In this work, we conducted standard cytotoxic assays and metabolomics in two lineages of cells, FN1 and HUV-EC-C, used as models to study wound healing induced by silver nanoparticles. The cells were exposed to a planned set of doses of AgNP during defined incubation times, and the effect on the average cellular growth and cytotoxicity was evaluated. Based on these first exploratory assays, relevant information on the intervals of AgNPs molar concentrations and incubation times were obtained and used to design the set of optimized conditions for the magnetic resonance experiments to study the main metabolic changes triggered by the nanoparticles treatment on each lineage. For such a purpose, both the cellular extracts and culture medium were evaluated in order to provide the fullfill metabolic profiling of cells summarized by the endometabolome and exometabolome, respectively 21. The first one provides a direct evaluation of the cellular metabolic status until methanol addition interrupts cell metabolism, lysis, and extraction of the metabolic components for analysis, which was performed by monitoring the levels of endogenous metabolites. The analyses of culture media components, such as aminoacids and some vitamins that were exogenously consumed by cells, in addition to regular and treated cell metabolites that were excreted, can be used to access indirectly the cellular metabolic status. The endometabolome and exometabolome analysis provide the most comprehensive and complete perspective about the metabolic scenery of cells exposed to silver nanoparticles and can reveal interrelated processes in the same metabolic pathway. Furthermore, it tells us about exclusive characteristics of a determinate kind of sample, for instance, the phospholipid composition of the cellular membrane 22.
Silver nanoparticles. The effect of silver nanoparticles depends on their interaction and uptake by cells, which depend on parameters such as shape, size, and surface potential of citric acid stabilized AgNP stock dispersion. This last parameter tends to be sensitive to ionic strength and agglomerate in higher ionic strength aqueous biological media. Also, its protecting molecular layer is weakly bound on nanoparticle surfaces. It tends to be easily replaced by molecules that can bind more strongly, commonly present in the culture medium, such as peptides and proteins, phosphate, and other components. Accordingly, experiments were conducted to evaluate the colloidal stability of nanoparticles in our control medium (DMEM culture medium supplemented with 10% v/v of FBS) necessary to support cellular metabolism during incubation. However, no significant change was observed by DLS.
Cytotoxicity assays. The CGK results summarized in Fig. 1 were organized by concentrations shown in separate panels, each containing the lineages differentiated by symbols, red circles, and green squares, to optimize the presentation of results. Analyzing these panels, we can first glimpse the impact of AgNP on the cellular viability by the impact on their proliferation rates 23. As can be confirmed in Fig. 1, both FN1 and HUV-EC-C lineages present a similar proliferation behavior up to 48 hours in the absence of AgNP, followed by an apparent reduction in the cell population that can be associated with the depletion of culture medium nutrients and the natural process of cellular death by competition for space. Also, it is possible to see by analyzing the data in that range that FN1 is more sensitive to the presence of silver nanoparticles, and the proliferation is more strongly impacted. Such a tendency can be qualitatively monitored by the average time-dependent curves estimated from the linear fits of the rates of expansion (𝜏 > 0) and reduction (𝜏 < 0) of cells populations incubated as a function of the silver molar concentration, thus separating the effects of natural cell death from that induced by AgNP cytotoxicity and defining the rates at which these phenomena occur.
When nanoparticles are added, the cytotoxic response exhibited by FN1 and HUV-EC-C for the first 48 hours, where the cell number should be naturally expanding, indicates a continuous decrease in the proliferation rates and an increasing tendency of cell death. As shown in Fig. 1, the cells from FN1 lineage present |𝜏| < 0 at [AgNPs] as low as 25 µM, whereas HUV-EC-C showed similar |𝜏| < 0 at [AgNP] of 50–75 µM, demonstrating much lower sensitivity to the cytotoxic effects of AgNP. As expected, for incubation times longer than 48 hours, the estimated values of 𝜏 are always negative for these two lineages suggesting a reduction in their respective populations by a combined effect of natural cell death and that induced by AgNP cytotoxicity. Furthermore, AgNPs showed pronounced cytotoxicity at concentrations higher than 75 µM, mainly above 100 µM, where both FN1 and HUV-EC-C cells exhibited quite similar behavior.
