Planar Boronic Graphene and Nitrogenized Graphene Heterostructure for Protein Stretch and Confinement

Single-molecule techniques such as electron tunneling and atomic force microscopy have attracted growing interests in protein sequencing. For these methods, it is critical to refine and stabilize the protein sample to a “suitable mode” before applying a high-fidelity measurement. Here, we show that a planar heterostructure comprising boronic graphene (BC3) and nitrogenized graphene (C3N) sandwiched stripe (BC3/C3N/BC3) is capable of the effective stretching and confinement of three types of intrinsically disordered proteins (IDPs), including amyloid-β (1–42), polyglutamine (Q42), and α-Synuclein (61–95). Our molecular dynamics simulations demonstrate that the protein molecules interact more strongly with the C3N stripe than the BC3 one, which leads to their capture, elongation, and confinement along the center C3N stripe of the heterostructure. The conformational fluctuations of IDPs are substantially reduced after being stretched. This design may serve as a platform for single-molecule protein analysis with reduced thermal noise.


Introduction
Protein sequencing at the single-molecule level is crucial for personalized medicines and the detection of post-translational modifications in proteins [1][2][3][4]. Recently, several single-molecule techniques such as atomic force microscopy (AFM) [5,6], quantum tunneling [7][8][9], and nanopore [3,10,11] have been proposed for protein sequencing, which allow the direct read-out of structural differences of individual amino acids. Although promised to be with high accuracy and low cost, a large gap still resides between these proof-of-principle methods and the ultimate sensitivity for the discrimination of 20 different amino acids. One major challenge is the noisy signals caused by the thermal fluctuations of amino acids [12,13]. Moreover, proteins usually possess coiled or folded conformations in a solution, which imposes difficulties in the analysis of the atomic structure of protein [13,14]. Therefore, to put the single-molecule protein sequencers into potential commercial use, the controllable manipulation (such as elongation) and confinement of the protein conformation are prerequisites.
Current nanochannel and nanopore sequencing techniques naturally provide steric confinement for analytes, and the single-molecular sensitivity can be realized with the crosssection of confinement in the same order of magnitude as the size of the amino acids [15,16]. However, the narrow cross-section inevitably causes a large entropy barrier that hampers protein capture into the nanoscale channel [15,16]. The non-specific interaction between the protein and the nanostructure can also affect the precision of measurement and induce clogging [4,12]. To address these issues, a planar two-dimensional (2D) heterostructure has recently been proposed for biomolecular capture, stretching, and confinement [17][18][19]. The planar 2D heterostructure can be fabricated by the seamless stitching of two 2D materials (for example, graphene and hexagonal boron nitride) with a similar lattice constant [20][21][22][23]. As the key mechanism for this heterostructure to manipulate protein conformation is the adsorption energy contrast for a protein molecule on different 2D materials [18,24], the performance should depend on the type of 2D material selected.
