm<sup>2</sup> vs. 5.1 mg/cm<sup>2</sup> ). On the other hand, the voltage delay (∆V) is regarded as an important indicator providing similar information to the IR drop. To discuss this discrepancy, we consider the voltage that reached zero for a current density at 1 mV/s as a reference. The ∆V values were then calculated by comparing the difference between the voltage recorded from each scanning rate and the reference; the corresponding data are visible in Figure S3. Under the voltage range of 0 V to 2.7 V and a scanning rate of 1 mV/s, the reference data for each free-standing electrode were 45 mV (HPAC electrode: 230 µm [7]), 22 mV (H@H electrode: 160 µm) and 12 mV (H-HPAC electrode: 100 µm [9]), respectively. With increase in scanning rate to 10 mV/s, it was found that the ∆V value was 128 mV, i.e., a 64% increase as compared to that of the H-HPAC (78 mV) [9]. This could be ascribed to the thickness of the H@H electrode, which was ~60% more than the H-HPAC electrode, prolonging the pathway for electron transportation. Δ μ μ μ Δ

**Figure 6.** (**a**) Cyclic voltammograms collected in the voltage range between 0 and 2.7 V at scanning rates from 1 to 10 mV/s and (**b**) dependence of the integral area on scanning rate, with the values acquired from (**a**) and Ref. [9] for H@H and H-HPAC, respectively.

Figure 7a presents the GCD profiles measured using the same voltage window as in the CV test but with current densities from 1 mA/cm<sup>2</sup> to 10 mA/cm<sup>2</sup> . Contributing to the ideal electric double-layer behavior and high reversibility, as shown in Figure 6a, the linear and symmetric charge–discharge behavior at each current density was observed. The specific capacitance discharged from the H@H electrode at the 100th cycle was 118.6 F/g at 1 mA/cm<sup>2</sup> with 99% Coulombic efficiency. This result was slightly higher than that outputted from the H-HPAC electrode (117.5 F/g [9]), implying that the specific capacitance was not appreciably affected by the thickness when applying a small current density. Additionally, the stable discharge capacitances of 110.2, 98.8, 81.6, and 73.9 F/g are compared in Figure 7b. The tendency for capacitance decay was the same as the H-HPAC, but the values were observably declined while the current densities were above 4 mA/cm<sup>2</sup> , which could also be attributed to the change in thickness. However, 96.4% of the recovery in capacitance after 100 cycles was obtained when the current density was returned to 1 mA/cm<sup>2</sup> . Figure 7c shows the EIS spectra that were recorded before and after rate-capability testing. It is recognized that charge transfer resistance (*R*ct) includes ionic and electronic resistances. The former is the resistance to the mobility of ionic electrolytes inside the textual pores of the electrode, while the latter comprises the intrinsic resistance of the electrode material and the contact resistance between the active layer and the current collector [34]. For the rate-capability testing, the same electrode conditions (i.e., composition, working area, and thickness) were used, based on the hypothesis that the electronic resistance should be no significant differences. Following repeated charging–discharging processes, the *R*ct value

increased from 18.5 Ohm to 29.1 Ohm. As reported, the diameter of solvated ions for TEA<sup>+</sup> and BF<sup>4</sup> − are 1.35 nm and 1.40 nm, respectively [35]. Hence, the bulky solvated ions would accumulate within the pores of H@H, further blocking the ionic transport and causing an increase in *R*ct as well as capacitance decay [36].

The energy and power densities calculated from the data presented in Figure 7b and the equations shown in Section 3.4 are plotted in Figure 7d. As indicated, the values ranged from 30.0 Wh/kg (@ 1 mA/cm<sup>2</sup> ) to 18.7 Wh/kg (@ 10 mA/cm<sup>2</sup> ) for the former, and from 88.1 W/kg (@ 1 mA/cm<sup>2</sup> ) to 881.7 W/kg (@ 10 mA/cm<sup>2</sup> ) for the latter. To compare with the H-HPAC electrode, ~74% of the power density was outputted by the H@H electrode for all current densities. On the basis of the same electrolyte and similar voltage window, the H@H electrode exhibits reasonable energy and power densities, comparable with the results reported previously (Figure 7d) [9,15,37–43]. Considering the cycling stability, the accelerated experiment conducted in the voltage ranged from 1.35 to 2.7 V (i.e., 50% of the state of discharge) and the current of 5 mA was used; the corresponding result is displayed in Figure 7e. The initial capacitance discharge to 1.35 V was ~0.5 F. After 30,000 cycles, about 76.6% of capacitance retention and ≥99.5% of Coulombic efficiency were found, respectively. Although the *R*ct value increased from 32.8 Ohm to 45.1 Ohm (see Figure S4), the H@H electrode provided a better lifespan than that of the H-HPAC electrode (capacitance retention: 76% after 10,000 cycles) [9]. It is reported that increasing the mass loading or electrode thickness leads to a decrease in capacitance and the rate capability of the electrode materials, which is related to the decreased accessible surface area, increased electrical resistance, prolonged ion transport channels, and poor electrolyte wetting [4]. Besides, the variety of heteroatom dopants and their corresponding amounts, as well as the porosity characteristics, within HPAC were also significant influences [17,44]. The comparison of the electrochemical performance of H@H and that reported for HPAC in symmetric supercapacitors using 1 M TEABF4/PC electrolyte is listed in Table S1. The variations in the electrochemical performance of H@H can be attributed to the following. First, even though the SSA value and pore parameters were lower than those for H-HPAC, the increased thickness of the H@H electrode would compensate for their active sites of capacitive storage, because the specific capacitance generated from the H@H electrode at 1 mA/cm<sup>2</sup> was slightly enhanced. Second, the number of nitrogen species doped in the H@H was increased by up to 152% in comparison with the H-HPAC, so the GCD profiles of the first and last five cycles in Figure 7f showed high symmetry, meaning that the overall resistance was not significant, even when increasing the thickness by 60% and after 30,000 cycles.