In order to isolate the cytotoxic effects induced by AgNP and the impact on the cellular morphology of the two lineages, some conditions were assumed. First, the incubation time within which the control cells were proliferating should be considered mainly, i.e., from 0 to 24 hours and from 24 to 48 hours, where 0 hours is the moment after 12 hours of incubation for cell stabilization in the absence of silver nanoparticles. Each FN1 and HUV-EC-C cell was assumed to be disc-shaped. The average diameter of the disc was measured in the Neubauer chamber using a microscope and a scale considering up to 10 cells for statistics. Such measurement was carried out after incubation for 24 and 48 hours and different concentrations of silver nanoparticles. In the Figs. 2 and 1 of Supplementary Information are plotted the obtained values for the average surface areas of cells calculated from their diameters.
Silver nanoparticles are considered low toxicity materials, but the optical microscopy images shown in the upper panels of Fig. 2 after 24 and 48 hours of incubation reveal cellular detachment and the presence of apoptotic bodies, as well as a tendency to decrease the diameter and related surface area, indicating a significant impact on the normal cell metabolism. The intervals of concentrations in which these tendencies occur are different depending on the cell lineage, showing important and distinct cellular resiliency against cytotoxic effects, following the previous results obtained for CGK assays. Furthermore, on average, the surface area of cells tends to decrease as the concentration and time of AgNP treatment increases, as shown in Fig. 1b of the Supplementary Information.
Figure 2b,c present the main results for the Neubauer counter and MTT assays, respectively, as the plots of “Number of viable Ncells cells vs. [AgNP]” and “Absorbance at 570 nm vs. [AgNP]” for the FN1 and HUV-EC-C cell lineages investigated as a function of AgNP (silver) molar concentration and incubation times up to 48 h, corresponding to the phase at cellular expansion. The significant cytotoxicity of silver nanoparticles is apparent in all concentrations, especially for FN1 cells, which becomes more evident after 48 hours for the lowest doses (25 and 50 µM). The analyses of these plots provide the first demonstration of the combined dose-dependent cytotoxicity induced by AgNP that allowed us to define the low and high toxicity ranges, respectively, denominated sublethal and lethal doses, used to choose the appropriate range of concentrations to prepare the samples for magnetic resonance experiments. Nevertheless, the range of concentrations and incubation times in which they are observed are more or less well-defined and different depending on the cell lineage, reinforcing the premise that they present specific tolerances to the toxic effects of AgNP. For instance, in Fig. 2b, FN1 cells seem tolerant to [AgNP] doses up to 50 µM.
In comparison, a significant decrease in the cell population can be observed after 24 and 48 hours of incubation (8% and 36%) at concentrations equal to or higher than 50 µM, as compared to the control (0 µM). This behavior suggests the presence of dose- and time-dependent cytotoxic effects where the first concentration range can be considered sub-lethal doses. However, the [AgNP] dose of 75 µM induced a clear decrease in the cell population, respectively, of 56% and 63% after 24 and 48 hours, as compared to control, which seems to be less dependent on incubation time. Accordingly, nanoparticle concentrations above 75 µM were assumed to be lethal to the FN1 cell lineage. In contrast, the ranges of sublethal and lethal doses for HUV-EC-C cells seem to be shifted to lower silver nanoparticle concentrations. For example, 50 µM of AgNP reduced the cell population by 17% and 47% after 24 and 48 hours, compared with control, which also defines this range as the sub-lethal. Similar to FN1 cells, an apparent decrease in the HUV-EC-C cell population of, respectively, 55% and 62% after 24 and 48 hours were observed upon incubation with 75 µM of AgNP and the dependence on the incubation time, such that concentrations higher than that were assigned as lethal. The MTT results, summarized in Fig. 2c, follow the tendencies observed in Fig. 2b, demonstrating a dose and incubation time-dependent cytotoxicity. However, the extension to which the effects are related to the cellular metabolic events is still an open problem in the literature. Below, we address this issue for the first time using the magnetic resonance spectroscopy technique.