Boronic graphene (BC 3 ) and nitrogenized graphene (C 3 N) are two new types of graphene derivatives that have been successfully synthesized [25,26]. Both BC 3 and C 3 N exhibit excellent structural stability and share very similar honeycomb lattices [24,25], making them suitable to form planar heterojunctions. On the other hand, with differently doped heteroatoms (boron and nitrogen), BC 3 and C 3 N have demonstrated a distinct contrast of binding affinities for biomolecules [27], which can be harnessed for biomolecular manipulation. Owing to these features, we are highly motivated to design a BC 3 /C 3 N/BC 3 in-plane heterostructure ( Figure 1) for protein stretching and confinement. To study the interaction mechanism between the heterostructure and protein, three representative intrinsically disordered proteins (IDPs), including amyloid-β (Aβ 1-42 ), polyglutamine (polyQ 42 ), and α-synuclein (α-Syn 61-95 ) are taken as examples. Utilizing all-atom molecular dynamics (MD) simulations, we show that the disordered conformations of IDPs can be stretched into a linear manner along the C 3 N stripe sandwiched between two BC 3 domains. This highly regular and confined conformation might be suitable for analysis by single-molecule methods such as AFM [5,6] and quantum tunneling [7][8][9]. Moreover, the conformational fluctuations of proteins can be significantly reduced after being stretched and energetically confined on the C 3 N stripe. The periodic atomic charge distributions on BC 3 also induce the formation of high-density water clusters on the BC 3 surfaces, which may further provide steric hindrances to restrict the conformational fluctuation of IDPs. The insights from our study might benefit the improvement of the signal-to-noise ratio for single-molecule protein analysis. and induce clogging [4,12]. To address these issues, a planar two-dimensional (2D) heterostructure has recently been proposed for biomolecular capture, stretching, and confinement [17][18][19]. The planar 2D heterostructure can be fabricated by the seamless stitching of two 2D materials (for example, graphene and hexagonal boron nitride) with a similar lattice constant [20][21][22][23]. As the key mechanism for this heterostructure to manipulate protein conformation is the adsorption energy contrast for a protein molecule on different 2D materials [18,24], the performance should depend on the type of 2D material selected. Boronic graphene (BC3) and nitrogenized graphene (C3N) are two new types of graphene derivatives that have been successfully synthesized [25,26]. Both BC3 and C3N exhibit excellent structural stability and share very similar honeycomb lattices [24,25], making them suitable to form planar heterojunctions. On the other hand, with differently doped heteroatoms (boron and nitrogen), BC3 and C3N have demonstrated a distinct contrast of binding affinities for biomolecules [27], which can be harnessed for biomolecular manipulation. Owing to these features, we are highly motivated to design a BC3/C3N/BC3 in-plane heterostructure ( Figure 1) for protein stretching and confinement. To study the interaction mechanism between the heterostructure and protein, three representative intrinsically disordered proteins (IDPs), including amyloid-β (Aβ1-42), polyglutamine (polyQ42), and α-synuclein (α-Syn61-95) are taken as examples. Utilizing all-atom molecular dynamics (MD) simulations, we show that the disordered conformations of IDPs can be stretched into a linear manner along the C3N stripe sandwiched between two BC3 domains. This highly regular and confined conformation might be suitable for analysis by single-molecule methods such as AFM [5,6] and quantum tunneling [7][8][9]. Moreover, the conformational fluctuations of proteins can be significantly reduced after being stretched and energetically confined on the C3N stripe. The periodic atomic charge distributions on BC3 also induce the formation of high-density water clusters on the BC3 surfaces, which may further provide steric hindrances to restrict the conformational fluctuation of IDPs. The insights from our study might benefit the improvement of the signal-to-noise ratio for single-molecule protein analysis.

Method
We used molecular dynamics (MD) simulations to simulate a two-dimensional (2D) sandwich BC 3 /C 3 N/BC 3 planar heterostructure with a total size of 16.2 × 14.1 nm 2 [2] ( Figure 1). Among them, the width of the C 3 N stripe seamlessly spliced between the two BC 3 sheets is 1.2 nm. The force fields of BC 3 and C 3 N can be obtained by referring to the previous research [28], in which the lattice constants of both are 2.5 Å, and the boron and nitrogen atoms have partial charges of 0.378 e and −0.168 e, respectively. To maintain the overall and local charge neutrality of the planar heterostructure, the carbon atoms in BC 3 and C 3 N carry −0.126 e and 0.056 e, respectively [29].
Following the similar approach in our previous studies [17,24,[30][31][32], we performed a pre-equilibrium simulation of the conformation of each IDP fragment from solution to adsorption on a BC 3 nanosheet, and then extended the BC 3 nanosheet (along with the peptide) to construct the BC 3 /C 3 N/BC 3 planar heterostructure. After that, the entire system is placed in a box with a size of 16.2 × 14.1 × 5.0 nm 3 [3] and solvated with 100 mM KCl electrolyte, which contains approximately 66,000 atoms. In addition, to explore the difference in hydrophilic or hydrophobic properties between BC 3 and C 3 N, we constructed two systems of BC 3 or C 3 N in a water box with a size of 4.9 × 4.2 × 5.0 nm 3 [3]. Later, to characterize the interface behavior of water molecules on the planar heterostructure, we additionally constructed a heterostructure-water system with a box size of 16.2 × 14.1 × 5.0 nm 3 [3].