**Figure 7.** *Cont*.

μ μ

**Figure 7.** (**a**) Galvanostatic charge–discharge profiles, (**b**) rate capabilities, (**c**) electrochemical impedance spectra, (**d**) Ragone plots, (**e**) discharge capacitance as a function of cycle number of the H@H electrode in a coin-type symmetric cell, and (**f**) galvanostatic charge–discharge profiles of the first and last five cycles received from (**e**). The inset of (**c**) illustrates the equivalent circuit model used for the parameter fitting [45].

#### **3. Materials and Methods**

#### *3.1. Chemicals*

μ μ All reagents, including polyvinylpyrrolidone (PVP, (C6H9NO)n, average MW: 1,300,000, Sigma-Aldrich, St. Louis, MO, USA), potassium carbonate (anhydrous, K2CO3, 99%, Alfa Aesar, Heysham, UK), carbon black (Super P®, Timcal Ltd., Bodio, Switzerland), vaporgrown carbon nanofibers (VGCFs, 7 µm in length and 0.11 µm in diameter, Yonyu Applied Technology Material Co., Ltd., Tainan, Taiwan), and colloidal polytetrafluoroethylene (PTFE) dispersion (D1-E, Daikin Industries Ltd., Osaka, Japan), were adopted without further purification. Deionized (DI) water produced from a Milli-Q Integral water purification system (Millipore Ltd., Burlington, MA, USA) was utilized throughout the experiments.

#### *3.2. Preparation of Hydrogel-Derived Hierarchical Porous Activated Carbon (H-HPAC) Encapsulated H-HPAC (H@H) Composite Material*

To construct the H@H composite material, PVP powders and K2CO<sup>3</sup> were welldissolved in DI water individually. Here, the mass ratio between K2CO<sup>3</sup> and PVP was controlled at 2 as no hydrogel is formed when the ratio was less than 2, as demonstrated in our previous study [9]. The concentrations of PVP and K2CO<sup>3</sup> solutions were 14.3 wt.% and 40.7 wt.%, respectively. The volume ratio between PVP solution and K2CO<sup>3</sup> solution was 2. Then, 0.5 g of H-HPAC was carefully added to the PVP solution, whereas the resultant was vigorously stirred to ensure homogenous mixing. Following the addition of K2CO<sup>3</sup> solution dropwise to the PVP/H-HPAC solution, the black elastomer-like sample was completely obtained within 5 min (see Figure S1). The resulting composite hydrogel was dried in an oven at 120 ◦C for 12 h to completely evaporate the water molecules that were encapsulated within the matrix. The residues were then thermally pyrolyzed in a tube furnace at 900 ◦C for 2 h under an argon atmosphere with a flow rate of 200 mL/min, so the newly formed H-HPAC converted from the PVP would encapsulate the original H-HPAC. After repeated rinses with DI water, drying, and grinding procedures, the loose H@H powders can be obtained.

#### *3.3. Characterizations*

The crystalline structure of the as-prepared H@H composite material was identified using a powder X-ray diffractometer (XRD, D2 PHASER, Bruker AXS Inc., Karlsruhe, Germany) with a Cu target (λ = 1.541 Å) that was excited at 30 kV and 10 mA. The corresponding PXRD pattern was recorded in the range of 2*θ* from 10◦ to 70◦ at a scanning rate of 0.5 s/step. The Raman spectrum was collected between 1000 cm−<sup>1</sup> and 1800 cm−<sup>1</sup> by a confocal Raman microscope (inVia, Renishaw, UK) equipped with a 633 nm laser source. For morphological observations, the micrographs were acquired from the scanning electron microscope (SEM, JSM-IT200, JEOL Ltd., Tokyo, Japan) and a transmission electron microscope (TEM, JEM-2100, JEOL Ltd., Tokyo, Japan). To examine the textural properties,

the N<sup>2</sup> adsorption–desorption isotherm was measured at 77 K on a surface area and porosity analyzer (ASAP 2020 V3.00, Micromeritics Instrument Corporation, Norcross, GA, USA) after degassing under vacuum at 160 ◦C for 8 h. An elemental analyzer (FLASH 2000, Thermo Fisher Scientific Inc., Waltham, MA, USA) was applied for determining the percentages of carbon, nitrogen, and oxygen in the H@H composite material. The chemical environments were analyzed with X-ray photoelectron spectroscopy (XPS, PHI 5000 VersaProbe III, ULVAC-PHI, Inc., Kanagawa, Japan) with a beam size of 100 µm under Al K<sup>α</sup> radiation (λ = 8.3406 Å). Their corresponding high-resolution spectra were further deconvoluted by the Gaussian–Lorentzian fitting method using an XPSPEAK 4.1 software.

#### *3.4. Electrochemical Measurements*

The electrochemical tests throughout this study were conducted in the symmetric two-electrode configuration at ambient conditions. To prepare the free-standing H@H electrodes, the ingredients (80 wt.% of H@H, 5 wt.% of Super P®, 5 wt.% of VGCFs, and 10 wt.% of PTFE) were mechanically blended and repeatedly calendared. The as-prepared H@H electrodes with a thickness of 160 <sup>±</sup> <sup>7</sup> <sup>µ</sup>m and a mass loading of 7.5 <sup>±</sup> 0.8 mg/cm<sup>2</sup> were obtained after drying at 130◦C. To assemble the coin-type cells, 1 M TEABF4/PC and cellulose-based membrane (TF4535, NKK, Kochi, Japan) were used as the organic electrolyte and separator, respectively. The cyclic voltammograms (CVs) and electrochemical impedance spectroscopy measurements were recorded using a multichannel electrochemical workstation (VSP-3e, Bio-Logic, Seyssinet-Pariset, France). The electrochemical impedance spectra (EIS) were recorded at open circuit potential (OCP) from 100 kHz to 0.01 Hz with an AC potential amplitude of 5 mV. The galvanostatic charge–discharge (GCD) profiles and the cycling stabilities were evaluated through a computer-controlled system (CT-4008T-5V50mA, Neware Technology Limited, Shenzhen, China). To determine the specific capacitance (*C<sup>s</sup>* , F/g) of the H@H electrode in the symmetric supercapacitor, the value can be calculated from the GCD profiles by *C<sup>s</sup>* = 2 *It*/*mV*, where *I* is the applied current (A), *t* is the recorded discharge time (s), *m* is the mass of active material at one electrode (g), and *V* is the voltage window (volts). As for the energy density (*E*, Wh/kg) and power density (*P*, W/kg), they can be further acquired based on the equations *E* = *CsV* <sup>2</sup>/(2 <sup>×</sup> <sup>4</sup> <sup>×</sup> 3.6) and *<sup>P</sup>* = 3600 *<sup>E</sup>*/*t*, respectively [9,15].

