Transport of environmental natural organic matter coated silver nanoparticle across cell membrane based on membrane etching treatment and inhibitors

Environmental natural organic matters (NOMs) have great effects on the physicochemical properties of engineering nanoparticles, which may impact the transport of nanoparticles across plasma membrane and the cytotoxicity. Therefore, the kinetics, uptake pathway and mass of transporting into A549 cell membrane of silver nanoparticles (AgNPs) coated with citric acid (CA), tartaric acid (TA) and fulvic acid (FA) were investigated, respectively. CA, FA and TA enhanced the colloidal stability of AgNPs in culture medium and have greatly changed the surface plasmon resonance spectrum of AgNPs due to the absorption of CA, FA and TA on surface of AgNPs. Internalizing model showed that velocity of CA-, TA- and FA-nAg transporting into A549 cell were 5.82-, 1.69- and 0.29-fold higher than those of the control group, respectively. Intracellular mass of Ag was dependent on mass of AgNPs delivered to cell from suspension, which obeyed Logistic model and was affected by NOMs that CA- and TA-nAg showed a large promotion on intracellular mass of Ag. The lipid raft/caveolae-mediated endocytosis (LME) of A549 cell uptake of AgNPs were susceptible to CA, TA and FA that uptake of CA-, TA- and FA-nAg showed lower degree of dependent on LME than that of the control (uncoated AgNPs). Actin-involved uptake pathway and macropinocytosis would have less contribution to uptake of FA-nAg. Overall, transmembrane transport of NOMs-coated AgNPs differs greatly from that of the pristine AgNPs.

The increasing application of engineering nanoparticles will inevitably result in the accumulation of these engineering nanoparticles in environment and may result in potential ecological and health risks [1][2][3] . For example, the accumulation of silver nanoparticles (AgNPs) can inhibit embryo growth 1 and cause a series of cytotoxicity such as gene mutation 4 , inhibition of cell proliferation 5 , apoptosis 6,7 and necrosis 8 . Now in vitro cytotoxicity investigations are frequently used to explore the toxic mechanism of nanoparticles [9][10][11][12] . The cytotoxicity of hardly soluble nanoparticles such as AgNPs mainly caused by intracellular particles according to "Trojan-horse mechanism" 5,13 . Therefore, many studies have been carried out on quantitative or qualitative analysis of intracellular nanoparticles to reveal cellular uptake of nanoparticles. Qualitative methods (e.g. Transmission electron microscopy 14 , Scanning electron microscopy 15 , Light scattering microscopy 16 , Super-resolution fluorescence microscopy 17 , Atomic force microscopy 18 ) have been fully studied to directly observe intracellular nanoparticles. However, the quantitative methods of nanoparticles entering into cell are developed slowly compared to the qualitative methods 19 . The main challenge is how to erase the disruption of cell surface associated nanoparticles which are hard to be differentiated from intracellular nanoparticles 14,[19][20][21][22] . Therefore, selective removal methods of cell surface associated nanoparticles with etchants have been developed 20,21 . The etchant I 2 -KI was firstly used to selectively remove gold nanoparticle (AuNPs) from cells and the internalized mass of Au nanoparticle was successfully analyzed 20 . The etchant K 3 Fe(CN) 6 -Na 2 S 2 O 3 was proved to effectively remove silver nanoparticles

Results
Characterization of NOMs-coated AgNPs and their stability in culture medium. After treated with CA, FA or TA, the size of silver nanoparticle was little changed (25.1 nm to 27.7 nm in average) as shown in Fig. 1a. This was consistent with our previous study 37 . Table 1 showed that D H in culture medium (CM) with 1% FBS of CA-nAg was substantially lower than D H of nAg control (99 nm to 117 nm in average). CA was absorbed on the surface of AgNPs and resulted in a higher carbon contents (1.91% to 1.77%) and higher ratio of content of carbon to content of nitrogen (C:N, wt:wt) compared to nAg control (6.69 to 5.10) as shown in Table 1. Moreover, Table 1 showed that zeta potential value of CA-nAg in water (− 45.0 mV) was much negative than nAg control in water (− 24.2 mV) since CA absorbed on the surface of AgNPs.
