BaP agglomeration and adsorption on graphene
Models of component molecules used in MD simulations are shown in Additional file 1: Fig. S1. Before evaluating the joint effect of PS interactions with graphene and adsorbed BaPs, we firstly examined the molecular agglomeration of BaPs and their gaseous adsorption on graphene. For BaPs of relatively high concentrations initially dispersed in a gaseous simulation box, they were found to form agglomerates of irregular shapes shortly (Fig. 1). Continuous aggregation and disaggregation being accompanied with inner structural rearrangement occurred as reflected in fluctuations of the BaP-BaP interaction energy before reaching equilibrium (Additional file 1: Fig. S2). Relative humidity was considered by adding a number of water molecules into the simulation box, and similar nanostructures of BaP agglomerates were obtained (Additional file 1: Fig. S3). By comparing time sequences of typical snapshots and evolutions of the BaP-BaP interaction energy, humidity was found to promote and stabilize BaP agglomeration (Additional file 1: Fig. S3), possibly due to liquid bridging forces which enhanced the agglomeration velocity as previously reported [22].
Next, dispersed BaPs were mixed with graphene, and all BaPs adsorbed on graphene (Fig. 1). Depending on the BaP concentration, single- and multi-layered adsorption structures were formed. The whole process can be described as two stages, as revealed in time evolutions of the graphene-BaP interaction energy (Additional file 1: Fig. S4). The first stage featured with a sudden decrease of the graphene-BaP interaction energy was that dispersed BaPs rapidly diffused and got contact with graphene. Subsequently, the adsorbed BaPs rearranged their positions and orientations to acquire more favorable contacts with both the graphene and neighboring BaPs, as reflected in slower decreases of the interaction energy (Additional file 1: Fig. S4).
For BaPs in the particulate phase encountering graphene, interestingly, they were found to stimulate graphene scrolling into tubular nanostructures, inside which BaPs were encapsulated (Fig. 1). That was similar to previous findings of the graphene scrolling guided by carbon nanotubes [23], iron nanowires [24], fullerene clusters [25], even nanodroplets of ionic liquid [26]. The graphene scrolling on BaPs was driven by their van der Waals interactions, but cost a finite graphene bending energy. Thus, larger BaP agglomerates were easier to scroll graphene by acquiring more favorable contacts and costing lower energy of graphene deformation (Additional file 1: Fig. S5). Similar nanostructures were formed for BaP adsorption on GO (Fig. 1), and humidity barely affected the equilibrium structures of BaP adsorption on graphene (Additional file 1: Fig. S6).
Solubilization of BaPs by PS
Given high capacity of graphene adsorbing BaPs in atmosphere, we speculated that more BaPs are carried by inhaled graphene into alveoli, where the BaP’s bioavailability can be increased through interactions with PS [17]. Shown in Fig. 2a are typical snapshots of interactions between PS and graphene with adsorbed BaPs. For a bare nanosheet of graphene deposited at the PS layer, a number of PS molecules were extracted from the layer and formed inverse micelles on graphene. Obvious PS adsorption on graphene was experimentally confirmed from confocal imaging (Fig. 2b, details are found in the Methods). MD simulations described the process as the graphene adhesion stage and PS extraction stage (Additional file 1: Fig. S7a), each generating 11.7 kJ/mol and 7.9 kJ/mol decreases of the graphene-PS interaction energy (Additional file 1: Fig. S7b). When more BaPs were adsorbed by graphene, there was an increasing number of BaPs released from graphene and solubilized by PS (Fig. 2c, detailed processes are found in Additional files 2 and 3: Movies S1 and S2), generating more increase in the BaP-BaP interaction energy and decrease in the BaP-PS interaction energy (Fig. 2d and Additional file 1: Fig. S8). We also calculated changes in the graphene-BaP and graphene-PS interaction energies (Fig. 2e), and found that change in the graphene-BaP interaction energy increased only when the number of adsorbed BaPs was less than 240, which contrarily decreased as more BaPs were adsorbed to reach a multi-layered adsorption state. Less PS molecules were extracted by graphene-BaP complexes than bare graphene, as reflected in the lower decrease of the graphene-PS interaction energy (Fig. 2e).