#### **4. Conclusions**

In summary, this study presents an alternative concept for the construction of the hydrogel-derived HPAC (H-HPAC) encapsulated H-HPAC (H@H) composite material through the thermal pyrolysis of a PVP/H-HPAC hydrogel under an argon atmosphere at 900 ◦C. Compared to the original H-HPAC, the as-prepared H@H retains good purity, lower crystallinity, and high proportions of sp<sup>2</sup> -type carbons. However, H@H has a lower specific surface area and decreased pore parameters, but a substantial increase in the percentage of nitrogen species. Even with the notable change in the textural features, the symmetric supercapacitor assembled by the H@H electrode with a mass loading of 7.5 mg/cm<sup>2</sup> and organic electrolyte still exhibits good reversible capacitance, comparable rate capability, and excellent cyclability. The results presented in this study support the H@H as a promising electrode material for other electrochemical energy storage fields, such as metal–ion capacitors.

**Supplementary Materials:** The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/ijms23126774/s1.

**Author Contributions:** Conceptualization, T.-F.H.; methodology, T.-F.H., R.M.A. and H.-H.H.; validation, T.-F.H., R.M.A., H.-H.H., M.M.A. and S.G.M.; investigation, T.-F.H., R.M.A., H.-H.H., M.M.A. and A.A.; resources, T.-F.H. and C.-C.Y.; writing—original draft preparation, T.-F.H. and R.M.A., writing—review and editing, T.-F.H., R.M.A., M.M.A., S.G.M. and C.-C.Y.; visualization, T.-F.H., R.M.A., H.-H.H., M.M.A. and A.A.; supervision, T.-F.H.; funding acquisition, T.-F.H. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Ministry of Science and Technology (MOST) of Taiwan (grant number: MOST 110-2222-E-131-001-MY3) and Ming Chi University of Technology (grant number: VK003-6100-110).

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Data sharing is not applicable to this article.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


#### *Article* **Function of Graphene Oxide as the "Nanoquencher" for Hg 2+ Detection Using an Exonuclease I-Assisted Biosensor**

**Ting Sun <sup>1</sup> , Xian Li 1 , Xiaochuan Jin 1 , Ziyi Wu 1 , Xiachao Chen <sup>2</sup> and Jieqiong Qiu 1, \***


**Abstract:** Graphene oxide is well known for its excellent fluorescence quenching ability. In this study, positively charged graphene oxide (pGO25000) was developed as a fluorescence quencher that is water-soluble and synthesized by grafting polyetherimide onto graphene oxide nanosheets by a carbodiimide reaction. Compared to graphene oxide, the fluorescence quenching ability of pGO25000 is significantly improved by the increase in the affinity between pGO25000 and the DNA strand, which is introduced by the additional electrostatic interaction. The FAM-labeled single-stranded DNA probe can be almost completely quenched at concentrations of pGO25000 as low as 0.1 µg/mL. A simple and novel FAM-labeled single-stranded DNA sensor was designed for Hg 2+ detection to take advantage of exonuclease I-triggered single-stranded DNA hydrolysis, and pGO25000 acted as a fluorescence quencher. The FAM-labeled single-stranded DNA probe is present as a hairpin structure by the formation of T–Hg 2+–T when Hg 2+ is present, and no fluorescence is observed. It is digested by exonuclease I without Hg 2+ , and fluorescence is recovered. The fluorescence intensity of the proposed biosensor was positively correlated with the Hg 2+ concentration in the range of 0–250 nM (R <sup>2</sup> = 0.9955), with a seasonable limit of detection (3σ) cal. 3.93 nM. It was successfully applied to real samples of pond water for Hg 2+ detection, obtaining a recovery rate from 99.6% to 101.1%.

**Keywords:** positively charged graphene oxide (pGO); exonuclease I; fluorescence quencher; hairpin structure; T–Hg 2+–T

#### **1. Introduction**

Water-soluble mercury(II) ion (Hg 2+ ), as one of the most familiar environmental pollutants, is a toxic heavy metal that can exist in metallic, inorganic, and organic forms, especially in freshwater and marine ecosystems [1]. After prolonged exposure, it is extremely toxic to the brain, kidney, and other organs of organisms at very low mercury(II) concentrations [2]. The accumulation of heavy metals can occur in animal and human bodies via the food chain and damage the reproductive, gastrointestinal, and cardiovascular systems. Based on the guidelines of the United States Environmental Protection Agency (EPA), the Ministry of Health of the People's Republic of China (MOH), and the World Health Organization (WTO), the maximum mercury(II) concentration in drinking water should be as low as 10 nM [3].

Currently, many traditional techniques have been developed for Hg 2+ analysis and detection, including inductively coupled plasma–optical emission spectrometry (ICP-OES) [4], chemical vapor generation–inductively coupled plasma–optical emission spectrometry (CVG-ICP-OES) [5], inductively coupled plasma–mass spectrometry (ICP-MS) [6], cold vapor atomic absorption spectroscopy (CVAAS) [7], atomic fluorescent spectroscopy [8], and electrochemical methods. Hg 2+ can be detected at pM concentrations by most of the abovementioned methods. However, high-cost, complex sample preparation and professional

**Citation:** Sun, T.; Li, X.; Jin, X.; Wu, Z.; Chen, X.; Qiu, J. Function of Graphene Oxide as the "Nanoquencher" for Hg 2+ Detection Using an Exonuclease I-Assisted Biosensor. *Int. J. Mol. Sci.* **2022**, *23*, 6326. https://doi.org/10.3390/ ijms23116326

Academic Editor: Ana María Díez-Pascual

Received: 11 May 2022 Accepted: 3 June 2022 Published: 5 June 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

operation are also needed. Therefore, it is essential to explore rapid, specific, sensitive, cost-efficient, convenient, and real-time biosensors instead of traditional approaches for monitoring heavy metal ions.