Surface plasmon resonance (SPR) spectrum of NOMs-coated AgNPs suspensions in CM (20 μg ml −1 ) at 37 °C was shown in Fig. 1b. The maximum absorption wavelength (λ max ) of nAg control , CA-, FA-and TA-nAg were 402, 408, 406 and 404 nm, respectively. Figure 1c,d recorded the SPR spectrums of suspensions within 60 min and the trends of absorption value at λ max (A max ) within 60 min. Red-shift of λ max for all of suspensions depended on contact time and A max trended to decrease. Within 60 min, λ max of nAg control , CA-, FA-and TA-nAg was red-shift for 2 nm and their A max decreased for 15.6, 8.4, 6.6 and 7.0%, respectively. The ratio of real time absorbance to initial absorbance (A/A 0 ) at λ max over 60 min could be fitted well with the first order removal model, which is a simplified model for particles sedimentation in water solution 40 . The model details and fitting parameters were listed in SI (Table S1). However, Fig. 1e showed that D H of these AgNPs in CM were almost constant within 30 min. It implied that these AgNPs suspended stable in CM in dilute concentration.   The internalization process was expressed as follow 17,20,41 :   Table 2 presented the fitting parameters of Langmuir adsorption model. The dissociation factor k d of nAg control , CA-, FA-and TA-nAg was almost equal to 0. These implied that the dissociation process of AgNPs bound on A549 cell surface was weak. The larger the value of k a is, the higher the affinity of AgNPs with cell surface will be 17 . The association factor-k a of CA-nAg (1.006 μM −1 h −1 ) was similar to that of nAg control (1.002 μM −1 h −1 ). However, k a value of FA-nAg (352.2 μM −1 h −1 ) was much higher than that of nAg control and the value of TA-nAg (1.386 μM −1 h −1 ) was slightly higher than that of nAg control . These implied that FA and TA absorbed on surface of AgNPs could increase the affinity of AgNPs with cell surface while CA affected very little. Table 3 presented the fitting parameters of internalizing model. The value of k i and k out of CA-nAg was about 5.82-and 11.3-folds higher than that of nAg control , respectively. The value of k i and k out of TA-nAg was about 1.69-and 1.14-folds higher than that of nAg control , respectively. To FA-nAg, k i and k out was much lower than them of nAg control (around 0.29-and 0.38-folds to nAg control , respectively). Hence, the equilibrium time (t max ) of internalization and exocytosis of CA-and TA-nAg was shorted from 18 h of nAg control to 9 h and 13 h, respectively. The value of t max of FA-nAg (52 h) was much larger than that of nAg control . The value of M imax is frequently used in the evaluation of cellular uptake of nanoparticles with certain size or specific surface properties which could influence the association of nanoparticles on cell surface and the binding of nanoparticles with receptors 14,17,41 .    Table 3. Fitting parameters for nAg control , CA-nAg, FA-nAg and TA-nAg internalized into A549 cell with internalizing model (incubated AgNPs concentration was 10 μg ml −1 ).  Table 3. Definitely, the parameter of k i /k out indicated the value of dividing the rate of internalization by the rate of exocytosis. Accordingly, k i /k out was suggested to be a valuable parameter to reflect cellular uptake of nanoparticles. Table 3 also shows that the values of k out were greatly higher than those of k i for nanoparticles except TA-nAg. It was reported that K i values were less than K out values for citrate-and PVA-coated Au nanospheres, while K i and K out values for PAA-coated Au nanospheres were significantly higher than those in the former two 20 . The authors explained that these were mainly caused by the different amount of Au nanospheres adsorbed onto the cell surface. These were consistent with our results that internalization process of coated nanoparticles would be influenced by the different organic ligands. Figure  . This meant that CA-, FA-and TA-nAg were more stable in CM than nAg control . It was consistent with the results in Fig. 1d. As shown in Fig. 4b, the relationship of M i and M d was likely subject to Logistic model which was found to fit well with the uptake of magnetic iron nanoparticles by T98G and U251 cell 19 . The Logistic model was expressed as follow 19 :

Relationship of intracellular AgNPs with AgNPs delivering from suspension.