For single-component BaP agglomerates deposited at the PS layer, they were completely solubilized by PS (Additional file 1: Fig. S9a and Additional file 4: Movie S3), as evidencedby a rapid increase in the BaP-BaP interaction energy and a decrease in the BaP-PS interaction energy (Additional file 1: Fig. S9b). The BaP-PS interaction was more energetically favorable than the BaP-BaP interaction, and should be the driving force for BaP solubilization. That was similar to BaPs at the outer layer of adsorption on graphene, but different from those at the first adsorption layer, where the solubilization of BaPs was at the cost of losing contacts with graphene. Compared with pristine graphene, more BaPs adsorbed on GO were released and dissolved by PS (Fig. 2f). Upon release of BaPs, PS molecules were extracted and expelled more BaPs into the PS layer. The strong adsorption of PS molecules on GO was also experimentally observed through confocal imaging (Fig. 2g). Solubilization experiments were conducted to further investigate solubilization of BaPs by PS (Fig. 2h). Consistent with the simulation results, BaPs were highly solubilized by 51.2% in the Curosurf solution (details are found in the Methods). By contrast, the solubilization was higher for BaPs adsorbed on GO, due to the lower strength of BaP-GO interactions as revealed by MD simulations.
PS perturbation by graphene-BaP complexes
Early deposition of BaP agglomerates induced slight PS perturbation, which disappeared as dispersion of BaPs inside the layer (Additional file 1: Fig. S9a). Solubilized BaP molecules aligned vertically in the PS layer to acquire more favorable contacts with PS (Fig. 3a). We calculated the mean square displacement (Additional file 1: Fig. S10), and the diffusion coefficients of PS components and BaPs were thus obtained. Compared to the intact PS layer, deposition of BaPs, which diffused at a higher rate, barely affected the PS liquidity (Fig. 3b). We calculated surface tension of the PS layer (γ) as a function of the PS area (A), from which the PS layer compressibility, defined as , was estimated < 0.01 m/mN, similar to that previously measured by experiments [27]. However, deposition of 160 BaPs apparently increased the PS layer compressibility (Fig. 3c).
For a bare graphene nanosheet deposited at the PS layer, the PS perturbation was mainly reflected as extraction of PS molecules forming inverse micelles on graphene (Additional file 1: Fig. S7), but a distinct interface was formed between graphene-BaP mixtures and PS. As shown in the order parameter diagram for lipids surrounding the graphene-BaP complex (Fig. 3d), PS molecules around the edge of graphene climbed onto graphene, while those beneath the mixture were aligned in higher orders. We calculated the angles between surfaces of the PS layer and BaPs respectively above and below graphene (Fig. 3e and Additional file 1: Fig. S11). BaPs adsorbed on the lower graphene surface were partially detached and vertically immersed in the PS layer to increase the ordering of surrounding lipids. In contrast, BaPs on the upper graphene surface kept lying flat during the simulation. Graphene adsorbed with more BaPs showed similar PS perturbation (Additional file 1: Fig. S12).
We calculated the diffusion coefficients of PS components as affected by graphene and graphene-BaP complexes. Apparently, deposition of a bare nanosheet of graphene strongly reduced fluidity of PS (Fig. 3f), while BaPs solubilized in PS barely affected the PS diffusivity (Fig. 3b). For graphene-BaP mixtures deposited at the PS layer, the restraining effect of graphene on the PS diffusivity was alleviated by adsorbed BaPs. We interpret that it was the strong attraction between graphene and PS to restrain diffusion of surrounding PS molecules. When the original surface of graphene was covered by BaPs, the direct contact between graphene and PS was sterically hindered to alleviate the restraining effect on PS diffusivity.
Curled graphene encapsulating BaPs interacted with the PS layer in a different pathway. They were horizontally immersed in the PS layer to induce PS perturbation (Additional file 1: Fig. S13), similar to that of rod-like nanoparticles and carbon nanotubes as probed in our previous studies [28, 29]. The energy of both BaP-PS and graphene-PS interactions decreased, while the energy of BaP-BaP and graphene-BaP interactions slightly increased (Additional file 1: Fig. S13), suggesting that BaPs, especially those encapsulated inside curled graphene, were barely released due to segregation by curled graphene from contact with PS.