Fluorescence-based methods have been widely used as potential techniques for Hg2+ detection. In view of the sensitivity improvement for Hg2+ detection, an increasing number of researchers have paid attention to signal amplification in DNA-based strategies by exonucleases, including exonuclease I and III (Exo I and III) [9–13]. For example, Exo I is a 3′–5′ exonuclease that can cleave single-stranded DNA (ssDNA) without sequencedependence [14,15]. Hg2+ can promote the formation of a DNA duplex (dsDNA) via T–Hg2+–T formation, which is not allowed to be digested by Exo I. However, Exo I can hydrolyze ssDNA. Thus, there is potential for applying Exo I in Hg2+ detection on the basis of DNA-based signal amplification strategies.

However, traditional fluorescent DNA probes, such as Taqman probes, molecular beacons (MBs), and scorpions, cannot meet the requirements [16,17] because no free 3′ -OH is present. For this reason, a label-free assay has been developed based on the fluorescence "turn-on" caused by dye intercalation into special DNA structures [18–20]. However, such label-free dyes have a non-negligible fluorescent background.

To solve these problems, guanine bases [16] and nanostructures [21] have been used as quenchers for DNA probes. Various nanostructures, such as gold nanoparticles (AuNPs) [22,23], single-walled carbon nanotubes (SWCNTs) [24], graphene oxide (GO) [25], fullerene (C60) [26,27], multiwalled carbon nanotubes (MWCNTs) [9], and positive carbon dots (P-CDs) [28], have been successfully used as nanoquenchers for mercury(II) ion detection. As a two-dimensional (2D) material, GO exhibits high-efficiency fluorescence quenching, good water dispersibility, low cost, and various surface modifications. Therefore, it is frequently used in biosensors [29]. Previous studies have confirmed that ssDNA is labelled with a fluorescent dye, which can be quenched by GO due to fluorescence resonance energy transfer (FRET) [12,30]. This result is attributed to the hydrogen bond and π–π stacking caused by nucleobases and GO, which make FRET more efficient. The fluorescence quenching efficiency of GO is dependent on the GO quantity used. A high concentration of GO can limit its application in Hg2+ detection, e.g., in cells. Therefore, it is necessary to improve the fluorescence quenching ability and efficiency of GO. In addition, the fluorescence quenching efficiency of GO can be increased by partially reducing graphene oxide due to the increase in π–π stacking interaction [25].

In this work, positively charged graphene oxide (pGO25000) was synthesized by grafting polyetherimide (PEI) onto GO nanosheets by a carbodiimide reaction. The first use of positively charged pGO25000 as an efficient fluorescence quencher was demonstrated. Compared to GO, the fluorescence quenching efficiency of pGO25000 can be enhanced by the positively charged surface that allows attraction of the negatively charged DNA strands via electrostatic interaction. Based on the special property of pGO25000, a FAM–ssDNA probe was designed for the highly selective and ultrasensitive detection of mercury(II) ions with the assistance of the Exo I enzyme under mild conditions.

#### **2. Results and Discussion**

#### *2.1. Strategy for Ultrasensitive Detection of Hg2+*

A FAM–ssDNA probe was designed for the highly sensitive and highly selective detection of Hg2+, with pGO25000 being a fluorescence nanoquencher and Exo I being a special enzyme for hydrolyzing ssDNA in the 3′→5 ′ direction, as shown in Scheme 1. GO, as a fluorescence quencher, can quench fluorescent dye via FRET when GO and fluorescent dye are sufficiently close to each other. Considering the binding affinity of ssDNA to GO, which results from π–π stacking and hydrogen bonding between ssDNA and GO, the dye-labelled ssDNA probes were designed for Hg2+ detection based on the GO fluorescence quenching ability. Positively charged GO (PEI-GO) has been reported as a fluorescence quencher of anionic dyes (i.e., Merocyanine 540) via electrostatic interactions [31]. However, dye-labelled DNA probes have never been reported as fluorescence quenchers. In this work,

pGO25000 was synthesized by grafting PEI (M.W. = 25,000) onto GO nanosheets, which can selectively bind to ssDNA/dsDNA at very low concentrations. Thus, the fluorescence quenching efficiency was enhanced based on the additional electrostatic attraction between the phosphate group and positively charged PEI. The fluorescence was almost quenched for the FAM–ssDNA probe by pGO25000; however, FAM–dsDNA can be also quenched by pGO25000. To solve this problem, enzyme-based technology was applied in the FAM– ssDNA/pGO25000 system. Exo I is a sequence-independent 3′–5′ exonuclease that cleaves ssDNA. It has been reported that the digestion of Exo I is limited by binding to the targets to form the DNA duplex and G-quadruplex structures [11,14,15,32]. In this study, pGO25000 was synthesized as an efficient nanoquencher of fluorescence for the proposed strategy to detect Hg2+. After adding Exo I, the special enzyme efficiently digested the FAM–ssDNA in the direction of 3′ to 5′ , and the fluorescence was restored. However, if dsDNA was present because of the formation of the T–Hg2+–T construct after adding Hg2+, which could suppress the activity of Exo I, no fluorescence was restored. Therefore, the fluorescence "turn-on" indicated that no Hg2+ was present in the analytical sample, and vice versa. It is expected that this strategy could provide a novel method to detect Hg2+ with great sensitivity and high selectivity. ′ ′ ′ ′

**Scheme 1.** Schematic diagram of the designed fluorescence "turn-off" strategy for Hg2+ detection with the assistance of Exo I nuclease using a FAM–ssDNA probe.