where M imaxC was maximum value of M i , pg; EC 50 was M d for 50% of M imaxC , pg; p was the slope factor 19 . Table 4 presented the fitting parameters of Logistic model. M imaxC of CA-nAg and TA-nAg was 1.59 and 1.50 pg, respectively, higher than that of nAg control (1.02 pg). M imaxC of FA-nAg was 0.85 pg, less than that of nAg control . M 0C of CA-nAg and TA-nAg was 6.78 and 5.39 pg, much higher than that of nAg control (5.00 pg). M 0C of FA-nAg (5.39 pg) was slightly higher than that of nAg control .   Figure 5a presented the influences of inhibitors on the internalization of AgNPs under "crowded state" cells and their influences between "rare" state cells and "crowded" state cells were show in Fig. 5b. These inhibitors (cytochalasin D, EIPA, chlorpromazine and filipin) have few influences to the stability of AgNPs suspension and little decrease in A549 cell viability incubated for 1 h under the used concentrations as shown in Fig. S2. To "crowded state" cells, the inhibition rate of internalization of AgNPs by cytochalasin D was at least 90%, except for FA-nAg (76%) (Fig. 5a). The addition of EIPA also resulted in significant inhibition of internalization of the positive control (nAg control ), CA-, FA-and TA-nAg at the rate of 85, 76, 73 and 81%, respectively. Chlorpromazine have also caused a large extent of decrease in internalization of nAg control , CA-, FA-and TA-nAg with the rate of 58, 59, 61 and 64%, respectively. The inhibition rate of internalization of nAg control by filipin was 61%, significantly higher than filipin for CA-, FA-and TA-nAg with the rate of 32, 47 and 31%, respectively. The influences of inhibitors to "rare state" and "crowded state" cells were different (Fig. 5b). The inhibition rate for "rare state" cells was significantly lower than that for "crowded state" cells, including EIPA to FA-and TA-nAg, chlorpromazine to TA-nAg, and filipin to nAg control (Fig. 5b). However, some reverse phenomena were observed, including EIPA to CA-nAg, chlorpromazine to nAg control , and filipin to CA-nAg and TA-nAg (Fig. 5b).

Discussion
The size of FA-nAg was little changed as shown in Fig. 1a. Table 1 showed that carbon content of FA-nAg was 2.64%, higher than that of nAg control (1.77%). This suggested that FA had been absorbed on the surface of AgNPs. FA-nAg presented more negative value of zeta potential in water (− 42.7 mV) than nAg control (− 24.2 mV) ( Table 1). Moreover, FA could also lightly enhance D H of FA-nAg in CM comparing to nAg control in CM (121 nm to 117 nm in average) as shown in Table 1. However, carbon content of TA-nAg (1.47%) was lower than that of nAg control (Table 1). This suggested that TA decreased the amounts of PVP coated on AgNPs. Moreover, C:N of TA-nAg was much higher than that of nAg control (7.12 to 5.1). This suggested that TA had absorbed on surface of AgNPs replacing parts of coated PVP. D H of TA-nAg in CM was much lower than that of nAg control in CM (81 nm to 117 nm in average) for shrinking hydration layer in water solution (Table 1). D H of CA-, FA-and TA-nAg in CM kept steady within 30 min as nAg control in CM, implied that little aggregation happened in dilute AgNPs suspension (1 μg ml −1 ). However, red-shift for 2 nm observed in the SPR spectrum of CA-, FA-and TA-nAg as nAg control suggested that light aggregation happened in concentrated AgNPs suspension (20 μg ml −1 ) 43,44 . The concentrated suspension of CA-, FA-and TA-nAg in CM seemed to be more stability than nAg control since only 8.4, 6.6 and 7.0% decreased in A max of them, lower than 15.6% of nAg control . The much higher affinity of FA-nAg to A549 cell than others were found as shown in Table 2. Cho et al. found that poly(allyamine hydrochloride) or PAA coated AuNPs showed higher affinity to SK-BR-3 breast cancer cells than others (k a was 10 times to others) because of the positive charge of amino function group 20 . FA were comprised of aromatic, carboxylic acid and amino function group according to the characterization in our previous report 37 . Accordingly, the amino function group in FA could make the difference.