Deposition of a GO nanosheet with adsorbed BaPs induced pores in the PS layer (Fig. 3g and Additional file 5: Movie S4), regardless of BaP adsorption (Additional file 1: Fig. S14). Even for a bare GO nanosheet, it rapidly adhered to and pierced through the PS layer from one corner (Additional file 1: Fig. S15). Compared to pristine graphene, GO has a lower hydrophobicity to make it easier to repel PS molecules and get more favorable contacts with water, thus opening a pore underneath [13]. We calculated the average PS area under different conditions and found different trends of PS layer expansion induced by GO, graphene, GO-BaP and graphene-BaP complexes. First, no expansion of the PS layer was induced by graphene (Fig. 3h), while a marked increase of the PS area was induced by a GO nanosheet (Fig. 3i), suggesting formation of a hydrophilic pore. For graphene with adsorbed BaPs, there was a slight increase of the PS area (Fig. 3h), suggesting release of partial BaPs into the layer to expand it. More BaPs were released from GO, generating higher increase of the PS area (Fig. 3i). Upon release of BaPs, more PS molecules were extracted, which in turn expelled more BaPs off GO to generate additional decreases in both the GO-PS and BaP-PS interaction energies (Additional file 1: Fig. S16).
Graphene translocation across the PS layer
Graphene can enter cells through piercing through the membrane [30, 31], and we thus expected translocation of inhaled graphene across the PS layer as affected by graphene oxidation and BaP adsorption. To simulate inhalation, an external force was exerted on graphene to pull it along the PS layer normal direction at a constant velocity 0.1 nm/ns, close to the human inhalation velocity estimated before [12]. Translocation of GO was more energetically favorable than graphene (Fig. 4a, b), because PS molecules can adsorb on the hydrophobic graphene to impede its translocation, while the less hydrophobic GO acquired more favorable contacts with water through translocation. Adsorbed BaPs facilitated translocation of graphene (Fig. 4a), but the translocation of GO was contrarily retarded (Fig. 4b). That was because BaPs adsorbed on graphene segregated its favorable contact with PS (Fig. 4c, g), while the GO’s surface hydrophobicity was increased by adsorbed BaPs to enhance interactions with PS (Fig. 4d, g). During and after translocation, all BaPs adsorbed on GO were released into the PS layer (Fig. 4e, g), while partial BaP and PS molecules were carried by graphene to enter the subsequent fluid (Fig. 4f, g).
Extraction of PS molecules by suspended graphene as affected by oxidation and BaP adsorption
For the suspended graphene nanosheet adsorbed with 80 BaPs, a number of PS molecules were rapidly extracted from the layer (Fig. 5a). Adsorbed BaPs were repelled by extracted PS to form more compact stacking on graphene. In consequence of the destructive PS extraction, the PS layer ruptured at t = 10 ns (Fig. 5e). By contrast, less PS molecules were extracted by GO (Fig. 5b and Additional file 1: Fig. S17), and the PS layer ruptured 10 ns later (Fig. 5e). When 200 BaPs adsorbed on graphene to reach a saturated adsorption state, no PS was extracted in the finite simulation time, despite release of few BaPs into the layer (Fig. 5c and Additional file 1: Fig. S17). By contrast for GO, more BaPs were released, thus offering space to extract PS molecules (Fig. 6d, Additional file 1: Fig. S17 and Additional file 6: Movie S5), and the PS layer ruptured at t = 82 ns (Fig. 5e).
The destructive extraction of PS molecules was experimentally verified and quantitively measured using QCM-D (details are found in Methods). First, the cover of SUV layer on the Au crystal sensor was confirmed by the changes of frequency shift ∆f and dissipation shift ∆D (Additional file 1: Fig. S18). When graphene suspension flowed through the SUV layer, the ∆f showed an upward trend due the mass loss (Fig. 5f). By contrast, the ∆f increased less when GO suspension of the same concentration flowed through the layer (Fig. 5g), suggesting lower number of PS molecules extracted by GO. Moreover, we fluorescently labeled GUVs, and observed their interactions with graphene using LCSM (Fig. 5h, details are found in Methods). GUVs stably bound to graphene, and a tubular structure protruded from the area of vesicle attachment on graphene, strongly suggesting membrane perturbation induced by graphene.