#### *2.2. Characterization of pGO25000*

GO has good water solubility due to the abundance of hydrophilic groups (hydroxyl, carboxylic, epoxy) that have been introduced onto the surface of GO after a series of

chemical modification processes. Because pGO25000 was prepared by grafting PEI onto GO nanosheets, pGO25000 also has good dissolvability. To study the surface charge of GO and pGO25000, zeta potential analysis was performed at concentrations of 1 mg/mL GO and pGO25000 solution. The GO solution showed a negative zeta potential level of −37.6 Mv, as shown in Figure 1, while pGO25000 had a positive zeta potential level of 25 mV because the PEI linkers completely changed the pGO25000 surface charge; thus, positively charged GO (pGO25000) was obtained [33]. To investigate the structural change in the condensation reaction of pGO25000 synthesis, Raman spectroscopy, FT-IR analysis, and high-resolution XPS were performed. Figure 2A shows the Raman spectra of GO and pGO25000, and two bands located at approximately 1320 cm−<sup>1</sup> and 1596 cm−<sup>1</sup> can be attributed to the D and G bands of graphitic materials, respectively. It is well known that the defect level of graphene sheets can be evaluated by the peak intensity ratio of the D band to the G band (ID/IG), and a higher ID/I<sup>G</sup> commonly indicates an increase in the degree of disorder [34]. pGO25000 gave a higher ID/I<sup>G</sup> ratio of 2.30 compared with GO (1.85), which can be attributed to the condensation reaction by incorporation of PEI, reducing the oxygen functional group and increasing the sp<sup>3</sup> carbon form [35–37]. Figure 2B shows the FT-IR spectra of GO and pGO25000. Peaks located at ~1720 cm−<sup>1</sup> , ~1620 cm−<sup>1</sup> , ~1400 cm−<sup>1</sup> , and ~1090 cm−<sup>1</sup> can be assigned to the stretching vibrations of the C=O, C–C, C–OH, and C–O (epoxy) groups [35,38,39]. Of note, the FT-IR spectrum of pGO25000 showed that PEI was successfully grafted onto the GO surface. Compared to GO, the N–C=O peak at 1650 cm−<sup>1</sup> appeared with the disappearance of the C=O peak at 1720 cm−<sup>1</sup> in pGO25000. Meanwhile, the C–O (epoxy) peak was replaced by the C–N peak (1384 cm−<sup>1</sup> ) on pGO25000. The N–C=O and C–N groups were produced by the amine reacting with the COOH and C–O (epoxy) groups. The band at 1580 cm−<sup>1</sup> appeared first, which corresponded to the C=N stretch by Schiff's base reaction [40,41]. − − − − − − − − −

**Figure 1.** Charge analysis of GO and pGO25000.

As shown in Figure 3A, there was almost no N1s signal in the spectrum of GO, whereas the spectrum of pGO25000 presented a clear N1s peak. After calibration of the binding energy position with C1s (284.4 eV) in XPS spectra, the five main peaks of carbon bonding in the C1s XPS spectra of GO with binding energies at 283.7, 284.4, 286.1, 286.9, and 288.3 eV (Figure S1A) were attributed to the C=C, C–C, C–O (hydroxyl and epoxy), C=O, and C(O)O bonds, respectively [25,42]. After reacting with PEI, the signal at 285.3 eV (C–N bond) appeared along with the disappearance of the C(O)O bond signal, which indicated that the condensation reaction between the amino group and carboxyl group was completed. The peak at 286.0 eV (C–O) was dramatically decreased due to the epoxy reacting with PEI (Figure 3B). The N1s spectrum had fitted curves at 400.4, 399.1, and 398.2 eV (Figure 3C), corresponding to the binding energies of nitrogen atoms in NH<sup>3</sup> + , CONH, and PEI [40,43,44]. Compared to the O1s spectrum of GO (Figure S1B), pGO25000

was deconvoluted into four peaks (Figure 3D), three of which were similar to those of GO, i.e., C=O (530.5 eV), C–OH (531.7 eV), and C–O (532.4 eV) [43]; a new peak with a bonding energy of 530.9 eV appeared, corresponding to the CONH bond. These results indicated that pGO25000 was successfully obtained by GO reacting with PEI.

**Figure 2.** Characterization of GO and pGO25000 by (**A**) Raman spectroscopy and (**B**) FT-IR.

**Figure 3.** High-resolution XPS spectra of GO and pGO25000: (**A**) wide scan, (**B**) C1s spectrum of pGO25000, (**C**) N1s spectrum of pGO25000, and (**D**) O1s spectrum of pGO25000.

#### *2.3. Fluorescence Quenching of FAM–ssDNA by pGO25000*

To better understand the fluorescence quenching efficiency of positively charged GO depending on the pH values, a solution with a pH range of 7.5–9.0 was investigated. The fluorescence intensity of FAM–ssDNA in a 600 nM Hg2+ solution significantly increased and then slightly decreased with increasing pH values, as shown in Figure S2A. At pH 8.5, the maximum fluorescence signal was obtained. After adding 0.1 µg/mL pGO25000, there was a sharp reduction in fluorescence intensity due to fluorescence quenching caused by pGO25000, and there was no large difference under various pH values from 7.5 to 9.0. Hence, pH 8.5 was the optimum pH according to the ratio of the fluorescence intensity without pGO25000 and in the presence of pGO25000. It has been mentioned that fluorescence quenching by pGO25000 can be completed immediately; thus, fluorescence detection was instantly performed after adding pGO25000.