The value of k i reflect the rate of particles internalizing into cell 17,20,41 . Table 3 showed that value of k i of CA-nAg and TA-nAg was larger than that of nAg control . Harris et al. proved that hydrating layer hinders protein adsorption and subsequent internalization 45 . Table 1 showed that the thickness of hydrating layer of CA-nAg and TA-nAg was lower than that of nAg control (D H of them was 99, 81 and 117 nm, respectively). Therefore, the increase of rate of CA-nAg and TA-nAg internalizing into cell were related to the decrease of hydrating layer of these AgNPs. However, the value of k i of FA-nAg was still much lower than that of nAg control even though the similar D H (121 nm) with nAg control as shown in Tables 1 and 3.  Table 5 recorded the M i of AgNPs and AuNPs into cells that have reported in many literatures 8,20,33,[46][47][48][49][50] . M i of these AgNPs into cancer cells were about 2.1 to 10 pg at a same order of magnitude with the results in this study, despite of different size, surface functionalization or other experimental conditions 8,47,50 , which showed much difference from the normal cells (M i = 47 pg for Pk 15 cells) 48 . Comparing with the M i of AgNPs into cancer cells,  Table 2). However, M 0C of CA-and TA-nAg were much higher than that of M 0 . These implied that when suspension of CA-and TA-nAg come to concentrate, the associated particles could gather together tightly and form large cluster on cell surface due to their small size of D H . The clusters made higher value of appearance M 0 that was called "M 0C ". As a consequence, M imaxC of CA-and TA-nAg was higher than that of nAg control which was different from the situation of M imax . Ratio of internalized AgNPs transferring from M 0C (R T , R T = M imaxC /M 0C × 100%) was defined to reflect the efficiency of surface association AgNPs transferring to internalized AgNPs. R T of CA-nAg and TA-nAg (23.5% and 21.2%) were higher than R T of nAg control (20.4%), and R T of FA-nAg (15.8%) was lowest. Combined with the conclusion of k i (CA-nAg, TA-nAg > nAg control > FA-nAg), these implied that CA and TA-nAg showed stronger ability but FA-nAg presented weaker ability of transport across plasma membrane than nAg control .
Cytochalasin D is a cell permeable toxin which can disrupt actin filaments 51 . EIPA is a Na + /H + ion channel blocking agent that inhibits the macropinocytosis-mediated pathway 29,52 . Chlorpromazine can prevent from the formation of clathrin in cells and is used to depress the uptake pathway of CME 53,54 . The addition of these inhibitors resulted in the suppression of internalization of the pristine AgNPs-nAg control , indicated that the uptake pathway of AgNPs were mostly depended on actin and contributed a lot to both macropinocytosis and CME. Moreover, the inhibition rate of these inhibitors to NOMs-coated AgNPs was less affected by CA, FA and TA, while the inhibition rate of cytochalasin D and EIPA to FA-nAg were significantly less than them to the positive control-nAg control (Fig. 5a). Accordingly, actin-involved uptake pathway or macropinocytosis would have less contribution to uptake of FA-nAg than uptake of the pristine AgNPs-nAg control . It could be the reason of lower efficiency in uptake for FA-nAg than for nAg control (k i and k out of FA-nAg were far below them of the pristine AgNPs-nAg control as presented in Table 3).