The Gromacs software package [33] (version 5.1.4) was used for our MD simulations, and VMD [34] was used for trajectory visualization. The TIP3P model [35] was used for water molecules, the CHARMM36 force field [36] for proteins/peptides, and standard force fields for ions. Following the scheme used in many previous, similar researches [30,[37][38][39][40][41], we used the LINCS algorithm to constrain the covalent bonds with hydrogen atoms, with a time step of 2 fs. The particle mesh Ewald (PME) method [42] with a grid size of about 1 Å was used to calculate the long-range electrostatic interactions, while the smooth cut-off method was used for the van der Waals (vdW) interactions, with a cut-off distance of 1.2 nm. Periodic boundary conditions were used in all three-dimensional directions. The Parrinello-Rahman algorithm [43] was applied in the z-direction with a semi-isotropic pressure coupling of 1 bar, and the V-rescale thermostat [44] was used to control the simulation temperature at 300 K. Then, under the NPT ensemble, several independent 400 ns trajectories were generated for each system for data collection. In all simulations, except that the atoms in the two-dimensional planar heterostructure are frozen, all other atoms can move freely.

Result
Some representative snapshots ( Figure 2) were shown for the trajectories of Aβ, polyQ, and α-Syn on the BC 3 /C 3 N/BC 3 planar heterojunction, respectively. Taking Aβ (Figure 2A) as an example, starting from the curled and folded structure adsorbed on the BC 3 domain, Aβ first diffused to the C 3 N region quickly, arriving at t =~22 ns, which is consistent with our previous findings on the similar free and rapid diffusions of adsorbents on other two-dimensional material plane [24]. Finally, at t =~250 ns, Aβ was stretched on the C 3 N stripe and maintains a straightened conformation. Following the same procedure, we also checked the trajectories of polyQ ( Figure 2B) and α-Syn ( Figure 2C) and found that although they had slightly different dynamic behaviors in the initial diffusion and subsequent stretching phases, they all displayed the same straightened conformation as Aβ on the C 3 N stripe at the end. In addition, we analyzed the average end-to-end distance normalized by the residue number (L/N) for each protein ( Figure 2D). During the simulation, the L/N of Aβ, polyQ, and α-Syn increased from~1.0 Å to 2.2 Å, 2.2 Å, and 2.3 Å, respectively. It is worth noting that although the L/N values of these IDPs fluctuate sharply due to their high flexibility before being fully stretched, they maintain their respective highest values during the last 100 ns simulations, indicating that IDPs can be spontaneously stretched on the C 3 N stripe. It is also worth noting that the fluctuations of L/N are reduced after the peptides are stretched ( Figure 2D). To characterize the change of conformational fluctuations of peptides in the stretching process, we further conducted MD simulations of peptides in free solution, on the BC 3 surface, and the BC 3 /C 3 N/BC 3 surface, respectively. Figure 3 illustrates the probability distributions of L/N of Aβ peptides in different environments. The distribution of L/N of Aβ confined in the heterostructure has a much narrower width than those in other environments. Moreover, distributions of L/N of polyQ ( Figure S1) and a-Syn ( Figure S2) also demonstrate similar results. The above analyses indicate that the conformational fluctuations of proteins can be significantly reduced when they are confined on the C 3 N stripe BC 3 /C 3 N/BC 3 .