To demonstrate that the positively charged GO (pGO25000) is more efficient in fluorescence quenching for DNA probes, GO and pGO25000 were analyzed. The fluorescence of FAM–ssDNA was quenched by various concentrations of GO from 0.01 to 30 µg/mL, as shown in Figure S3B. As the amount of GO increased, the fluorescent signal was reduced, and fluorescence quenching was not efficient even if the concentration was enhanced to 30 µg/mL. However, the fluorescence was almost quenched by pGO25000 at 0.1 µg/mL (Figure 4). Thus, positively charged PEI plays a crucial role in the affinity between pGO25000 and FAM–ssDNA, which can promote the fluorescent dyes to be close to pGO25000, thus increasing the FRET efficiency to quench fluorescence. The fluorescence spectra of FAM–ssDNA quenched by pGO25000 in the range from 0 to 30 µg/mL with excitation at 495 nm are shown in Figure 4 and Figure S3A. The signal intensity of the FAM– ssDNA probes was moderately reduced with increasing pGO25000 concentration, clearly increased with a pGO25000 concentration higher than 0.1 µg/mL, and then decreased again until the concentration of pGO25000 was greater than 6 µg/mL. The fluorescence quenching ability of pGO25000 was induced by electrostatic and π–π stacking interactions, which were dependent on the concentration of pGO25000. This result indicated that pGO25000 had a better binding affinity with the FAM–ssDNA probes at very low concentrations from 0 to 0.1 µg/mL due to electrostatic interactions. As pGO25000 increased from 0.1 µg/mL to 6 µg/mL, fluorescence quenching was not efficient. This was possibly caused by the steric hindrance of PEI in pGO25000, which impeded the interaction of FAM–ssDNA with GO in pGO25000, reducing π–π stacking and hydrogen bonding between nucleobases and GO. However, the fluorescence quenching efficiency was improved when the pGO25000 concentration was more than 6 µg/mL. This phenomenon was observed because the electrostatic interaction between FAM–ssDNA and pGO25000 was increased, and fluorescence quenching was mainly dependent on the electrostatic interaction.

The influence of various metal ions on the fluorescence quenching ability of pGO25000 was also assessed by measuring the fluorescence intensity, and different metal ions were used, including Hg2+, K<sup>+</sup> , Sn2+, Al3+, Ni2+, Mn2+, Mg2+, Cu2+, and Co2+. Figure S4 shows that there was no influence on the fluorescence quenching ability. Therefore, pGO25000 was synthesized as an efficient nanoquencher of fluorescence.

#### *2.4. Fluorescence Detection of Hg2+*

Fluorescence quenching efficiency is dependent on the interaction between DNA and GO, which is determined by the length of DNA [45,46], GO surface modification [25], size of GO [30], and concentrations of DNA and GO [47]. Compared to GO, the positively charged modified GO (pGO25000) had a perfect fluorescence quenching ability at very low concentrations. However, there was no large change in the fluorescence quenching of FAM–ssDNA by 0.1 µg/mL pGO25000 with or without Hg2+. However, FAM–ssDNA exists in the hairpin structure due to the formation of T–Hg2+–T after adding Hg2+. It was indicated that fluorescence quenching was efficient for the same sequences of DNA with different structures. Based on this result, an Exo I-assisted strategy to detect Hg2+ is proposed.

π–π

π–π

To demonstrate the effect of pH values on T–Hg2+–T complex formation and the activity of Exo I, the Exo I/pGO25000-assisted FAM–ssDNA sensor was studied for Hg2+ detection, as demonstrated in Figure S2B. The fluorescent signal was significantly improved with increasing pH values in the absence of Hg2+. It is well known that FAM is a pHdependent dye, and the optimal pH value is >8.5 [48,49]. FAM–ssDNA can be digested by the 3′–5′ Exo I and releases FAM dyes in a range of pH 7.5~9.0; thus, the fluorescence response can be recovered without Hg2+. When Hg2+ was present, all samples at various pH values were fluorescence-quenched except for the conditions at pH 7.5 and 9.0. Thus, the FAM–ssDNA probe could be subjected to the conditions at pH 8.0 and 8.5. The fluorescence signal was not restored because of the FAM–hairpin DNA structure formation, which was caused by the T–Hg2+–T construction, which prevented digestion by Exo I. However, at pH 7.5 and 9.0, the fluorescence was slightly recovered, and the reason could be attributed to the overactivity of Exo I under these conditions. As a result, 10 mM Tris–HNO<sup>3</sup> buffer (40 mM NaNO3) with a pH value of 8.5 was used during Hg2+ detection.

#### *2.5. Sensitivity of Hg2+ Detection*

The Hg2+ concentration has a large effect on the FAM–ssDNA probe with the pGO25000/ Exo I-assisted strategy. The sensitivity of this proposed method was determined using various concentrations of Hg2+ solution. Figure 5A shows that the emission signal was gradually reduced by excitation at 495 nm with increasing Hg2+ concentrations from 0 to 800 nM, and fluorescence was nearly quenched when the concentration of Hg2+ was greater than 600 nM. In a certain range, the higher the concentration of Hg2+, the more efficient the fluorescence quenching. This occurred because Exo I activity was restricted by the DNA hairpin structure, which was formed by the T–Hg2+–T complexes, and the FAM dye could not be released from the DNA strand, which was quenched by pGO25000. Without Hg2+ added, the fluorescence signal was perfectly recovered after digestion by Exo I. It was attributed to the destruction of the interaction between the DNA strand and pGO25000 during the hydrolysis of DNA; thus, the FAM dye was released from the DNA strand and kept far away from pGO25000; then, the fluorescence was restored.

The relative fluorescence intensity (F/F0) decreased proportionally as the Hg2+ concentration increased. It showed excellent analytical performance with a linear relationship in the range of 0 to 250 nM Hg2+, following a linear correlation equation described as y = <sup>−</sup>0.0031x + 0.9804 (y represents F/F0, x represents the concentration of Hg2+ in solution, R<sup>2</sup> = 0.9955). Based on the 3σ slope, the limit of detection (LOD) for the Exo I/pGO25000-assisted FAM–ssDNA sensor was estimated to be 3.93 nM, which was far below the largest permissible dose of Hg2+ in potable water (10 nM) by the U.S. Environ-

mental Protection Agency (EPA) [21,50]. The obtained result indicates that this biosensor strategy has potential applications in the quantitative analysis of Hg2+ at certain concentrations. Various nanomaterials were used as the nanoquencher of DNA probes, which were designed for Hg(II) detection, and the results are shown in Table 1. The sensitivity and fluorescence quenching efficiency of the nanoquenchers were compared using the previously presented analytical methods. Note that the proposed scheme presented a lower LOD and higher sensitivity compared with that using GO, SWCNTs, or MWCNTs as the nanoquencher, and pGO25000 had a more efficient fluorescence quenching ability compared to C60.