Filipin, a drug can bind with sterol, is known as an inhibitor of LME 30,55 . The addition of filipin resulted in 61% decrease in the internalization of the pristine AgNP-nAg control . This indicated that LME was also involved in the internalization of AgNPs. However, inhibition rate of CA-nAg, FA-nAg and TA-nAg by filipin was 32, 47 and 31%, respectively that were consistently lower than the rate to the pristine AgNPs-nAg control (61%). These indicated that CA, FA or TA would change the way of uptake of AgNPs mainly through depressing the contribution of LME. The results were highly accordance with yielded ROS level by these AgNPs that nAg control could arouse ROS level for twofold but 1.3-fold for FA-nAg and no significant influence for CA-nAg and TA-nAg (Fig. 6). LME is a unique pathway for nanoparticles that the intracellular vesicle could escape from lysosomal and uptake the nanoparticles into cytoplasm 26,28 . Huk et al. found that AgNPs caused much higher level of cytotoxicity since it could have reached into nucleus and mitochondria 4 . It implied that the higher dependence on LME of AgNPs would have higher opportunity to reach into nucleus or mitochondria and cause more damage to cells. www.nature.com/scientificreports/ Figure 7 illuminated the reason about lower inhibition rate of "rare" state than "crowded" state. The "rare state" means that mass of cell surface associated AgNPs 42 were far below M 0 while "crowded state" means that mass of associated AgNPs were enough to reach M 0 . To the addition of inhibitor to "rare state" cells, the M s would be higher than M s without inhibitors, because M i would be reduced but M d might be rarely impacted according to Eq. (4). Thus, more free sites on cell surface would be occupied and result in exceeding M i . Therefore, the calculated inhibition rate would be lower than what it should be. This may be called as the "waning and waxing" phenomenon. To the "crowded state", this phenomenon could be ignored since M s was a constant.
However, some reverse phenomenon happened because of certain properties of endocytosis pathway. LME inhibited by filipin is known to be a receptor-specific uptake and usually form 50 to 80 nm caveolae in size 26,28 . As shown in Table 4, it can be concluded that CA-nAg and TA-nAg could form clusters on cell surface under concentrated CA-nAg and TA-nAg suspensions. Accordingly, the formed clusters were not appropriate in size to be trapped by caveolae. The inhibition rate of filipin to "crowded state" cell of CA-nAg and TA-nAg was less than the inhibition rate to their "rare state" cell.
In summary, CA treatment reduced D H and enhanced the colloidal stability of CA-nAg in CM comparing to the pristine AgNPs-nAg control . Consequently, the increase for both of k i and k out but decrease in M i were found. TA treatment reduced D H and enhanced colloidal stability of TA-nAg in CM compared to the pristine AgNPs-nAg control . Consequently, the increase for k i , k out and M i were found. FA enhanced the stability of FA-nAg in CM, but much decrease for k i , k out and M i were found which resulted from less dependent on actin involved uptake pathway and macropinocytosis than the pristine AgNPs-nAg control . In addition, intracellular mass of these AgNPs were dependent on M d , which obeyed Logistic model. According to the internalization model and Logistic model, CA and TA-nAg showed stronger ability but FA-nAg presented weaker ability of transport  www.nature.com/scientificreports/ across plasma membrane than the pristine AgNPs-nAg control . Moreover, uptake of CA-, TA-and FA-nAg was less dependent on LME comparing to the pristine AgNPs-nAg control , which resulted from cell surface association state of AgNPs that affected by NOM.

Methods
Natural organic matters. The NOM used in this study include CA, TA and FA. CA and TA were purchased from Sinopharm Chemical Reagent Co., LTD. FA was extracted from the sediments of Xuanwu Lake at Nanjing, China, and the properties were presented in our previous report 37 . Preparation and characterization of NOM coated nAg. The pristine silver nanoparticles of 20 nm were synthesized according to the previous reports with minor modification (more details seen the Supporting Information, SI) 56 . The obtained silver was marked as p-nAg for the subsequent treatment. NOMs-treated nAg was made according to our previous study 37 . Briefly, the obtained p-nAg suspension was treated with solutions containing citric acid, tartaric acid and fulvic acid Concentrations of CA and TA in solution were set as 10 mM and FA were 200 mg l −1 (more details seen the Supporting Information, SI).