Biomolecules 2021, 11, x FOR PEER REVIEW 4 of 11 be spontaneously stretched on the C3N stripe. It is also worth noting that the fluctuations of L/N are reduced after the peptides are stretched ( Figure 2D). To characterize the change of conformational fluctuations of peptides in the stretching process, we further conducted MD simulations of peptides in free solution, on the BC3 surface, and the BC3/C3N/BC3 surface, respectively. Figure 3 illustrates the probability distributions of L/N of Aβ peptides in different environments. The distribution of L/N of Aβ confined in the heterostructure has a much narrower width than those in other environments. Moreover, distributions of L/N of polyQ ( Figure S1) and a-Syn ( Figure S2) also demonstrate similar results. The above analyses indicate that the conformational fluctuations of proteins can be significantly reduced when they are confined on the C3N stripe BC3/C3N/BC3.   To reveal the stretching mechanism of IDPs on the BC3/C3N/BC3 planar heterostructure, we adopted Aβ as an example to examine the details of the interaction between peptides and BC3/C3N/BC3 (Figure 4). During the simulation, the average number of contact atoms per Aβ residue with BC3 decreased from ~9 to ~4, while the average number of contact atoms per Aβ residue with C3N increased from 0 to ~8 ( Figure 4A). Here, we defined a contact when any atom in BC3/C3N/BC3 was within 4.0 Å of any heavy atom of the protein. As shown in the scatter plot of Figure 4B, in the stretching process, the peptide contacted more with C3N while less with BC3, accompanied by a gradual increase in the magnitude of interaction energy. According to the above results, the peptide interacted more strongly with C3N than BC3, which drove its stretching on BC3/C3N/BC3. To further understand the physical mechanism of the driven process, we scanned the interaction energy ΔE between the peptide and BC3/C3N/BC3 by moving the stretched conformation of Aβ horizontally and rigidly from the C3N stripe to the BC3 domain. As shown in Figure  4C, when the peptide was on the C3N stripe, the van der Waals interaction (vdW) energy presented a narrow energy well with a depth of −1.6 kcal per mol per residue (black dotted line). It is also worth noting that the average value of Coulomb interaction energy between Aβ and the BC3/C3N/BC3 was almost zero (red line) due to the local charge neutrality of the planar heterostructure. Moreover, we repeated all of the above analyses for the polyQ ( Figure S3) and α-Syn ( Figure S4) on BC3/C3N/BC3. Similarly, the straightening and restriction of polyQ and α-Syn on the narrow C3N stripe were also mainly due to the differences in the vdW interactions between the polypeptides and C3N and BC3. To reveal the stretching mechanism of IDPs on the BC 3 /C 3 N/BC 3 planar heterostructure, we adopted Aβ as an example to examine the details of the interaction between peptides and BC 3 /C 3 N/BC 3 (Figure 4). During the simulation, the average number of contact atoms per Aβ residue with BC 3 decreased from~9 to~4, while the average number of contact atoms per Aβ residue with C 3 N increased from 0 to~8 ( Figure 4A). Here, we defined a contact when any atom in BC 3 /C 3 N/BC 3 was within 4.0 Å of any heavy atom of the protein. As shown in the scatter plot of Figure 4B, in the stretching process, the peptide contacted more with C 3 N while less with BC 3 , accompanied by a gradual increase in the magnitude of interaction energy. According to the above results, the peptide interacted more strongly with C 3 N than BC 3 , which drove its stretching on BC 3 /C 3 N/BC 3 . To further understand the physical mechanism of the driven process, we scanned the interaction energy ∆E between the peptide and BC 3 /C 3 N/BC 3 by moving the stretched conformation of Aβ horizontally and rigidly from the C 3 N stripe to the BC 3 domain. As shown in Figure 4C, when the peptide was on the C 3 N stripe, the van der Waals interaction (vdW) energy presented a narrow energy well with a depth of −1.6 kcal per mol per residue (black dotted line). It is also worth noting that the average value of Coulomb interaction energy between Aβ and the BC 3 /C 3 N/BC 3 was almost zero (red line) due to the local charge neutrality of the planar heterostructure. Moreover, we repeated all of the above analyses for the polyQ ( Figure S3) and α-Syn ( Figure S4) on BC 3 /C 3 N/BC 3 . Similarly, the straightening and restriction of polyQ and α-Syn on the narrow C 3 N stripe were also mainly due to the differences in the vdW interactions between the polypeptides and C 3 N and BC 3 .