**Figure 5.** (**A**) Fluorescence emission spectra upon addition of various Hg2+ concentrations ranging from 0 to 800 nM in Tris–HNO<sup>3</sup> buffer (10 mM, pH 8.5) containing 40 mM NaNO<sup>3</sup> . The Hg2+ concentrations were as follows: 0, 2, 5, 10, 30, 50, 120, 150, 200, 250, 300, 400, 600, and 800 nM. (**B**) Linear response of the relative fluorescence intensity (F/F<sup>0</sup> , F: fluorescence-detected, with various Hg2+ concentrations from 0 to 800 nM, F<sup>0</sup> : fluorescence-initial, without Hg2+). An amount of 50 nM FAM–ssDNA was used for each reaction, and the reactions were performed in Tris–HNO<sup>3</sup> buffer (10 mM, 40 mM NaNO<sup>3</sup> , pH 8.5). *E*x = 495 nm, *E*m = 520 nm.


**Table 1.** Comparison of different nanomaterials as nanoquencher for Hg(II) detection.

#### *2.6. Selectivity of Hg2+*

′ ′

Confirming the selectivity for Hg2+ is the key point to evaluate the performance of the developed DNA biosensor. To estimate the selectivity of Exo-I/pGO25000 by the FAM– ssDNA sensing system, experiments were performed to detect the fluorescence intensities of the Exo I/pGO25000-assisted FAM–ssDNA probes in the presence or absence of Hg2+ (600 nM) solution mixed with other metal ions (K<sup>+</sup> , Fe2+, Sn2+, Al3+, Ni2+, Mn2+, Mg2+ , Cu2+, and Co2+) at a concentration of 6 mM, as shown in Figure 6. Hg2+ produced a remarkable decrease in fluorescence intensity, indicating that only Hg2+ could bind to two thymine bases, and T–Hg2+–T mismatched base pairs formed, resulting in the formation of a hairpin structure; thus, degradation by Exo I was hindered. As a result, the FAM–

**Figure 6.** Selectivity of Hg2+ detection by the FAM–ssDNA probe. The fluorescence intensities of the Exo I (2 U)/pGO25000 (0.1 µg/mL)-assisted FAM–ssDNA probes were recorded without or with Hg2+ (600 nM) and mixed with other possible interfering metal ions (K<sup>+</sup> , Fe2+, Sn2+, Al3+, Ni2+, Mn2+ , Mg2+, Cu2+, and Co2+, 6 mM). *E*x = 495 nm, *E*m = 520 nm. Each sample was repeated three times.

hairpin DNA remained for fluorescence quenching by pGO25000. The obtained results also revealed that the proposed method could still detect Hg2+, which mixed with other

#### *2.7. Application in Real Samples*

μ To assess the approach feasibility, the Exo I/pGO25000-assisted FAM–ssDNA biosensor strategy was used for the Hg2+ analysis of pond water samples collected from a pond at Zhejiang Sci-Tech University. The impurities of pond water were removed by filtration using NY 0.22 µm, and three concentrations of Hg2+ (25 nM, 50 nM, 200 nM) were diluted by the pond water, which were individually detected by the sensor system. The recoveries of Hg2+ in pond water were in the range of 99.6%–101.1%, and all relative standard deviations (RSDs) were as low as 5% (*n* = 3) (Table 2). It is revealed that the quantitative detection of Hg2+ can be performed by this sensitive biosensor method. Meanwhile, it illustrated that the proposed sensor has good feasibility and accuracy to measure the Hg2+ concentration in real samples.

**Table 2.** Average recoveries of Hg2+ in the pond water samples (*n* = 3).


#### **3. Experimental**

#### *3.1. Chemicals and Materials*

The oligonucleotide (FAM–ssDNA probe: 5′ -FAM-TATCGTGCTCCCCTGCTCGTTA) was purified by HPLC and purchased from Sangon Biotech Co., Ltd. (Shanghai, China). The oligonucleotide stock solution (50 µM) was prepared with double distilled water (ddH2O). Exonuclease I (Exo I, 20 U/µL) was provided by Bio Basic Inc. (Canada). A 10 mM Tris–HNO<sup>3</sup> buffer including 40 mM NaNO<sup>3</sup> (pH 8.5) was used in the experiments. Hg2+ solutions with different concentrations were prepared from a standard mercury ion solution (5.0 mM), which was purchased from Aoke Biology Research Co., Ltd. (Beijing, China). All metal salts and tris(hydroxymethyl)aminomethane (Tris) were obtained from Macklin (Shanghai, China) and Aladdin (Beijing, China). Graphene oxide (GO) was supplied by XFNano Material Tech Co., Ltd. (Nanjing, China). PEI (average MW = 25,000) was purchased from Sigma-Aldrich. All other chemicals and reagents used in this work were of analytical grade and used without further purification. The pond water was prefiltered with an NY 0.22 µm filter.

#### *3.2. Apparatus*

Fluorescence spectra were measured by an F-4600 fluorescence spectrometer (Hitachi, Japan) at RT (room temperature, approximately 27 ◦C). The emission spectra data were acquired from 511 nm to 620 nm after illuminating at the maximum excitation wavelength of 495 nm, and the fluorescence intensity was measured at 520 nm (Emmax). Both the excitation and emission slit widths were set at 5 nm and 10 nm, respectively. High-resolution X-ray photoelectron spectroscopy (XPS) was performed using a Thermo Scientific K-Alpha XPS spectrometer (Kratos Analytical, Manchester, UK). The charge polarity and density of GO/pGO25000 colloids were obtained by zeta potential measurement (Malvern Zetasizer Nano ZS90, Great Malvern, UK). The Raman spectrum was recorded by an Optosky ATP3007 (Xiamen, China) Raman spectrometer with a 785 nm excitation lase. A Nicolet iS50 spectrometer (Thermo Fisher Scientific, Waltham, USA) using KBr pellets was used for Fourier-transform infrared (FT-IR) characterization.