The sizes of the pristine and NOM-coated AgNPs were characterized by Transmission Electron Microscope (TEM, JEM-2100 (HR), Japanese JEOL Corporation). The carbon and nitrogen contents of these samples were determined using element analyser (EA, CHN-O-Rapid, Germany Heraeus Corporation). The stability of the pristine and NOM-coated AgNPs suspensions in Ham's F-12K (Kaighn's) Medium (1×, Gibco) which is used as grow up medium (CM) for A549 cell in this study supported with 1% fetal bovine serum (FBS, Hyclone) and 1% antibiotics (penicillin streptomycin sol, Gibco) were characterized by UV-Vis spectrometer (UH5300, Japanese Hitachi Corporation) at 37 °C. Briefly, 5 mg of the pristine and NOM-coated AgNPs was put in 5 ml 1% FBS supported CM. The suspensions were diluted with 1% FBS supported CM to a final concentration of 20 μg ml −1 . The absorbance from 300 to 600 nm (step by 2 nm) of suspensions were detected at 0, 10, 20, 30, 40, 60 min, respectively. Hydrodynamic diameter (D H , nm) and zeta potential value (ZP, mV) of the suspensions (diluted to 1 μg ml −1 ) were monitored over 30 min at 37 °C by nanoparticle size analyzer (90Plus, Brookhaven Instruments Corporation). The initial D H was determined based on the average value of dynamic light scattering (DLS) data within 3 min.
Etching AgNPs bound on the cell surface. Etching method was proposed and verified by Gray B. Braun for removing the absorbed the pristine and NOM-coated AgNPs on cell surface which disrupt the quantitative of intracellular AgNPs 21 . Actually, etching method failed to clean the well-plate touching side of adherent cell where solvent was hard to infiltrate while AgNPs could be transferred from top side surface due to the fluidity of cell membrane. An etching method integrated adherent and suspended cell to remove the absorbed AgNPs on cell surface for the quantitative analysis of the intracellular AgNPs was developed and the related experiments with results and discussion were described in SI. The low cytotoxicity of our used etchants to A549 cell in short etching time was shown in Fig. S3a. The high efficiency of the etching method to remove association AgNPs on A549 cell surface was also proved in Fig. S3b. Therefore, based on the quantitative analyses of particles internalized into cell.
Kinetics of the pristine and NOM-coated AgNPs uptake by A549 cell. A549 cells were seeded in 12 well-plate for 24 h prior to exposure with the pristine and NOM-coated AgNPs. At the following day, the CM was removed, and then rinsed with PBS for twice. After the addition of 10 μg ml −1 AgNPs suspension, cells were incubated for 0, 1, 2, 4, 6, 8 and 12 h, and then treated with the etchant and collected.

Relationship of intracellular AgNPs with AgNPs delivering from suspension. After incubation
for 24 h of seeded cells in 12 well-plate, the CM were removed and cells were rinsed with PBS for twice. Designed concentrations of AgNPs suspension were added and cells were incubated for 1 h. Cells were treated with the etchant and collected. Concentrations of AgNPs were set as: 0-35 μg ml −1 for nAg control , 0-50 μg ml −1 for CA-nAg, 0-40 μg ml −1 for FA-nAg and 0-60 μg ml −1 for TA-nAg. Fig. S4 showed that cell viability of AgNPs was above 80% even the incubated concentration of AgNPs up to 100 μg ml −1 , which implied that concentration of AgNPs used in this experiment were harmless to A549 cell. Cellular uptake pathway of NOM-coated AgNPs on A549 cell. Some literatures report that cellular uptake of particles are dependent on their aggregation 32 or aggregation behavior 57 on cell surface. Therefore, the gathering state of AgNPs on cell surface would affect the cellular uptake pathway utilized by AgNPs. Accordingly, cellular uptake pathway for NOM-coated AgNPs to A549 cell were studied under two levels of AgNPs density on cell surface (rare state and crowded state).