Nuclear magnetic resonance and statistical analysis. Nuclear magnetic resonance spectroscopy is a powerful tool for characterizing molecular species. Each molecule has a spectral signature that can be used to identify them; if well-resolved enough, spectra can be obtained. This signature implies removing any contribution that can decrease the homogeneity of the applied main magnetic field and shimming procedure for each measurement. Cell extract and corresponding culture medium samples of cells after 24 and 48 hours of incubation with AgNP were measured to assess the changes in the endometabolome and exometabolome. The 1H-NMR spectroscopy results obtained for the FN1 and HUV-EC-C cells exposed to four different AgNP concentrations and the control (no AgNP) are shown in Fig. 3 and Fig. 4, respectively.
Each metabolite has characteristic peaks in the spectrum, whose integrated area is proportional to the relative metabolite concentration or level. Even though most high-resolution lines have been successfully identified, no apparent differences can be observed directly in their corresponding intensities, as expected for changes in their concentrations induced by the treatment with AgNP. The metabolite level is more suitable for the multi- and univariate statistical methods to access the biochemical cell’s response induced by the treatment with silver nanoparticles. The first step of the multivariate method is to identify the tendencies of clustering among groups of samples with similar metabolomes. This analysis was performed using supervised PLS-DA, and the results are shown in the scores plots shown in Fig. 5 and Fig. 6. The second step was analyzing the VIP scores and identifying the metabolites responsible for such tendencies 24.
Let us first look at the results for 24 hours of AgNP exposition. In Fig. 5a,b, one can notice in the scores plots that in the combination Comp. 1 vs. Comp. 2, clusters of points are ascribed to distinct groups in separated quadrants of Hotelling’s ellipses with different distances among their respective distribution centers. For instance, the separation between the control groups and the groups treated with 50 µM of AgNP is larger than the distance between 75 and 100 µM. Furthermore, groups in the right quadrants are pretty separated from those in the left one. However, inspecting the combination Comp. 1 vs. Comp. 3 in Fig. 5c,d at the same incubation time, the previous scenery seems to be inverted, as can be seen in the scores plots of both cell extracts and culture medium samples. Overall, the isolated contribution of each component of the generated model tends to decrease as the order of components increases 25,26. The percentual values in the brackets of each component evidence that observation. Thus, the combination of the first two consecutive components, Comp. 1 vs. Comp. 2, is expected to have higher statistical differences in the system than the combination of the first with the third, Comp. 1 vs. Comp. 3, when the differences tend to decrease. This rule holds, and the combination Comp. 2 vs. Comp. 3 presents even fewer differences than the previous ones (not shown). Therefore, we can conclude that in the first case of combination Comp. 1 vs. Comp. 2 of the Fig. 5a,b, the variation in the levels of the metabolites is intense for the first dose of treatment, 50 µM, as compared with control, and tends to diminish for the higher doses, 75 and 100 µM. However, the 75 and 100 µM doses also keep differences in the levels of their respective metabolites but are lower in magnitude than the control and cells treated with 50 µM AgNPs. In summary, for 24 hours of treatment, the results obtained for cells and culture media indicated that most changes in FN1 metabolites occur in the first dose of treatment, decreasing for the higher ones, probably indicating a homeostatic response.
Now, we can look at the results after 48 hours of treatment with FN1 cells with AgNP. PLS scores can be interpreted using the same arguments presented previously. For instance, the combination of Comp. 1 vs. Comp. 2 in Fig. 5e suggests that the dose inducing higher modifications in metabolites levels is 25 µM rather than 50 µM used within 24 hours of treatment. However, if we inspect the same combination of components for culture media samples in Fig. 5f, the doses of 25 and 50 µM seem to be very similar, suggesting the induction of almost the same metabolomic effect on the cells, but that now can be detected indirectly in samples of culture media. The correlation between Comp. 1 vs. Comp. 3 at the same doses of 25 and 50 µM is now quite similar for cell extract (Fig. 5g) but more differentiated in the culture medium samples (Fig. 5h). Based on this behavior, we could argue that longer incubation times, 48 hours, and lower doses can impact the metabolism nearly in the same way as higher doses and shorter time of incubation, 24 hours. Similar changes were first detected in cell extracts and then in culture media of FN1 cells upon treatment with AgNP.