To further investigate the influence of the C 3 N stripe width on the efficiency of protein stretching, we also constructed BC 3 /C 3 N/BC 3 heterostructures with the stripe widths ranging from 0.6 nm to 1.8 nm and scanned the vdW interaction energy between the elongated Aβ peptide and each heterostructure, respectively ( Figure S5). Figure S5B shows the potential wells of C 3 N stripes with different widths, and Figure S5C shows the "potential-depth" and the "potential-width" at a half-minimum of each well as a function of the C 3 N stripe width. For both curves of the depths and widths of potential wells, the inflection points occur when the C 3 N layer width is equal to 1.2 nm. After this deflection point, the increasing stripe width would lead to a slower decrease of the depth of the potential well, and a faster increase of the width of the potential well. It is noteworthy that a potential well with a deeper depth and narrower width can lead to a better performance of the protein stretching. Considering the optimization of both the depth and the width of the potential well, the C 3 N layer width of 1.2 nm is recommended for efficient protein stretching. To further investigate the influence of the C3N stripe width on the efficiency of protein stretching, we also constructed BC3/C3N/BC3 heterostructures with the stripe widths ranging from 0.6 nm to 1.8 nm and scanned the vdW interaction energy between the elongated Aβ peptide and each heterostructure, respectively ( Figure S5). Figure S5B shows the potential wells of C3N stripes with different widths, and Figure S5C shows the "potentialdepth" and the "potential-width" at a half-minimum of each well as a function of the C3N stripe width. For both curves of the depths and widths of potential wells, the inflection points occur when the C3N layer width is equal to 1.2 nm. After this deflection point, the increasing stripe width would lead to a slower decrease of the depth of the potential well, and a faster increase of the width of the potential well. It is noteworthy that a potential well with a deeper depth and narrower width can lead to a better performance of the protein stretching. Considering the optimization of both the depth and the width of the potential well, the C3N layer width of 1.2 nm is recommended for efficient protein stretching.
Moreover, in practice, the interface between BC3 and C3N may not be perfect, as shown in Figure 1, and rough domains with mixed units of BC3 and C3N might exist at the interface. One simple way to illustrate this effect is to use the 1.2-nm-width C3N stripe Moreover, in practice, the interface between BC 3 and C 3 N may not be perfect, as shown in Figure 1, and rough domains with mixed units of BC 3 and C 3 N might exist at the interface. One simple way to illustrate this effect is to use the 1.2-nm-width C 3 N stripe in our current simulation as the base and use the width of the 1.0-nm-to 1.4-nm-C 3 N stripes as the upper and lower bounds, as the adsorption potential well of the stripe with rough interfaces should be in-between the wells of these two widths. As shown above in Figure  S5B, the potential wells of the 1.0-nm-width, 1.2-nm-width, and 1.4-nm-width C 3 N stripes share similar shapes, with potential-well depths of −1.4 kcal/mol, −1.6 kcal/mol, and −1.7 kcal/mol respectively. Thus, the roughness of the interface is not likely to affect much on the potential well for protein stretching and confinement.
In addition, it is known that the behavior of water on the surface of 2D materials also has an important influence on the interaction between protein and 2D materials in the aqueous environment. Therefore, we further analyzed the behaviors of interfacial water on BC 3 and C 3 N respectively, and the influence of these interfacial waters on the confinement of IDPs on the planar heterostructure ( Figure 5). From the water density maps ( Figure 5A,B) along the normal (Z) directions of 2D materials, the first water solvation shells of BC 3 and C 3 N were both located on Z = ±0.35 nm. More interestingly, we found a periodic enhancement of water density clusters between the BC 3 layer and its first solvation shell, which were not observed clearly on the C 3 N layer. The presence of these water clusters on BC 3 might be attributed to the strong but nonuniform partial charge distributions (+0.378 e for a boron atom and −0.126 e for a carbon atom). While on C 3 N, the partial charges (−0.168 e for a carbon atom and +0.056 e for a nitrogen atom) were much smaller, and not strong enough to induce noticeable water clusters between the 2D layer and the first solvation shell. Meanwhile, we further calculated the two-dimensional water density map along the surface (XY direction) of BC 3 /C 3 N/BC 3 (only water molecules within 0.25 nm in the Z direction of the 2D plane were counted). As shown in Figure 5C,D, these water clusters were periodically distributed on the BC 3 domain, which could form a steric hindrance ( Figure 5E) that hindered the diffusion of IDPs' residues to the BC 3 domain. These interfacial water molecules might further strengthen the restriction on the linear conformation of IDPs.