Rare state. Rare state means the level of AgNPs associated with cell surface was much less than maximum capacity to accept AgNPs on A549 cell surface. Firstly, A549 cells were seeded in 12 well-plate. The cells were pretreated with inhibitors (their final concentrations were 5 μM for cytochalasin D, 5 μM for EIPA, 30 μM for chlor-Scientific Reports | (2021) 11:507 | https://doi.org/10.1038/s41598-020-79901-y www.nature.com/scientificreports/ promazine and 0.5 μg ml −1 for filipin, respectively) for 30 min in incubator. Then, cells were rinsed with PBS for 1 time and exposure to 10 μg ml −1 of pristine or NOMs-coated AgNPs suspension with inhibitor (kept the same concentration), and incubated for 1 h in incubator. Finally, cells were treated with etching method and collected.
Crowded state. Crowded state means the mass of AgNPs associated with cell surface was enough to reach maximum capacity to accept AgNPs on A549 cell surface. Similar to above, after pretreated with inhibitor, cells were rinsed with PBS for 1 time and exposure to pristine or NOMs-coated AgNPs suspension with inhibitor. To make 20 pg per cell AgNPs associated on cell surface before cellular uptake starting, cells were firstly incubated for 1 h at 4 °C. Then, cells were put in incubator for next 1 h. Finally, cells were treated with etching method and collected. Final concentration of nAg control , CA-nAg, FA-nAg and TA-nAg for "crowded state" were 20, 70, 50 and 30 μg ml −1 , respectively. Cells exposed to the pristine silver nanoparticles with inhibitor were set as positive control (marked as nAg control ). Cells exposed to the pristine and NOMs-coated AgNPs without any inhibitor were set as negative control. Cells incubated with inhibitor and without pristine and NOMs-coated AgNPs were set as blank control. The experiment was run with triplicate.
TEM observation of intracellular pristine and NOM-coated AgNPs. TEM has frequently been used to observe the localization of nanoparticles 8,14 . A549 cells were centrifuged and rinsed by PBS after 6 h exposure to 75 μg ml −1 nAg control , CA-nAg, FA-nAg and TA-nAg or 10 μg ml −1 CA-nAg and TA-nAg. The harvested cells were prefixed in 2.5% glutaraldehyde at 4 °C overnight and washed with PBS three times. Subsequently, the cells were stained with 1% osmic acid followed by gradient dehydration with ethanol and acetone. Then, the samples were embedded in epoxy resin, sectioned, and post stained with lead citrate and uranyl acetate before TEM observation. Finally, cells were observed using the TEM.

ROS level.
The cells were seeded in 12 well-plate for 24 h prior to exposure with AgNPs. Seeding density was 5 × 10 5 cells per well. Cells were exposure to 1 ml 75 μg ml −1 AgNPs suspension for 24 h. Then, the plates were rinsed with PBS for twice and loaded with 10 μM DCFH-DA in CM for 20 min in incubator. Thereafter, cells were rinsed with CM for three time and treated with 0.2 ml EDTA-trypsin solution. The suspended cells were collected with PBS and the fluorescence was recorded with a flow cytometer (MoFlo XDP, Beckman Coulter) by reading 5 × 10 4 cells at FL1 channel (excitation 485 nm, emission 535 nm).
Elemental analysis. The cells (cell number: 5 × 10 5 ) were digested with concentrated HNO 3 . Concentrations of Ag were measured using an inductively coupled plasma optical emission spectrometry (ICP-OES, Optima 5300, Perkin-Elmer SCIEX, USA). The calibration standard solutions were diluted from obtained by the dilution of the standard stock solutions (Custom Assurance Standard) purchased from SPEX CertiPreP (1000 mg l −1 , Lot number: 28-232CR) with 2% HNO 3 (V/V). The relative percentage differences of parallel samples were within 20%, or the experiments were repeated. Data analysis. The significant differences were analysed by independent-sample T tests in SPSS statistic 17.0. First-order removal model and Logistic model were fitted with the trends of AgNPs sedimentation in CM and the mass reliable internalization process, respectively in origin 9.1. Langmuir absorption model and internalizing kinetic model were fitted with cell surface association process and internalizing kinetic of AgNPs, respectively, in Matlab R2016a.