#### *3.3. Preparation of pGO25000*

Positively charged GO (pGO25000) was synthesized by grafting PEI (MW = 25,000) onto GO nanosheets via a condensation reaction between amino groups and carboxyl groups. Briefly, 120 mg of GO, 1.0 g of PEI, and 300 mg of EDC were sequentially dissolved in 40 mL of ddH2O. Then, the pH value of this mixture was adjusted to pH 7.0 by adding a certain amount of diluted HCl (1.0 M) and stored at 4 ◦C for 24 h. Then, the mixture was dialyzed against ddH2O for two weeks to remove foreign ions. After diluting this mixture with ddH2O and ultrasonication, the pGO solution (1 mg/mL) was successfully prepared.

#### *3.4. Fluorescence Quenching Assay*

A 50 nM FAM–ssDNA probe was incubated in 200 µL of Tris–HNO<sup>3</sup> buffer (10 mM, 40 mM NaNO3, pH 8.5) with different concentrations of pGO25000/GO (0, 0.0001, 0.001, 0.005, 0.01, 0.02, 0.05, 0.1, 1, 3, 6, 15, and 30 µL/mL). The mixtures were immediately measured by a fluorescence spectrometer, and the fluorescence emission was monitored in the wavelength range from 511 to 620 nm. Each sample solution was repeated and measured at least three times.

#### *3.5. Study of Exo I Activity in Hg2+ Detection*

A 0.01 nmole FAM–ssDNA probe (50 µM, 0.2 µL) and 1 µL of Hg2+ solution (600 nM) were added to a 0.5 mL centrifuge tube, followed by dilution to 10 µL with Tris–HNO<sup>3</sup> buffer (10 mM, 40 mM NaNO3, pH 8.5). The FAM–ssDNA probe/Hg2+ solution was added to 2 U of Exo I and incubated for 30 min at 27 ◦C (RT) or 5 min at 40 ◦C in an oven. Then, pGO25000 (0.1 µL/mL) was mixed with the FAM–ssDNA probe/Hg2+/Exo I mixture and diluted with Tris–HNO<sup>3</sup> buffer (10 mM, 40 mM NaNO3, pH 8.5) to a final volume of 200 µL.

Exo I nuclease was denatured by heating at 80 ◦C for 15 min before the fluorescence test. To evaluate the sensitivity of Hg2+ detection, the final concentrations of Hg2+ were 0, 2, 5, 10, 30, 50, 120, 150, 200, 250, 300, 400, 600, and 800 nM. To evaluate the selectivity of Hg2+ detection, 6 µM K<sup>+</sup> , Na<sup>+</sup> , Cu2+, Mn2+, Ni2+, Pb2+, and Fe3+ and 600 nM Hg2+ were used. The fluorescence emission was analyzed from 511 to 620 nm for all samples with an excitation wavelength of 495 nm, and the maximum fluorescence intensity was measured at 520 nm. To assess the application of the DNA sensor in the real samples, a 50 nM FAM–ssDNA probe was incubated in 10 µL of Tris–HNO<sup>3</sup> buffer (10 mM, 40 mM NaNO3, pH 8.5) with three different concentrations of Hg2+ solution (25 nM, 50 nM, 200 nM), which were prepared from the pond water. Each sample solution was analyzed three times.

#### **4. Conclusions**

In summary, positively charged GO (pGO25000) was synthesized by modification of the GO surface with PEI and used in fluorescence quenching of a DNA probe. Compared to GO, the fluorescence quenching efficiency of pGO25000 was dramatically improved due to the additional electrostatic interaction induced by PEI. Electrostatic attraction plays a vital role in the interaction between pGO25000 and DNA strands, which increases the affinity of pGO25000 to the DNA strands. As a result, when pGO25000 is at a very low concentration (0.1 µg/mL), it possesses a higher and more efficient fluorescence quenching ability compared to GO. In view of the perfect fluorescence quenching efficiency of pGO25000, the pGO25000 and FAM–ssDNA probes were designed for Hg2+ detection with the assistance of Exo I, and fluorescence was specifically prevented by adding Hg2+. The FAM–ssDNA probe was formed in a hairpin structure in the presence of Hg2+ due to the formation of T–Hg2+–T complexes, which imposed restrictions on the degradation by Exo I. However, the fluorescence was recovered without Hg2+ because of the hydrolysis of FAM–ssDNA caused by Exo I. Therefore, the limit of detection of 3.93 nM was obtained. Compared to other metal ions, the Exo I/pGO25000-assisted FAM–ssDNA sensor has a great selectivity for Hg2+. In addition, the designed strategy is applicable for Hg2+ detection in real samples with satisfactory recoveries. In consideration of the efficient fluorescence quenching property of pGO25000, as well as the absolute quantification ability by the Exo I-assisted FAM–ssDNA sensor, the biosensor mechanism can be applied in more toxic substances and in gene mutation analysis (e.g., SNP).

**Supplementary Materials:** The following supporting information can be downloaded at: https:// www.mdpi.com/article/10.3390/ijms23116326/s1.

**Author Contributions:** Conceptualization, J.Q.; methodology, X.J. and Z.W.; software, X.L.; validation, T.S. and X.L.; formal analysis, X.J.; investigation, Z.W.; resources, X.L.; data curation, Z.W.; writing—original draft preparation, T.S.; writing—review and editing, J.Q. and X.C.; visualization, J.Q.; supervision, J.Q.; project administration, J.Q.; funding acquisition, J.Q. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by Zhejiang Sci-Tech University grant number 16042017-Y.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Not applicable.

**Acknowledgments:** We thank all members of the Xiachao Chen Laboratory for many productive discussions.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**