The PLS scores plots shown in Fig. 6a-h addressing cells from HUV-EC-C lineage reveal that genetic factor plays an essential role in the way cells metabolome respond to dose and incubation time with AgNP, as evidenced by the cytotoxic results. For instance, the combination Comp. 1 vs. Comp. 2 for 24 hours of treatment shows that the distribution centers of control groups and 50 µM are now close to each other, whereas 75 and 100 µM are more distant from the metabolome of cell extract samples (Fig. 6a). This situation changes for the culture medium samples and at the same time of exposition. The distributions of groups for cells treated with 50 and 75 µM are close to each other and more distinct from the control and 100 µM groups (Fig. 6b). Such a metabolic profile indicated that HUV-EC-C cells tend to be more resilient in changing the levels of their metabolites when treated with AgNP doses starting with 50 µM and 24 hours of incubation than the FN1 in the same experimental conditions. On the other hand, the proximity of the distribution centers of the 50 and 75 µM groups seems to indicate a different aspect of cell metabolism that can be now accessed from the data from the culture medium samples (Fig. 6b) compared with data from the cell extract samples (Fig. 6a).
In general, when we used statistical multivariate methods such as PCA and PLS-DA, this last one the choosed to deal our data, we present the axis in the graphics as PCx vs. PCy (abbreviation for principal components) for PCA method and Comp.x vs. Comp.y (just components) for PLS-DA. There is a difference between the methods related with the statistic used in each one: variances for PCA and co-variances for PLS-DA. Indeed, PCA has zero covariances and maximizes variances organized in descend order, so receiving the name de principal components for the axis in the scores plots. The axis in scores plot of PLS-DA do not have this denomination because it is calculated in another mathematical way, receiving just the name "component". Anyway, it is hard to explain, but is a general way to describe the axis in PLS results. Besides, all graphics of scores plot were built in this way. We already published one paper and it was gone alright.
The information brought by the combination of Comp. 1 vs. Comp. 3 is complementary to that revealed by the combination of Comp. 1 vs. Comp. 2, corroborating that the metabolites levels of control and cell exposed to 50 µM of AgNP can exhibit only minor differences (Fig. 6c), that are also present in the treated and control groups culture medium samples (Fig. 6d). Additionally, the exposition to both doses seems to impact nearly in the same way the metabolites levels in the cells extracts and those found in the culture medium, except for culture medium data of FN1 at the same incubation time that revealed that the 50 µM molar concentration impacted much more incisively the composition. Considering that all experimental parameters were kept the same for the cultures of both cell lineages, such differences were tentatively attributed to the genetic origin of cells.
Among all the VIP scores obtained in the multivariate data analysis, each one was ascribed to a specific metabolite. Only those with statistical significance (p < 0.05) were retained in the metabolic evaluations. We used parametric and nonparametric methods chosen according to the data normality distribution 27,28. Data were numerically tabulated, and heat maps were used to demonstrate the percentage shifts of increase (positive %) and decrease (negative %) in the levels of the significant VIP scores, i.e., the discriminant metabolites. For this purpose, the metabolites levels of the control samples were adopted as reference (no shift). The obtained results are summarized in Tables 2 and 3. Accordingly, those data provide a comprehensive impact of AgNP cytotoxicity on the endometabolome and exometabolome of the treated as compared with the respective control cells.
Moreover, the set of significantly impacted identified metabolites provide a clue on the possible metabolic pathways. It characterized the dynamic of biochemical transformations in the cells, demonstrating their metabolic response to the treatments with AgNP. For instance, it was possible to infer the cell lipid membrane composition and oxidative metabolism occurring in the mitochondria 21,22, as discussed below.