Conclusions
In this work, we applied molecular dynamics (MD) simulation to study the stretching process of several representative IDPs on the 2D sandwiched BC3/C3N/BC3 planar heterostructure. The IDPs could be spontaneously straightened and then restricted along the C3N stripes. Moreover, we have shown that the conformational fluctuations of IDPs were significantly reduced when IDPs were confined on the C3N stripe. The protein stretching and confinement were mainly driven by the stronger adsorption potential of C3N than that of BC3. Additionally, we found that the interfacial water molecules on the BC3 surface where the 2D materials are located at Z = 0. As shown in the structure below, the calculation was performed along the zigzag direction, while only water molecules inside the red square were calculated. (C) Two-dimensional water density map on the BC 3 /C 3 N/BC 3 surface; water molecules within ±0.25 nm in the Z direction were counted. (D) Magnified map of water density at the boundary of BC 3 and C 3 N. (E) Straightened Aβ peptide on BC 3 /C 3 N/BC 3 with interfacial water surrounded; only water molecules within 0.5 nm of both the heterostructure and protein are shown.

Conclusions
In this work, we applied molecular dynamics (MD) simulation to study the stretching process of several representative IDPs on the 2D sandwiched BC 3 /C 3 N/BC 3 planar heterostructure. The IDPs could be spontaneously straightened and then restricted along the C 3 N stripes. Moreover, we have shown that the conformational fluctuations of IDPs were significantly reduced when IDPs were confined on the C 3 N stripe. The protein stretching and confinement were mainly driven by the stronger adsorption potential of C 3 N than that of BC 3 . Additionally, we found that the interfacial water molecules on the BC 3 surface might further act as steric hindrances to enhance the restriction of protein. This linearly confined structure on the 2D surface may be feasible for scanning tunneling microscope (STM) [45,46] and atomic force microscope (AFM) [5,6] to identify the amino acids in the protein. Furthermore, compared with other "hard" nano-confinements such as nanochannels and nanopores, this heterostructure provides "soft" confinement that is based on the adsorption potential difference between the two 2D nanomaterials. Considering the large entropy barrier for stretching a coiled or folded protein to a linear conformation, this soft, energetic confinement allows some protein residues to temporarily move outside the center C 3 N stripe while "regulating" the protein conformation, thus smoothly overcoming the entropy barrier in the stretching process. Therefore, this heterostructure holds the potential to be coupled with nanopore-and nanochannel-sensing methods to ease the clogging problem. The delivery of the stretched protein samples from this heterostructure to a nanochannel can be driven by a pressure-driven flow, for example. Last but not least, C 3 N and BC 3 have been reported to possess higher biocompatibility than graphene [28]. Our work may also offer insight into the design of biocompatible nanodevices.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/10 .3390/biom11121756/s1, Figure S1: Analysis of end-to-end distances for polyglutamine (polyQ 42 ) in solution, on boronic graphene (BC 3 ), and the nitrogenized graphene (C 3 N) stripe. Figure S2: Analysis of end-to-end distances for α-Synuclein (α-Syn 61-95 ) in solution, on BC 3 , and the C 3 N stripe, Figure S3: Detailed analysis of the polyQ 42 stretching process on the BC 3 /C 3 N/BC 3 heterostructure, Figure S4: Detailed analysis of the α-Syn 61-95 stretching process on the BC 3 /C 3 N/BC 3 heterostructure, Figure S5: Influence of the C 3 N stripe width on the potential well for the straightened Aβ peptide.